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 |
|---|---|---|---|---|---|
Retinol plays a vital role in the immune response to infection , yet proteins that mediate retinol transport during infection have not been identified . Serum amyloid A ( SAA ) proteins are strongly induced in the liver by systemic infection and in the intestine by bacterial colonization , but their exact functions remain unclear . Here we show that mouse and human SAAs are retinol binding proteins . Mouse and human SAAs bound retinol with nanomolar affinity , were associated with retinol in vivo , and limited the bacterial burden in tissues after acute infection . We determined the crystal structure of mouse SAA3 at a resolution of 2 Å , finding that it forms a tetramer with a hydrophobic binding pocket that can accommodate retinol . Our results thus identify SAAs as a family of microbe-inducible retinol binding proteins , reveal a unique protein architecture involved in retinol binding , and suggest how retinol is circulated during infection .
Retinol plays a vital role in the physiological response to microbial challenge . Retinol is derived from dietary vitamin A and can be converted enzymatically to retinoic acid , which complexes with nuclear receptors to regulate gene transcription programs in cells ( Germain et al . , 2006 ) . In this way , retinol promotes the maturation of innate immune cells ( Lawson and Berliner , 1999; Stephensen , 2001; Spencer et al . , 2014 ) , governs the differentiation of adaptive immune cells ( Mucida et al . , 2007; Hall et al . , 2011 ) , and facilitates the regeneration of epithelial barriers damaged by infection ( Osanai et al . , 2007 ) . A hallmark of vitamin A deficiency in humans is a markedly increased susceptibility to infection ( Sommer 2008 , Underwood , 2004 ) , underscoring the broad impact of retinol on immunity . As a small lipid-soluble compound , retinol cannot freely circulate but is instead transported among cells and tissues by specialized retinol binding proteins . Serum retinol binding protein ( RBP ) facilitates transport of retinol among the intestine , which is the site of retinol acquisition , the liver , which is the major site of retinoid storage , and other tissues that require retinol for their physiological functions ( Blaner , 1989 ) . Despite the increased requirement for retinol , serum RBP is markedly reduced following microbial challenge ( Rosales et al . , 1996 ) , leaving open the question of how retinol is transported among tissues during infection . Serum amyloid A ( SAA ) proteins are a family of proteins that are expressed in the intestinal epithelium ( Eckhardt et al . , 2010; Reigstad and Bäckhed , 2010 ) and liver ( Uhlar and Whitehead , 1999 ) and circulate in the serum ( Whitehead et al . , 1992 ) . SAA family members are encoded in the genomes of virtually all vertebrates and are highly conserved among species , suggesting essential biological functions ( Uhlar et al . , 1994 ) . Expression of SAAs is strongly induced by microbial exposure . SAAs are induced in intestinal epithelial cells by the microbiota ( Ivanov et al . , 2009; Reigstad et al . , 2009; Eckhardt et al . , 2010; Reigstad and Bäckhed , 2010 ) , and have been implicated in promoting Th17 cell development in response to specific microbiota components ( Ivanov et al . , 2009 ) . Similarly , liver and serum SAAs are markedly elevated following systemic bacterial or viral infection ( Meek and Benditt , 1986; Chiba et al . , 2009 ) . Although it has been proposed that SAAs generally contribute to inflammation and immunity ( Eckhardt et al . , 2010 ) , the exact functions of SAAs remain poorly defined . Interestingly , SAAs have characteristics that suggest they could bind hydrophobic ligands . First , all SAAs are predicted to form amphipathic helices with a hydrophobic face that could interact with non-polar molecules ( Stevens , 2004 ) . Second , SAAs circulate in the serum associated with high-density lipoprotein ( HDL ) , which transports lipid-bound lipoproteins amongst tissues ( Whitehead et al . , 1992 ) . However , the identity of potential SAA ligand ( s ) remains unclear . Here , we show that mouse and human SAAs are retinol binding proteins . We demonstrate that SAA expression in mice requires dietary vitamin A , that mouse and human SAAs bind tightly to retinol , and that SAA recovered from serum following bacterial infection is associated with retinol . We find that Saa1/2−/− mice , which harbor deletions of both the Saa1 and Saa2 genes , have higher bacterial burdens in spleen and liver following an acute bacterial infection , supporting an essential role for SAAs in the response to microbial challenge . Finally , we provide structural insight into the binding interaction by solving the mouse SAA3 crystal structure , which reveals a tetrameric assembly with a hydrophobic binding pocket that can accommodate retinol . These studies thus identify SAAs as a family of retinol binding proteins and reveal a new protein architecture supporting retinol binding . Our findings suggest that SAAs mediate retinol transport during microbial challenge and thus constitute a key component of the physiological response to infection .
Initially we uncovered a relationship between SAA expression and dietary vitamin A status in mice . Microarray analysis disclosed that mice fed a vitamin A-deficient diet exhibited lower abundances of serum amyloid A ( Saa ) 1 and 2 transcripts in the intestine as compared to mice fed a vitamin A-replete diet ( Figure 1—figure supplement 1; Table 1 ) . Real-time quantitative PCR and immunofluorescence analysis verified that expression of small intestinal SAA1 , 2 , and 3 was reduced in mice fed a vitamin A-deficient diet ( Figure 1A , B; Table 1 ) . Liver expression of SAA1 and 2 was also reduced in mice fed a vitamin A-deficient diet , although the reduction in expression was less pronounced than in the intestine ( Figure 1C , D ) . This is likely because dietary vitamin A deficiency does not completely deplete stored retinoids in the liver ( Liu and Gudas , 2005 ) . We also observed elevated expression of intestinal Saa1 and Saa2 following addition of retinol directly to the epithelial surface of small intestinal explants , and of liver Saa1 and Saa2 after intraperitoneal supplementation with retinoic acid ( Figure 1—figure supplement 2 ) . These findings support the idea that retinoids directly impact Saa expression . Addition of retinol or retinoic acid to cultured HepG2 cells ( a human liver cell line ) enhanced expression of SAA1 and 2 in the presence of IL-1β and IL-6 ( Figure 1E , F ) , suggesting that the impact of dietary vitamin A on SAA expression is due to cell-intrinsic effects of retinoids . Collectively , these results show that full expression of SAAs in the intestine and liver requires dietary vitamin A . 10 . 7554/eLife . 03206 . 003Table 1 . Primers used in Q-PCR analysisDOI: http://dx . doi . org/10 . 7554/eLife . 03206 . 003Primer namePrimer sequencemouse SAA1 F5ʹ-CATTTGTTCACGAGGCTTTCCmouse SAA1 R5ʹ-GTTTTTCCAGTTAGCTTCCTTCATGTmouse SAA2 F5ʹ-TGTGTATCCCACAAGGTTTCAGAmouse SAA2 R5ʹ-TTATTACCCTCTCCTCCTCAAGCAmouse SAA3 F5ʹ-CGCAGCACGAGCAGGATmouse SAA3 R5ʹ-CCAGGATCAAGATGCAAAGAATGhuman SAA1 F5ʹ-GGCATACAGCCATACCATTChuman SAA1 R5ʹ-CCTTTTGGCAGCATCATAGThuman SAA2 F5ʹ-GCTTCCTCTTCACTCTGCTCThuman SAA2 R5ʹ-TGCCATATCTCAGCTTCTCTGmouse 18S F5ʹ-CATTCGAACGTCTGCCCTATCmouse 18S R5ʹ-CCTGCTGCCTTCCTTGGAmouse Gapdh F5ʹ-TGGCAAAGTGGAGATTGTTGCCmouse Gapdh R5ʹ-AAGATGGTGATGGGCTTCCCGhuman Gapdh F5ʹ-CCTGGTCACCAGGGCTGCTTTTAAChuman Gapdh R5ʹ-GTCGTTGAGGGCAATGCCAGCC10 . 7554/eLife . 03206 . 004Figure 1 . SAA expression requires dietary vitamin A . ( A ) Mice were maintained on a normal ( Vit . A+ ) diet or a vitamin A-deficient ( Vit . A− ) diet as in Figure 1—figure supplement 1 . Ileal Saa expression was quantified by Q-PCR ( primer sequences are given in Table 1 ) . N = 3–5 mice per condition . ( B ) Ileal sections were stained with anti-SAA antibody ( ‘Materials and methods’ ) and anti-rabbit IgG-Cy3 ( red ) , and counterstained with DAPI ( blue ) . Scale bar = 50 µm . ( C ) Q-PCR determination of Saa expression levels in livers of mice on a normal or vitamin A-deficient diet . N = 5 mice/condition . ( D ) Liver sections were stained with anti-SAA antibody and anti-rabbit IgG-Cy3 ( red ) , and counterstained with DAPI ( blue ) . Scale bars = 50 µm . ( E and F ) Analysis of SAA expression in HepG2 cells . Cells were cultured in retinoid-free medium and then treated with IL-1β and IL-6 and/or 1 μM retinol ( E ) or 100 nM retinoic acid ( F ) . SAA expression was determined by Q-PCR . N = 3 independent experiments . Mean ± SEM is plotted . nd , not detected . *p < 0 . 05; **p < 0 . 01; ***p < 0 . 001 . p values were determined by two-tailed Student's t test . DOI: http://dx . doi . org/10 . 7554/eLife . 03206 . 00410 . 7554/eLife . 03206 . 005Figure 1—figure supplement 1 . Intestinal Saa1 and Saa2 are differentially regulated by dietary vitamin A . Affymetrix Mouse Genome 430 2 . 0 arrays were used to compare transcript abundance in small intestines from germ-free ( gf ) and conventional ( cv ) mice , and from mice fed a normal diet vs a vitamin A-deficient diet . Differentially expressed transcripts were identified as outlined in ‘Materials and methods’ , revealing 329 differentially-expressed genes between the germ-free and conventional groups , and 138 differentially-expressed genes between the vitamin A+ and vitamin A− groups . A Venn diagram representation of the experimental results is shown at left . 19 genes were differentially expressed in both comparisons and are displayed as a heatmap in which expression level is defined by Z-score ( defined in ‘Materials and methods’ ) . Saa1 and 2 are highlighted in blue . Other vitamin A-sensitive genes identified by this screen include granzymes A and B ( gene symbols: GzmA and GzmB ) . There is a known role for retinoic acid in CD8+ T cell differentiation ( Allie et al . , 2013 ) , which could explain the lowered abundance of granzyme transcripts in the vitamin A-deficient mice . Additionally , the array data reveal unexpected vitamin A-sensitive expression of three glycosyltransferases–fucosyltransferase 2 ( Fut2 ) , β1 , 6-N acetylglucosaminyltransferase ( Gcnt1 ) , and β1 , 3-galactosyltransferase 5 ( B3galt5 ) –that participate in mucin glycan synthesis ( Thomsson et al . , 2002; Brockhausen et al . , 2009 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03206 . 00510 . 7554/eLife . 03206 . 006Figure 1—figure supplement 2 . Retinoid supplementation stimulates Saa expression in intestine and liver . ( A ) Five centimeter explants from the distal small intestine ( ileum ) of vitamin A-replete mice were cultured for 6 hr in the presence of 0 . 1% DMSO or 1 µM retinol in 0 . 1% DMSO . Saa transcript abundance was determined by Q-PCR . N = 6 mice/condition . ( B ) Vitamin A-depleted mice were treated with retinoic acid administered by intraperitoneal injection daily over the course of three days . Saa transcript abundance was determined by Q-PCR . Note that injection of DMSO vehicle alone resulted in increased Saa expression . N = 4–17 mice/condition . Mean ± SEM is plotted . *p < 0 . 05 , **p < 0 . 01 . p values were determined by the Mann–Whitney test in ( A ) and two-tailed Student's t test in ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03206 . 006 Transcriptional control by retinoids is frequently observed in proteins that function in retinoid transport and metabolism ( Noy , 2000 ) . Because SAAs have predicted hydrophobic binding surfaces ( Stevens , 2004 ) and are induced by retinol , we hypothesized that they might be retinol binding proteins . We therefore tested for retinol binding activity of recombinant human SAA1 ( hSAA1 ) , mouse SAA1 ( mSAA1 ) , and mouse SAA3 ( mSAA3 ) using fluorometric binding assays that exploit the unique spectral properties of retinol . ( Note that we were unable to express recombinant mSAA2 ) . Retinol exhibits intrinsic fluorescence that is enhanced upon binding to proteins through energy transfer from tryptophan residues , and this fluorescence change can be used to quantify binding ( Cogan et al . , 1976 ) ( Figure 2A ) . We did fluorometric titrations to determine apparent dissociation constants ( Kds ) for the all-trans isomer of retinol , and extracted Kds of 259 , 169 , and 145 nM for retinol binding to hSAA1 , mSAA1 , and mSAA3 , respectively ( Figure 2B , E ) . These values are similar to binding affinities calculated for human serum retinol binding protein ( hRBP ) ( Cogan et al . , 1976 ) ( Figure 2—figure supplement 1 ) . Thus , SAAs bind retinol tightly , with affinities similar to that of a known retinol binding protein . 10 . 7554/eLife . 03206 . 007Figure 2 . Human and mouse SAAs bind retinol . ( A ) Retinol exhibits intrinsic fluorescence that is enhanced upon binding to proteins through energy transfer from tryptophan residues . All-trans-retinol was titrated into mSAA3 and fluorescence emission was monitored following excitation at 348 nm . The chemical structure of retinol is shown . ( B ) All-trans-retinol was titrated into hSAA1 , mSAA1 , mSAA3 , human transferrin ( hTfr; negative control ) , and apolipoprotein A1 ( ApoA1; negative control ) . Binding was quantified by monitoring retinol fluorescence at 460 nm following excitation at 348 nm as in ( A ) . Plots are representative of five independent experiments . ( C ) Retinoic acid lacks intrinsic fluorescence , but can quench intrinsic protein fluorescence due to energy transfer from tryptophan residues ( Cogan et al . , 1976 ) . All-trans-retinoic acid was titrated into mSAA3 and fluorescence quenching was monitored following excitation at 296 nm . The chemical structure of retinoic acid is shown . ( D ) All-trans-retinoic acid was titrated into hSAA1 , mSAA1 , mSAA3 , hTfr , and ApoA1 . Fluorescence emission was monitored at 334 nm with excitation at 296 nm as in ( C ) . Plots are representative of three independent experiments . ( E ) Kds were calculated from the binding assay data plotted in ( B ) and ( D ) and were derived from three independent experiments . nd , not determined . Additional ligand binding measurements are provided in Figure 2—figure supplements 1 and 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 03206 . 00710 . 7554/eLife . 03206 . 008Figure 2—figure supplement 1 . Retinol and retinoic acid binding to human retinol binding protein 4 ( hRBP4 ) . For comparison , we calculated binding affinities of retinol and retinoic acid to hRBP4 using the fluorescence binding assays described in Figure 2 . The results are consistent with published values ( Cogan et al . , 1976 ) and are similar to the binding affinities calculated for human and mouse SAAs in Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 03206 . 00810 . 7554/eLife . 03206 . 009Figure 2—figure supplement 2 . Additional ligand binding studies on human and mouse SAAs . ( A–C ) Retinyl acetate ( A ) , β-carotene ( B ) , and retinyl palmitate ( C ) were titrated into hSAA1 , mSAA1 , mSAA3 , hTfr ( negative control ) and ApoA1 ( negative control ) , and fluorescence quenching was monitored at 334 nm with excitation at 296 nm . Plots are representative of three independent experiments . The Kds are averages of the values derived from the three experiments . ( D ) Competitive inhibition of retinol binding by cholesterol was quantified . Saturating concentrations of retinol were added to hSAA1 , mSAA1 , and mSAA3 , and fluorescence quenching was monitored as in ( A–C ) . 10 μM cholesterol was added into the assay and inhibition of fluorescence quenching by retinol was monitored . Values are the average ± SEM of triplicate experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 03206 . 009 Retinoic acid lacks intrinsic fluorescence but can quench inherent protein fluorescence due to energy transfer from tryptophan residues ( Cogan et al . , 1976 ) . We therefore measured retinoic acid binding using a modified fluorescence assay that monitored quenching of protein fluorescence ( Figure 2C ) . Titration of all-trans retinoic acid yielded Kds of 268 and 224 nM for retinoic acid binding to hSAA1 and mSAA3 , respectively ( Figure 2D , E ) , which are similar to binding affinities calculated for human RBP binding to retinoic acid ( Cogan et al . , 1976 ) ( Figure 2—figure supplement 1 ) . There was weak binding of retinoic acid to mSAA1 and we were unable to calculate a Kd for the interaction ( Figure 2D , E ) . Thus , while hSAA1 and mSAA3 bind both retinol and retinoic acid , mSAA1 selectively binds retinol . mSAA1 also showed weak binding to other retinoids , including β-carotene and retinyl acetate , while hSAA1 bound these compounds with Kds of 497 and 347 nM , respectively , and mSAA3 bound β-carotene with a Kd of 159 nM ( Figure 2—figure supplement 2A , B ) . All SAA isoforms bound weakly to retinyl palmitate ( Figure 2—figure supplement 2C ) . Since long chain retinyl esters ( such as retinyl palmitate ) are the major form of stored retinoid in the liver ( Vogel et al . , 1999 ) , this suggests that SAAs do not transport retinoids for storage . Although a role for SAAs in cholesterol transport and metabolism has been proposed ( van der Westhuyzen et al . , 2005 ) , we found that cholesterol was unable to competitively inhibit retinol binding to SAAs ( Figure 2—figure supplement 2D ) . To test whether SAAs also associate with retinol in vivo , we sought to purify SAAs from mouse tissues and assay for the presence of associated retinol . SAAs were difficult to purify from the mouse intestine due to the presence of large amounts of contaminating protein , even under conditions where expression of SAAs was maximally induced . However , SAAs constitute a high proportion of serum protein during acute systemic infection ( McAdam and Sipe , 1976; Zhang et al . , 2005 ) . We were therefore able to use size exclusion chromatography to recover a SAA-enriched fraction from the sera of mice infected intraperitoneally with Salmonella typhimurium for 24 hr ( Figure 3—figure supplement 1A–C ) . Mass spectrometry revealed that the SAA-enriched protein fraction was devoid of other known retinol binding proteins ( Figure 3—figure supplement 1D ) . Liquid chromatography tandem mass spectrometry ( LC-MS/MS ) indicated the presence of retinol in the SAA-enriched fraction ( Figure 3 , Figure 3—figure supplement 2A–C ) in a molar ratio of ∼1 mol retinol/4 mol SAA ( Figure 3 , inset ) . In these analyses , retinoic acid was not detected , and retinol was not detected in the equivalent serum fraction recovered from Saa1/2−/− mice ( Eckhardt et al . , 2010 ) ( Figure 3 ) , suggesting that the retinol was preferentially associated with SAA . 10 . 7554/eLife . 03206 . 010Figure 3 . Serum SAA is associated with retinol in vivo . Wild-type or Saa1/2−/− mice were infected intraperitoneally with S . typhimurium and serum was collected 24 hr later . The serum was fractionated by size exclusion chromatography and a major SAA-containing fraction from wild-type mice was identified by Western blot ( Figure 3—figure supplement 1 ) . The SAA-containing fraction and the equivalent serum fraction from Saa1/2−/− mice were hexane-extracted and analyzed by LC-MS/MS against retinol and retinoic acid standards . Additional support for the identification of retinol is provided in Figure 3—figure supplement 2 . The LC-MS/MS chromatograms of daughter ion 93 are shown . mol SAA/mol retinol is shown in the inset . Data are representative of duplicate experiments with triplicate samples in each experiment . DOI: http://dx . doi . org/10 . 7554/eLife . 03206 . 01010 . 7554/eLife . 03206 . 011Figure 3—figure supplement 1 . Size-exclusion chromatography and mass spectrometry analysis of SAA-containing serum fractions . ( A ) Wild-type and Saa1/2−/− mice were challenged intraperitoneally with S . typhimurium and serum was collected 24 hr later . The serum was concentrated and the proteins were fractionated into two major peaks by size exclusion chromatography on a Superdex 75 HiLoad 16/60 column . ( B ) SAA was detected in peak 1 by Western blot . ( C ) SDS-PAGE and Western blot analysis of serum peak 1 . SDS-PAGE reveals four major protein bands that correspond to SAA bands as determined by Western blot . ( D ) Mass spectrometry analysis was performed to identify other proteins in serum peak 1 . No other retinoid binding proteins were detected in the peak . DOI: http://dx . doi . org/10 . 7554/eLife . 03206 . 01110 . 7554/eLife . 03206 . 012Figure 3—figure supplement 2 . Mouse SAA is associated with retinol in the serum following infection . ( A–C ) Wild-type mice were infected by intraperitoneal delivery of S . typhimurium and serum was collected and processed for retinoids as described in ‘Materials and methods’ . ( A ) LC-MS profiles of retinol standard and the SAA-enriched fraction from wild-type mouse serum , prepared as described in Figure 3—figure supplement 1 . The retinol peaks were further analyzed by spectroscopy and display maxima at 325 nm , which is characteristic of retinol . ( B and C ) Retinoids were extracted from unprocessed serum ( B ) or from the SAA-enriched serum fraction ( C ) of S . typhimurium infected wild-type mice and analyzed by LC-MS/MS . Retinol was detected by analyzing two daughter ions ( 93 , 119 ) . The inset in the top left panel shows the derivation of the daughter ions from the parent retinol . We noted a modest decrease in the elution time of retinol between the experiments shown in ( B ) and ( C ) ; however , this difference was observed in both the experimental sample and the retinol standard . DOI: http://dx . doi . org/10 . 7554/eLife . 03206 . 012 SAAs lack sequence homology to the two known families of retinol binding proteins: cellular retinol binding proteins ( CRBP ) and serum retinol binding proteins ( RBP ) ( Blaner , 1989; Noy , 2000 ) . Thus , the three-dimensional structures of CRBP and RBP proteins ( Newcomer et al . , 1984; Cowan et al . , 1993 ) provide no direct insight into the structural basis for retinol binding by SAAs . To understand how SAAs bind retinol , we therefore determined the three-dimensional structure of recombinant mSAA3 by X-ray crystallography . The protein was crystallized in a P62 space group with two subunits in the asymmetric unit , and the structure was determined to a resolution of 2 Å by single-wavelength anomalous dispersion ( SAD ) phasing using a selenomethionyl-derivatived crystal ( Figure 4A; Table 2 ) . The crystal structure reveals that mSAA3 is highly α-helical ( Figure 4A ) , as predicted on the basis of its primary sequence ( Figure 4—figure supplement 1 ) ( Stevens , 2004 ) . The structure is very similar to the recently determined structure of human SAA1 . 1 ( Lu et al . , 2014 ) , an isoform that has a marked tendency to form pathogenic amyloid fibrils ( Yu et al . , 2000 ) . Like the SAA1 . 1 structure , the mSAA3 structure consists of four α-helices , designated α1-4 from the N- to the C-termini , forming a cone-shaped four-helix bundle with a comparatively longer α1 . The helices form two sets of antiparallel helices , α1-α2 and α3-α4 , connected by a very short loop ( Figure 4A ) . The monomer is stabilized by an extensive network of hydrogen bonding interactions among conserved residues and water molecules in the interior of the monomer . As in the SAA1 . 1 structure , the C-terminal tail wraps around the helix bundle making a number of hydrogen bonding interactions that add to monomer stability , underscoring the importance of the C-terminal tail . 10 . 7554/eLife . 03206 . 013Figure 4 . Structure of mSAA3 and molecular contacts within the tetrameric unit . ( A ) Structure of the mSAA3 monomer ( side view ) , with helices and termini labeled . ( B ) Top view of the tetrameric mSAA3 structure . Chains forming dimer pairs are colored cyan and magenta . In ( C ) , helices α1-4 and the N- and C-termini of two monomers are labeled . Residues that make dimer contacts are shown as green sticks while residues involved in tetramer stabilization are shown as yellow sticks . ( D and E ) Magnified regions of a dimer interface ( D ) and tetramer interface ( E ) are shown . Views are slightly rotated so that the interactions can be clearly visualized . Crystal structure data collection and refinement statistics are provided in Table 2; alignments of mouse and human SAAs are shown in Figure 4—figure supplement 1; parameters from the protein Interfaces , Surfaces , and Assemblies ( PISA ) analysis showing a tetrameric state are provided in Table 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 03206 . 01310 . 7554/eLife . 03206 . 014Figure 4—figure supplement 1 . Sequence alignment of mouse and human SAAs . Secondary structure is based on the mSAA3 crystal structure and is indicated above the sequence . α-helices are shown as solid red bars while loops and non-helical secondary structure are marked as solid black lines . DOI: http://dx . doi . org/10 . 7554/eLife . 03206 . 01410 . 7554/eLife . 03206 . 015Table 2 . Crystal structure data collection and refinement statisticsDOI: http://dx . doi . org/10 . 7554/eLife . 03206 . 015Data collection Space groupP62 Cell dimensions ( Å ) a = b = 78 . 33 , c = 62 . 32α = β = 90o , γ = 120o Wavelength ( Å ) 0 . 9794 Rsym or Rmerge ( % ) 8 . 4 Resolution ( Å ) *50–2 . 05 ( 2 . 09–2 . 05 ) I/σI19 . 19 ( 3 . 23 ) Completeness ( % ) 99 . 8 ( 97 . 3 ) Redundancy6 . 2 ( 5 . 4 ) Refinement No . reflections12 , 206 Resolution ( Å ) *39 . 17–2 . 06 ( 2 . 14–2 . 06 ) Rwork/Rfree0 . 17/0 . 21 ( 0 . 16/0 . 19 ) No . atoms Protein1608 Ligand/ion3 Water61 R . m . s . deviations Bond lengths ( Å ) 0 . 0077 Bond angles ( ° ) 0 . 932*Highest resolution shell is shown in parenthesis . Size exclusion chromatography and cross-linking experiments showed that mSAA3 forms a tetramer in solution ( Figure 5A , B ) . Consistent with these findings , analysis of the mSAA3 crystal structure using the Protein Interfaces , Surfaces and Assemblies ( PDBePISA ) server ( http://www . ebi . ac . uk/msd-srv/prot_int/ ) yielded a tetrameric quaternary structure ( Figure 4B; Table 3 ) . This is in contrast to the hexameric structure derived for SAA1 . 1 ( Lu et al . , 2014 ) . There are several potential reasons for the discrepancy in the oligomeric structures . First , it has been suggested that different SAAs can adopt different oligomeric states ( Wang et al . , 2002 , 2011 ) . SAA1 . 1 was also observed to produce a ∼43 kDa species in solution ( Lu et al . , 2014 ) , suggesting that SAA1 . 1 may in part adopt a tetrameric state . Second , SAA1 . 1 has a more hydrophobic N-terminus than mSAA3 , which is thought to be a determinant of amyloidogenicity ( Yu et al . , 2000; Lu et al . , 2014 ) . Third , the crystallized SAA1 . 1 protein retained the hexa-histidine tag ( Lu et al . , 2014 ) , which may have contributed to the difference in oligomeric state . 10 . 7554/eLife . 03206 . 016Figure 5 . Mouse SAA3 is tetrameric in solution . ( A ) Size exclusion chromatography profile of purified mouse SAA3 on a Superdex 75 10/300 GL column . Elution of standards ( BioRad ) is shown in the top panel and elution of SAA3 is shown in the bottom panel . mSAA3 elutes at a position consistent with a tetramer ( monomer is 12 . 2 kDa ) . ( B ) Cross-linking analysis of mSAA3 . Purified mSAA3 was cross-linked with glutaraldehyde and analyzed by SDS-PAGE . DOI: http://dx . doi . org/10 . 7554/eLife . 03206 . 01610 . 7554/eLife . 03206 . 017Table 3 . Parameters from the Protein Interfaces , Surfaces , and Assemblies ( PISA ) analysisDOI: http://dx . doi . org/10 . 7554/eLife . 03206 . 017ParameterValueMultimeric state4CompositionA2B2Dissociation pattern2 ( AB ) Surface area , Å219485 . 8Buried area , Å26125 . 7ΔGintrinsic , kcal/mol−79 . 8ΔGdiss , kcal/mol11 . 1TΔSdiss , kcal/mol12 . 6 The mSAA3 tetramer is formed by two sets of tightly associated dimers ( Figure 4B , C ) . The dimers pack against each other through an aromatic interface formed by W71 , as well as a non-polar interaction involving V75 residues ( Figure 4C , E ) . The dimer is formed by two identical chains oriented pseudo-anti-parallel to each other , and is held together by two pairs of tight hydrogen bond interactions between K74 and D78 residues on oppositely oriented α3 helices ( Figure 4C , D ) . The interaction is also supported by a set of aromatic interactions between F99 and W103 of the α4 helices ( Figure 4C , D ) . These interactions result in tightly held α3 helices composed primarily of non-polar residues , thus forming the inside hollow pocket of the tetramer . The hollow pocket is surrounded by the remaining helices , creating a hydrophobic interior that is protected from the external aqueous environment . Retinol is a highly apolar lipid-like molecule consisting of a β-ionone ring , an isoprenoid tail , and a hydroxyl group . Thus , it requires a non-polar environment for transport among tissues and within cells . Structures of known retinol binding proteins , including serum RBP , indicate that these proteins consist mainly of β-sheet secondary structures forming a β-barrel tertiary structure , with the retinol molecule held in an interior non-polar binding pocket ( Newcomer et al . , 1984 ) . In contrast , mSAA3 is α-helical and oligomerizes to form a hollow , largely non-polar interior that could serve as a binding pocket for a non-polar small molecule ( Figure 6A ) . We were unable to obtain mSAA3 crystals with the bound retinol ligand as retinol is highly unstable ( Barua and Furr , 1998 ) and the crystals required several weeks to grow . However , a ligand docking analysis using SwissDock ( Grosdidier et al . , 2011 ) indicated that retinol can be docked in this hydrophobic pocket with favorable free energy ( ∼ −7 kcal/mol ) and FullFitness values ( Zoete et al . , 2010 ) ( ∼ −2400 kcal/mol ) ( Figure 6B–D ) . Consistent with this prediction , introducing a Trp71Ala ( W71A ) mutation in the mSAA3 hydrophobic core reduced the affinity of mSAA3 for retinol ( Figure 6E ) . Thus , the mSAA3 structure supports our biochemical data showing a retinol binding function for SAAs and explains how mSAA3 could bind retinol . 10 . 7554/eLife . 03206 . 018Figure 6 . The mSAA3 tetramer forms a hollow hydrophobic binding pocket that can accommodate retinol . ( A ) A surface rendering of the tetramer showing the interior cavity , with the electrostatic potential displayed using a color gradient ranging from negative ( red ) to neutral ( white ) to positive ( blue ) . The orientation is similar to that in Figure 4B . ( B–D ) Different views of a retinol molecule docked in the putative ligand-binding pocket . ( B ) A semi-transparent surface representation of the protein in the same orientation as Figure 4B , with a cartoon trace . Retinol atoms are represented as green spheres . The views in ( C ) and ( D ) are rotated by approximately 90° in the horizontal plane relative to ( A ) and ( B ) , and ( B ) is rotated by approximately 90° in the vertical plane relative to ( C ) . In ( D ) , a surface model of the protein is shown , sliced close to the binding pocket . Retinol atoms are shown as sticks . ( E ) Wild-type or Trp71Ala ( W71A ) mutant mSAA3 was assayed for retinol binding as described in Figure 2 . Representative plots and Kds were calculated from the binding assay data and were derived from five independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 03206 . 018
Numerous biochemical , physiologic , and epidemiologic studies have shown that vitamin A and its derivative retinol are essential for the development of robust immunity ( Stephensen , 2001 ) . Retinol and retinoic acid are critical for the development of innate and adaptive immunity ( Lawson and Berliner , 1999; Mucida et al . , 2007; Hall et al . , 2011; Spencer et al . , 2014 ) , and also promote maintenance and repair of epithelial barriers ( Osanai et al . , 2007 ) . However , a prominent response to acute infection is the marked decline in serum RBP ( Rosales et al . , 1996 ) , which paradoxically occurs at a time of increased demand for retinol to support development of immunity and barrier defense . Thus , it has been unclear how retinol is transported among tissues following an acute microbial challenge . We propose that SAAs fulfill this role , supported by several lines of evidence . First , SAAs are strongly induced by microbial exposure at sites of retinol uptake ( intestine ) and storage ( liver ) , and are present at high levels in the circulation following microbial challenge ( Chiba et al . , 2009; Ivanov et al . , 2009; Reigstad et al . , 2009 ) . Second , we have shown that SAAs bind retinol at nanomolar affinity in vitro , and that serum SAAs circulate in association with retinol . Third , the three dimensional structure of mSAA3 exhibits a hollow hydrophobic binding pocket , providing structural insight into how SAAs bind retinol . Although the precise tissue targets of circulating retinol-bound SAAs remain under investigation , several observations support the idea that SAAs promote immunity to infection . First , Saa1/2−/− mice exhibit increased susceptibility to chemically-induced colitis in mice ( Eckhardt et al . , 2010 ) , suggesting that SAAs contribute to intestinal immunity . Second , studies in zebrafish show that commensal microbiota stimulate neutrophil migration through induction of SAA ( Kanther et al . , 2013 ) . Third , we found that intraperitoneal infection of Saa1/2−/− mice with S . typhimurium resulted in higher bacterial loads in liver and spleen as compared to wild-type mice ( Figure 7A , B ) , suggesting that SAAs also contribute to systemic immunity . 10 . 7554/eLife . 03206 . 019Figure 7 . Saa1/2−/− mice have higher bacterial burdens following S . typhimurium infection . 10 week old wild-type and Saa1/2−/− mice were inoculated intraperitoneally with 10 , 000 cfu of S . typhimurium . Livers ( A ) and spleens ( B ) were collected after 24 hr and analyzed for bacterial counts by dilution plating . Combined results from two independent experiments are shown . Each point represents one mouse and geometric means are indicated . Dotted line indicates limit of detection . **p < 0 . 01 using the Mann–Whitney test . DOI: http://dx . doi . org/10 . 7554/eLife . 03206 . 019 SAA4 is an SAA isoform that is expressed in the livers of healthy , non-infected mice and humans ( de Beer et al . , 1991 , 1994 , 1995 ) . SAA4 circulates at concentrations that are markedly lower than those observed for SAA1 and 2 following acute infection ( de Beer et al . , 1995 ) but are similar to the concentrations of RBP in uninfected individuals ( Willett et al . , 1985; Friedman et al . , 1986 ) . SAA4 is 54–56% homologous to SAA1 , 2 , and 3 , and retains the hydrophobic amino acids that are predicted to line the hydrophobic binding pocket in SAA3 . Further , homology modeling using the mouse SAA3 structure and the mouse SAA4 sequences yields a SAA4 model with a highly similar predicted structure ( alignment score of 0 . 1 and a Global Model Quality Estimate of 0 . 73 ) . Thus , the possibility that SAA4 is a retinol binding protein that functions to transport retinol in healthy , non-infected animals will be a subject for future investigation . SAA1 , 2 , and 3 are markedly induced in the intestinal epithelium by the microbiota ( Ivanov et al . , 2009; Eckhardt et al . , 2010; Reigstad and Bäckhed , 2010 ) , and thus our findings may provide insight into how the intestinal microbiota regulates host immunity and inflammation . Previous studies have suggested that SAAs promote Th17 cell development in response to specific components of the microbiota , such as segmented filamentous bacteria ( Ivanov et al . , 2009 ) . Consistent with a function for SAAs in retinol binding and transport , retinol/retinoic acid is required to elicit Th17 cell responses to infection and mucosal vaccination ( Hall et al . , 2011 ) . A key question is whether there are tissue-specific effects of intestinal epithelial SAAs , or whether the intestinal SAAs enter the circulation with bound retinol acquired directly from the diet . For example , intestinal epithelial SAAs could be involved in the direct delivery of retinol from epithelial cells to underlying immune cells in the lamina propria , or from epithelial cells to mucosal lymphoid tissues . Altogether , our results provide insight into the biological function of SAAs , reveal a new protein architecture that supports retinol binding , and suggest how retinol is transported among cells and tissues during infection . These findings may prove useful in designing new strategies for enhancing resistance to infection and/or controlling inflammation during disease .
C57BL/6 wild-type mice were maintained in the barrier at the University of Texas Southwestern Medical Center . Saa1/2−/− mice were obtained from Dr Frederick C de Beer ( Eckhardt et al . , 2010 ) at the University of Kentucky and were maintained in the barrier at the University of Texas Southwestern Medical Center . 6–12 weeks old mice were used for all experiments . Experiments were performed using protocols approved by the Institutional Animal Care and Use Committees of the UT Southwestern Medical Center . Recombinant mSAA1 and 3 were expressed and purified as described below . hSAA1 protein was from PeproTech ( Rocky Hill , NJ ) and resuspended as recommended . The protein is a consensus SAA molecule corresponding to human apo-SAA1α except for the presence of an N-terminal methionine and substitution of asparagine for aspartic acid at position 60 and arginine for histidine at position 71 . Anti-SAA antiserum was raised against purified recombinant mSAA1 . Retinol , retinoic acid , β-carotene , retinyl acetate , retinyl palmitate , and cholesterol were from Sigma-Aldrich ( St . Louis , MO ) and were reconstituted into ethanol , DMSO , or dioxane , depending on the experiment . IL-1β and IL-6 were from Invitrogen . Vitamin A-deficient ( TD . 09838 ) and control ( ∼20 , 000 IU vitamin A/kg; TD . 09839 ) diets were purchased from Harlan Laboratories ( South Easton , MA ) . At day 10 of gestation , pregnant females were placed on the standard diet or the vitamin A-deficient diet ( Hall et al . , 2011 ) . Mothers and pups were maintained on the diets until weaning , and pups stayed on the diet for two additional months prior to sacrifice . The protocol was adapted from a previously described procedure ( Hall et al . , 2011 ) . A total of 250 μg of all-trans-retinoic acid ( Sigma-Aldrich ) was resuspended in 30 μl of biotechnology performance certified DMSO ( Sigma-Aldrich ) . The suspension was administered daily by intraperitoneal injection to vitamin A-deficient mice over the course of 3 days . 24 hr following the third injection , mice were sacrificed and tissues were harvested . Control mice received an injection of the DMSO vehicle . Terminal ileum ( 5 cm ) was collected from mice post-sacrifice and flushed with a solution of phosphate-buffered saline with penicillin ( 100 units/ml ) and streptomycin ( 100 μg/ml ) . Ileal segments were cultured on equilibrated cell culture plate inserts ( PIHA03050; Millipore ) at 37°C and 95% oxygen for 6 hr in Dulbecco's modified Eagle's medium ( 4 g/l glucose and L-glutamine; Invitrogen , Carlsbad , CA ) supplemented with 10% charcoal-stripped heat-inactivated fetal bovine serum ( Gibco , Carlsbad , CA ) , 10% NCTC135 media ( Sigma ) , 25 mM HEPES , 100 units/ml penicillin , 100 μg/ml streptomycin , and either 0 . 1% DMSO or 1 μM retinol in 0 . 1% DMSO . After culture for 6 hr , segments were flash-frozen and processed for total RNA extraction . Total RNAs were isolated from mouse ileum using the Qiagen Midi-Prep RNA isolation kit . For each condition , RNA was isolated from two independent groups of five to eight mice . The RNAs in each group were pooled and used to generate biotinylated probes for microarray analysis . Probes were hybridized to Affymetrix Mouse Genome 430 2 . 0 GeneChips in the University of Texas Southwestern Microarray Core . To identify genes that are differentially expressed between germ-free and conventional mice , we performed two-way comparisons between germ-free and conventional groups , with germ-free samples designated as baseline . Raw data were imported into Affymetrix ( Santa Clara , CA ) GeneChip software for analysis , and previously established criteria were used to identify differentially expressed genes ( Cash et al . , 2006 ) . Briefly , a twofold difference was considered significant if three criteria were met: ( 1 ) the GeneChip software returned a difference call of increased or decreased; ( 2 ) the mRNA was called present by GeneChip software in either germ-free or conventional samples; and ( 3 ) the difference was observed in duplicate microarray experiments . We performed a similar analysis to identify genes that are differentially regulated between mice fed a normal diet vs those fed a vitamin A-deficient diet . Finally , we identified 19 genes that were differentially regulated by colonization status and by dietary vitamin A content . Signal intensity data for this group of 19 genes were converted to Z-scores ( z = ( x − μ ) /σ , where x = signal intensity , μ = mean signal intensity for all samples , and σ = SD across all samples ) , which were visualized as heatmaps using Java TreeView software . Total RNA was isolated from homogenized tissues or cells using the Qiagen RNeasy RNA isolation kit . Random primed cDNAs were assayed by SYBR Green-based real-time PCR using SAA-specific primers as given in Table 1 . Signals were normalized to 18S rRNA or Gapdh . Zinc-fixed , paraffin embedded tissue sections were stained with anti-SAA antiserum raised against purified recombinant mSAA1 and detected using a goat anti-rabbit IgG Cy3 conjugate ( Biomeda ) . Tissues were counterstained with DAPI and images were captured on a Zeiss AxioImager M1 Microscope . HepG2 cells were purchased from ATCC . Cells were maintained in 1X DMEM , 10% FBS ( or charcoal stripped FBS ) , 1X Penstrep , 1X glutamax , and 1X sodium pyruvate . Cells were maintained at 5% CO2 . Prior to addition of retinoids , the cells were grown overnight in DMEM containing 10% charcoal-stripped FBS ( to removes retinoids ) and were treated with 1 μM retinol or 100 nM retinoic acid , 10 ng/ml of IL-1β , and 10 ng/ml IL-6 . Genes encoding mouse SAA1 and SAA3 ( minus the signal sequence ) were cloned into the pET28 ( a ) + expression vector between NdeI and BamHI restriction endonuclease sites , with an N-terminal hexa-histidine tag followed by a thrombin cleavage site and a C-terminal stop codon . Proteins were expressed in Escherichia coli BL21-CodonPlus ( DE3 ) -RILP cells ( Stratagene , La Jolla , CA ) by induction with 0 . 4 mM isopropyl-β-D-galactoside ( IPTG ) for ∼3 hr at 25°C for mSAA1 , and at 37°C for mSAA3 . Cells were harvested by centrifugation at 4500×g for 25 min at 4°C and re-suspended in lysis buffer ( 50 mM NaH2PO4 , 500 mM NaCl , 10 mM imidizole for mSAA1 and 500 mM NaCl , 50 mM Tris pH 8 . 0 , 10 mM imidazole , 15 mM β-mercaptoethanol for mSAA3 ) . After sonication , decyl maltopyranoside ( DM ) ( Avanti Polar Lipids , Alabaster , AL ) was added to a final concentration of 40 mM and incubated for ∼3 hr at 4°C . The mixture was pelleted by centrifugation at 10 , 000×g for 30 min and the supernatant loaded onto a Ni2+ metal affinity column ( Qiagen , Valencia , CA ) pre-equilibriated with 4 mM DM in lysis buffer . Non-specific contaminants were washed away with 25 mM imidazole in DM buffer and the protein was eluted in DM buffer containing 300 mM imidazole . All subsequent buffers do not include detergent in order to completely remove detergent . The eluate was desalted with a HiTrap desalting column ( GE Life Sciences , Pittsburgh , PA ) into a 200 mM NaCl buffer . Thrombin ( Roche , Basel , Switzerland ) was then added ( ∼1 unit/1 . 2 mg protein ) and incubated overnight at 4°C . Undigested protein was removed by passing the overnight digest over Ni2+ affinity matrix , collecting only the flow through . The eluate was concentrated in a 3 K cutoff Amicon Ultra centrifugal device ( Millipore , Billerica , MA ) and further purified by size exclusion chromatography on either a HiLoad Superdex 200 or a Superdex 75 ( 10/30 ) column ( GE Life Sciences , Pittsburgh , PA ) , in 100 mM NaCl , 20 mM Tris pH 8 , 15 mM a β-mercaptoethanol and 5% glycerol . For crystallization , mSAA3 was concentrated to ∼3 mg/ml . The mSAA3-W71A mutant was generated using the QuickChange II site-directed mutagenesis kit ( Agilent Technologies , Santa Clara , CA ) . Protein expression , purification , and retinol binding assays were done as for wild-type mSAA3 . Steady state fluorescence was measured using a QuantaMaster 40 spectrofluorometer ( Photon Technology International , Edison , NJ ) and FelixGX software program . For retinol titrations , samples were excited at 348 nm and emissions monitored at 460 nm . For retinoic acid and other retinoids , samples were excited at 296 nm and tryptophan quenching was monitored by emission at 334 nm . Experiments on mSAA1 and mSAA3 were done in 25 mM Tris pH 8 . 0 , 100 mM NaCl , and 4 mM decyl maltopyranoside ( DM ) . Experiments with hSAA1 , human ApoA1 , and human transferrin were conducted in PBS . For cholesterol competition assays , saturating concentrations of retinol were added to hSAA1 , mSAA1 , and mSAA3 , and fluorescence quenching was monitored by emission at 334 nm . 10 μM cholesterol was added to the assay and inhibition of fluorescence quenching by retinol was monitored . All assays were done using a protein concentration of 0 . 5 μM . Salmonella enterica Serovar Typhimurium ( SL1344 ) was grown overnight in Luria Broth at 37°C . Mice were infected intraperitoneally with 1 × 104 organisms per mouse . Mice were sacrificed after 24 hr and tissues and serum were collected for experiments . Serum was pooled from 3–5 mice and 500 μl was separated by size-exclusion chromatography on a Superdex 75 HiLoad 16/60 column ( GE Life Sciences ) . Peak fractions were analyzed by SDS-PAGE and stained with Coomassie Blue . Duplicate samples were analyzed by Western blot with anti-SAA antibody to identify peak fractions containing SAA protein . Retinoid extraction was modified and scaled from a previously described procedure ( McClean et al . , 1982 ) . SAA-containing fractions purified by size exclusion chromatography were pooled , added to an equal volume of 1:1 1-butanol:acetonitrile , and vortexed for 60 s . 20 μl of 20 . 6 M K2HPO4 was added for each 1 ml of pooled fractions . Samples were then vortexed 30 s and 5 ml of hexane per 1 ml sample was added . Samples were vortexed for another 30 s and centrifuged at 1 , 000×g for 5 min and the top organic phase was dried in a nitrogen evaporator ( Organomation Associates , Berlin , MA ) . Samples were prepared the day before the assay and stored at 80°C . Standard solutions were resuspended in ethanol and prepared fresh for every use . Standard curves were generated by spiking retinol or retinoic acid into 1 ml of 20 mM Tris pH 8 . 0 , 100 mM NaCl and processed as for serum samples . Samples were resuspended in 200 μl of acetonitrile before injection . Compound levels were monitored by LC-MS/MS on an AB/Sciex ( Framingham , MA ) 4000 Qtrap mass spectrometer coupled to a Shimadzu ( Columbia , MD ) Prominence LC after a 20 μl injection . The compounds were detected using electrospray ionization ( ESI ) with the mass spectrometer in MRM ( multiple reaction monitoring ) mode by following the precursor to fragment ion transition 269 . 2 → 93 . 1 and 269 . 2 → 119 for retinol ( pos . mode; [M-H2O]+ ) and 301 . 2 → 123 . 1 for retinoic acid ( pos . mode; M+H+ ) . An Agilent ( Santa Clara , CA ) Eclipse XDB C18 column ( 150 × 4 . 6 mm , 5 micron packing ) was used for chromatography with the following conditions: mobile phase A: acetonitrile:methanol:H2O:formic acid ( 55:33:12: . 01 ) ; mobile phase B: acetonitrile:formic acid ( 100:0 . 01 ) . Over a total run time of 18 min , the following gradient was applied: 0 to 3 min 50% B; 3 to 10 min gradient to 100% B; 10 to 17 min 100% B; 17 to 18 min gradient to 50% B . Stoichiometries of the SAA-retinol association were determined by quantifying serum retinol and serum SAA . Total serum retinol was calculated based on peak areas from the mass spectrometer analysis in samples compared to a retinol standard curve . Serum SAA was quantified by Western blot analysis with anti-SAA antiserum and densitometry . Crystals were grown by sitting-drop vapor diffusion at 20°C by mixing equal volumes of protein and reservoir . An initial hit was obtained in 30–40% 2-methyl-2 , 4-pentanediol ( MPD ) , 0 . 1 M sodium acetate pH 4 . 5 after more than a month . Further refinement yielded better crystals at 75–80% MPD . Crystals were directly flash frozen , the MPD serving as a cryoprotectant . Crystals were of space group P62 with cell dimensions a = 78 . 327 Å , c = 62 . 319 Å and two subunit copies in the asymmetric unit . Selenomethionyl-derivatived crystals were grown the same way as the native protein crystals , except the culture media used was defined media with selenomethionine additive ( Molecular Dimensions , Altamonte Springs , FL ) and purification buffers after the last Ni2+ column elution contained 10 mM DTT ( instead of β-mercaptoethanol ) and 0 . 5 mM EDTA . Data were collected at 100 K under a nitrogen gas stream at the Advanced Photon Source ( APS ) beamlines 19ID or 23IDD of the Argonne National Laboratory . Single-wavelength anomalous dispersion ( SAD ) data were collected at the selenium K edge ( 0 . 9793 Å ) . Diffraction data were processed with the HKL2000/3000 package ( Otwinowski and Minor , 2013 ) . Heavy atom substructure as well as initial phases were obtained using the SAD pipeline in the PHENIX crystallographic software package ( Afonine et al . , 2012 ) . This was followed by manual model building in Coot ( Emsley et al . , 2010 ) interspersed with iterative rounds of rigid body , simulated annealing and individual isotropic B-factor refinement and finally TLS refinement in PHENIX . The structure was determined to a resolution of 2 Å . Data collection and refinement statistics are summarized in Table 2 . Glutaraldehyde ( Sigma ) was added to varying final concentrations ( 0 . 05% , 0 . 005% , and 0 . 0005% wt/vol ) to purified mSAA3 ( 0 . 5 mg/ml in PBS ) . The reaction mixtures were incubated for ∼30 min on ice , quenched with 0 . 1 M Tris pH 8 , and analyzed by SDS-PAGE . Statistical differences were calculated by the unpaired two-tailed Student's t test or Mann–Whitney test using GraphPad Prism software . Results are expressed as the mean ± standard error of the mean ( SEM ) . | Vitamins are nutrients that organisms require in order to survive and grow . If an organism is unable to synthesize a vitamin in sufficient quantities , it is essential that it obtain the vitamin through its diet instead . Vitamin A is found in foods such as eggs , animal liver and carrots , and a diet that is lacking in this vitamin can cause blindness and an increased risk of microbial infections . Vitamin A is not a single compound , but rather a collection of compounds with similar molecular structures . One of these is retinol , which plays a vital role in the body's response to microbial infection . Retinol must bind to specific proteins to be able to move through the bloodstream and be transported around the body . Serum retinol binding protein transports ingested retinol from the intestine to the liver and other tissues . However , during microbial infection—when retinol transport is particularly important—the amount of this protein dramatically decreases; as such it is unclear how retinol is transported when the body is under attack from pathogens . It had been suggested that Serum Amyloid A ( SAA ) proteins , a family of proteins made by some liver and intestinal cells , could be involved in the response to infection , because these proteins' levels increase during infection . However , their exact functions were unknown . Derebe , Zlatkov et al . found that mice fed a diet poor in vitamin A produced fewer SAA proteins in their liver and intestinal cells . However , treating the cells with retinol or the molecule it is broken down into—called retinoic acid—caused more SAAs to be made . Derebe , Zlatkov et al . also discovered that SAAs are associated with retinol in blood samples taken from mice infected with salmonella; and that both mouse and human SAAs bind tightly to retinol . Combined , this evidence suggests that SAAs are the retinol binding proteins that transport retinol during infections . Derebe , Zlatkov et al . went on to solve the crystal structure of a mouse SAA protein , and showed that four SAA molecules bind together to form a ‘pocket’ that can hold a retinol molecule . Future work will focus on understanding exactly how the transport of retinol by SAAs affects the development of immunity to infections . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics",
"immunology",
"and",
"inflammation"
] | 2014 | Serum amyloid A is a retinol binding protein that transports retinol during bacterial infection |
The unfolded protein response ( UPR ) is an essential cell signaling system that detects the accumulation of misfolded proteins within the endoplasmic reticulum ( ER ) and initiates a cellular response in order to maintain homeostasis . How cells detect the accumulation of misfolded proteins remains unclear . In this study , we identify a noncanonical interaction between the ATPase domain of the ER chaperone BiP and the luminal domains of the UPR sensors Ire1 and Perk that dissociates when authentic ER unfolded protein CH1 binds to the canonical substrate binding domain of BiP . Unlike the interaction between chaperone and substrates , we found that the interaction between BiP and UPR sensors was unaffected by nucleotides . Thus , we discover that BiP is dual functional UPR sensor , sensing unfolded proteins by canonical binding to substrates and transducing this event to noncanonical , signaling interaction to Ire1 and Perk . Our observations implicate BiP as the key component for detecting ER stress and suggest an allosteric mechanism for UPR induction .
The endoplasmic reticulum ( ER ) is an essential eukaryotic organelle responsible for a number of processes including folding and maturation of secretory proteins destined for the extracellular space . The sudden requirement for processing large quantities of secretory proteins can be immense and results in over burdening the folding machinery within the ER , leading to accumulation of misfolded proteins and ER stress ( Malhotra and Kaufman , 2007; Walter and Ron , 2011; Wang and Kaufman , 2012 ) . The unfolded protein response ( UPR ) is a cell signaling system that detects the presence of misfolded proteins within the ER and carries out a varied cellular response to maintain homeostasis ( Malhotra and Kaufman , 2007; Hetz et al . , 2011; Walter and Ron , 2011; Wang and Kaufman , 2012 ) . Both Ire1 and Perk are UPR sensor proteins possessing luminal domains that are involved in detecting the presence of unfolded protein , although the precise mechanism is unclear ( Malhotra and Kaufman , 2007; Hetz et al . , 2011; Wang and Kaufman , 2012; Carrara et al . , 2013 ) . Early studies within the field provide evidence for the role of BiP ( ER Hsp70 chaperone ) in UPR activation by binding to the luminal domains and maintaining them in an inactive state ( Bertolotti et al . , 2000; Liu et al . , 2000; Okamura et al . , 2000; Ma et al . , 2002; Zhou et al . , 2006 ) . An alternative model proposed a direct recognition of misfolded proteins by Ire1 ( Credle et al . , 2005; Gardner and Walter , 2011; Promlek et al . , 2011 ) . In this study , we set out to glean new insights into the mechanism of mammalian UPR activation , by initially reconstituting the mechanistic events in vitro , using recombinant human Ire1 and Perk luminal domain proteins , in the presence of authentic and relevant ER unfolded protein CH1; and assessing whether BiP is involved in this process . By primarily using biophysical/biochemical techniques , we discover a direct noncanonical interaction between the ATPase domain of BiP to the luminal domains of Ire1 and Perk , clearly indicating a UPR signaling role . This interaction is unaffected by nucleotide binding to BiP . We further show that unfolded protein CH1 binds to the canonical BiP substrate binding domain; this relieves the interaction between BiP and luminal domains of Ire1 and Perk . Moreover , our data indicate that this model is consistent in cells . Overall , our observations suggest a novel allosteric model for UPR induction that involves BiP as the key component for detecting ER stress .
We initially expressed and purified full-length human Ire1 and Perk luminal domains along with full-length BiP encompassing both the ATPase domain and the substrate binding domains . To assess the role of BiP in UPR sensing , we first examined whether there was an interaction between the luminal domains of Ire1 and Perk to full-length BiP . Using microscale thermophoresis , we found that both Ire1 and Perk luminal domains bound to BiP with binding affinities of Kd = 1 . 33 μM and 1 . 92 μM , respectively , consistent with a typical transient protein–protein interaction ( Figure 1A–B ) . To dissect the molecular basis of the interaction between Ire1 and Perk with BiP , we tested if this interaction was mediated by the canonical substrate binding domain , as it is the case for this chaperone , or the ATPase domain . Both BiP's ATPase and substrate binding domains were expressed and purified separately and assessed for their ability to bind to Ire1 and Perk luminal domains . Analysis of binding by thermophoresis revealed that in contrast to full-length BiP , we measured no binding between luminal domains and BiP's substrate binding domain ( Figure 1C–D ) . Surprisingly , we discovered that BiP's ATPase domain bound to both Ire1 and Perk luminal domains with binding affinities of Kd = 1 . 97 μM and 2 . 05 μM , which were almost identical to full-length BiP ( Figure 1C–D ) . These results reveal that BiP binding to the luminal domains of Ire1 or Perk does not involve BiP's substrate binding domain . Rather , the interaction between BiP with Ire1 and Perk luminal domains is entirely mediated by BiP's ATPase domain ( Figure 1C–D ) . Such an interaction with BiP ATPase domain is novel , and suggests that the BiPATPase domain interaction with Ire1 or Perk is distinct from the classical chaperone-substrate interaction , and may serve some key signaling functions that we set out to elucidate . To confirm that the noncanonical interactions detected by microscale thermophoresis were robust , we developed an independent assay . Pull down assays with purified proteins recapitulated our previous results: His6-tagged full-length BiP and BiP ATPase domain proteins were able to form an interaction with luminal domain of Ire1 and Perk , while His6-tagged BiP substrate binding domain did not bind to Ire1 and Perk luminal domains ( Figure 1E ) . Thus , we have here recapitulated the interaction between full-length BiP and Ire1 and Perk luminal domains . Moreover , we confirm that Ire1 and Perk luminal domains bind to BiP's ATPase domain . 10 . 7554/eLife . 03522 . 003Figure 1 . Noncanonical binding of BiP ATPase domain to Ire1 and Perk . ( A–B ) Microscale thermophoresis ( MST ) analysis showing sigmoidal binding curves for interaction between full-length BiP and the complete luminal domains ( region I–V ) of ( A ) Ire1 luminal domain ( Kd = 1 . 33 μM ) and ( B ) Perk luminal domain ( Kd = 1 . 92 μM ) . ( C–D ) MST binding curves of interaction between BiP sub-domains ( ATPase and substrate binding domain ) and ( C ) Ire1 luminal domain ( ATPase Kd = 1 . 97 μM; no binding to substrate binding domain ) and ( D ) Perk luminal domain ( ATPase Kd = 2 . 05 μM; no binding to substrate binding domain ) . ( E ) Pull down assay showing BiP-luminal domain complexes using His6-tagged BiP proteins and luminal domains of Perk and Ire1 visualized by coomassie brilliant blue stained SDS PAGE gel . Ire1 and Perk luminal domains bind to full-length BiP and BiP ATPase domain . No binding to BiP substrate binding domain was observed . DOI: http://dx . doi . org/10 . 7554/eLife . 03522 . 003 ATP–ADP cycling is an important part of BiP's chaperone activity ( Mayer et al . , 2003 ) . To further characterize the novel interaction between BiP , Ire1 , and Perk , we next assessed if the formation of the complex between BiP ATPase domain and the UPR luminal domain was affected by the presence of nucleotides . Using thermophoresis , we measured the binding of full-length BiP to luminal domains of Ire1 and Perk in the presence of 10 mM ATP , ADP , AMPPNP , and also in the absence of nucleotide . We observed the affinity of interaction was very similar both in the presence of the various nucleotides , and when nucleotide was absent . Therefore , the addition of ATP , AMPPNP , and ADP had no effect upon the formation of the full-length BiP-Perk and full-length BiP-Ire1 luminal domain complexes ( Figure 2A–C ) . This indicates that the interaction between BiP ATPase domain and Ire1 or Perk is unrelated to the interaction between BiP and its canonical substrates . 10 . 7554/eLife . 03522 . 004Figure 2 . The noncanonical binding of BiP ATPase domain to Ire1 and Perk is independent of nucleotides . ( A–B ) MST analysis showing sigmoidal binding curves for full-length BiP interaction with ( A ) Ire1 luminal domain and ( B ) Perk luminal domain in the presence of 10 mM ATP , ADP , AMPPNP and in the absence of nucleotides . ( C ) List of Kd values ( μM ± SE ) for Ire1 and Perk luminal domain interactions with full-length BiP in the presence of nucleotides for the binding curves represented in A and B . Binding between luminal domains and BiP was not affected by the presence of the various nucleotides . DOI: http://dx . doi . org/10 . 7554/eLife . 03522 . 004 The luminal domain of yeast Ire1 has been divided into five subregions based upon a series of deletion mutant's ability to activate UPR ( Kimata et al . , 2004 , 2007 ) . We designed a range of constructs based upon this classification in order to determine more precisely the interaction site between Ire1 and Perk luminal domains with BiP's ATPase domain ( Table 1 ) . One point to note , is that using this assignment ( Kimata et al . , 2004 , 2007 ) yeast Ire1 luminal domain possesses an extended region I , whilst the equivalent region in human Ire1 is essentially absent . The implication for human Ire1 is that both regions I and II are very close together and map onto the equivalent of yeast region II . Next , we measured the binding affinities for BiP's ATPase domain interacting with human Ire1 and Perk luminal domains comprising of various regions between I and V . We found that binding affinities for region II–IV were essentially the same as those measured for binding to the full-length luminal domain construct region I–V , indicating the core interaction between UPR luminal domains and BiP maps to the luminal domain region II–IV and ATPase domain , respectively ( Figure 3A–C ) . To reinforce the in vitro data , we assessed whether the interaction between Ire1 luminal domain regions II–IV and BiP occurs in cells . Principally , we conducted a co-immunoprecipitation experiment using an Ire1 region V deletion mutant ( Ire1ΔV; Δ390–430 ) with a BiP construct possessing an N-terminal HA-tag , in the absence and presence of ER stress , and then compared with full-length Ire1 ( Figure 3D ) . In the absence of ER stress , immunoblotting with Ire1-specific antibody clearly indicates the presence of bands that are consistent for both Ire1 ΔV mutant and full-length protein co-immunoprecipitating with BiP . Upon ER stress , both Ire1 ΔV mutant and full-length Ire1 protein display a reduced level of interaction with BiP ( Figure 3D ) . We also assessed direct binding with BiP ATPase domain to both Ire1 full-length and ΔV mutant—observing an interaction consistent with in vitro data ( Figure 3E ) . These results suggest that BiP interacts with Ire1 region II–IV in cells and that region V is dispensable for this interaction to occur , reinforcing the previous in vitro analysis ( Figure 3D , E ) . Ire1 luminal domain region II–IV has been suggested to be important for stress sensing in yeast ( Credle et al . , 2005 ) . Here , we identified region II–IV as the binding site for human BiP-luminal domain protein interaction . This suggests that BiP is likely to be important in mammalian ER stress sensing . 10 . 7554/eLife . 03522 . 005Table 1 . Construct sizes for all BiP , Ire1 and Perk in vitro constructs used in this studyDOI: http://dx . doi . org/10 . 7554/eLife . 03522 . 005ProteinResidue rangeBiP FL28–654BiP ATPase28–405BiP SBD422–654Ire1 I–V24–440Ire1 I–IV24–390Ire1 II–V32–440Ire1 II–IV32–390Ire1 LD for cross-link experiment32–390Perk I–V54–509Perk I–IV54–403Perk II–V105–509Perk II–IV54–40310 . 7554/eLife . 03522 . 006Figure 3 . Core interaction between BiP ATPase and luminal domains occurs via region II-IV of luminal domains . ( A–B ) MST binding curves of interaction between BiP ATPase and different length constructs of ( A ) Ire1 luminal domain and ( B ) Perk luminal domain . ( C ) List of Kd values ( μM ± SE ) for BiP ATPase interaction with the various Ire1 and Perk luminal domain constructs ( based on regions I–V ) for binding curves represented in A and B . The luminal domain region II–IV , is solely responsible for binding to BiP proteins and regions I and V are dispensable in this interaction . ( D ) Co-immunoprecipitation experiment in which HEK293T cells were co transfected with either Ire1 mutant lacking region V ( Ire1ΔV; Δ390–430 ) or full-length Ire1 , along with HA-tagged BiP , in the absence or presence of ER stress ( TM = 5 μM; 4 hr tunicamycin ) . Immunoprecipitating with HA peptide and then immunoblotting with Ire1 specific antibody reveals an interaction between BiP and both full-length and mutant Ire1 that is missing region V ( Ire1 ΔV ) , which is reduced after ER stress . This interaction in cells reinforces the in vitro data . ( E ) Co-immunoprecipitation experiment similar to ( D ) , but cells were co transfected with HA-tagged BiP ATPase and were not subjected to ER stress with tunicamycin . DOI: http://dx . doi . org/10 . 7554/eLife . 03522 . 006 The identification of a noncanonical interaction between BiP's ATPase domain and the luminal domains of Ire1 and Perk raises the possibility that this interaction might have a signaling function in initiating UPR activation . In order to probe the significance of this interaction in UPR signaling , we set out to reconstitute ER stress in our system . Therefore , we critically assessed the effect of unfolded peptide substrates , and hence ER stress , upon the BiP-luminal domain complexes using our assays . First , we assessed the previously described yeast Ire1-specific unfolded peptide mimic , ΔEspP ( Gardner and Walter , 2011 ) . ΔEspP has been shown to bind directly to yeast Ire1 luminal domain and to subsequently cause luminal domain oligomerisation ( Gardner and Walter , 2011 ) . Although , we measured weak binding between ΔEspP and human luminal domain proteins of Ire1 and Perk ( Figure 4—figure supplement 1A–C ) , surprisingly we detected no effect upon full-length BiP-Ire1 and full-length BiP-Perk luminal domain complexes upon addition of peptide ( Figure 4—figure supplement 1D–F ) . To reconstitute ER stress in vitro , we next turned to an authentic and relevant ER-resident unfolded protein to assess the effects of misfolded proteins on the BiP-Ire1 and BiP-Perk complexes . The intrinsically unfolded immunoglobulin constant heavy chain domain ( CH1 ) , which is disordered in the absence of its cognate binding partner CL , is a relevant , ER localized unfolded protein substrate ( Feige et al . , 2009; Marcinowski et al . , 2011 ) . First , we examined if unfolded protein CH1 binds to the luminal domains of Ire1 and Perk . Surprisingly , we measured no interaction between CH1 and luminal domains suggesting that Ire1 and Perk luminal domains are not directly involved in detecting unfolded proteins ( Figure 4A–B ) . Next , we assessed the interaction of CH1 to full-length BiP . As expected ( Marcinowski et al . , 2011 ) , we found that CH1 bound to full-length BiP with a binding affinity of Kd = 8 . 7 μM . To identify what region of BiP , CH1 specifically bound to , we used ATPase and substrate binding domain proteins for interaction analysis . As expected for an Hsp70 chaperone ( Marcinowski et al . , 2011 ) , we found that CH1 bound only to the substrate binding domain of BiP with a similar binding affinity to that of full-length BiP , suggesting that BiP's substrate binding domain was solely responsible for binding to CH1 unfolded protein . 10 . 7554/eLife . 03522 . 007Figure 4 . Unfolded protein CH1 binds to canonical BiP substrate binding domain without binding to UPR luminal domains . ( A ) MST binding curves for CH1 binding to full-length BiP ( Kd = 8 . 7 μM ) , BiP's ATPase domain ( no binding ) , BiP's substrate binding domain ( Kd = 5 . 1 μM ) , Ire1 luminal domain ( no binding ) and Perk luminal domain ( no binding ) . ( B ) Pull down experiment showing CH1 binding to both full-length and substrate binding domain of BiP only , with no interaction observed to luminal domains , reaffirming the data for CH1 interactions using MST in part A . DOI: http://dx . doi . org/10 . 7554/eLife . 03522 . 00710 . 7554/eLife . 03522 . 008Figure 4—figure supplement 1 . Assessing the role of unfolded protein peptide mimic ( ΔEspP ) in UPR stress sensing . ( A–C ) ITC binding curves for ΔEspP titration into ( A ) Ire1 luminal domain ( Kd = 6 . 4 μM; N = 1 . 97 ) , ( B ) Perk luminal domain ( Kd = 9 . 3 μM; N = 1 . 98 ) , ( C ) full-length BiP ( no binding ) . ( D ) SEC MALS analyses showing Ire1 luminal domain is dimeric in solution ( MW = 104 . 5 kDa ) and the addition of ΔEspP has no effect on Ire1 luminal domain oligomeric state ( MW = 103 . 2 kDa ) . ( E ) SEC MALS analyses showing Perk luminal domain is dimeric in solution ( MW = 67 . 8 kDa ) and the addition of ΔEspP has no effect on Perk luminal domain oligomeric state ( MW = 67 . 6 kDa ) . ( F ) Pull down assay to test the effect of ΔEspP on His6-tagged full-length BiP-luminal domain complexes . ΔEspP has no visible effect upon BiP-luminal domain complexes . DOI: http://dx . doi . org/10 . 7554/eLife . 03522 . 008 Having established that CH1 binds exclusively to BiP's substrate binding domain and that BiP's ATPase domain is responsible for its interaction with the luminal domains; we evaluated the effects of CH1 binding to the BiP-luminal domain complex using our pull down assay . The addition of CH1 to the complex caused complete dissociation of both full-length BiP-Ire1 and full-length BiP-Perk luminal domain complexes ( Figure 5A ) . The dissociation of the BiP-Ire1 and BiP-Perk complexes by CH1 reveals that unfolded protein binding to BiP substrate binding domain , causes BiP ATPase domain to dissociate from BiP-luminal domain complex . Thus , when BiP is engaged in a signaling complex with Ire1 or Perk , its substrate binding domain remains available to interact directly with misfolded protein . This interaction dissociates the complex and initiates UPR signaling . To confirm these findings , we incubated CH1 with BiP and then subsequently added Ire1 and Perk luminal domains . Consistent with our prediction , we found that BiP engaged with misfolded protein CH1 , is unable to bind to the luminal domains of Ire1 or Perk ( Figure 5B ) . Thus , the binding of CH1 and luminal domains to BiP are mutually exclusive ( Figure 5B ) . 10 . 7554/eLife . 03522 . 009Figure 5 . The unfolded protein CH1 dissociates the noncanonical interaction between BiP ATPase domain and the luminal domain of Ire1 or Perk . ( A ) Pull down assay assessing the effects upon addition of unfolded protein CH1 to His6-tagged full-length BiP-luminal domain complexes . CH1 disrupts BiP-luminal domain interaction and causes the complexes to dissociate . ( B ) When His6-tagged full-length BiP is initially incubated with CH1 and then subsequently Ire1 and Perk luminal domains are added , we see no binding between luminal domains and BiP indicating that luminal domains and CH1 binding to BiP are mutually exclusive . DOI: http://dx . doi . org/10 . 7554/eLife . 03522 . 009 Thus far , the in vitro data suggest an allosteric/conformational change engendered by unfolded protein binding to BiP's substrate binding domain , causing dissociation of BiP , via it's ATPase domain , from Ire1 and Perk . To test if this model occurs in cells , we co transfected Ire1 and Perk , with HA-tagged full-length BiP , and BiP deletion mutant lacking the substrate binding domain; encompassing BiP ATPase domain only , consistent with earlier in vitro constructs used . We compared the levels of phosphorylated Ire1 and Perk , as indicators of UPR signaling , when expressed with full-length BiP or BiP ATPase domain , in both unstressed and ER stressed cells . We observed that in ER stressed cells expressing full-length BiP with Ire1 or Perk , exhibited significantly greater levels of phosphorylation when compared to cells expressing BiP ATPase domain with Ire1 or Perk ( Figure 6A , B ) . In agreement with our prediction , cells that lacked the substrate binding domain of BiP were unable to efficiently respond to ER stress and displayed significantly less phosphorylation—a result of attenuated UPR signaling . These data suggest that allosteric regulation occurs in cells , consistent with our in vitro model . 10 . 7554/eLife . 03522 . 010Figure 6 . BiP deletion mutants , lacking the substrate binding domain , attenuate UPR signaling . ( A ) Ire1 was co expressed with either full-length BiP ( HA–BiPFL ) or BiP ATPase domain , lacking the substrate binding domain ( HA–BiPATPase ) , and challenged with tunicamycin ( 5 μM ) over 0 hr and 4 hr time points in Ire1−/− cells . S724 phosphorylated Ire1 was measured as an indicator of UPR signaling ( Ali et al . , 2011 ) using pIre1s724 antibody ( ICR ) . Cells expressing full-length BiP and Ire1 were fully able to respond to induced ER stress , whilst expression with BiP ATPase domain attenuated UPR signaling . EV = empty vector . ( B ) Similar to ( A ) , but expressing Perk full-length in Perk−/− cells and using pPerk antibody ( Santa Cruz ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03522 . 010 The dissociation of BiP from Ire1 and Perk luminal domains upon ER stress leads to further downstream signaling ( Bertolotti et al . , 2000; Liu et al . , 2000; Okamura et al . , 2000 ) . It is highly likely that some sort of oligomerisation event occurs ( Walter and Ron , 2011 ) , preceding from a dimer state , which makes signal propagation more efficient , although exactly how this happens is not clearly understood . To understand luminal domain oligomerisation and how BiP affects this process , if at all; we analyzed the effects of cross linking upon Ire1 luminal domain protein in the absence and presence of His6-tagged BiP ATPase domain protein . The BiP ATPase domain interacts directly with Ire1 luminal domain , but perhaps more importantly—for clarity of results—it exists as a monomer both in absence and presence of cross linker . Thus , making the identification of the Ire1-BiP multimer bands an easier task . Upon addition of cross linker , Ire1 luminal domain forms two distinct species: a dimer and tetramer state . When we added BiP ATPase in a 1:1 ratio with Ire1 luminal domain protein , and then subjected the mixture to cross linking , we see a reduction in size of the tetramer band; concomitantly , there appears a band that corresponds to a trimer in size ( Figure 7A ) . Immunoblotting against His6 peptide reveals that the dimer band in the 1:1 mixture sample contains BiP ATPase; similarly , the trimer band also contains BiP ATPase . The tetramer band is exclusively made up of Ire1 luminal domain protein , and consequently does not show up when immunoblotting ( Figure 7B ) . The data suggest that Ire1 luminal domain dimer and tetramer formation is being impeded by BiP , which binds in a 1:1 hetero dimer and 2:1 hetero trimer interaction , with Ire1 . The caveat here is that since full-length BiP may also form dimers , we cannot rule out a 2:2 association between BiP and Ire1 . Nonetheless , the data do suggest that BiP may act to impede Ire1 luminal domain dimer and tetramer formation . 10 . 7554/eLife . 03522 . 011Figure 7 . BiP impedes Ire1 LD dimer and tetramer formation . ( A ) Ire1 luminal domain ( LD; regions II–IV ) and His6-tagged BiP ATPase domain proteins , both individually and in 1:1 molar ratio mixture , were visualized as control lanes on a 4–12% Bis-Tris SDS-PAGE gel . The same proteins were then subjected to EGS cross linker for 1 hr , after which the reaction was quenched and samples were visualized along side control lanes . In the presence of cross linker , Ire1 LD forms dimer and tetramer species; when in a mixture with BiP ATPase , there is a reduction in the corresponding tetramer band ( * ) . Also , a band appears that is consistent in size with a trimer species . ( B ) Samples from ( A ) were immunoblotted using anti-His6 antibody , which detects His6-tagged BiP ATPase domain protein . Since BiP ATPase protein is monomeric in the absence or presence of cross linker , BiP ATPase forms hetero dimer and hetero trimer with Ire1 LD . The binding of BiP to Ire1 reduces the size of the tetramer ( * ) that is exclusively formed by Ire1 LD , leading to the conclusion that BiP inhibits Ire1 LD tetramer formation , by preventing formation of the dimer species . DOI: http://dx . doi . org/10 . 7554/eLife . 03522 . 011
In this study , we identify a noncanonical interaction between BiP ATPase and luminal domains of Ire1 and Perk that is independent of nucleotide binding , and hence BiP's chaperone function . Furthermore , we discover that this noncanonical interaction is dissociated by canonical substrate binding to BiP via the substrate binding domain . Thereby , suggesting an allosteric mechanism for UPR induction ( Figure 8 ) . 10 . 7554/eLife . 03522 . 012Figure 8 . Allosteric model of UPR induction . In the absence of misfolded protein , BiP interacts with UPR luminal domains , which acts to repress the UPR signal . Upon ER stress , unfolded protein binds to the canonical BiP substrate binding domain , which in turn causes the noncanonical BiP ATPase-luminal domain interaction to dissociate , ultimately leading to UPR signal activation/propagation . DOI: http://dx . doi . org/10 . 7554/eLife . 03522 . 012 There have been several models proposed that describe how misfolded proteins are detected and how this leads to UPR signal activation . An early BiP dependent model describes the central role of BiP in this process ( Bertolotti et al . , 2000; Liu et al . , 2000; Okamura et al . , 2000; Ma et al . , 2002 ) . In this model , the interaction between luminal domains and BiP represses UPR signaling . Upon ER stress , BiP releases from the sensors , Ire1 and Perk , leading to activation . Experimental evidence for this comes from a series of studies that show an interaction between BiP and Ire1 in unstressed cells that dissociate in response to ER stress ( Bertolotti et al . , 2000; Liu et al . , 2000; Okamura et al . , 2000; Ma et al . , 2002; Oikawa et al . , 2009 ) . However , there have been a number of issues with this model . It was originally thought that this interaction was regulated by nucleotide binding ( Bertolotti et al . , 2000 ) ; thereby , suggesting a chaperone substrate type interaction—a non-productive association for UPR signaling—thus complicating analysis . But perhaps more significantly , the model failed to delineate a precise mechanism of BiP release upon accumulation of misfolded proteins , leading to the idea that BiP is competitively titred from luminal domains ( Ron and Walter , 2007 ) : a notion that has some weakness , not least because UPR is sensitive in response to ER stress . The present study clarifies this interaction by indicating that it occurs solely via the ATPase domain of BiP , and is unaffected by nucleotides , clearly suggesting a UPR significant role . Previous studies conducted in cells may have been unable to separate this specific ATPase interaction from the chaperone substrate interaction that may occur due to induction of stress and expression of recombinant protein; a process that would no doubt require some BiP acting in a chaperone capacity , and only by analysis in a clean in vitro system that this becomes apparent . The binding interaction analysis between Ire1 luminal domains ( regions I–V ) and BiP , to our knowledge is the first biophysical measurements giving specific affinities for association . Surprisingly , we see the previously implicated BiP binding region , Ire1 luminal domain region V , is completely dispensable for association to occur . A key study ( Kimata et al . , 2004 ) suggested that this region , an area proximal to the ER membrane , was important for BiP binding and that region V deletion mutants were perfectly able to respond to ER stress , leading to the idea that BiP release was not the principal determinant for Ire1 activity—but acts as a first step , followed by unfolded protein binding to luminal domains in a two step mechanism ( Kimata et al . , 2004 , 2007 ) . Structural descriptions of yeast Ire1 luminal domain supported this view ( Credle et al . , 2005 ) . The formation of a groove upon dimerization that resembles an MHC type fold suggested that unfolded proteins bound to luminal domains directly , bypassing BiP for UPR activation . This model has recently gained prominence ( Gardner and Walter , 2011; Promlek et al . , 2011 ) ; a study showed direct binding of unfolded peptide mimics to yeast Ire1 luminal domain in vitro ( Gardner and Walter , 2011 ) . The binding of these peptides was observed to cause oligomerisation of luminal domains , with the implication that this leads to UPR activation . In our study , we initially used such peptides to mimic ER stress , observing binding directly to luminal domains , but curiously not to BiP . Moreover , the peptide mimic in our system had no impact upon the noncanonical interaction between BiP ATPase and luminal domains . We noted that certain peptide mimics required addition of a number of charged residues to make it soluble and were derived from non-ER signal peptides . Our attention turned to using CH1 substrate , a previously characterized authentic ER unfolded protein ( Feige et al . , 2009; Marcinowski et al . , 2011 ) . Using this substrate , we measure robust binding to both full-length BiP and to the substrate binding domain , but not to the ATPase domain of BiP . Furthermore , we do not observe an interaction between CH1 and luminal domains in vitro . Therefore , the suggestion is that unfolded proteins bind exclusively to BiP's substrate binding domain , a notion that is well accepted for Hsp70 chaperones and for ER isoform BiP ( Marcinowski et al . , 2011 , 2013 ) . Indeed the recent crystal structure of isolated Hsp70 substrate binding domain with unfolded peptide mimics ( Zhang et al . , 2014 ) indicates that the substrate binding domain alone is also capable to bind unfolded peptide substrates . Interestingly , a study attempted to reconcile the involvement of BiP in UPR activation with direct binding of unfolded protein to luminal domains ( Pincus et al . , 2010 ) . In this model BiP sequesters inactive Ire1 . Upon high levels of ER stress , unfolded proteins bind to Ire1 causing the formation of a higher order active complex , which then recruits inactive monomers from BiP in a competitive fashion . Experimental evidence for this comes from differences in activation between wild-type Ire1 and mutant version of Ire1 ( Ire1 ΔV a . k . a Ire1bipless ) that is unable to bind BiP . The mutant Ire1 was unable to offer any buffering capacity since it was thought to be lacking the BiP binding region , and hence sensitized Ire1 to low unfolded protein load ( Pincus et al . , 2010 ) . The idea that Ire1 , BiP , and unfolded proteins exist in some dynamic equilibrium is certainty plausible—however , our data suggest that region V is dispensable for BiP binding to occur . Surprisingly , we do not observe CH1 unfolded protein directly interacting with luminal domains in any of our assays , thus providing evidence against direct association of unfolded protein—a role that is usually reserved for molecular chaperones . Interestingly , one noticeable feature of least some of the previous UPR models mentioned is that they involve a level of competition between components ( Ron and Walter , 2007; Pincus et al . , 2010 ) ; our present model suggests an allosteric mechanism at the heart of UPR induction . In summary , we identify a noncanonical interaction between BiP ATPase and luminal domains of Ire1 and Perk that dissociates when unfolded protein binds to the canonical substrate binding domain of BiP . Thus , implicating BiP as a central player in detecting ER stress and suggesting a novel allosteric mechanism for UPR induction .
All human BiP , Ire1 and Perk proteins used in this study were expressed in Escherichia coli BL21 ( DE3 ) cells ( Invitrogen , UK ) as fusion proteins with an N-terminal His6-tag followed by a PreScission Protease cleavage site . The constructs used are summarized in Table 1 . All proteins were purified by Co2+-NTA affinity using HiTrap TALON crude columns ( Clontech , CA ) in buffer A ( 50 mM HEPES ( pH 7 . 5 ) , 200 mM NaCl and 10% glycerol ) and eluted in the presence of 250 mM imidazole . Initial lysis and Co2+-NTA affinity purifications steps of BiP were supplemented with 5 mM ATP and 10 m MgCl2 . Unless otherwise specified , the His6-tag was removed by overnight incubation with PreScission Protease followed by an additional Co2+-NTA affinity step to remove any uncleaved protein . Proteins were further purified by anion-exchange using a HiTrap Q HP column ( GE Healthcare , UK ) and size-exclusion chromatography on a HiLoad 16/60 Superdex 200 column in buffer B ( 50 mM HEPES [pH 7 . 5] , 75 mM NaCl , 10% glycerol , and 1 mM TCEP ) . CH1 protein was expressed as previously described ( Marcinowski et al . , 2011 ) . Soluble ∆EspP ( MKKHKRILALCFLGLLQSSYSAAKKKK ) was purchased from AltaBiosciences ( Gardner and Walter , 2011 ) . All pull down experiments were carried out in 5 ml gravity flow columns . 50 μl of TALON resin pre-equilibrated with buffer B was incubated with 50 μl of purified BiPhis protein at 25 μM for 1 hr at RT . The resin was washed with 1 ml of buffer B to remove any unbound BiPhis . BiPhis was replaced by buffer B in control experiments . Then , 200 μl of purified untagged Ire1 and perk LD or CH1 proteins at 500 μM were added and incubated for 1 hr at RT . The resin was extensively washed with a total of 5 ml of buffer B in 500 μl volumes . For competition pull-downs , 200 μl of Ire1 LD , Perk LD , CH1 or ∆EspP at 500 μM in buffer B were then added , incubated for a further 1 hr at RT and washed as previously with buffer B . Buffer B was supplemented with 10 mM ATP , ADP or AMPPNP plus 10 mM MgCl2 , and 30 mM KCl where specified . Finally , the resin was resuspended with 50 μl of buffer , spun at 10000×g for 5′ and the resulting supernatant was analyzed on a 4–12% gradient SDS-PAGE gel . MST experiments were carried out using a Monolith NT . 115 instrument ( NanoTemper Technologies , Germany ) . Buffer B was used for all experiments and where specified additional 10 mM ATP , ADP or AMPPNP; and 10 mM MgCl2 , 30 mM KCl were included . Proteins were labeled using the Monolith NT Protein labeling Kit Red-NHS at 50 nM concentration and mixed with equal volumes of sixteen twofold serial dilutions of the unlabeled binding partner . Experiments were carried out in standard treated capillaries with 100% LED power and 80% IR-laser at 25°C . NanoTemper Analysis 1 . 2 . 101 software was used to fit the data with a nonlinear solution of the law of mass action and Kd values were determined . Each measurement was repeated in three independent experiments and Kd values were averaged . Standard error ( SE ) values are shown . The homobifunctional protein cross linker ethylene glycolbis ( succinimidylsuccinate ) ( EGS ) ( Thermo Scientific Pierce , MA ) was solubilised in DMSO at a final concentration of 20 mM . BiP , Ire1 and BiP-Ire1 complex were diluted to a final concentration of 50 μM with the reaction buffer ( 50 mM Hepes pH 8 . 0 , 50 mM NaCl , 5% glycerol and 5 mM DTT ) . Proteins were incubated with 50-fold molar excess of EGS for 1 hr . The reaction was then quenched for 15 min adding Tris buffer at a final concentration of 50 mM . Samples were first diluted to a final concentration of 10 μM with reaction buffer and then to 5 μM with Laemmli buffer ( Sigma ) . Samples were boiled for 10 min and loaded in NuPAGE 4–12% Bis-Tris pre-cast polyacrylamide gel . For the western blot , gel was transferred to nitrocellulose membrane ( Invitrogen's iBlot ) and blocked overnight at 4°C in PBST ( PBS in presence of 0 . 1% Tween 20 ) + 5% non fat dry milk . Primary anti-His antibody was added to PBST + 2% non fat dry milk in concentration of 1:10000 ( Sigma ) for 1 hr at room temperature . The membrane was then washed three times in PBST buffer and incubated with anti mouse-HRP antibody . Secondary antibody was diluted ( 1:10000 ) in PBST + 2% non fat dry milk and was incubated for 1 hr at room temperature . Followed by another three washes , blots were visualized by Millipore Luminata Crescendo Western HRP substrate and developed on Amersham Hyperfilm ECL . Human Embryonic Kidney cell ( HEK293T ) was cultured in Dulbecco's Modified Eagle Medium supplemented with 10% Fetal Bovine Serum , 2 mM L-Glutamine , 50 U Penicillin/50 μg Streptomycin/ml , 50 μM 2-Mercaptoethanol , and non-essential amino acid ×1 . A day before transfection 1 , 000 , 000 cells/well ( 5 ml ) were plated on 60 mm tissue culture plates . DNA containing either pcDNA3 . 1 ( empty vector control ) or Ire1ΔV , HA-BiP , up to a concentration of 6 μg total , were mixed with Fugene 6 reagent ( Promega , WI ) in ratio 1:3 , and then used to transfect cells . After 48 hr , cells were lysed by 450 μl non-denaturing lysis buffer ( +HALT Proteases Inhibitors Coctail , Pierce ) , scraped and centrifuged . Supernatant was co-immunoprecipitated by anti-HA agarose , mixed with Laemmli buffer , boiled and run on Tris-Glycine 4–12% gel . Immunoblotting—Gels were transferred to nitrocellulose membrane and blocked in TBST buffer plus 5% Marvel Dried Milk 1 hr in RT . Next , anti-Ire1 ( Abcam , UK ) and anti-HA were added to blocking buffer ( TBST + 1% milk powder ) and incubated 1 hr in RT . After that membranes were washed three times in TBST buffer and incubated with secondary antibody in TBST + 2% milk: anti-rabbit ( Cell Signaling ) for Ire1 and anti mouse for anti-HA , respectively . After 1 hr incubation in RT and another three washes , blots were visualized by Millipore Luminata Crescendo Western HRP substrate and developed on Amersham Hyperfilm ECL . Ire1−/− and Perk−/− MEF cells ( gift from Prof David Ron ) were cultured in Dulbecco's Modified Eagle Medium supplemented with 10% Fetal Bovine Serum , 2 mM L-Glutamine , 50 U Penicillin/50 μg Streptomycin/ml , 50 μM 2-Mercaptoethanol , and Non-Essential Amino Acid ×1 . A day before transfection , 500000 cells/well ( 2 ml ) were plated on 6-well plate . DNA containing either pcDNA3 . 1 ( empty vector control ) or Ire1 , Perk , HA-BiP , HA-BiP ATPase domain , up to a concentration of 3 μg total , were mixed with Fugene HD reagent ( Promega ) in ratio 1:6 , and then used to transfect cells . After 24 hr for Ire1−/− , and 48 hr for Perk−/− , cells were induced by 5 µM tunicamycin dissolved in DMSO ( 0 . 5% vol/vol ) and harvested after 0 hr and 4 hr . Next , cells were lysed by 250 μl non-denaturing lysis buffer ( +HALT Proteases Inhibitors Cocktail , Pierce ) , scraped and centrifuged . Supernatants were then mixed with 2× Laemmli sample buffer , boiled and run on Tris-Glycine 4–12% gel . For Perk general antibody analysis , cells were immunoprecipitated by Dynabeads Protein G ( Life Technologies ) . Immunoblotting—gels were transferred to nitrocellulose membrane and blocked in TBST buffer plus 5% Marvel Dried Milk 1 hr in RT . Next , anti-pPerk ( Santa Cruz , CA ) , anti-Perk ( Cell Signaling ) , anti Ire1 ( abcam ) , anti p-Ire1s724 ( Prof Ian Collins , ICR ) , and anti-HA antibody ( Life Technologies ) were added to blocking buffer ( TBST + 1% milk powder ) and incubated 1 hr in RT . After that membranes were washed three times in TBST buffer and incubated with secondary antibody in TBST + 2% milk: anti-rabbit ( Cell Signaling ) for anti-pPERK and PERK , and anti mouse ( GE ) for anti-HA . After 1 hr incubation in RT and another three washes , blots were visualized by Millipore Luminata Crescendo Western HRP substrate and developed on Amersham Hyperfilm ECL . Complex formation between ∆EspP and BiP , Ire1 or Perk was measured by ITC using a MicroCal VP-ITC system . ∆EspP interaction with Ire1 or Perk was carried out in buffer B; ∆EspP and BiP interaction was carried out in buffer B plus 10 mM ADP , 10 mM MgCl2 , and 30 mM KCl . All experiments were performed at 25°C . The sample cell contained BiP , Ire1 or Perk at approximately 40 μM concentrations and the syringe contained ∆EspP at approximately 450 μM concentrations . Heat of dilution , as determined by titrating ∆EspP into the buffer alone , was subtracted from the raw titration data before analysis . Data were fit by least-squares procedure assuming a one-site binding model using Microcal Origin ( version 7 . 0 ) . Kd values were averaged over three measurements , standard error values are indicated . To measure the absolute MW of protein species , SEC MALS was carried out using an Agilent 1260 system equipped with a miniDAWN TREOS ( Wyatt Technologies ) Light Scattering detector and an Optilab T-rEX ( Wyatt Technologies ) Refractive Index detector . Briefly , 100 μl Ire1 or Perk at 100 μM was mixed with excess ∆EspP ( 100 μl at 500 μM ) for 15′ at RT . In control experiments , 100 μl of buffer B was added instead . Samples were run on a Superdex 200 PC 3 . 2/30 column ( GE Healthcare ) pre-equilibrated with buffer B . Data were analyzed with the ASTRA software ( Wyatt Technologies , CA ) . | Proteins perform many essential tasks in cells , but to be able to work they first have to correctly fold into a specific three-dimensional shape . Within the cell , many proteins are folded with the help of ‘chaperone’ proteins . If any proteins fold incorrectly , the normal workings of the cell can be disturbed , which may damage the cell . This is more likely to happen if a cell suddenly requires a large number of proteins to be made , which can overwhelm the chaperone proteins . In humans and other eukaryotic organisms , many proteins are folded in a compartment within the cell called the endoplasmic reticulum . Inside this compartment there is a system called the unfolded protein response that detects misfolded proteins and boosts the cell's capacity to re-fold them . As part of this system , two sensor proteins detect when misfolded proteins are present , but it is not clear how they do so . It has been suggested that a chaperone protein called BiP may be able to activate these sensor proteins in order to turn on the unfolded protein response . In this study , Carrara et al . studied the sensor proteins and BiP using an artificial set-up in the laboratory . The experiments show that both of the sensor proteins can bind to a section of the BiP chaperone called the ATPase domain . However , in the presence of an unfolded protein , BiP stopped interacting with the sensor proteins , which could allow the sensor proteins to activate the unfolded protein response . The experiments also show that BiP must bind to the unfolded protein to activate the unfolded protein response . Carrara et al . 's findings suggest that BiP has a dual role in cells: to sense unfolded proteins by binding to them , and then to activate the sensor proteins that trigger the unfolded protein response . Together , these results suggest a new model for how cells detect and respond to misfolded proteins within the endoplasmic reticulum , and may provide new targets for therapies to treat diseases caused by defects in protein folding . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2015 | Noncanonical binding of BiP ATPase domain to Ire1 and Perk is dissociated by unfolded protein CH1 to initiate ER stress signaling |
Histone modifiers play essential roles in controlling transcription and organizing eukaryotic genomes into functional domains . Here , we show that Set1 , the catalytic subunit of the highly conserved Set1C/COMPASS complex responsible for histone H3K4 methylation ( H3K4me ) , behaves as a repressor of the transcriptome largely independent of Set1C and H3K4me in the fission yeast Schizosaccharomyces pombe . Intriguingly , while Set1 is enriched at highly expressed and repressed loci , Set1 binding levels do not generally correlate with the levels of transcription . We show that Set1 is recruited by the ATF/CREB homolog Atf1 to heterochromatic loci and promoters of stress-response genes . Moreover , we demonstrate that Set1 coordinates with the class II histone deacetylase Clr3 in heterochromatin assembly at prominent chromosomal landmarks and repression of the transcriptome that includes Tf2 retrotransposons , noncoding RNAs , and regulators of development and stress-responses . Our study delineates a molecular framework for elucidating the functional links between transcriptome control and chromatin organization .
The packaging of eukaryotic DNA with histones into chromatin provides ample opportunities for chromatin-modifying factors to exert extensive control over many aspects of genome-based processes ( Kouzarides , 2007 ) . In particular , enzymes catalyzing the covalent posttranslational modifications of histones are increasingly seen as critical regulators of transcription and the assembly of chromatin into various functional domains ( Henikoff and Shilatifard , 2011; Badeaux and Shi , 2013 ) . Two of the better understood posttranslational modifications of histones are acetylation and methylation . Whereas acetylation of histones by histone acetyltransferases ( HATs ) is generally associated with gene activation ( Rando and Chang , 2009 ) , deacetylation of histones by histone deacetylases ( HDACs ) tends to correlate with gene repression ( Yang and Seto , 2008 ) . Coordinated activities among HATs result in region-wide hyperacetylated chromatin states , leading to the formation of euchromatin domains supporting active transcription , and conversely , hypoacetylated chromatin states catalyzed by HDACs give rise to heterochromatin domains refractory to transcription ( Grunstein , 1998; Grewal and Jia , 2007 ) . In contrast , histone methylation is associated with either transcriptional activation or repression , and hence , with euchromatin or heterochromatin ( Huisinga et al . , 2006; Henikoff and Shilatifard , 2011 ) . Two well-characterized methylation marks occurring on two closely spaced residues near the amino-terminal tail of histone H3 exemplify this pattern ( Grewal and Jia , 2007 ) . Methylation at lysine 4 of histone H3 ( H3K4me ) and at lysine 9 ( H3K9me ) distinguishes euchromatin and heterochromatin , respectively ( Litt et al . , 2001; Noma et al . , 2001 ) . However , studies from the fission yeast Schizosaccharomyces pombe and other systems show that the euchromatic and heterochromatic landscapes are somewhat fluid , with islands of H3K9me transiently assembled within euchromatin at certain meiotic genes and the 3′ ends of convergent genes ( Cam et al . , 2005; Huisinga et al . , 2006; Gullerova and Proudfoot , 2008; Zofall et al . , 2012; Tashiro et al . , 2013 ) . Conversely , the RNA interference ( RNAi ) and exosome machineries , certain HATs and an active RNA polymerase II ( Pol II ) have been documented to contribute directly to the assembly of heterochromatin ( Volpe et al . , 2002; Djupedal et al . , 2005; Kato et al . , 2005; Buhler et al . , 2007; Xhemalce and Kouzarides , 2010; Reyes-Turcu et al . , 2011; Yamanaka et al . , 2013 ) . These observations point to the potential roles for other chromatin-modifying factors normally associated with euchromatin in heterochromatin assembly . In particular , the Saccharomyces cerevisiae homolog of Set1 ( KMT2 ) responsible for H3K4 methylation ( H3K4me ) has been implicated in transcriptional silencing at a number of genetic elements ( Nislow et al . , 1997; Krogan et al . , 2002; Berretta et al . , 2008; Camblong et al . , 2009; Kim and Buratowski , 2009; van Dijk et al . , 2011 ) . Set1 forms the catalytic engine of a highly conserved chromatin-modifying complex termed Set1C or COMPASS ( Shilatifard , 2012 ) . Set1C subunits have been shown to be recruited to active Pol II genes and provide the H3K4me signature for the gene-rich euchromatin ( Krogan et al . , 2003; Ng et al . , 2003 ) . H3K4me can exist in a mono- ( H3Kme1 ) , di- ( H3K4me2 ) , or tri- ( H3K4me3 ) methylated form ( Kusch , 2012 ) . The three forms of H3K4me have different distributions , with H3K4me3 and H3K4me2 enriched at gene promoters and gene bodies , respectively ( Cam et al . , 2005; Pokholok et al . , 2005 ) . H3K4me1 is enriched at the 3′ end of Pol II genes in budding yeast and at enhancers in mammals ( Pokholok et al . , 2005; Heintzman et al . , 2007 ) . Gene expression profiling analyses ascribe the repressor function of Set1C to H3K4me2 and/or H3K4me3 ( Margaritis et al . , 2012; Weiner et al . , 2012 ) . We have recently discovered a role for the S . pombe Set1 in the transcriptional repression and genome organization of long terminal repeat Tf2 retrotransposons and heterochromatic repeats that are dependent and independent of the Set1C complex and H3K4 methylation ( Lorenz et al . , 2012; Mikheyeva et al . , 2014 ) . In this study , we investigate the regulatory control of the fission yeast transcriptome by Set1 and its associated Set1C subunits . By systematically analyzing the transcriptomes of H3K4me mutants and mutant strains deficient in each of the Set1C subunits , we find that even though loss of H3K4me generally results in derepression , Set1 exerts its repressive function on most of its targets largely independently of the other Set1C subunits and H3K4me . Intriguingly , genome-binding profiles showed that Set1 localization is not linearly correlated with the levels of transcription at its target loci . In addition to localization at active Pol II genes , Set1 localizes to repetitive elements and repressed loci associated with development and stress-response pathways . Furthermore , we demonstrate that the conserved stress-response ATF/CREB Atf1 transcription factor mediates the recruitment of Set1 and modulates the levels of H3K4me3 at the centromere central cores and ribosomal DNA array . We show that Set1 coordinates with the class II HDAC Clr3 to mediate the assembly of H3K9me-associated heterochromatin and genome-wide repression of diverse transcripts , including Tf2 retrotransposons , noncoding RNAs , and developmental and stress-response genes . Our study illuminates a surprising cooperation between two histone-modifying enzymes with seemingly opposing activities in imposing genome-wide repression over the transcriptome and organizing the genome into euchromatin and heterochromatin .
Set1 is the catalytic engine of the Set1C complex that includes seven other subunits ( Roguev et al . , 2003 ) . Except for Shg1 , Set1 and six S . pombe subunits ( Swd1 , Swd2 , Swd3 , Spp1 , Ash2 , Sdc1 ) have orthologs in S . cerevisiae and humans ( Roguev et al . , 2003; Shevchenko et al . , 2008; Shilatifard , 2012 ) . Loss of individual Set1C complex subunits affects differentially the levels and states of H3K4me in S . pombe ( Roguev et al . , 2003; Mikheyeva et al . , 2014 ) . We performed expression profiling analyses in mutant strains deficient in H3K4me or lacking individual subunits of the Set1C complex . Whereas loss of set1 resulted in significant derepression of nearly 1000 of ∼42 , 000 tiling microarray probes ( average log2 fold-change vs wild-type >1 . 5 , p < 0 . 05 ) , H3K4me null mutants H3K4R ( histone H3 lysine 4 substituted with arginine ) or set1FH3K4me− ( H3K4me abolished by Set1 C-terminal FLAG epitope insertion ) ( Lorenz et al . , 2012; Mikheyeva et al . , 2014 ) affected ∼100 probes ( Figure 1A ) . Profiling analysis of other Set1C subunits showed a wide range of effects on transcriptional repression , with fewer than 100 probes significantly changed versus wild-type in ash2Δ to ∼300 in spp1Δ . Similar to the other H3K4me mutants , most probes affected in Set1C subunit mutants corresponded to upregulated transcripts , consistent with previous observations in budding yeast showing that loss of H3K4me tends to result in derepression ( Margaritis et al . , 2012; Weiner et al . , 2012 ) . Importantly , our results show that the major repressive function of Set1 in S . pombe occurs largely distinct from H3K4me and the Set1C complex . Variations among Set1C/H3K4me mutants in the proportion of affected probes corresponding to sense , antisense , and intergenic transcripts were also observed ( Figure 1B ) , with equal proportions of differentially expressed probes among the three classes of transcripts seen in set1Δ , H3K4R , and set1FH3K4me− mutants . Loss of ash2 primarily resulted in increased sense transcription , and loss of shg1 , spp1 , or swd3 predominantly affected intergenic transcripts . 10 . 7554/eLife . 04506 . 003Figure 1 . Set1/COMPASS subunits act primarily as transcriptional repressors . ( A ) Counts and ( B ) percentage of probes by matching feature strand/position of differentially expressed probes from custom 44 , 000-probe tiling microarrays . Significantly changed probes were defined as absolute log2 fold-changes ≥ 1 . 5 , false discovery rate ( FDR ) -adjusted p values <0 . 05 from duplicate arrays . ( C ) Hierarchical clustering of differentially expressed probes ( absolute log2 fold-change vs wild-type ≥1 . 5 , p < 0 . 05 ) in Set1C/H3K4me mutant strains . Probes showing significant expression changes in the indicated mutant versus wild-type strains were clustered using the HOPACH algorithm . The bottom panel shows the positions of probes matching repetitive centromeric , subtelomeric ( 100 , 000 bp end sequences of all chromosomes ) , Tf2 retrotransposons , the sense or antisense strands of annotated protein coding genes , or intergenic long noncoding RNAs ( lncRNAs ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 00310 . 7554/eLife . 04506 . 004Figure 1—source data 1 . Gene ontology ( GO ) enrichment in Set1C/COMPASS mutant expression profiling microarrays . GO term mappings were obtained from www . pombase . org . Enrichment analysis was performed using the R/Bioconductor GOstats package for known transcripts displaying statistically significant changes in the indicated mutant vs wild-type strain ( absolute log2 fold-change > 1 . 5 , FDR-adjusted p-value < 0 . 05 ) . Only significantly enriched GO terms ( p < 0 . 05 ) are included . See file header for complete column descriptions . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 00410 . 7554/eLife . 04506 . 027Figure 1—source data 2 . Comparative analysis of common enriched GO terms in Set1C/COMPASS mutant expression profiling microarrays . p-value data for Sense strand gene sets from Figure 1—source data 1 were retabulated to facilitate comparison of GO enrichment between Set1C/COMPASS mutants . min_Pvalue denotes the minimum p-value across all experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 027 Because Set1C/H3K4me mutants displayed varying degrees of transcriptional effects , we performed two-dimensional hierarchical clustering of all differentially expressed probes to gain further insights into their functional relationships . Despite their functions being linked to H3K4me , transcriptional profiles clustered broadly into four distinct groups ( Figure 1C , upper panel ) . The loss of ash2 and sdc1 , which affected a higher proportion of sense strand probes than in other mutants ( Figure 1B ) , shared a subset of upregulated transcripts with significant gene ontology ( GO ) enrichment for terms common to stress response , including ‘response to stress’ ( p ≈ 10−3 , ash2Δ; p ≈ 10−18 , sdc1Δ ) , ‘oxidoreductase activity’ ( p ≈ 10−3 , ash2Δ; p ≈ 10−11 , sdc1Δ ) , and ‘generation of precursor metabolites and energy’ ( p ≈ 10−5 , ash2Δ , p ≈ 10−3 , sdc1Δ ) ( Figure 1—source data 1A ) . The profiles of shg1 and spp1 mutants formed the second group of predominantly upregulated probes corresponding to diverse intergenic regions and antisense transcripts sharing comparatively weak GO enrichment . The group consisting of swd1Δ , swd2Δ , swd3Δ , set1FH3K4me− , and H3K4R mutants included smaller subsets of differentially expressed probes ( Figure 1C , upper panel ) , with modestly significant GO enrichment for upregulated transcripts related to stress response and carbohydrate metabolism ( Figure 1—source data 1A ) . The profile of set1Δ forms its own distinct group , containing a large set of upregulated transcripts including Tf2 retrotransposons , pericentromeric repeats , and long noncoding RNAs ( lncRNAs ) that were little affected in the other Set1C and H3K4me mutants ( Figure 1C , lower panel; Figure 1—source data 1B ) . These results suggest that loss of individual Set1C subunits produces different effects on the transcriptome that could not be fully accounted for by their known contributory roles to H3K4 methylation . While H3K4me is known to be enriched at transcriptionally active loci ( Cam et al . , 2005; Pokholok et al . , 2005 ) , we consistently observed transcriptional derepression in the set1Δ mutant at non-active , stress-response genes or heterochromatic repeats . We therefore performed genome-wide mapping of Set1 to gain insights into its repressor function . Consistent with its documented recruitment to active Pol II genes ( Ng et al . , 2003 ) , Set1 is enriched at sites that correspond to highly active Pol II promoters , including those of the housekeeping gene act1 and the ribosomal protein rps102 ( Figure 2A ) . Surprisingly , despite little enrichment of Pol II at certain lowly expressed genes ( e . g . , scr1 ) and repressed developmental genes ( e . g . , ste11 ) , noticeable Set1 binding was detected at the promoters of these genes ( Figure 2B; Figure 2—figure supplement 1 ) . Set1 localization at active and repressed targets was not hampered by the loss of H3K4me or its catalytic activity . Indeed , the inability of the set1FH3K4me− to methylate H3K4 appears to enhance its association with chromatin . To discern the relationship between Set1 binding and the transcriptional status of its targets , we ranked 290 protein-coding genes with significant Set1 binding ( chromatin immunoprecipitation ( ChIP ) fold enrichment ≥2 at three or more adjacent probes ) according to their expression levels ( Figure 2C , left panel ) . While transcript abundance generally correlated with Pol II occupancy levels ( Figure 2C , middle panel ) and 80% of promoter regions enriched for Set1 corresponded to actively transcribed genes ( Figure 2—figure supplement 2 ) , transcript abundance or Pol II occupancy levels did not linearly correlate with the levels of Set1 binding ( Figure 2C , right panel ) . Functional differences between high-abundance and low-abundance Set1-bound genes were assessed by GO analysis of genes rank-ordered by expression levels into quintiles ( Figure 2D ) . Whereas highly expressed genes occupied by Set1 were enriched with expected GO terms associated with rapid exponential growth ( ribosome , translation , glycolysis ) , Set1-bound genes with low abundance transcripts ( excluding heterochromatic noncoding RNAs due to limited GO annotation ) were enriched for terms related to stress response , cell wall and membrane-bound protein biogenesis , and Pol II transcription factor function ( Figure 2—source data 1 ) . Thus , our results suggest that Set1 localization at chromatin is not solely dependent on active Pol II , and that Set1 localization at lowly expressed or repressed loci might be functionally distinct from its canonical role at active Pol II genes . 10 . 7554/eLife . 04506 . 005Figure 2 . Set1 localizes to lowly expressed and repressed loci . ( A and B ) Enrichment of Set1 and RNA polymerase II ( Pol II ) determined by chromatin immunoprecipitation ( ChIP ) –chip displaying significant Set1 enrichment at highly transcribed genes ( A ) and repressed genes ( B ) . Positions of genomic features on forward ( top ) and reverse strands ( bottom ) , top panel . Black bars denote protein coding gene open reading frames ( ORFs ) ; white , associated untranslated regions ( UTRs ) ; orange , noncoding RNAs . Pol II ChIP–chip data was derived from Chen et al . ( 2008 ) . ( C ) Set1 enrichment relative to transcript abundance and Pol II occupancy . Comparisons of RNA-seq expression levels ( blue ) , Pol II ChIP-seq enrichment ( green ) and Set1 ChIP–chip enrichment ( red ) at loci showing significant Set1 enrichment ( N = 290 transcripts with nonoverlapping annotated features ) . Processed RNA-Seq FPKM data were obtained from Rhind et al . ( 2011 ) and Pol II ChIP-seq data from Zaratiegui et al . ( 2010 ) . The horizontal red line denotes mean expression for all Schizosaccharomyces pombe transcripts ( Rhind et al . , 2011 ) . ( D ) Gene ontology ( GO ) analysis of Set1-bound transcripts by expression level quintile . Representative GO terms were significantly enriched ( p ≤ 1 × 10−5 , hypergeometric test ) and found exclusively in quintiles of highly expressed ( top panel ) versus lowly expressed genes ( bottom panel ) . See Figure 2—source data 1 for a complete list of all significantly enriched GO terms/quintile . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 00510 . 7554/eLife . 04506 . 006Figure 2—source data 1 . Gene ontology ( GO ) enrichment of Set1-localized transcripts ( ChIP-chip ) by target expression level . Set1-targeted transcripts ( see Figure 2C ) were rank ordered by absolute expression level and divided into quintiles . GO analysis of each quintile was performed as for Figure 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 00610 . 7554/eLife . 04506 . 007Figure 2—figure supplement 1 . Set1 localization at active and repressed loci . Localization of FLAG-set1 or Set1 mutants deficient in H3K4me ( set1FH3K4me− ) , or lacking the catalytic domain ( set1-SETΔ ) at ( A ) the housekeeping gene act1 , ( B and C ) repressed genes scr1 and ste11 , ( D ) pericentromeric ( cen ) , or ( E ) rDNA array was assessed by chromatin immunoprecipitation ( ChIP ) followed by qPCR . Relative ChIP fold enrichment to input ( whole cell extract ) was calculated using the 2−ΔΔCt method after normalization by primers corresponding to mitochondrial DNA ( Lorenz et al . , 2012 ) . ( SD , error bars; n = 3 qPCR replicates . ) Untagged corresponds to a wild-type strain that did not express any FLAG tagged protein . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 00710 . 7554/eLife . 04506 . 008Figure 2—figure supplement 2 . Distribution of Set1-localized versus all Schizosaccharomyces pombe transcripts by absolute expression level . Histogram showing number of genes by expression level ( green bars ) , overlaid with Set1-bound transcripts ( red bars ) . The red vertical line denotes mean log2 FPKM , all S . pombe transcripts; black lines denote quintiles of Set1-bound genes with RNA-Seq transcripts . Processed RNA-Seq FPKM data were obtained from ( Rhind et al . , 2011 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 008 A number of low-abundance transcripts shown to be enriched for Set1 in genome-wide binding profiling ( e . g . , ste11 ) have previously been shown to be targets of the highly conserved ATF/CREB transcription factor Atf1 . In addition to localizing to its targets before their activation ( Eshaghi et al . , 2010 ) , which is important for subsequent proper response to environmental stresses ( Chen et al . , 2003 ) , Atf1 contributes to heterochromatic silencing at the silent mating-type locus ( Jia et al . , 2004 ) . We performed genome-wide binding profiling of Atf1 and compared it with that of Set1 to gain insights into the mechanism of Set1 recruitment to chromatin . We observed colocalization of Atf1 and Set1 at centromeric tRNA clusters flanking the euchromatin/heterochromatin boundaries of centromere II and the inner imr repeats of the central core ( Figure 3A , upper panel ) . Similar colocalization patterns were detected at centromeres I and III ( Figure 3—figure supplement 1 , upper panels ) . We also detected colocalization of Atf1 and Set1 at the intergenic region of the rDNA and the promoter of the developmental regulator ste11 ( Figure 3B , C , upper panel; Figure 3—figure supplement 2 ) . We assessed the loss of atf1 on Set1 activity by mapping distributions of H3K4me3 at these loci in wild-type and atf1Δ cells . In wild-type cells , H3K4me3 signals could be detected throughout the centromere central cores and the rDNA array but were little enriched at the ste11 promoter ( Figure 3A , B , C; Figure 3—figure supplement 1 , bottom panels ) . Loss of atf1 resulted in a sizeable reduction of H3K4me3 levels throughout the central cores and rDNA array . Moreover , genome-wide analysis identified many loci displaying reduced H3K4me3 in atf1Δ compared with wild-type ( Figure 3—source data 1 ) . The repressed status of the ste11 gene was not noticeably affected by atf1Δ ( Figure 3—figure supplement 4 ) and hence has little effect on the status of H3K4me3 . However , we noticed that several repressed genes whose promoters are occupied by Atf1 exhibited increased H3K4me3 levels in atf1Δ cells ( Figure 3—figure supplement 3 ) , probably owing to the loss of Atf1-mediated repression . 10 . 7554/eLife . 04506 . 009Figure 3 . Atf1 mediates recruitment of Set1 to centromeres , rDNA , and ste11 and contributes to H3K4 methylation . ( A ) Colocalization of Atf1 and Set1 ( upper panels ) at centromere II , ( B ) rDNA array , and ( C ) the promoter of the developmental regulator ste11 . Enrichment of H3K4me3 ( A–C , lower panels ) and Set1 ( D ) at the aforementioned loci in wild-type and atf1Δ cells . Enrichment of Set1 , Atf1 and H3K4me3 at indicated loci ( A–C ) was done by chromatin immunoprecipitation ( ChIP ) –chip . ( E ) Set1 and Atf1 regulate a common set of targets . Venn diagram of Atf1 and Set1 ChIP–chip peaks . Peaks were deemed overlapping if found within 1 kb of each other . The p value was determined by a hypergeometric test with population size N = 3667 Schizosaccharomyces pombe intergenic regions . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 00910 . 7554/eLife . 04506 . 010Figure 3—source data 1 . Differential enrichment of H3K4me3 levels in atf1Δ vs . wild-type cells . Comparative statistical analysis of H3K4me3/input ChIP-chip enrichment levels in wild-type vs . atf1Δ microarray experiments was performed using the R/Bioconductor limma package ( see Materials and Methods ) . Shown are significantly changed microarray probes , probe chromosomal position , corresponding genomic feature , log2 fold change in wild-type vs . atf1Δ experiments and FDR-adjusted p-value . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 01010 . 7554/eLife . 04506 . 011Figure 3—figure supplement 1 . Colocalization of Set1 and Atf1 at centromeres I and III . Colocalization of Atf1 and Set1 ( upper panels ) at centromeres I and III ( upper panels ) . Reduced H3K4me3 levels at centromere central cores in atf1Δ cells ( lower panels ) . Enrichment of Set1 , Atf1 , and H3K4me3 was analyzed by chromatin immunoprecipitation ( ChIP ) –chip . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 01110 . 7554/eLife . 04506 . 012Figure 3—figure supplement 2 . Enrichment of Atf1 at repressed loci . Confirmation of Atf1 binding at the rDNA array , ste11 , and pericentromeric heterochromatin ( dg ) was carried out by chromatin immunoprecipitation ( ChIP ) followed by qPCR . ChIP fold enrichment was calculated relative to input after normalization by primers corresponding to the act1 promoter . ( SD , error bars; n = 3 triplicates . ) DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 01210 . 7554/eLife . 04506 . 013Figure 3—figure supplement 3 . Atf1 acts as a transcriptional repressor . ( A and B ) Distributions of Atf1 and Set1 at ( A ) fbp1 and ( B ) srk1 ( upper panels ) . Increased H3K4me3 levels at fbp1 and srk1 in atf1Δ cells ( lower panels ) . Enrichment of Set1 , Atf1 , and H3K4me3 was determined by chromatin immunoprecipitation ( ChIP ) –chip . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 01310 . 7554/eLife . 04506 . 014Figure 3—figure supplement 4 . Derepression of ste11 in mutants deficient in both atf1 and set1 . Expression changes on forward and reverse strands at the ste11 locus in atf1Δ ( blue dashed lines ) , set1Δ ( red solid lines ) , and atf1Δ set1Δ ( dotted purple lines ) mutants . Tiling microarray probes corresponding to both forward and reverse strands from each window were binned into ∼600 bp windows , and log2 fold-changes of mutant versus wild-type from duplicate arrays for each mutant strain in each window were averaged . Data smoothing was performed using a three-consecutive-probe window moving average . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 014 To determine whether reduced H3K4me3 levels at the centromere central cores and the rDNA array partly reflect the failure of Atf1 to recruit Set1 , we assessed Set1 localization at these loci by ChIP . We found that Set1 enrichment at these loci , including the ste11 gene , was reduced in atf1Δ cells ( Figure 3D ) . At ste11 , Atf1 and Set1 appear to act primarily in parallel pathways to keep ste11 expression repressed , as appreciable upregulation of ste11 expression was seen only in mutants deficient for both atf1 and set1 ( Figure 3—figure supplement 4 ) . Comparing Atf1 and Set1 localization at the genome scale revealed 217 and 261 distinct bound loci for Atf1 or Set1 , respectively , with more than one-third co-occupied by both proteins ( p < 0 . 001 , Fisher's exact test ) ( Figure 3E ) . Collectively , our results suggest that Set1 recruitment to certain repressed loci is mediated in part by Atf1 , which in turn is important for proper maintenance of H3K4me levels and , depending on genomic context , transcriptional repression . To better understand the repressive function of Set1 , we sought to identify factors that cooperate with Set1 in heterochromatic silencing . The class II HDAC Clr3 has been shown to contribute to transcriptional silencing of heterochromatin ( Grewal et al . , 1998; Yamada et al . , 2005 ) Tf2 retrotransposons ( Hansen et al . , 2005; Cam et al . , 2008 ) , and stress-response genes ( Lorenz et al . , 2012 ) . These classes of genetic elements are also regulated by Set1 , suggesting a possible functional link between Clr3 and Set1 . To explore this idea , we constructed a mutant strain deficient for both set1 and clr3 ( set1Δ clr3Δ ) . We observed that in contrast to wild-type or single set1Δ or clr3Δ mutant strains , a double mutant set1Δ clr3Δ strain exhibited a significant synthetic slow-growth phenotype and sensitivity to the tubulin inhibitor thiabendazole ( Figure 4A ) , suggesting defects in chromosome segregation . Importantly , the set1FH3K4me− clr3Δ double mutant , in which set1 has no H3K4me activity , exhibited only slight defects . Derepression of a reporter gene inserted within the pericentromeric repeats has been observed in mutants deficient for either set1 ( Kanoh et al . , 2003 ) or clr3 ( Grewal et al . , 1998 ) . We observed additional derepression of the reporter gene in mutants deficient for both set1 and clr3 ( Figure 4B ) . Defects in heterochromatic silencing result in transcriptional derepression of both the forward and reverse strands of pericentromeric repeats ( Volpe et al . , 2002; Moazed , 2011; Alper et al . , 2013 ) . We performed expression analysis using tiling microarrays to assess transcription on both strands in set1 and clr3 mutant strains . Modest increases in transcript levels were found on both strands associated with the pericentromeric dg and dh repeats in single set1Δ and clr3Δ mutants . However , in the set1Δ clr3Δ double mutant , the increase was not only synergistic but occurred throughout the entire pericentromeric region ( Figure 4C ) . 10 . 7554/eLife . 04506 . 015Figure 4 . Set1 and the class II HDAC Clr3 cooperates in heterochromatic silencing and heterochromatin formation . ( A ) Serial dilution analysis ( SDA ) of set1 and clr3 mutant strains in nonselective ( N/S ) media or in the presence of the tubulin inhibitor thiabendazole ( TBZ ) , ( B ) uracil minus media ( −Ura ) or in the presence of the uracil counter selective drug 5-fluoroorotic acid ( 5-FOA ) . ( C ) Transcription of forward and reverse strands at centromere II in indicated mutant strains was analyzed by microarrays . ( D ) H3K9 dimethylation ( H3K9me2 ) in strains deficient for set1 and clr3 at the pericentromeric dg repeat . H3K9me2 enrichment at the dg repeat in indicated strains was carried out by chromatin immunoprecipitation ( ChIP ) and quantified by qPCR . ( E ) H3K9me2 distribution across the entire centromere II in wild-type and set1Δ clr3Δ strains . H3K9me2 at centromere II was assayed by ChIP–chip . ( F ) siRNA levels in wild-type , set1 and clr3 mutant strains . Detection of siRNAs was carried out by a northern blot using a probe specific for pericentromeric dg repeats . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 01510 . 7554/eLife . 04506 . 016Figure 4—figure supplement 1 . Pol II and Swi6 localization at pericentromeres in set1 and clr3 mutants . ( A ) Pol II and ( B ) Swi6 levels at the pericentromeric repeat dg in wild-type , set1Δ , clr3Δ , or set1Δ clr3Δ mutants were analyzed by chromatin immunoprecipitation ( ChIP ) followed by qPCR . ChIP fold enrichment was calculated relative to input after normalization by primers corresponding to the act1 promoter . ( SD , error bars; n = 3 triplicates . ) DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 01610 . 7554/eLife . 04506 . 017Figure 4—figure supplement 2 . H3K9me2 defects at centromeres I and III , mating type locus and subtelomeric regions in a strain deficient for both set1 and clr3 . ( A ) H3K9me2 distribution across major heterochromatin domains including centromeres I and III , ( B ) subtelomeres I , and ( C ) the silent mating type region was assayed by chromatin immunoprecipitation ( ChIP ) –chip in wild-type and set1Δ clr3Δ strains . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 017 Heterochromatin assembly is characterized by the establishment of histone H3 lysine 9 methylation ( H3K9me ) and HP1/Swi6 proteins bound to H3K9me ( Nakayama et al . , 2001 ) . H3K9me/Swi6 is thought to provide a platform for the recruitment of histone modifiers such as HDACs which could restrict the accessibility of Pol II ( Yamada et al . , 2005 ) . We performed chromatin immunoprecipitation ( ChIP ) followed by quantitative PCR ( qPCR ) to monitor the levels of H3K9me , Swi6 , and Pol II at the pericentromeric dg repeats in the set1 and clr3 mutants . Similar to previous observations ( Yamada et al . , 2005 ) , the loss of clr3 resulted in increased levels of H3K9me2 and Pol II and a decrease in Swi6 enrichment ( Figure 4D; Figure 4—figure supplement 1 ) . Loss of set1 resulted in a slight increase of Pol II localization ( Xhemalce and Kouzarides , 2010 ) and did not diminish H3K9me2 and Swi6 levels at the dg repeats . In contrast , there was a dramatic reduction in the levels of H3K9me2 and Swi6 accompanied by further increase of Pol II occupancy in the double mutant lacking both set1 and clr3 . We extended our analysis of H3K9me2 genome-wide and found that H3K9me2 levels in set1Δ clr3Δ mutant were reduced across the entire pericentromeric region ( Figure 4E ) . H3K9me2 defects in the double mutant were seen at other centromeres and heterochromatin domains , including the silent mating type region and subtelomeres ( Figure 4—figure supplement 2 ) . The RNAi machinery is known to contribute to the assembly of pericentromeric heterochromatin , in part by acting in cis to generate siRNAs ( Volpe et al . , 2002; Noma et al . , 2004 ) . We found that whereas loss of clr3 or set1 resulted in an increase of siRNAs ( Sugiyama et al . , 2007 ) , the level of siRNAs was dramatically reduced in the double mutant ( Figure 4F ) . Thus , our results reveal compensatory mechanisms by Set1 and Clr3 acting in parallel pathways to maintain heterochromatin at major chromosomal landmarks in S . pombe . To assess the extent of functional cooperation between Set1 and Clr3 in controlling transcription genome-wide , we performed comparative transcriptome analysis in set1 and clr3 mutant cells . While the majority of the differentially expressed probes in the set1Δ mutant corresponded to increased expression , loss of clr3 resulted in 792 probes changing significantly in comparison with wild-type , with approximately equal numbers corresponding to upregulated and downregulated transcripts ( Figure 5A ) . Intriguingly , cells lacking both set1 and clr3 displayed differential expression of nearly 2900 probes , 2343 of which were upregulated . Loss of H3K4me in a clr3 null background ( set1FH3K4me− clr3Δ ) did not produce such a drastic change to the transcriptome compared with set1Δ clr3Δ , but only reduced the proportion of downregulated transcripts seen in the single clr3Δ mutant . Similar proportions of probes corresponding to the sense or antisense strands of known transcripts were differentially expressed across set1Δ and set1Δ clr3Δ mutants , with the exception of clr3Δ cells , which displayed an increased proportion of sense strand probes ( Figure 5B ) . Hierarchical clustering showed that transcripts downregulated in set1Δ tended to be downregulated further in set1Δ clr3Δ ( Figure 5—figure supplement 1 ) , and transcripts that were upregulated in set1Δ ( i . e . , Tf2s and subtelomeric regions ) were further upregulated in the double mutants ( Figure 5C; Figure 5—figure supplement 2 ) . Most notably , loss of both set1 and clr3 resulted in significant expression changes within protein-coding gene regions for a large subset of genes displaying negligible change in individual set1 or clr3 mutants ( Figure 5C ) . Upregulated transcripts include well-characterized developmental and stress-response regulatory proteins that include fbp1 , mei2 and ste11 ( Figure 5—figure supplement 3 ) . Gene ontology analysis suggested that most of the upregulated transcripts in set1Δ clr3Δ are associated with stress-response processes that include the Tor2-Mei2-Ste11 pathways ( Figure 5D; Figure 5—source data 1 ) . These pathways are known to be activated during the meiotic development program ( Otsubo and Yamamoto , 2012 ) . In this regard , we noted that compared with wild-type or single mutant strains , the set1Δ clr3Δ double mutant exhibited considerable meiotic defects ( Figure 5—figure supplement 4 ) . Collectively , our results disclose unexpected coordination between Set1 and Clr3 in ensuring genome-wide repression of the fission yeast transcriptome and proper developmental control . 10 . 7554/eLife . 04506 . 018Figure 5 . Upregulation of a large fraction of the transcriptome in a strain deficient for both set1 and clr3 . ( A ) Counts and ( B ) percentage of probes matching feature strand/position in indicated mutant strains were analyzed similarly to Figure 2A and B . ( C ) Hierarchical clustering of significantly changed protein coding genes in set1 and clr3 mutant gene expression profiles ( n = 346 ) . Sense strand probes from two microarray experiments were averaged and clustered as in Figure 2C . ( D ) Gene ontology ( GO ) analysis of upregulated transcripts in set1 and clr3 mutant gene expression microarrays . Representative GO terms from biological process ( ‘BP’ ) , molecular function ( ‘MF’ ) , and cellular component ( ‘CC’ ) ontologies displaying most significant enrichment ( right panel ) and corresponding number of upregulated genes ( left panel ) in indicated mutant strains; all enriched terms are listed in Figure 5—source data 1 p values , hypergeometric test . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 01810 . 7554/eLife . 04506 . 019Figure 5—source data 1 . Gene ontology ( GO ) term enrichment in set1/clr3 mutant expression profiling microarrays . GO term enrichment analysis was performed similar to Figure 1—source data 1 for the sets of significantly changed sense strand transcripts in the indicated mutant vs . wild-type experiment ( see Figure 5D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 01910 . 7554/eLife . 04506 . 020Figure 5—figure supplement 1 . Representative genes whose expression requires set1 and clr3 . ( A ) Expression changes on forward and reverse strands at hem14 , ( B ) med7 , ( C ) and naa15 gene loci in clr3Δ ( purple dashed lines ) , set1Δ ( red solid lines ) , and clr3Δ set1Δ ( dotted orange lines ) mutants . Expression analysis was performed similarly to Figure 3—figure supplement 2 . Positions of genomic features on forward ( top ) and reverse strands ( bottom ) , top panel . Black bars denote protein coding gene open reading frames ( ORFs ) ; white , associated untranslated regions ( UTRs ) ; gray , noncoding RNAs; orange tRNA . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 02010 . 7554/eLife . 04506 . 021Figure 5—figure supplement 2 . Synergistic upregulation of Tf2s and subtelomeric regions in strain deficient for both set1 and clr3 . ( A ) Expression changes on forward and reverse strands at the Tf2 retrotransposons and ( B ) the chromosome I left subtelomere in clr3Δ ( purple dashed lines ) , set1Δ ( red solid lines ) , and clr3Δ set1Δ ( dotted orange lines ) mutants . Expressions were from tiling array analysis similar to Figure 3—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 02110 . 7554/eLife . 04506 . 022Figure 5—figure supplement 3 . Set1 and Clr3 cooperate to control genes involved in the core environmental stress response . ( A ) Expression changes on forward and reverse strands at fbp1 , ( B ) mei2 , and ( C ) ste11 gene loci in clr3Δ ( purple dashed lines ) , set1Δ ( red solid lines ) , and clr3Δ set1Δ ( dotted orange lines ) mutants . Expressions were from tiling array analysis similar to Figure 3—figure supplement 2 . Positions of genomic features on forward ( top ) and reverse strands ( bottom ) , top panel . Black bars denote protein coding gene open reading frames ( ORFs ) ; white , associated untranslated regions ( UTRs ) ; gray , noncoding RNAs . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 02210 . 7554/eLife . 04506 . 023Figure 5—figure supplement 4 . Cooperation between Set1 and Clr3 in development . Diploid cells homozygous for wild-type ( WT ) , set1Δ , clr3Δ , or set1Δ clr3Δ were streaked onto EMM medium to induce meiotic entry and allowed to complete meiosis at 26°C for four days . Cells were subsequently exposed briefly to iodine vapour , which efficiently stains meiotic products ( haploid spores ) dark brown . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 023
Recent transcriptome studies of chromatin mutants in S . cerevisiae reveal that loss of set1 or any of the other four core Set1C subunits ( Swd1 , Swd3 , Bre2/Ash2 , Sdc1 ) produces comparable expression profiles ( Margaritis et al . , 2012 ) . Furthermore , loss of set1 has only a modest effect on the transcriptome , mainly towards derepression that could fully be accounted by the loss of H3K4me ( Margaritis et al . , 2012; Weiner et al . , 2012 ) . Similar to these studies , our current study shows that complete loss of H3K4me ( i . e . , H3K4R , set1FH3K4me− mutants ) in S . pombe has only a slight impact on the transcriptome , with most differentially expressed transcripts upregulated . However , there are important differences . Except for the expression profiles of H3K4R and set1FH3K4me− mutants , the profiles among S . pombe Set1C subunit mutants are notably disparate , which could not be fully explained by their roles as subunits of Set1C or contributions to H3K4me ( Roguev et al . , 2003 ) . For example , Ash2 and Sdc1 are thought to form heterodimers that together with Swd1 and Swd3 constitute the core of the Set1C complex ( Roguev et al . , 2001; Dehe et al . , 2006; Southall et al . , 2009; Kim et al . , 2013 ) . Yet , while their expression profiles are most similar to each other , there are even differences between them , with the sdc1 mutant displaying stronger derepression for a subset of genes involved in response to oxidative stress than those seen in the ash2 mutant ( Figure 1C ) . These similarities and differences might reflect their association with other chromatin modifiers such as the Lid2 complex , not present in budding yeast ( Roguev et al . , 2003; Shevchenko et al . , 2008 ) . Most importantly , the expression profile of set1Δ is strikingly different from those of other Set1C/H3K4me mutants , displaying more than eight times the number of upregulated probes relative to those of swd3 or H3K4R mutants . Our findings show that unlike the results reported for S . cerevisiae , Set1 in S . pombe not only exerts more regulatory influence over the transcriptome , but also mediates its repressive function largely independently of the other Set1C subunits and H3K4 methylation—probably , as a consequence of the uncoupling of Set1 protein stability from H3K4me levels ( Mikheyeva et al . , 2014 ) . Interestingly , S . pombe Set1 has been reported as a component of at least two complexes: a large ∼1 MDa complex similar in size to that of S . cerevisiae Set1C and a smaller complex ( ∼800 kDa ) containing a shorter version of Set1 ( Roguev et al . , 2003 ) . Thus , Set1 might mediate its repressive nonH3K4me function via a distinct form of Set1 different from the form associated with the canonical Set1C complex . Our study reveals extensive functional interactions across the genome between Set1 and the stress-response transcription factor Atf1 at stress-response genes and major chromosomal landmarks , including the tandem rDNA array and centromeres . At the rDNA array and centromere central cores , Atf1 mediates Set1 recruitment and modulates H3K4me3 levels that might contribute to proper chromatin organization rather than transcriptional repression itself . At loci of stress response and developmental regulators such as ste11 , Atf1 and Set1 appear to act in parallel pathways that contribute to the repression of ste11 as loss of both atf1 and set1 resulted in significant derepression of ste11 ( Figure 3—figure supplement 4 ) . The transcriptional activation of Atf1 is controlled by phosphorylation mediated by the stress-activated mitogen-activated protein kinase ( MAPK ) Sty1 pathway ( Shiozaki and Russell , 1996; Lawrence et al . , 2007 ) . It is likely that co-occupancy of Set1 and Atf1 at the promoters of certain developmental and stress-response regulators not only helps keep these genes in a poised transcriptional off-state , but might also contribute to their rapid transcriptional activation in response to proper developmental or environmental stress signals . Pol II activity is known to be required for transcriptional silencing and heterochromatin assembly at pericentromeric repeats ( Djupedal et al . , 2005; Kato et al . , 2005 ) . Other factors associated with active Pol II transcription including components of the Mediator complex have also been shown to contribute to heterochromatin formation ( Oya et al . , 2013 ) . Our study identifies an important role for Set1 in the assembly of heterochromatin domains such as those present at pericentromeres ( Figure 6 ) . Set1 represses transcription on both the forward and reverse strands of the pericentromeric repeats and cooperates with Clr3 to assemble H3K9me-associated heterochromatin . Importantly , this heterochromatic activity of Set1 appears to be independent of its canonical H3K4me function associated with the Set1C complex , consistent with previous observations for the general lack of H3K4me within H3K9me heterochromatin ( Noma et al . , 2001; Cam et al . , 2005 ) . Set1-mediated heterochromatin assembly might involve Set1 methylating a nonhistone substrate similar to that of SUV39H1/Clr4 methylating Mlo3 , an RNA processing and nuclear export factor that also contributes to RNAi-mediated heterochromatin assembly ( Zhang et al . , 2011 ) . The only known nonhistone target of Set1 is the kinetochore protein DAM1 in S . cerevisiae ( Zhang et al . , 2005 ) . However , the S . pombe dam1 ortholog does not appear to be the target of Set1-mediated heterochromatic silencing as repression of Tf2 retrotransposons and pericentromeric heterochromatin is maintained in dam1 mutant cells ( Mikheyeva and Cam , unpublished data ) . 10 . 7554/eLife . 04506 . 024Figure 6 . Model for Set1 functions at euchromatin and heterochromatin domains . At euchromatin domains , the Set1C/COMPASS complex is recruited to active Pol II genes and provides the H3K4me marks . Set1 is also recruited to certain lowly expressed and repressed genes associated with developmental and stress-response pathways in part by Atf1 , other transcription factors ( TFs ) , and probably transcriptionally poised Pol II . Set1 acts in a parallel pathway with the histone deacetylase ( HDAC ) Clr3 to impose transcriptional repression at these loci . At a heterochromatin domain such as the pericentromeric region , Atf1 and probably other unidentified TFs mediate the recruitment of Set1 to sites enriched for tRNAs known to act as boundary elements . Set1 coordinates with Clr3 in the establishment of SUV39H1/Clr4-mediated H3K9me/HP1 ( HP: Swi6 and Chp2 ) heterochromatin and suppression of bidirectional transcription independently of H3K4me and the other Set1C subunits . Set1-mediated silencing could occur via methylation of nonhistone substrate ( s ) through the same or different pathways from those of RNAi ( i . e . , RITS , Rdp1 , Dicer ) or the exosome ( not shown ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04506 . 024 In addition to heterochromatic repeats , a significant fraction of the transcriptome is under repressive control by Set1 and Clr3 . Such genome-wide repressive effect strongly suggests that Set1 behaves largely as a bona fide repressor . At developmental and stress-response loci such as ste11 , Set1 may act in concert with transcription factors , including Atf1 together with Clr3 and other HDACs , to keep the target genes repressed in a steady-state condition . However , unlike heterochromatin , the chromatin states of these loci probably support a transcriptionally poised Pol II and in response to appropriate environmental signals enable Pol II to rapidly upregulate transcription .
Null mutants of Set1C subunits were constructed using a kanamycin cassette ( Bahler et al . , 1998; Mikheyeva et al . , 2014 ) . Double mutants were generated by standard genetic cross methods ( Moreno et al . , 1991 ) . Liquid cultures were grown at 30°C in standard rich media supplemented with 75 mg/l adenine ( YEA ) . ChIP assays were performed as previously described ( Lorenz et al . , 2012 ) . ChIP enrichment was quantified by qPCR analysis . ChIP–chip was carried out as previously described using Agilent tiling microarrays ( Cam et al . , 2005 ) . ChIP–chip analysis was performed using the R/Bioconductor ringo package ( Toedling et al . , 2007 ) . Preprocessing was carried out by loess normalization . ChIP-enriched regions were defined as three or more adjacent microarray probes with fold-enrichment greater than a two-Gaussian null distribution threshold ( greater than twofold enrichment ) . Between-array analysis of H3K4me3 in wild-type and atf1Δ experiments was performed using the limma ( linear models for microarray data ) package after interarray quantile normalization . Antibodies used for ChIP and ChIP–chip assays were anti-FLAG Set1 ( M2; Sigma-Aldrich , St . Louis , MO ) , anti-Atf1 ( sc-53172; Santa Cruz Biotechnology , Inc . , Dallas , Texas ) , anti Pol II ( ab5408; Abcam , Cambridge , MA ) , anti-H3K4me3 ( 07-473; Millipore , Billerica , MA ) , anti-H3K9me2 ( ab1220; Abcam ) , and anti-Swi6 ( Nakayama et al . , 2000 ) . Small RNAs were purified from 50 ml culture of logarithmically growing cells using the Ambion mirVana miRNA/siRNA isolation kit ( Life Technologies , Grand Island , NY ) . Small RNAs ( 60 µg ) were loaded onto a 15% denaturing polyacrylamide gel and run at 300 V until the bromophenol blue dye reached the bottom of the gel ( ∼1 . 5 hr ) . Northern transfer was done overnight by capillary blotting in Tris-borate-EDTA buffer at room temperature onto Hybond-N+ membrane ( GE Healthcare , Pittsburgh , PA ) . The membrane was subsequently UV crosslinked twice at 1200 J . Hybridization was carried out in 10 ml ULTRAhyb-Oligo buffer ( Life Technologies ) at 40°C overnight with a 32P-labeled RNA probe specific to pericentromeric dg repeats . The RNA probe was generated by in vitro transcription using a T7 RNA polymerase system and 50 µCi of [α-32P]UTP . Detection of the siRNA signals was carried out using the Storm 820 molecular imager ( Molecular Dynamics; GE Healthcare ) . Transcriptional profiling analysis was done as previously described ( Lorenz et al . , 2012 ) . Briefly , RNA was extracted from batch cultures of mid-exponential phase ( OD595 ∼ 0 . 3–0 . 6 ) from mutant and isogenic wild-type strains , reverse-transcribed into cDNA , and labeled with either Alexa Fluor 555 ( wild-type sample ) or Alexa Fluor 647 ( mutant sample ) using Superscript Indirect cDNA labeling system ( Life Technologies ) . Equal amounts of labeled cDNA ( 200–300 ng ) from wild-type and mutant samples were mixed and hybridized on a custom 4 × 44k probe Agilent tiling microarray as previously described ( Cam et al . , 2005 ) . For hierarchical clustering using the R/Bioconductor hopach package ( van der Laan and Pollard , 2003 ) , interarray quantile normalization was performed using the limma package , and transcripts with more than one differentially expressed probe were averaged . The cosine angle function was used for the clustering distance metric . Gene Ontology ( GO ) enrichment was performed as previously described ( Lorenz et al . , 2012 ) . Datasets associated with transcriptional profiling and ChIP–chip experiments in this study can be accessed at the Gene Expression Omnibus under accession number GSE63301 . | Genes can be turned on or off at different times in an organism's life . In humans , yeast and other eukaryotes , this is mainly controlled by the way DNA is packaged with proteins—known as histones—in a structure called chromatin . Genes that are switched on , or only temporarily switched off , are associated with areas of the genome where the chromatin is loosely packed . In contrast , genes that remain switched off for long periods of time are found in regions—known as heterochromatin—where the chromatin is tightly packed . There are many enzymes that can modify histones to change the structure of chromatin . One enzyme—called Set1—adds a methyl tag to chromatin , which is known to be associated with genes being switched on . However , Lorenz et al . found that Set1 also has other roles in modifying chromatin in the yeast Schizosaccharomyces pombe . The experiments found that Set1 helps to keep genes switched off and that this role is largely independent of its ability to add the methyl tag to chromatin . Set1 is recruited to many sites across the genome by another protein called Atf1 , which is involved in the cell's response to environmental stresses . Lorenz et al . believe that this helps to put these genes in a ‘poised’ off state so that they are ready to be switched on rapidly if needed . Set1 also works with another protein that removes acetyl tags—which encourage chromatin to be less tightly packed—from histones . Together , both proteins contribute to the assembly of heterochromatin and keep genes involved in development and stress responses switched off when they are not required . Collectively , these experiments reveal unexpected and important insights into how Set1—which plays critical roles in many aspects of human health including aging and cancer—works in cells . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"genetics",
"and",
"genomics"
] | 2014 | Heterochromatin assembly and transcriptome repression by Set1 in coordination with a class II histone deacetylase |
Traumatic brain injury ( TBI ) is a major cause of death and disability worldwide . Unfavorable TBI outcomes result from primary mechanical injuries to the brain and ensuing secondary non-mechanical injuries that are not limited to the brain . Our genome-wide association study of Drosophila melanogaster revealed that the probability of death following TBI is associated with single nucleotide polymorphisms in genes involved in tissue barrier function and glucose homeostasis . We found that TBI causes intestinal and blood–brain barrier dysfunction and that intestinal barrier dysfunction is highly correlated with the probability of death . Furthermore , we found that ingestion of glucose after a primary injury increases the probability of death through a secondary injury mechanism that exacerbates intestinal barrier dysfunction . Our results indicate that natural variation in the probability of death following TBI is due in part to genetic differences that affect intestinal barrier dysfunction .
Traumatic brain injury ( TBI ) is the leading cause of death for people under the age of 44 in the United States ( Harrison-Felix et al . , 2009; Coronado et al . , 2011 ) . Death following TBI is not only due to primary injuries , that is , mechanical injuries that occur at the moment of impact to the brain , but also to secondary injuries , that is , non-mechanical injuries that evolve over time in response to primary injuries ( Masel and DeWitt , 2010; Blennow et al . , 2012; Prins et al . , 2013 ) . Because secondary injuries are non-mechanical and are delayed relative to primary injuries they may be sensitive to therapeutic interventions . For example , secondary injuries to the intestine can rapidly follow primary injuries to the brain , and interventions that block intestinal injuries may prevent bacterial translocation and subsequent sepsis ( Hang et al . , 2003; Feighery et al . , 2008; Jin et al . , 2008; Bansal et al . , 2009 , 2010 ) . However , present understanding of the cellular and molecular mechanisms that underlie secondary injuries is not yet sufficient to develop therapeutic interventions ( Menon , 2009; Xiong et al . , 2013 ) . We have used a Drosophila melanogaster model to investigate the mechanisms underlying secondary injuries that cause death following traumatic injury . Our fly model uses the high-impact trauma ( HIT ) device , consisting of a metal spring with a stationary end attached to a board and a free end positioned over a polyurethane pad , to inflict traumatic injury ( Katzenberger et al . , 2013 ) . A plastic vial containing unanesthetized flies is connected to the free end . When the spring is deflected and released , the vial rapidly strikes the pad , and a mechanical force is delivered to the flies as they impact the vial wall . A high-speed movie shows that a strike from the HIT device causes flies to hit the vial wall multiple times with their head and body , probably inflicting traumatic injury to multiple organs , including the brain ( Balsiger et al . , 2014 ) . Closed-head TBI may result from impacts to the head or body that cause the fly brain to ricochet and deform against the head capsule , similar to what happens to humans in falls and car crashes ( Davceva et al . , 2012 ) . Accordingly , flies treated with the HIT device display phenotypes consistent with brain injury , including temporary incapacitation followed by ataxia , gradual recovery of mobility , neurodegeneration over time , and death within 24 hr ( Katzenberger et al . , 2013 ) . However , as in polytraumatic injuries in humans ( e . g . , blast injuries ) , damage to organs other than the brain may contribute to morbidity ( Scott et al . , 2006 ) . Therefore , we provisionally use the term traumatic injury to refer to the primary injury . One goal of this study is to identify the injured body part or parts that cause death within 24 hr . We quantify death following traumatic injury by determining the percentage of flies that die within 24 hr of the primary injury , which we define as the mortality index at 24 hr ( MI24 ) . Previously , we found that genotype and age at the time of traumatic injury affect the MI24 ( Katzenberger et al . , 2013 ) . Younger flies have a lower MI24 than older flies , suggesting that aging-related processes promote death following traumatic injury . In addition , genotype can affect the MI24 many fold , indicating the existence of genes that suppress or enhance the secondary injury mechanisms that cause death following traumatic injury . We also found that the innate immune response is activated shortly after primary injuries . In flies , the Toll and Immune deficiency ( Imd ) innate immune response pathways are responsible for defense against pathogens such as bacteria ( Lemaitre and Hoffmann , 2007 ) . Both pathways upregulate the transcription of antimicrobial peptide ( AMP ) genes , which encode small , secreted peptides that contribute to the elimination of pathogens . The innate immune response pathways are also activated in response to various types of stress , including oxidative stress and tissue damage . Here , we further investigate the roles of aging , genotype , and the innate immune response in mortality following traumatic injury . We performed a genome-wide association ( GWA ) study that implicates specific genes in affecting the probability of death following traumatic injury in young flies . Several of the genes have functions related to septate junctions , which are similar to tight junctions in vertebrates ( Furuse and Tsukita , 2006 ) . Septate junctions and tight junctions serve as barriers in the intestine , brain , and other tissues that prevent pathogen invasion and restrict the paracellular transport of materials . These junctions are constructed of transmembrane proteins such as Claudins , which interact between neighboring cells , and intracellular proteins such as PDZ ( PSD-95 , Discs-large , ZO-1 ) domain proteins , which interact with the cytoplasmic tail of transmembrane proteins . Mutation of the septate junction-associated , PDZ domain-containing protein Big Bang ( BBG ) permits bacteria from the intestinal lumen to cross the intestinal epithelial barrier and activate the innate immune response ( Bonnay et al . , 2013 ) . In addition , aging-related death in flies is highly correlated with intestinal barrier dysfunction and activation of the innate immune response ( Rera et al . , 2012 ) . In light of these links among tissue barrier dysfunction , the innate immune response , and aging-related death , we investigated the role of tissue barrier dysfunction in death following traumatic injury . Our findings indicate that traumatic injury to the brain is a major cause of death in our model and that mortality from brain injury is dependent on genetic and environmental effects on intestinal barrier permeability .
To investigate the role of genotype in determining the MI24 , we analyzed the D . melanogaster Genetic Reference Panel ( DGRP ) , a collection of wild-type , fully sequenced , isogenic fly lines ( called RAL lines ) ( Mackay et al . , 2012 ) . Figure 1A shows the MI24 data for 179 RAL lines that were treated with the standard injury protocol . 60 young flies ( 0–7 day old ) were placed in a vial and subjected to four strikes from the HIT device with 5 min between strikes . Following a 10 min recovery period after the last strike , flies were transferred to a new vial containing molasses food and were incubated at 25°C . The number of dead flies was counted after 24 hr . To control for death not due to injuries during this time , a vial of flies not subjected to injury was equivalently processed . Every experiment consisted of at least three independent trials , and the MI24 represents the average percent death for flies with injuries minus the average percent death for flies without injuries . We found that the MI24 had a continuous distribution among the RAL lines , over a wide range from 6 . 7 ± 0 . 8 to 57 . 5 ± 1 . 7 ( Figure 1A and Supplementary file 1 ) . Similarly , we found that the MI24 had a continuous distribution among a collection of 53 wild-type African lines , from 24 . 8 ± 9 . 8 to 68 . 0 ± 6 . 4 ( Figure 1—figure supplement 1 and Supplementary file 2 ) . These data indicate that the MI24 is influenced by genotype and that genetic variants affecting this parameter occur among natural populations of Drosophila . 10 . 7554/eLife . 04790 . 003Figure 1 . The MI24 greatly varies among 0–7 day old RAL flies and is associated with SNPs in grh . ( A ) Average and standard deviation of the MI24 for 179 RAL lines . Supplementary file 1 lists MI24 values for each of the RAL lines . Colored dots represent SNPs in grh associated with the MI24 and are used to indicate the RAL lines that contain the SNPs . ( B ) Schematic diagram of the intron-exon structure of grh with the location of the four SNPs associated with the MI24 ( St Pierre et al . , 2014 ) . Numbered boxes indicate exons and lines indicate introns . Note that spacing of the colored dots is not drawn to scale . ( C ) SNPs affect the MI24 . Average and standard error of the mean ( SEM ) of the MI24 for RAL lines ( light gray bars ) and progeny from crosses between RAL lines and RAL892 ( dark gray bars ) . *p < 0 . 05 , **p < 0 . 01 , and ***p < 0 . 001 , one-tailed t test comparison . DOI: http://dx . doi . org/10 . 7554/eLife . 04790 . 00310 . 7554/eLife . 04790 . 004Figure 1—figure supplement 1 . The MI24 greatly varies among 0–7 day old wild-type African lines . Average and standard deviation of the MI24 for 53 African lines . Supplementary file 2 lists MI24 values for each of the African lines . DOI: http://dx . doi . org/10 . 7554/eLife . 04790 . 004 To further assess the effect of genotype on the MI24 , we crossed the RAL line that had the highest MI24 ( RAL892 ) to other RAL lines and determined the MI24 of 0–7 day old progeny . We found that progeny from crosses between RAL892 and RAL lines with a low MI24 ( RAL352 , RAL907 , and RAL774 ) had an intermediate MI24 ( Figure 1C ) . In contrast , progeny from crosses between RAL892 and RAL lines with a high MI24 ( RAL707 , RAL73 , RAL799 , and RAL161 ) maintained a high MI24 . Variation in the MI24 could be due to many genes or environmental factors . However , the fly lines were cultured under the same conditions ( temperature , humidity , diet , light/dark cycle , and density ) , which limited the contribution of environmental factors . Thus , the continuous distribution over a wide range of the MI24 among wild-type fly lines and the intermediate MI24 of progeny from crosses between fly lines with significantly different MI24s suggest that the probability of death following traumatic injury is a quantitative trait affected by many genes ( Falconer and Mackay , 1996 ) . To identify genes that affect the MI24 , we carried out GWA analysis using the MI24 data shown in Figure 1A and ∼2 . 5 million single nucleotide polymorphisms ( SNPs ) among the RAL lines ( Mackay et al . , 2012 ) . This analysis revealed that 216 unique SNPs located in or near 98 genes were associated with the MI24 at a discovery significance threshold of p < 10−5 ( Supplementary file 3 ) . However , despite the small p-values , some of the associations may be false positives because the minor allele frequency cut-off of the DGRP Freeze 1 algorithm was 4 lines , allowing the p-value to be driven by a few extreme lines . Reanalysis using the DGRP Freeze 2 algorithm that has a minor allele frequency cut-off of 10 lines revealed significant associations of SNPs in only 10 of the 98 genes ( Huang et al . , 2014 ) . The discrepancy between the Freeze 1 and 2 analyses is illustrated by SNPs in grainyhead ( grh ) , which were significantly associated with the MI24 in the Freeze 1 analysis but were not identified in the Freeze 2 analysis because they occurred in fewer than 10 lines ( Figure 1A and Supplementary file 1 ) . At the time that we obtained the MI24 data for the RAL lines , the Freeze 2 algorithm had not been developed , so we moved forward based on the Freeze 1 data , initially focusing on grh because it contained the SNP that was most significantly associated with the MI24 ( p = 1 . 15 × 10−10 ) as well as three other significant SNPs . The four SNPs in grh were located in a 523 bp region of intron 10 suggesting that they have similar effects on the regulation of grh expression ( Figure 1B ) ( St Pierre et al . , 2014 ) . Alternatively , since three of the four SNPs ( red , blue , and yellow dots in Figure 1A ) are shared by three lines ( RAL73 , RAL161 , and RAL892 ) , linkage disequilibrium may account for their significant association with the MI24 . These data indicate that flies carrying particular grh alleles are more likely to die within 24 hr of a traumatic injury than flies lacking these alleles . Grh encodes a transcription factor crucial for many aspects of development , including epithelial barrier formation ( Nüsslein-Volhard et al . , 1984; Paré et al . , 2012 ) . In humans , one of the three grh orthologs , Grainyhead-like 2 ( Grhl2 ) activates the expression of claudin genes , and in mice , Grhl3 knockout reduces the expression of claudin genes ( Yu et al . , 2008; Werth et al . , 2010; Senga et al . , 2012 ) . In flies , misexpression of grh in a tissue that normally lacks septate junctions is sufficient to induce expression of septate junction proteins ( Narasimha et al . , 2008 ) . Thus , we hypothesized that the four SNPs in grh affect the function of septate junctions by altering the expression of genes encoding septate junction proteins . In support of this hypothesis , SNPs in bbg and scribbled ( scrib ) , which encode PDZ domain-containing , septate junction-associated proteins , were also associated with the MI24 ( Supplementary files 3 , 4 ) ( Bilder and Perrimon , 2000; Bonnay et al . , 2013 ) . bbg remained significantly associated with the probability of death ( p = 2 . 36 × 10−6 ) when the data from Figure 1A were reanalyzed using the DGRP Freeze 2 algorithm ( Huang et al . , 2014 ) . Additional support for the hypothesis comes from the finding that direct mechanical damage to the brain in rodent TBI models causes disruption of the intestinal barrier and a decrease in expression of tight junction proteins ( Hang et al . , 2003; Feighery et al . , 2008; Jin et al . , 2008; Bansal et al . , 2009 , 2010 ) . Lastly , gastrointestinal dysfunction is a common complication in TBI patients , and disruption of intestinal tight junction barriers can trigger systemic diseases ( Krakau et al . , 2006; Suzuki , 2013 ) . Thus , we tested this hypothesis by examining the permeability of tissue barriers following traumatic injury . Functionality of intestinal barrier can be ascertained in flies using a dye permeability assay in which flies are fed a nonabsorbable blue dye ( Rera et al . , 2011 , 2012 ) . If the intestinal barrier is functional , the dye remains in the digestive tract ( Figure 2A ) . In contrast , if the intestinal barrier is disrupted , the dye crosses the barrier into the hemolymph and disperses throughout the body , a process referred to as ‘Smurfing’ . Hemolymph is extracellular fluid in the open circulatory system of insects that contacts all internal organs and carries substances such as nutrients and metabolic waste to and away from cells , respectively ( Handke et al . , 2013 ) . We found that treatment of 0–7 day old w1118 flies ( a common laboratory strain ) with the standard injury protocol caused 23 . 3 ± 2 . 1% of the flies to Smurf within 24 hr of the primary injury ( Figure 2C ) , whereas only 0 . 5 ± 0 . 2% of untreated flies Smurfed in the same time period . These data indicate that traumatic injury increases the permeability of the intestinal barrier in flies . 10 . 7554/eLife . 04790 . 005Figure 2 . Traumatic injury causes intestinal barrier dysfunction . ( A ) Flies that were fed molasses food with blue dye . In flies without traumatic injury ( left ) , the dye was confined to the gut ( arrow ) . In some flies with traumatic injury ( right ) , the dye leaked out of the intestine into the hemolymph and dispersed throughout the body , producing a ‘Smurf’ phenotype ( Rera et al . , 2011; Rera et al . , 2012 ) . ( B ) A fly that was fed molasses food with blue dye and received brain injury from head compression . ( C ) Average and SEM of the MI24 ( light gray bars ) and SI24 ( dark gray bars ) for the indicated fly lines . The MI24 and SI24 were not significantly different for any of the fly lines ( p > 0 . 32 , one-tailed t test ) . The correlation coefficient ( r ) between the MI24 and SI24 was 1 . 0 . DOI: http://dx . doi . org/10 . 7554/eLife . 04790 . 00510 . 7554/eLife . 04790 . 006Figure 2—figure supplement 1 . Incapacitated flies had a significantly higher MI24 than non-incapacitated flies ( p = 0 . 007 , one-tailed t test ) . Average and standard deviation of the MI24 for 599 total flies , of which 535 were non-incapacitated and 64 were incapacitated . DOI: http://dx . doi . org/10 . 7554/eLife . 04790 . 006 To determine whether direct injury to the brain affects intestinal permeability , we inflicted closed-head injuries and monitored Smurfing as a reporter of intestinal permeability . We found that brain injury caused by compressing the head of 0–7 day old w1118 flies from eye-to-eye using forceps was sufficient to cause Smurfing within 24 hr ( Figure 2B ) . Of the 540 treated flies , 15 . 4% Smurfed and died within 24 hr of the primary injury , whereas only 0 . 5% Smurfed but did not die and 4 . 6% died but did not Smurf . In contrast , of the 540 untreated flies , 0 . 7% Smurfed and died within 24 hr , none Smurfed but did not die , and 0 . 4% died but did not Smurf . These data are consistent with the observed link between brain injury and intestinal barrier dysfunction in rodents , and they support the conclusion that increased intestinal permeability of flies subjected to the HIT device is due to brain injury ( Hang et al . , 2003; Feighery et al . , 2008; Jin et al . , 2008; Bansal et al . , 2009 , 2010 ) . In addition , these data suggest a causal link between increased intestinal permeability and death following TBI . In mammals , TBI not only disrupts the intestinal barrier but also the BBB ( Alves , 2014 ) . Therefore , we used a fluorescence assay to examine the effect of traumatic injury on integrity of the BEB as a reporter of the BBB ( DeSalvo et al . , 2011; Pinsonneault et al . , 2011 ) . Septate junctions are essential for creating the BEB , which restricts the transport of molecules between the retina and the hemolymph ( Banerjee et al . , 2008 ) . We used intra-thoracic injection to introduce tetramethylrhodamine-conjugated dextran molecules ( MW = 10 , 000 ) into the hemolymph of the fly . If the BEB is intact , the molecules accumulate along the border of the eye forming a hemolymph exclusion line ( Figure 3A ) ( Pinsonneault et al . , 2011 ) . In contrast , if the BEB is disrupted , the molecules cross the barrier and disperse throughout eye ( Figure 3B ) . We subjected 1–4 day old w1118 flies to the standard injury protocol , injected them in the thorax with fluorescent molecules , waited 2 hr , and examined the pattern of fluorescence in the eyes . We found that relative to untreated flies , a significantly greater percentage of HIT device-treated flies had fluorescence throughout the eye , indicating that the BEB is disrupted following traumatic injury ( Figure 3C ) . Furthermore , because permeability of the BEB is a reporter of permeability of the BBB , these data suggest that traumatic injury also causes BBB disruption in flies ( DeSalvo et al . , 2011 ) . 10 . 7554/eLife . 04790 . 007Figure 3 . Traumatic injury causes BEB disruption . ( A ) A w1118 fly without traumatic injury that was injected with tetramethylrhodamine-dextran molecules . Note the accumulation of fluorescence at the border of the eye ( arrow ) , which reflects an intact BEB . ( B ) A w1118 fly with traumatic injury that was injected with tetramethylrhodamine-dextran molecules . Note the fluorescence throughout the eye , which reflects BEB permeability . ( C ) Average and SEM of the percent of HIT device-treated ( + ) and untreated ( − ) w1118 flies with a permeable BEB . Traumatic injury ( TI ) significantly increased the percent of flies with a permeable BEB ( p = 0 . 0017 , one-tailed t test ) . ( D ) The dye penetration scale from − to ++++ for RAL flies . Arrows indicate the fluorescent pseudopupil . ( E ) The percent of flies with ( + ) or without ( − ) traumatic injury in each scale category for the indicated RAL lines . At least 85 flies were examined for each condition . DOI: http://dx . doi . org/10 . 7554/eLife . 04790 . 007 To investigate a causal link between intestinal barrier dysfunction and death following traumatic injury , we determined the overlap between flies that Smurfed and flies that died within 24 hr of the primary injury . In the case of w1118 flies , the percentage of flies that Smurfed within 24 hr of the primary injury , which we define as the Smurfing Index at 24 hr ( SI24 ) , was statistically similar to the MI24 ( Figure 2C ) . Moreover , there was almost complete overlap between flies that Smurfed and flies that died; less than 1% of flies Smurfed but did not die within 24 hr and less than 1% of flies that did not Smurf died within 24 hr . To determine the generality of these findings , we examined eight RAL lines: three of which had a low MI24 ( RAL352 , RAL907 , and RAL774 ) and five of which had a high MI24 ( RAL707 , RAL73 , RAL799 , RAL161 , and RAL892 ) . In all cases , we found that there was no significant difference between the SI24 and MI24 for each fly line ( Figure 2C ) . The almost perfect correlation between the SI24 and MI24 suggests that intestinal barrier dysfunction is closely linked with death following traumatic injury . As we have previously reported , about 10% of flies subject to a single strike from the HIT device are temporarily incapacitated , lying motionless on their back or side with no evident physical damage before gradually recovering motor activity within 5 min ( Katzenberger et al . , 2013 ) . This phenotype is similar to the symptoms of a concussion in humans ( Giacino et al . , 2014 ) . It is also similar to temporary paralysis observed in bang-sensitive Drosophila mutants following mild mechanical shock that disrupts normal electrical activity in the brain ( Fergestad et al . , 2008; Parker et al . , 2011; Burg and Wu , 2012 ) . These observations are consistent with the idea that at least a fraction of the flies subjected to the HIT device suffer a brain injury that temporarily disturbs normal neuronal function resulting in temporary paralysis . These observations also raise the question of whether this presumptive brain injury contributes to the observed mortality of treated flies . To address this question , we determined the correlation between temporary incapacitation and the MI24 following traumatic injury . Individual flies were subjected to a single strike from the HIT device and were scored as incapacitated if they were motionless immediately after injury . As previously observed , incapacitated flies did not die immediately ( Katzenberger et al . , 2013 ) . Of the 600 flies examined , 64 ( 10 . 7% ) were incapacitated , and all , except one , recovered mobility within 5 min . Nonetheless , incapacitated flies had an ∼fivefold higher MI24 than non-incapacitated flies , indicating that injuries that cause incapacitation also contribute significantly to the cause of death within 24 hr ( Figure 2—figure supplement 1 ) . Although we cannot rule out other possibilities , these data taken together with data in Figure 2 are consistent with a model in which death following traumatic injury inflicted by the HIT device is dependent on intestinal barrier dysfunction , which is evoked by damage to the brain via an unknown mechanism . To investigate a causal link between BEB/BBB and death following traumatic injury , we used the RAL lines to examine the correlation between BEB dysfunction and the MI24 . Because the RAL lines have red eyes , the BEB disruption phenotype is different from that described earlier for w1118 flies , necessitating a modification of the BEB protocol ( DeSalvo et al . , 2011 ) . Flies with an intact BEB had eyes with no fluorescence , whereas flies with a permeable BEB had eyes with a fluorescent pseudopupil that ranged in intensity . Using the scale shown in Figure 3D , we qualitatively scored the intensity of fluorescence in various RAL lines before and after subjecting flies to injury . We found that the percent of flies with a leaky BEB following traumatic injury , that is , those scored + to ++++ , was comparable among different RAL lines , as was the distribution of flies among the scale categories , irrespective of whether a particular line had a low MI24 ( RAL352 and RAL774 ) or a high MI24 ( RAL161 and RAL892 ) ( Figure 3E ) . In addition , the percent of RAL352 and RAL774 flies with BEB dysfunction following injury was substantially higher than their respective MI24 values . These data indicate that BEB/BBB dysfunction does not correlate with the probability of death following traumatic injury . The correlation between intestinal barrier disruption and the MI24 suggested the hypothesis that death following traumatic injury is triggered by paracellular leakage of factors such as bacteria or food components from the intestinal lumen to the hemolymph . The Drosophila gut commonly contains bacterial species in the Lactobacillus and Acetobacter genera ( Buchon et al . , 2013 ) . To determine if traumatic injury permits bacteria to leak across the impaired intestinal barrier , we quantified bacterial levels in the hemolymph . We extracted hemolymph from 0–7 day old w1118 flies 1 hr after they were subjected to the standard injury protocol and determined the bacterial count by spreading a given amount of hemolymph on LB plates ( Liu et al . , 2012 ) . We found that HIT device-treated flies had >400-fold more bacteria in the hemolymph than untreated flies ( Figure 4A and Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 04790 . 008Figure 4 . Traumatic injury causes an increase in the amount of bacteria and glucose in the hemolymph of w1118 flies . ( A ) Average and SEM of the number of bacterial colonies per microliter of hemolymph from flies without ( − ) or with ( + ) traumatic injury ( TI ) . Flies without traumatic injury had 0 . 8 ± 0 . 9 bacterial colonies per microliter of hemolymph . ( B ) Average and SEM of glucose concentration at times after traumatic injury . A significant increase in glucose concentration occurred between 2 and 8 hr , and a significant decrease in glucose concentration occurred at 24 hr . ( C ) Average and SEM of glucose concentration in flies fed either molasses food or water for the indicated amount of time after traumatic injury . Molasses food significantly increased the glucose concentration . In contrast , water significantly decreased the glucose concentration . *p < 0 . 05 , **p < 0 . 01 , and ***p < 0 . 001 , one-tailed t test comparison between flies without ( − ) and with ( + ) traumatic injury . DOI: http://dx . doi . org/10 . 7554/eLife . 04790 . 00810 . 7554/eLife . 04790 . 009Figure 4—figure supplement 1 . Traumatic injury causes bacteria to leak into the hemolymph . LB plates spread with hemolymph from w1118 flies not subjected to traumatic injury ( left ) or subjected to traumatic injury ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04790 . 009 To determine if traumatic injury permits glucose to leak across the impaired intestinal barrier , we determined the concentration of glucose in the hemolymph . Hemolymph was extracted from 0–7 day old w1118 flies at various time points after treatment with the standard injury protocol and the glucose concentration was determined using a colorimetric-based enzymatic assay ( Tennessen et al . , 2014 ) . We found that the hemolymph glucose concentration of injured flies was significantly higher than that of untreated flies between 2 and 8 hr after injury ( Figure 4B ) . Collectively , the bacteria and blood glucose data indicate that traumatic injury disrupts paracellular barriers formed by septate junctions , allowing the escape of factors at least as large as a bacterium from the intestinal lumen into the hemolymph . To address whether ingested food is the source of increased glucose in the hemolymph after traumatic injury , we determined the glucose concentration of hemolymph from flies fed molasses food or water after treatment with the standard injury protocol . We found that traumatic injury significantly increased the glucose concentration of hemolymph of flies fed molasses food but not flies fed water ( Figure 4C ) indicating that ingested molasses food is the source of increased glucose in the hemolymph following traumatic injury . As leakage of bacteria from the intestine could contribute to death following traumatic injury , we tested this possibility by eliminating endogenous bacteria in the gut and elsewhere by feeding flies a mixture of antibiotics in molasses food . It was previously shown that the mixture of antibiotics does not interfere with Imd pathway activation and is not toxic to flies ( Liu et al . , 2012 ) . After feeding antibiotics to 0–2 day old flies for 5 day , we treated the resulting 5–7 day old flies with the standard injury protocol . Some of the flies were used to determine the MI24 , and others were used to determine the effectiveness of the antibiotic treatment . For w1118 and RAL fly lines , PCR analysis of bacterial 16S rDNA levels using primers that recognize most bacterial species revealed that antibiotic treatment eliminated the endogenous bacteria ( Figure 5B ) ( Weisburg et al . , 1991; Liu et al . , 2012; Wong et al . , 2013 ) . Colony counts of whole fly extracts spread on LB plates yielded the same conclusion ( Figure 5—figure supplement 1 ) . Nonetheless , we found that antibiotics did not significantly affect the MI24 ( Figure 5A ) . Thus , bacteria do not significantly contribute to mortality following traumatic injury . Similarly , we found that the SI24 of antibiotic-fed flies was not significantly different from the SI24 of flies without antibiotics ( Figure 5C ) indicating that bacteria are also not involved in primary or secondary mechanisms that cause intestinal barrier dysfunction . 10 . 7554/eLife . 04790 . 010Figure 5 . Endogenous bacteria do not affect the probability of death following traumatic injury . ( A ) Average and SEM of the MI24 for the indicated fly lines fed for 5 day on molasses food ( light gray bars ) or on molasses food containing antibiotics ( dark gray bars ) before being subjected to the standard injury protocol . ( B ) Levels of bacteria in fly lines shown in panel A , as detected by PCR analysis for bacterial 16S rDNA and fly actin as a loading control . DNA extracted from flies fed ( + ) or not fed ( − ) antibiotics was used as a template . Indicated on the left are DNA size markers in basepairs . ( C ) Average and SEM of the SI24 for w1118 flies with the indicated treatments . Flies fed antibiotics had an SI24 that was not significantly different than flies without antibiotics ( p = 0 . 36 , one-tailed t test ) . Flies fed water after the primary injury had an SI24 that was significantly lower than flies fed molasses food after the primary injury ( p < 0 . 0001 , one-tailed t test ) . Flies fed 1 . 2 M sucrose after the primary injury had an SI24 that was significantly higher than flies fed water after the primary injury ( p = 0 . 0027 , one-tailed t test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04790 . 01010 . 7554/eLife . 04790 . 011Figure 5—figure supplement 1 . Antibiotic treatment of flies eliminates endogenous bacteria . LB plates spread with extracts from whole RAL892 flies fed food without antibiotics ( left ) or with antibiotics ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04790 . 011 We previously observed that expression of the innate immune response is activated in flies following traumatic injury ( Katzenberger et al . , 2013 ) . Leakage of bacteria across the intestinal barrier would be one way that traumatic injury could trigger the innate immune response . If so , this response should be dampened in antibiotic-fed flies . We used qRT-PCR to quantify AMP gene expression following traumatic injury in flies treated with antibiotics compared to controls . We found that 2 hr after treatment of 0–7 day old w1118 flies with the standard injury protocol , both antibiotic-fed flies and flies without antibiotics had higher levels of expression of the AMP genes Attacin C ( AttC ) , DiptB ( Diptericin B ) , and Metchnikowin ( Mtk ) than equivalently treated flies that were not subjected to the standard injury protocol ( Figure 6A and Figure 6—figure supplements 1 , 3 ) . Similar results were observed for some of the RAL lines . These data indicate that activation of the innate immune response following traumatic injury is not solely due to a bacteria-dependent mechanism but is triggered by other injury-associated factors as well . 10 . 7554/eLife . 04790 . 012Figure 6 . Analyses of the role that bacteria play in activation of the innate immune response by traumatic injury and the role that the level of activation of the innate immune response plays in causing death following traumatic injury . ( A ) Average and SEM of AttC expression normalized to actin expression in HIT device-treated flies relative to untreated flies . This analysis was performed with flies fed food without antibiotics ( light gray bars ) or with antibiotics ( dark gray bars ) . Expression levels were determined 2 hr after treatment with the standard injury protocol . Analogous data are shown for DiptB and Mtk in Figure 6—figure supplements 1 , 3 , respectively . ( B ) Average and SEM of AttC expression normalized to actin expression in antibiotic-fed flies ( dark gray bars ) and flies without antibiotics ( light gray bars ) 2 hr after treatment with the standard injury protocol . Analogous data are shown for DiptB and Mtk in Figure 6—figure supplements 2 , 4 , respectively . *p < 0 . 05 , one-tailed t test . DOI: http://dx . doi . org/10 . 7554/eLife . 04790 . 01210 . 7554/eLife . 04790 . 013Figure 6—figure supplement 1 . Level of DiptB expression normalized to actin expression in HIT device-treated flies relative to untreated flies for flies without antibiotics ( light gray bars ) and antibiotic-fed flies ( dark gray bars ) . Expression levels were determined 2 hr after treatment with the standard injury protocol . Average and SEM of at least three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 04790 . 01310 . 7554/eLife . 04790 . 014Figure 6—figure supplement 2 . Level of DiptB expression normalized to actin expression 2 hr after treatment with the standard injury protocol in flies without antibiotics ( light gray bars ) antibiotic-fed flies ( dark gray bars ) . Average and SEM of at least three independent experiments . *p < 0 . 05 , one-tailed t test . DOI: http://dx . doi . org/10 . 7554/eLife . 04790 . 01410 . 7554/eLife . 04790 . 015Figure 6—figure supplement 3 . Level of Mtk expression normalized to actin expression in HIT device-treated flies relative to untreated flies for flies without antibiotics ( light gray bars ) and antibiotic-fed flies ( dark gray bars ) . Expression levels were determined 2 hr after treatment with the standard injury protocol . Average and SEM of at least three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 04790 . 01510 . 7554/eLife . 04790 . 016Figure 6—figure supplement 4 . Level of Mtk expression normalized to actin expression 2 hr after treatment with the standard injury protocol in flies without antibiotics ( light gray bars ) and antibiotic-fed flies ( dark gray bars ) . Average and SEM of at least three independent experiments . *p < 0 . 05 , ***p < 0 . 001 , one-tailed t test . DOI: http://dx . doi . org/10 . 7554/eLife . 04790 . 016 We also found that the antibiotic-fed flies had significantly lower levels of expression of AttC , DiptB , and Mtk than flies without antibiotics ( Figure 6B and Figure 6—figure supplements 2 , 4 ) . Thus , antibiotics reduce the level of activation of the innate immune response but do not affect the MI24 , indicating that death following traumatic injury is not influenced by activation of the innate immune response . On the other hand , the level of AMP gene expression does correlate with the MI24 . For example , the correlation coefficient ( r ) between AttC expression and the MI24 for the eight RAL lines examined was 0 . 75 ( p = 0 . 03 ) for antibiotic-fed flies and 0 . 74 ( p = 0 . 03 ) for flies without antibiotics . The respective r-values for DiptB were 0 . 52 ( p = 0 . 19 ) and 0 . 51 ( p = 0 . 20 ) and for Mtk were 0 . 81 ( p = 0 . 02 ) and 0 . 63 ( p = 0 . 09 ) . These data indicate that the level of expression of some AMP genes , for example , AttC and possibly Mtk , is predictive of death following traumatic injury . Possibly , AMP gene expression level may somehow reflect the extent of intestinal barrier permeability following injury and thus be a predictor of subsequent mortality . As mortality following traumatic injury did not appear to be influenced by leakage of bacteria through a disrupted intestinal barrier , we went on to inquire whether leakage of an ingested food component influenced mortality by reducing the amount of food in the gut before or after traumatic injury . For the ‘before’ treatment , we cultured 0–6 day old w1118 flies in vials with water-soaked filter paper for 24 hr , subjected flies to the standard injury protocol , transferred them to vials containing molasses food , and determined the MI24 . We found that the MI24 of flies fed water before traumatic injury did not differ significantly from that of molasses food-fed flies ( Figure 7A ) . For the ‘after’ treatment , flies were cultured on molasses food prior to the standard injury protocol and then transferred to vials containing water-soaked filter paper . We found that feeding the flies water rather than molasses food after traumatic injury significantly reduced the MI24 . Feeding water rather than molasses food after injury also reduced the MI24 of RAL lines with a low ( RAL774 ) or high ( RAL707 ) MI24 ( Figure 7B ) . These data indicate that death following traumatic injury is dependent on a secondary injury mechanism that involves ingestion of molasses food after the primary injury . However , the present data do not rule out the possibility that water provides protection against death following traumatic injury . In either case , since death following traumatic injury was not completely eliminated by substituting water for molasses food , these data indicate that other independent mechanisms also contribute to mortality after traumatic injury . Consistent with these observations and the correlation between the SI24 and mortality after injury , we found that flies fed water after traumatic injury had a significantly lower SI24 than molasses food-fed flies ( Figure 5C ) . These data indicate that ingestion of molasses food after a primary injury promotes intestinal barrier disruption . 10 . 7554/eLife . 04790 . 017Figure 7 . Food ingested after the primary injury affects the MI24 . ( A ) Average and SEM of the MI24 for flies fed water ( W ) or molasses food ( M ) for 24 hr before or after the primary injury , for example , W–M means water for 24 hr before the primary injury and molasses food for 24 hr after the primary injury . ( B ) Average and SEM of the MI24 for flies of the indicated genotype and food treatments . ( C–D ) Average and SEM of the MI24 for flies fed the indicated molar ( M ) concentrations of ( C ) sucrose , ( D ) glucose , and ( E ) fructose for 24 hr after the primary injury . *p < 0 . 05 , **p < 0 . 01 , ***p < 0 . 001 , ****p < 0 . 0001 , one-tailed t test . DOI: http://dx . doi . org/10 . 7554/eLife . 04790 . 017 Our finding that traumatic injury causes glucose levels to increase in the hemolymph suggested that molasses , which is predominantly sucrose but also contains significant amounts of glucose and fructose , is the primary component of molasses food that promotes death following traumatic injury ( Dionex , 2003 ) . In support of this hypothesis , genes in the hexosamine biosynthesis pathway ( HBP ) , which functions as a sensor and regulator of glucose levels , were associated with the MI24 ( Supplementary files 3 , 4 ) ( Marshall , 2006 ) . One gene , super sex combs ( sxc ) , encodes an O-linked N-acetylglucosamine ( O-GlcNAc ) transferase ( OGT ) that uses one of the major HBP end products , UDP-N-acetylglucosamine ( UDP-GlcNAc ) , as a substrate for post-translational modification of proteins involved in insulin production and utilization ( Copeland et al . , 2008; Sinclair et al . , 2009 ) . Another gene , polypeptide GalNAc transferase 2 ( pgant2 ) , encodes a polypeptide N-acetylgalactosaminyltransferase that uses the other major HBP end product , UDP-N-acetylgalactosamine ( UDP-GalNAc ) , as a substrate for post-translational modification of proteins that remain to be identified ( Zhang and Ten Hagen , 2010 ) . In addition , SNPs in microRNA-14 ( mir-14 ) , which regulates insulin production in neurosecretory cells , and CG7882 , which encodes a protein similar in sequence to human glucose transporters ( GLUTs ) that facilitate transport of glucose across plasma membranes , were associated with the MI24 ( Thorens and Mueckler , 2010; Varghese et al . , 2010 ) . Consequently , we investigated the possibility that sugar levels contribute to mortality following traumatic injury by determining the MI24 of 0–7 day old w1118 flies that were transferred to vials with sugar-soaked filter paper ( sucrose , glucose , or fructose ) following the standard injury protocol . The concentrations of sugars tested were based on molasses food , which is 10% molasses vol/vol ( approximately 0 . 33 M sugar , that is , 0 . 13 M sucrose , 0 . 1 M glucose , and 0 . 1 M fructose ) ( Dionex , 2003 ) . We found that flies cultured on any of the sugars at 0 . 4 M had an MI24 that was not significantly different from the MI24 of flies cultured on water ( Figure 7C–E ) . In contrast , flies cultured on 0 . 6 M sucrose , 1 . 0 M glucose , or 0 . 8 M fructose had an MI24 that was significantly higher than the MI24 of flies cultured on water . The effective concentrations of the individual sugars are higher than they are in molasses suggesting that sugars in combination have a synergistic effect on the MI24 or that another component of molasses food is important . Figures 7C–E also indicate that the effect of sugars on the MI24 is saturable . For example , flies cultured on 0 . 8 , 1 . 0 , or 1 . 2 M sucrose had statistically similar MI24 values . We also found that flies fed 1 . 2 M sucrose after a primary injury had a significantly higher SI24 than flies fed water after a primary injury ( Figure 5C ) , indicating that sugar ingested after a traumatic injury promotes intestinal barrier disruption . Together , these data indicate that ingestion of sugar beyond a certain threshold after traumatic injury increases the probability of death by exacerbating intestinal barrier dysfunction and that the secondary injury mechanism by which sugar promotes death is saturable . The increased concentration of glucose in the hemolymph that results from traumatic injury could be exacerbated by reduced transport of glucose from the hemolymph into cells . Therefore , we investigated the effect of traumatic injury on insulin signaling , a major regulator of glucose transport into cells ( Hazelton and Fridell , 2010 ) . To do this , we examined expression of gene targets of the insulin signaling pathway . Insulin signaling through the insulin receptor ( InR ) leads to phosphorylation and activation of the Akt kinase and inactivation of the FOXO transcription factor ( Teleman , 2009 ) . If insulin signaling is impaired , FOXO is activated and expression of FOXO target genes such as InR , Lipase 4 ( Lip4 ) , Ecdysone-inducible gene L2 ( Impl2 ) , and 4E-BP ( Thor ) , increases ( Wang et al . , 2005 ) . We used qRT-PCR to determine expression levels of FOXO target genes in Smurfed and non-Smurfed 0–7 day old w1118 flies 2 hr after treatment with the standard injury protocol and culturing on molasses food . We found that Smurfed flies had a small but significant increase in expression of three of the four FOXO target genes compared with non-Smurfed flies ( Figure 8A ) . In contrast , Smurfed and non-Smurfed RAL flies with low ( RAL774 ) or high ( RAL892 ) MI24 had similar levels of FOXO target gene expression when cultured on molasses food after the primary injury ( Figure 8B , C ) , as did Smurfed and non-Smurfed w1118 flies when cultured on 1 . 2 M sucrose after injury ( Figure 8D ) . Thus , impaired insulin signaling does not appear to explain the increase in glucose concentration in hemolymph caused by traumatic injury nor the subsequent mortality . 10 . 7554/eLife . 04790 . 018Figure 8 . In general , FOXO target gene expression is not affected in response to traumatic injury . mRNA expression level of the indicated genes normalized to actin expression in non-Smurfed ( light gray bars ) and Smurfed ( dark gray bars ) flies 2 hr after the standard injury protocol . ( A ) w1118 flies cultured on molasses food after the primary injury , ( B ) RAL774 flies cultured on molasses food after the primary injury , ( C ) RAL892 flies cultured on molasses food after the primary injury , and ( D ) w1118 flies cultured on 1 . 2 M sucrose after the primary injury . InR , Lip4 , Impl2 , and Thor are FOXO target genes , and TAF1 is not a FOXO target gene . *p < 0 . 05 , **p < 0 . 01 , one-tailed t test . DOI: http://dx . doi . org/10 . 7554/eLife . 04790 . 018
TBI in humans can result from a strong mechanical jolt to the body or head that causes the brain to collide against the rigid skull ( Davis , 2000; Masel and DeWitt , 2010 ) . Initial or primary damage from impact forces results in brain dysfunction manifested by a variety of symptoms , including loss of consciousness , seizure , and other behavioral and cognitive impairments . Subsequently , additional non-mechanical secondary injuries can arise over time in response to the primary injuries resulting in further pathological consequences , which in the most severe cases includes death . The biological pathways linking primary and secondary injuries and the proximate cause of death following TBI are poorly understood at present . We have recently developed a traumatic injury model in Drosophila that replicates many of the key features of TBI in humans ( Katzenberger et al . , 2013 ) . Here , we investigate the link between primary and secondary injuries in this model and determine which pathophysiological manifestations correlated with death 24 hr after injury . Our results suggest that subjecting flies to collision impact by the HIT device results in primary TBI that triggers secondary injuries , including damage to the intestinal barrier . The resulting increased intestinal permeability is highly correlated with mortality within 24 hr after injury and is very likely to be one of the main causative factors of death . Because TBI in humans is diagnosed based on symptoms rather than by any precise medical test , and because flies subjected to mechanical impact injury by our HIT device do not exclusively contact the vial wall with their head , it is difficult to prove rigorously that we have caused a TBI-like injury in flies . Nonetheless , our results are consistent with the supposition that at least some fraction of the flies subjected to the HIT device have suffered an injury analogous with TBI in humans . They exhibit diagnostic features associated with human TBI , notably temporary loss of motor activity with flies lying incapacitated on their back or side , with no evidence of external mechanical damage ( Figure 2—figure supplement 1 ) . Motor activity recovers gradually over a 5-min period , although ataxia persists for a longer period . In addition to these immediate impairments , over a more extended period , the injured flies also manifest a shortened lifespan , behavioral deficits , and onset of neurodegeneration in the central brain ( Katzenberger et al . , 2013 ) . These phenotypes are all indicative of brain dysfunction , which is the defining characteristic of TBI . Indeed , screens in Drosophila for reversible , conditional loss of motor activity , resulted in the isolation of mutants with defects in neuronal excitability and synaptic transmission , which were further enriched for neurodegenerative phenotypes ( Siddiqi and Benzer , 1976; Littleton et al . , 1998; Palladino et al . , 2002 , 2003; Fergestad et al . , 2006; Gnerer et al . , 2006; Babcock et al . , 2015 ) . Activation of the innate immune response in the brain following treatment by the HIT device is a further indication of injury to the brain ( Katzenberger et al . , 2013 ) . Moreover , because flies become paralyzed immediately after mechanical injury , this must be a primary injury response . The high-speed movie shows that flies subjected to the HIT device could sustain primary injuries to the brain through multiple mechanisms: acceleration-deceleration forces on the brain due to contact of the head with the vial , coup contrecoup forces on the brain due to rebounding of the spring , or sheer forces on the brain due to unrestricted head movements that accompany contact of the body with the vial ( Balsiger et al . , 2014 ) . Thus , although we cannot rule out other more complicated interpretations , as a working hypothesis , we believe our data indicate that some percentage of flies subjected to mechanical impact by the HIT device suffer a brain injury that shares significant features with TBI in humans . In this study , we focused on investigating the underlying causes of mortality within 24 hr after subjecting flies to TBI . Since flies that die within this period do not do so immediately after the primary injury , there must be some secondary effect that amplifies the initial injury to cause death . Unexpectedly , our results point to intestinal barrier dysfunction as a physiological consequence following TBI that is a major factor in subsequent mortality . Four lines of evidence support this conclusion . ( 1 ) GWA analysis for variation in the MI24 uncovered genes linked to the function of septate junctions ( Figure 1A and Supplementary file 1 ) , including grh , which encodes a transcription factor required for epithelial barrier formation , and bbg and scrib , which encode PDZ domain-containing , septate junction-associated proteins ( Bilder and Perrimon , 2000; Narasimha et al . , 2008; Bonnay et al . , 2013 ) . ( 2 ) There was a very high correlation between the MI24 and the onset of Smurfing , a reporter of increased intestinal permeability ( Figure 2C ) . ( 3 ) There was a strong correlation between the MI24 and intestinal leakage of glucose that was ingested after a primary injury ( Figures 4B , C , 7 ) . ( 4 ) There was a significant correlation between activation of the innate immune response by leakage of bacteria from the intestine and the MI24 ( Figure 6 and Figure 6—figure supplements 1–4 ) . Our results also provide evidence that intestinal barrier dysfunction is secondary to brain injury inflicted by the HIT device . In particular , leakage of glucose from the intestine , which is delayed relative to the time of injury by the HIT device ( Figure 4B ) , and the increased probability of intestinal permeability by ingestion of food after injury ( Figure 5C ) both indicate that intestinal permeability is a secondary response to the initial mechanical injury . Moreover , we found that direct injury to the brain via a crushing injury , is sufficient to trigger intestinal barrier dysfunction ( Figure 2B ) . This conclusion is supported by the observation that flies immediately incapacitated following injury by the HIT device , which are the ones most likely to have suffered a brain injury , had a significantly higher probability of death , which is associated with intestinal barrier dysfunction , than non-incapacitated flies ( Figure 2—figure supplement 1 ) . Physiological events associated with death following traumatic injury in flies are shared with TBI in mammals . Gastrointestinal dysfunction , including increased intestinal permeability , is frequently observed in TBI patients ( Krakau et al . , 2006 ) . Moreover , increased intestinal permeability occurs in rodent TBI models in which injury is inflicted exclusively to the brain , demonstrating that increased intestinal permeability can result from direct mechanical injury to the brain ( Hang et al . , 2003; Feighery et al . , 2008; Jin et al . , 2008; Bansal et al . , 2009 , 2010 ) . While increased intestinal permeability is linked to death in critically ill patients and correlates with the severity of injury in trauma patients , it has not yet been linked to death in TBI patients ( Doig et al . , 1998; Faries et al . , 1998; Reintam et al . , 2009; Tude Melo et al . , 2010; Piton et al . , 2011 ) . On the other hand , patients with severe TBI have significantly higher blood glucose levels than patients with moderate or mild TBI , and hyperglycemia is highly predictive of death following TBI ( Rovlias and Kotsou , 2000; Salim et al . , 2009; Tude Melo et al . , 2010; Harun Harun et al . , 2011; Prisco et al . , 2012; Alexiou et al . , 2014; Elkon et al . , 2014; Yuan et al . , 2014 ) . In addition , patients with diabetes mellitus , a disease characterized by insulin resistance , have an increased risk of death following TBI ( Ley et al . , 2011; Lustenberger et al . , 2013 ) . However , thus far , modulating blood glucose levels by intensive insulin treatment in humans has had no effect on the probability of death within 6 months of a primary injury ( Bilotta et al . , 2008; Yang et al . , 2009 ) . Thus , additional research is still need to understand the mechanistic relationship between blood glucose levels and the probability of death following TBI . Because key features appear to be conserved between flies and mammals , further studies using the fly TBI model should help provide important new information . Taken together , these data support the conclusion that primary TBI triggers secondary intestinal barrier dysfunction ( Figure 9 ) . Consequent leakage of sugars across the impaired intestinal barrier causes further impairment of this barrier and ultimately death through an unknown proximal event . 10 . 7554/eLife . 04790 . 019Figure 9 . Genetic , cellular , and molecular data presented in this study suggest a model for the pathway of events following TBI . Intestinal barrier dysfunction may play an important role in promoting death following TBI ( indicated in red ) . TBI induces additional physiological changes ( indicated in black ) that do not cause death but may contribute to other outcomes such as neurodegeneration . Bacteria-independent and bacteria-dependent pathways that activate the innate immune response are indicated by ‘ind . ’ and ‘dep . ’ , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 04790 . 019 Physiological phenotypes of flies that die from TBI are shared with those of flies that die from old age . Rera et al . ( 2012 ) found that , regardless of chronological age , a few days prior to death , w1118 flies show increased intestinal permeability and increased activation of the innate immune response . They also found that w1118 flies had reduced insulin signaling , which we observed with w1118 flies cultured on molasses food , but , paradoxically , not with other fly lines or under other conditions ( Figure 8 ) . In addition , we previously found that the probability of death following TBI increases with age in flies , and others have found that the probability of death within one or 6 months of a primary injury increases with age in TBI patients ( Susman et al . , 2002; Hukkelhoven et al . , 2003; Dhandapani et al . , 2012; Katzenberger et al . , 2013 ) . Taken together , these data suggest that as flies and humans age , events such as breakdown of the intestinal epithelial barrier progressively become more severe until they are sufficient to cause death . Thus , as has been suggested for neuropathologies in human TBI ( Smith et al . , 2013 ) , TBI may cause death in young flies by triggering secondary injury mechanisms that parallel the defects that otherwise occur as part of the normal aging process . Older flies would then be more susceptible to death following TBI than younger flies because some critical physiological mechanism such as breakdown of the intestinal epithelial barrier would already be compromised to the point where it would require less of a subsequent insult to push the impairment beyond the threshold for lethality . Our results demonstrate that the probability of death following TBI is a quantitative trait likely to be affected by many genes . This discovery was possible because we were able to examine many genetically diverse wild-type fly lines under conditions where factors known to affect the probability of death following TBI , that is , the force , number , and timing of primary injuries , age at the time of the primary injury , and environmental conditions , were kept constant ( Figure 1A and Figure 1—figure supplement 1 ) . Furthermore , through GWA analysis , we were able to identify 216 SNPs in 98 genes that are significantly associated with the probability of death following TBI ( Supplementary file 3 ) . Presumably , these SNPs create a physiological state that alters the severity of primary and/or secondary injuries . For example , SNPs in genes that affect the function of septate junctions ( grh , bbg , and scrib ) or glucose homeostasis ( sxc , pgant2 , miR-14 , and CG7882 ) may generate intestinal epithelial barriers that are sensitive to disruption by cellular and molecular mechanisms triggered by TBI . Genotype is likely to play an equally important role in humans since variation in several genes has already been shown to be associated with the probability of death of severe TBI patients ( Dardiotis et al . , 2010; Garringer et al . , 2013; Failla et al . , 2015 ) . These data suggest that SNPs in orthologs of genes such as grh will correlate with the probability of death of TBI patients and could be used for genetic susceptibility testing to identify individuals at high risk of death following TBI . Nevertheless , it remains to be determined whether any of the 98 genes identified by our GWA study directly affect the probability of death following TBI . We attempted to determine this for grh by examining the MI24 of existing grh mutant fly lines . We found that the MI24 ranged from 18 . 9 ± 2 . 8 to 31 . 6 ± 0 . 9 for 0–7 day old flies that were heterozygous for different grh mutant alleles , two lethal P-element alleles and three lethal EMS alleles . However , determining whether and how these grh mutations affect the MI24 requires comparison with isogenic flies that are wild-type for grh , which unfortunately do not exist . Moreover , we do not yet know how the identified SNPs affect grh expression or function . For example , if the SNPs increase grh expression or alter only a subset of grh functions then grh loss-of-function mutations would not be expected to have the same effect on the MI24 . Use of the CRISPR technology in future experiments will help resolve these issues ( Gratz et al . , 2013 ) . Our studies have also led to the identification of additional physiological events evoked by TBI , including BBB dysfunction ( Figure 3 ) , leakage of bacteria from the intestinal lumen ( Figures 4A , 5 ) , and activation of the innate immune response ( Figure 6 and Figure 6—figure supplements 1–4 ) . Although occurrence of these events is not correlated with the probability of death following TBI , we hypothesize that they are not benign . One likely possibility is that they are associated with unfavorable long-term TBI outcomes in flies and humans that survive for an extended period after injury . For example , since activation of the innate immune response in the brain causes neurodegeneration in flies and is a common feature of human neurodegenerative diseases , it may be an important factor for neurodegeneration in the fly TBI model and for chronic traumatic encephalopathy , a form of neurodegeneration , in TBI patients ( Tan et al . , 2008; Arroyo et al . , 2011; Chinchore et al . , 2012; Petersen et al . , 2012 , 2013; Cao et al . , 2013; Baugh et al . , 2014 ) . In support of this idea , the β-lactam antibiotic ceftriaxone has neuroprotective effects in a rat TBI model ( Wei et al . , 2012; Goodrich et al . , 2013; Cui et al . , 2014 ) . It will be interesting to determine if reducing the level of activation of the innate immune response by feeding flies antibiotics reduces the severity of neurodegeneration following TBI . Finally , we found that the amount of sugar ingested immediately after a primary injury greatly affects the probability of death ( Figure 7 ) . There is considerable evidence that diet after a primary injury influences the outcomes of TBI patients ( Greco and Prins , 2013; Scrimgeour and Condlin , 2014 ) . For example , zinc supplementation reduces the probability of death of severe TBI patients , and the amount of nutrition in the first 5 days after a primary injury affects the probability of death of severe TBI patients ( Young et al . , 1996; Härtl et al . , 2008 ) . Our results in Drosophila , suggest that limiting sugar intake immediately after TBI in humans may be worth investigating as a therapeutic option to reduce the probability of death . Moreover , evolutionary conservation of the intestinal response to TBI between flies and humans suggests that elucidation of the underlying genotype- and age-dependent mechanisms in flies will have clinical relevance . In summary , these studies have shown that key phenotypic manifestations of TBI and the underlying physiological mechanisms are shared between Drosophila and humans . By exploiting the many experimental advantages offered by a Drosophila TBI model , it should be possible to obtain novel information to gain further insight into the biology of TBI and ultimately derive new therapeutic strategies to limit its deleterious outcomes in humans .
Flies were maintained on molasses food at 25°C unless otherwise stated . Molasses food contained 30 g Difco granulated agar ( Becton-Dickinson , Sparks , MD ) , 44 g YSC-1 yeast ( Sigma , St . Louis , MO ) , 328 g cornmeal ( Lab Scientific , Highlands , NJ ) , 400 ml unsulphured Grandma's molasses ( Lab Scientific ) , 3 . 6 l water , 40 ml propionic acid ( Sigma ) , and tegosept ( 8 g Methyl 4-hydroxybenzoate in 75 ml of 95% ethanol ) ( Sigma ) . Water and sucrose , glucose , and fructose ( all from Sigma ) vials were prepared immediately before use by placing a circular piece of Whatman filter paper ( GE Healthcare Bio-Sciences , Pittsburgh , PA ) at the bottom of the vial to absorb 200 μl of liquid . Molasses food with antibiotics contained 100 μg/ml ampicillin ( Fisher Scientific , Fair Lawn , NJ ) , 50 μg/ml vancomycin ( Sigma ) , 100 μg/ml neomycin ( Sigma ) , and 100 μg/ml metronidazole ( Sigma ) in standard molasses food , as described by Liu et al . ( 2012 ) . The DGRP collection of flies was obtained from the Bloomington Stock Center , the African collection was provided by John Pool ( UW-Madison ) , and grh mutants were provided by Melissa Harrison ( UW-Madison ) ( Mackay et al . , 2012; Bastide et al . , 2014 ) . The MI24 was determined as described in Katzenberger et al . ( 2013 ) . Fly lines not treated with the HIT device had ≤1 . 15% death within the 24 hr period examined . SNPs associated with the probability of death following TBI were identified using the DGRP Freeze 1 and 2 web tools ( Mackay et al . , 2012; Huang et al . , 2014 ) . Intestinal permeability was determined using the Smurf assay , as described by Rera et al . ( 2011 ) , ( 2012 ) . BEB permeability was determined using the fluorescence assay described by Pinsonneault et al . ( 2011 ) , except that tetramethylrhodamine-conjugated dextran ( Life Technologies , Grand Island , NY ) was used as the probe . Hemolymph was extracted from flies by centrifugation , as described by Tennessen et al . ( 2014 ) , except that flies were decapitated rather than punctured and hemolymph was extracted from heads and bodies . Also , the glass wool was packed tightly to block passage of solid material and the collected hemolymph was thoroughly mixed to resuspend bacterial that may have pelleted during centrifugation . Glucose concentration in hemolymph was performed using the glucose oxidase ( GO ) assay ( Sigma ) , as described by Tennessen et al . ( 2014 ) . Bacterial counts in whole flies were performed as described by Liu et al . ( 2012 ) . Bacterial counts in hemolymph were determined by diluting 1 μl of hemolymph into 50 μl of LB , spreading the whole sample on an LB plate , and counting the number of colonies after 2 day at 25°C . Quantitative real-time reverse transcription PCR ( qRT-PCR ) was performed on total RNA extracted from whole flies as described in Petersen et al . ( 2012 ) . PCR of 16S rDNA was performed using 1 μg of total DNA extracted from flies and primers that amplify 16S rDNA from most eubacteria ( Weisburg et al . , 1991 ) . PCR primer sequences are listed in Supplementary file 5 . | Traumatic brain injury ( TBI ) caused by a violent blow to the head or body and the resultant collision of the brain against the skull is a major cause of disability and death in humans . Primary injury to the brain triggers secondary injuries that further damage the brain and other organs , generating many of the detrimental consequences of TBI . However , despite decades of study , the exact nature of these secondary injuries and their origin are poorly understood . A better understanding of secondary injuries should help to develop novel therapies to improve TBI outcomes in affected individuals . To obtain this information , in 2013 researchers devised a method to inflict TBI in the common fruit fly , Drosophila melanogaster , an organism that is readily amenable to detailed genetic and molecular studies . This investigation demonstrated that flies subjected to TBI display many of the same symptoms observed in humans after a brain injury , including temporary loss of mobility and damage to the brain that becomes worse over time . In addition , many of the flies die within 24 hr after brain injury . Now Katzenberger et al . use this experimental system to investigate the secondary injuries responsible for these deaths . First , genetic variants were identified that confer increased or decreased susceptibility to death after brain injury . Several of the identified genes affect the structural integrity of the intestinal barrier that isolates the contents of the gut—including nutrients and bacteria—from the circulatory system . Katzenberger et al . subsequently found that the breakdown of this barrier after brain injury permits bacteria and glucose to leak out of the intestine . Treating flies with antibiotics did not increase survival , whereas reducing glucose levels in the circulatory system after brain injury did . Thus , Katzenberger et al . conclude that high levels of glucose in the circulatory system , a condition known as hyperglycemia , is a key culprit in death following TBI . Notably , these results parallel findings in humans , where hyperglycemia is highly predictive of death following TBI . Similarly , individuals with diabetes have a significantly increased risk of death after TBI . These results suggest that the secondary injuries leading to death are the same in flies and humans and that further studies in flies are likely to provide additional new information that will help us understand the complex consequences of TBI . Important challenges remain , including understanding precisely how the brain and intestine communicate , how injury to the brain leads to disruption of the intestinal barrier , and why elevated glucose levels increase mortality after brain injury . Answers to these questions could help pave the way to new therapies for TBI . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"neuroscience"
] | 2015 | Death following traumatic brain injury in Drosophila is associated with intestinal barrier dysfunction |
Conflict over parental investment between parent and offspring is predicted to lead to selection on genes expressed in offspring for traits influencing maternal investment , and on parentally expressed genes affecting offspring behaviour . However , the specific genetic variants that indirectly modify maternal or offspring behaviour remain largely unknown . Using a cross-fostered population of mice , we map maternal behaviour in genetically uniform mothers as a function of genetic variation in offspring and identify loci on offspring chromosomes 5 and 7 that modify maternal behaviour . Conversely , we found that genetic variation among mothers influences offspring development , independent of offspring genotype . Offspring solicitation and maternal behaviour show signs of coadaptation as they are negatively correlated between mothers and their biological offspring , which may be linked to costs of increased solicitation on growth found in our study . Overall , our results show levels of parental provisioning and offspring solicitation are unique to specific genotypes .
The close interaction between mother and offspring in mammals is fundamental to offspring development and fitness . However , parent and offspring are in conflict over how much parents should invest in their young where offspring typically demand more than is optimal for the parent ( Trivers , 1974; Godfray , 1995 ) , and the existence of this genetic conflict has been demonstrated in empirical research ( Kölliker et al . , 2015 ) . The resulting selection pressures are predicted to lead to the evolution of traits in offspring that influence parental behaviour ( and thus investment ) . Conversely , parental traits should be selected for their effects on offspring traits that influence parental behaviour indirectly ( Kilner and Hinde , 2012 ) . The correlation between parental and offspring traits has been the focus of coadaptation models where specific combinations of demand and provisioning are selectively favoured ( Wolf and Brodie III , 1998; Kölliker et al . , 2005 ) . The fundamental assumption underlying predictions about the evolution of traits involved in parent-offspring interactions is that genetic variation in offspring exists for traits that indirectly influence maternal investment and vice versa . However , it remains to be shown whether specific genetic variants in offspring indirectly influence maternal behaviour . In an experimental mouse population , we demonstrate that genes expressed in offspring modify the quality of maternal behaviour and thus affect , indirectly , offspring fitness . To investigate the genetics of parent-offspring interactions we conducted a cross-fostering experiment between genetically variable and genetically uniform mice , using the largest genetic reference panel in mammals , the BXD mouse population . We generated families of genetically variable mothers and genetically uniform offspring by cross-fostering C57BL/6J ( B6 ) litters , in which no genetic variation occurs between animals of this strain , to mothers of a given BXD strain . Conversely , a BXD female’s litter was cross-fostered to B6 mothers ( Figure 1 ) . Thus , we can analyse the effects of genetic variation in mothers or offspring while controlling for genetic variation in the other . This cross-fostering design has been successfully utilized in previous studies on family interactions because it breaks the correlation between maternal and offspring traits . Here , different families , or naturally occurring variation of maternal and offspring trait combinations across different broods , are assumed to represent distinct evolved strategies ( Agrawal et al . , 2001; Hager and Johnstone , 2003; Meunier and Kölliker , 2012 ) . From birth until weaning at 3 weeks of age we recorded offspring and maternal body weights and behaviour , following Hager and Johnstone ( 2003 ) . 10 . 7554/eLife . 11814 . 003Figure 1 . Experimental cross-foster design . Females of different lines of the BXD strain ( light to grey mice ) adopt B6 offspring ( dark ) and B6 females ( dark ) adopt offspring born to females of different BXD lines . A total of 42 BXD lines with three within-line repeats plus the corresponding B6 families were set up for the experiment . DOI: http://dx . doi . org/10 . 7554/eLife . 11814 . 003
While in the previous section we have focused on analysing traits within foster families , we now turn to the correlation between traits of biological families , i . e . mothers and their biological offspring . This correlation has been analysed in coadaptation models , which make specific predictions about how parental and offspring traits are correlated , and in empirical work ( Kölliker et al . , 2000; Agrawal et al . , 2001; Curley et al . , 2004; Lock et al . , 2004; Kölliker et al . , 2005; Hinde et al . , 2010; Meunier and Kölliker , 2012 ) . Prior experimental studies have found both a positive ( Parus major; Kölliker et al . , 2000: Nicrophorus vespilloides; Lock et al . , 2004 ) and negative correlation between offspring solicitation and parental traits ( Sehirus cinctus; Agrawal et al . , 2001 ) . In our study , we found a negative correlation . When we measured short-term provisioning , we found a negative correlation between BXD offspring short-term weight gain and the corresponding provisioning of their biological ( BXD ) mothers on day 10 , as well as a negative correlation between BXD offspring solicitation and the corresponding provisioning of their biological ( BXD ) mothers on day 14 ( GLM , F1 , 32 = 4 . 77 , P = 0 . 036; r = -0 . 34 and GLM , F1 , 28 = 8 . 046 , P = 0 . 008 , r = -0 . 48; Figure 5 and Supplementary file 2 , d ) . Thus , our results suggest that mothers who are generous providers produce young that solicit less maternal resources than offspring born to less generous mothers . Such a negative correlation is predicted to occur when maternal traits are predominantly under selection as long as parents respond to offspring demand ( which we have shown above; Kölliker et al . , 2005 ) . One scenario to explain this negative correlation might be that each BXD line , i . e . genotype , is characterized by a unique ( to this line , everything else being equal ) combination of offspring and maternal behaviours where higher maternal provisioning is correlated with lower offspring solicitation . This may be due to the cost of increased solicitation ( reflected in reduced bodyweight for the effort expended ) for which we found evidence in our study . Bodyweight is indeed negatively correlated with the level of offspring solicitation ( GLM , F1 , 66 = 20 . 57 , P < 0 . 001 e . g . day 10 , r = -0 . 39 , and day 14 , r = -0 . 44; Figure 6 and Supplementary file 2 , e ) . 10 . 7554/eLife . 11814 . 011Figure 5 . Correlation between offspring and maternal traits in biological BXD families . The first panel shows the correlation between BXD offspring short-term weight change per pup and provisioning of their corresponding biological BXD mother on day 10 per pup . The second panel shows the correlation between the level of BXD offspring solicitation per pup on day 14 and their mother’s provisioning per pup . DOI: http://dx . doi . org/10 . 7554/eLife . 11814 . 01110 . 7554/eLife . 11814 . 012Figure 5—source data 1 . Source data for Figure 5 as uploaded to GeneNetwork , showing BXD offspring short-term weight change on day 10 per pup , BXD maternal provisioning on day 10 , BXD offspring solicitation on day 14 per pup and BXD maternal provisioning on day 14 for the BXD lines . DOI: http://dx . doi . org/10 . 7554/eLife . 11814 . 01210 . 7554/eLife . 11814 . 013Figure 6 . Correlation between per pup offspring solicitation and corresponding body weight in BXD lines on day 10 and day 14 , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 11814 . 01310 . 7554/eLife . 11814 . 014Figure 6—source data 1 . Source data for Figure 5 as uploaded to GeneNetwork , showing BXD offspring solicitation on day 10 per pup , BXD offspring weight on day 10 per pup , BXD offspring solicitation on day 14 per pup and BXD offspring weight on day 14 per pup for the BXD lines . DOI: http://dx . doi . org/10 . 7554/eLife . 11814 . 014
Our study of the genetics underlying family interactions has revealed that genes expressed in offspring can indirectly influence the quality of maternal behaviour and thus offspring fitness . At the same time , we detected specific loci in maternal genotype that indirectly modify offspring traits , which shows that IGEs can be an important component of the genetic architecture of complex traits ( Bijma and Wade , 2008 ) . We note that while postnatal cross-fostering controls for postnatal maternal effects , prenatal maternal effects can only be addressed ( to some degree ) by embryo transfer ( e . g . Cowley et al . , 1989 ) , a procedure that is impractical in genetics experiments . Potentially , this may strengthen or weaken , for example , effects of BXD genotype on B6 maternal phenotype . At the same time , pre-natal maternal effects may also contribute to a phenotypic correlation between parent and biological offspring traits , as reported above on coadaptation . In-utero effect variation due to differences in BXD genotype may therefore contribute to differences in BXD offspring behaviour , in turn affecting the behaviour of their adoptive mothers . We now need to investigate the candidates identified here and how their effects on parental and offspring traits are integrated into the gene networks determining individual development . By controlling for genetic variation in either mothers or offspring we have been able to show that levels of maternal provisioning and offspring solicitation are unique to specific genotypes ( here each BXD line ) and that solicitation is costly . The ability to conduct complex systems genetics analyses in experimental systems of parent offspring interactions will enable us to concentrate now on understanding the underlying pathways involved , and how they are modified by social environmental conditions that determine adult phenotypes and associated reproductive success .
We used mice of the BXD recombinant inbred population , which consists of experimentally tractable and genetically defined mouse lines capturing a large amount of naturally occurring genetic variation , which underlies variation at the phenotypic level ( e . g . Chesler et al . , 2005; Hayes et al . , 2014 ) . The BXD panel incorporates ~5 million segregating SNPs , 500 , 000 insertions and deletions , and 55 , 000 copy-number variants . These lines are used for complex systems genetics analyses integrating massive phenotype and gene expression data sets obtained across years and studies ( e . g . Andreux et al . , 2012; Ashbrook et al . , 2014 ) . The 42 BXD strains used in this study ( 1 , 11 , 12 , 14 , 24 , 32 , 34 , 38–40 , 43–45 , 48a , 49–51 , 55 , 56 , 60–64 , 66–71 , 73a , 73b , 73–75 , 83 , 84 , 87 , 89 , 90 , 98 , 102 ) were obtained from Professor Robert W . Williams at the University of Tennessee Health Science Centre , Memphis , TN . C57BL/6J ( B6 ) mice were obtained from Charles River , UK . Three within-line repeats plus the corresponding 42 B6 families with three within-line repeats were set up for the experiment ( Figure 1 ) . Sample size was determined considering power analyses and logistical aspects . Mapping power is maximised with increasing number of lines whereas within-line repeats n increase confidence of line average phenotypes , which , however , rapidly diminishes as n exceeds four ( Belknap , 1998 ) . We have modelled power and effect sizes following ( Belknap , 1998 ) : n = ( Zα + Zβ ) 2 / S2QTL / S2Res ) . Zα and Zβ are Z values for a given α and β; S2QTL is the phenotypic variance due to a QTL and S2Res is the residual variance . With power ( 1-β ) of 80% , α of 0 . 05 we estimated that with 45 lines we can detect QTL at genome-wide significance explaining ~ 16% of trait variance , which is sufficient mapping power given effect sizes of prior work . We have modelled the relationship between power and number of replicates using qtlDesign ( Sen et al . , 2007 ) . Everything else being equal power can be optimized by maximizing the number of genotypes ( i . e . lines ) and reducing replicates , even with varying degrees of heritability . Thus , we set up three replicates using 45 lines , and line averages are mapped , although for some traits and some lines this number may be lower due to lower breeding success . Outliers have been retained as they represent distinct genotypes , evinced by outliers for different traits being from the same line . Interval mapping ( Haley and Knott , 1992 ) relies on 3795 informative SNP markers across all chromosomes , except Y , as implemented in GeneNetwork ( GN ) ( Hager et al . , 2012 ) . The BXD strains were genotyped using the MUGA array in 2011 , along with genotypes generated earlier using Affymetrix and Illumina platforms ( Shifman et al . , 2006 ) , and mm9 is used . Loci are identified in GN by the computation of a likelihood statistic score and significance was determined using 2000 permutations of the phenotype data . Candidates were identified within the region defined by using GeneNetwork ( http://www . genenetwork . org ) and further information combined from QTLminer ( Alberts and Schughart , 2010 ) , Entrez genes ( http://www . ncbi . nlm . nih . gov/gene ) and Mouse Genome Informatics ( Eppig et al . , 2015 ) . Mice were maintained under standard laboratory conditions in the same room , exclusively used for the experiment in individually ventilated cages ( IVC Tecniplast Green line ) , and given chow and water ad lib . Humidity ranges between 50% and 65% relative humidity , temperature between 20°C and 21°C . All animals were kept on a reverse dark light cycle with 12h red light ( active phase ) and 12h white light . Cross-fostering of entire litters took place within 24h of birth of corresponding B6 and BXD females and analyses used trait values per pup to adjust for differences in adoptive litter size , which is a significant covariate for provisioning and solicitation as of course a mother nursing a large litter will overall provide more than a mother nursing a small litter . Both mothers and litters were weighed at birth and once weekly , for three weeks until weaning , i . e . at the end of week 1 , week 2 and week 3 , respectively , to enable the calculation of growth during these periods . In addition , we recorded maternal and offspring behaviour on postnatal days 6 , 10 and 14 when we simulated maternal departure to standardize observation conditions ( Hager and Johnstone , 2003 , 2005 ) . After a 4h separation , mother and litter were re-joined and maternal and offspring behaviours recorded simultaneously over 15 min , using scan sampling every 20 s ( Martin and Bateson , 2007 ) . Provisioning is measured using an established protocol ( Hager and Johnstone , 2003 , 2005 , 2007 ) as maternal and offspring weight change after reunification with pups over the following two hours . Because rodents are nocturnal all observations occurred under red light , i . e . the active phase . Maternal behaviour was recorded as the sum of nursing , suckling and nest building . Nursing is defined as attending the litter , sitting on the nest and suckling up to half the litter while suckling refers to the entire litter being suckled at the same time . This distinction was used as sometimes it cannot be ascertained whether pups are suckling or not because of the position of the mothers in the nest . Nestbuilding behaviour is gathering nesting material and constructing a nest . Pup solicitation behaviour in mice is defined as pups attempting to suck and following the mother , but individual pups were not distinguished . All procedures were approved by the University of Manchester Ethics Committee . | Genes encode instructions that can influence the behaviour and physical traits of the individual that carries them . Individuals of the same species can carry different versions ( or variants ) of the same gene , leading to a variety of traits in the population . However , a gene expressed in one individual can also alter the traits of another individual . This is known as an indirect genetic effect . For example , a gene in a mother that affects her ability to provide care may influence how her offspring develop . Researchers have predicted that offspring should be able to manipulate their mothers to try and gain more care than the mothers are willing to give . Furthermore , the offspring born to mothers who respond to this begging are predicted to save energy and beg less . However , few gene variants that indirectly modify the behaviour of mothers or offspring have so far been identified . Ashbrook et al . used mice to test the idea that genetic variation in particular locations in the offspring’s genome can affect maternal behaviour , and vice versa . In the first experiment , mother mice with different gene variants fostered litters of mouse pups that were all genetically identical . In the other experiment , genetically identical mothers fostered litters of mouse pups with different gene variants . Ashbrook et al . identified several locations in the offspring’s genome that modified the behaviour of their foster mothers . Furthermore , the experiments also show that genetic variation among the mothers influenced the development of their offspring , independent of the genes carried by the offspring . The next steps are to identify the specific genes underlying the changes in behaviour , and the molecular and genetic pathways by which they impact indirectly on the traits of other individuals . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"genetics",
"and",
"genomics"
] | 2015 | Genetic variation in offspring indirectly influences the quality of maternal behaviour in mice |
Selective transcription of individual protein coding genes does not occur in trypanosomes and the cellular copy number of each mRNA must be determined post-transcriptionally . Here , we provide evidence that codon choice directs the levels of constitutively expressed mRNAs . First , a novel codon usage metric , the gene expression codon adaptation index ( geCAI ) , was developed that maximised the relationship between codon choice and the measured abundance for a transcriptome . Second , geCAI predictions of mRNA levels were tested using differently coded GFP transgenes and were successful over a 25-fold range , similar to the variation in endogenous mRNAs . Third , translation was necessary for the accelerated mRNA turnover resulting from codon choice . Thus , in trypanosomes , the information determining the levels of most mRNAs resides in the open reading frame and translation is required to access this information .
In trypanosomatids , RNA polymerase II dependent transcription of protein coding genes is polycistronic ( Van der Ploeg , 1986; Worthey et al . , 2003 ) . Individual monocistronic mRNAs are produced after co-transcriptional processing by trans-splicing a 39 nucleotide , capped , exon to the 5’ end and linked endonucleolytic cleavage and polyadenylation of the upstream mRNA ( Sutton and Boothroyd , 1986; Murphy et al . , 1986; Ullu et al . , 1993; Matthews et al . , 1994 ) . This mechanism of mRNA transcription has co-evolved with the structure of the genome; protein coding genes occur in tandemly arrayed clusters containing up to hundreds of genes in one transcription unit ( Ivens et al . , 2005; Berriman et al . , 2005 ) . The transcription start sites for these gene clusters have been identified ( Kolev et al . , 2010 ) and are marked by histone variants ( Siegel et al . , 2009 ) and although the nature of promoters remains incompletely defined , a GT-rich promoter element has been identified that can both initiate transcription and cause histone variant deposition ( Wedel et al . , 2017 ) . There is remarkable synteny between the genomes of trypanosomatid species that diverged hundreds of millions of years ago ( El-Sayed et al . , 2005 ) suggesting a functional organisation of genes within polycistronic transcription units . In most cases , this is not due to the presence of an equivalent of prokaryotic operons but appears to be linked to the distance from the transcription start site . For example , genes up-regulated during heat shock tend to be distal to the transcription start site so that they continue to be transcribed by elongating RNA polymerase II for >60 min after transcription initiation has stopped as a consequence of the stress ( Kelly et al . , 2012 ) . As yet , there is no convincing evidence for selective regulation of RNA polymerase II transcription of individual genes or transcription units . In the absence of regulation based on selective transcription of individual genes , the disparate levels of individual mRNAs within a trypanosome cell is dependent solely on post-transcriptional processes . In trypanosomatids , some very abundant RNA polymerase II transcribed mRNAs are encoded by tandemly arrayed multigene families so one pass of the RNA polymerase transcribes multiple copies of the genes in series producing increased numbers of pre-mRNAs . For example , α- and β-tubulins are encoded by a tandem array of 19 copies of each gene ( Ersfeld et al . , 1998 ) . However , the majority of genes are single copy and are expressed in the cell at different levels ( Kolev et al . , 2010 ) so mechanisms for regulation of mRNA abundance must exist and these must act post-transcriptionally . It is probable that the abundance of each mRNA is regulated at several levels during its production/degradation cycle: the rate of maturation against the decay rate of the pre-mRNA in the nucleus; the rate of export to the cytoplasm against the half-life of the mature mRNA in the nucleus; the half-life of the mature mRNA in the cytoplasm ( Haanstra et al . , 2008 ) . Most work aimed at understanding determinants of mRNA levels has investigated cis- and trans-acting factors that modulate the cytoplasmic half-life of the minority of mRNAs that alter in response to a developmental or an environmental trigger . In Trypanosoma brucei there are a series of developmental forms that are adapted to different host niches and the variation in mRNA expression levels between the mammalian bloodstream form ( BSF ) and the insect midgut procyclic form ( PCF ) has been used to assay for factors necessary for a decreased half-life in one developmental form . Cis-elements have been identified in the 3’UTRs that differentially affect stability and/or rate of translation of mRNAs in specific developmental stages; the best characterised being those in procyclin mRNAs ( Furger et al . , 1997; Hehl et al . , 1994 ) , cytochrome oxidase subunits ( Mayho et al . , 2006 ) , and a short stem loop , necessary and sufficient for response to external purine concentration ( Fernández-Moya et al . , 2014 ) . Manipulation of these elements rarely quantitatively recapitulates the observed variations in mRNA levels in vivo whereas transfer of an entire 3’UTR can ( Webb et al . , 2005a ) , this may reflect a combinatorial mechanism necessary for complete developmental regulation . Trans-acting factors that affect mRNA levels have been identified . The most spectacular examples of these include the RNA recognition motif-containing proteins RPB6 and RPB10 . Over expression of RBP6 is sufficient to drive PCFs through several successive developmental transitions ( Kolev et al . , 2012 ) and over expression of RBP10 causes a PCF transcriptome to switch to BSF ( Mugo and Clayton , 2017 ) . Other RNA binding proteins have been shown to have specific effects on sets of mRNAs , for example RBP42 binds within the open reading frame ( ORF ) of many mRNAs that encode proteins involved in energy metabolism ( Das et al . , 2012 ) . Similarly , ZPF3 binds the LII cis-element in the EP1 procyclin mRNA 3’ UTR and displaces a negative regulator of translation leading to increased EP1 protein abundance ( Walrad et al . , 2009; Walrad et al . , 2012 ) . These observations provide strong evidence for co-ordinated regulation of mRNA cohorts and have led to a model in which the fate of cytoplasmic mRNAs is regulated by a combinatorial mechanism dependent on a set of interactions between RNA-binding proteins and their cognate sites in individual mRNAs . In this model , the information for the range of possible fates for any mRNA is contained in discrete cis-elements . However , the model is derived from a consideration of a small subset of mRNAs prone to dramatic regulation and further , it fails to consider work that demonstrated a codon bias in highly expressed genes in T . brucei and related trypanosomatid species ( Horn , 2008; Seward and Kelly , 2016 ) . The use of favoured codons in highly expressed genes correlated with cognate tRNA gene numbers and led to the suggestion that translational selection was likely to act on highly abundant genes . Moreover , selection was also acting in proportion to mRNA abundance to reduce the biosynthesis requirements of those transcripts ( Seward and Kelly , 2016 ) . More recently , it has been shown that codon choice is a major determinant of mRNA levels in yeast ( Presnyak et al . , 2015 ) operating through a system that monitors the rate of translation elongation involving displacement of the RNA helicase DHH1 which in turn affects the activity of the NOT1 deadenylase complex ( Radhakrishnan et al . , 2016 ) . Further , codon choice is also an important determinant of maternal mRNA turnover in zebra fish ( Mishima et al . , 2016 ) and Drosophila ( Bazzini et al . , 2016 ) . Here , we provide evidence that codon use is also central to the regulation of mRNA levels in trypanosomes . A novel codon usage statistic , the gene expression codon adaptation index ( geCAI ) , was developed that maximises the relationship between codon usage and the measured transcript abundance for the full range of mRNAs present in a cell . Then geCAI predictions were tested using differently coded GFPs and mRNA levels were predicted with >90% accuracy over a ~ 25 fold dynamic range . An investigation of the mechanism of how geCAI determined mRNA levels showed that mRNAs with low geCAI scores had shorter half-lives and that turnover was prevented if translation was blocked . These observations and measurements show that for most mRNAs the majority of the information determining mRNA levels resides in the ORF and is accessed by the translating ribosome .
All subsequent experiments used procyclic form trypanosomes and GFP as a reporter . A range of GFP transgenes were integrated into the tubulin locus and were expressed after endogenous transcription by RNA polymerase II ( Figure 2 ) . The transgene-derived GFP mRNA contained an α-tubulin 5’UTR , the GFP ORF and the actin 3’UTR ( Supplementary file 2A ) . Different GFP ORFs were selected from a pre-existing library ( Kudla et al . , 2009 ) and others were newly synthesised ( Supplementary file 3 ) . On analysing transgenic cell lines , it was immediately apparent that different GFP ORFs produced a range of fluorescence levels ( Figure 2—figure supplement 1 ) . Prior to further experiments , the use of fluorescence as a proxy for GFP protein was investigated by comparing GFP expression levels measured using flow cytometry with estimates from western blotting . Three independent clones of cell lines expressing GFPs with different codon use ( eGFP , GFP 065 , GFP 226 and GFP 102; Supplementary file 3 ) were used . GFP levels were measured by flow cytometry of live cells and lysates from the same cultures were used to estimate GFP expression using densitometric scanning of a western blot and comparison with a standard curve of recombinant GFP ( Figure 2—figure supplement 2 ) . The Pearson’s correlation coefficient between the measurement of GFP by flow cytometry and estimates of GFP expression from western blotting was R2 = 0 . 965; validating the use of GFP fluorescence as a measure of total GFP protein . In initial experiments using transgenes with different ORFs , the expression of GFP measured by flow cytometry did not correlate as well as might be expected with CAI ( Carbone et al . , 2005 ) or tAI ( Tuller et al . , 2010 ) scores ( described in more detail below ) . We decided to re-visit the calculation of the association between codon scores and expression levels . Building on the observation that altering codon use affected both protein and mRNA levels for the MS2bp-GFP-nls transgene described above , the approach taken was to make an assumption that codon use directly affected mRNA levels for most constitutively expressed genes and to derive a codon statistic for genes that best correlated with their measured mRNA abundance levels . To estimate the abundance of individual mRNAs , an RNA-Seq analysis of poly ( A ) enriched RNA was performed with three biological replicates of mid-log phase PCF T . brucei cells ( Kelly et al . , 2017 ) ( EBI Array Express E-MTAB-3335 ) . The level of each mRNA was then determined and expressed as the mean of the transcripts per million transcripts ( TPM ) across all three replicates ( Supplementary file 4 ) . To develop a set of numerical codon values that could explain mRNA levels , an approach was used in which the codon usage statistic learnt a codon value by maximising the Spearman’s rank correlation coefficient between the measured mRNA level of a set of transcripts and the codon usage statistics for the cognate ORFs . Spearman’s rank correlation coefficient was used to avoid distributional assumptions about the hidden distribution of per-gene translational efficiencies and the observed distribution of mRNA abundance estimates . To calculate the codon values , the measured expression levels of mRNAs from 5136 single copy genes were used . This set of genes represented the genome with the following exclusions: ( i ) mRNAs with >2 fold differential expression when BSF and PCF trypanosomes were compared ( Kelly et al . , 2017 ) as the mRNA abundance are likely to be modulated by cis-acting RNA binding proteins and thus obscure the codon score learning process . ( ii ) mRNAs encoded by single copy genes without a homologue in at least one other kinetoplastid species , this was a mechanism to exclude any potentially spurious ORFs that may not encode proteins . ( iii ) mRNAs encoded by multicopy genes families as the accuracy of TPM calculations is compromised by uncertainty of copy number and associated errors in allocation of many sequence reads to individual genes . The calculation of codon weights Ci proceeded by randomly generating a set of codon scores S were Ci [0 , 1] . Here , a score of 1 means that a codon has the maximum positive effect on mRNA levels and a score of 0 the minimum effect . The geCAI score of an ORF was evaluated as the geometric mean of the scores for all codons in that ORF . The Spearman’s rank correlation coefficient between geCAI scores for the 5136 ORFs and mRNA abundances was then computed . A Markov chain Monte Carlo algorithm was then employed to introduce stochastic changes into the codon score matrix and the entire set of genes was rescored and the Spearman’s rank correlation coefficient recalculated . If in a given generation of the Markov chain , a matrix was found that produced a higher correlation coefficient than the current best matrix , then it replaced the current best matrix and formed the basis for subsequent generations of stochastic modification . One thousand chains each starting from a different random starting matrix were initiated and allowed to run for 5000 generations . In all cases , stationary phase was reached between 1500 and 2000 generations ( Figure 3A ) . The gene expression codon adaptation index ( geCAI ) values for each codon were calculated as the median of the values obtained from the 1000 chains ( Figure 3B and Table 1 ) . When mRNA abundance expressed as TPM was plotted against the ORF geCAI scores ( Supplementary file 4 ) , the Spearman’s rank correlation coefficient for the 5136 mRNAs was ρ = 0 . 55 ( Figure 3C ) . A mid-log phase PCF cell contains approximately 50000 mRNA molecules ( Dhalia et al . , 2006 ) and so 20 TPM is equivalent to one mRNA molecule/cell , so the range of transcript abundance for the vast majority is between ~1 and 40 mRNAs/diploid gene pair/cell . The use of geCAI allows a prediction of a range of expression level for an mRNA based on the sequence of the ORF alone , for example an ORF with a geCAI score of 0 . 35 will be expressed at 2 . 5 mRNAs/cell on average with the vast majority of mRNAs in a range between 1 and 5 . Several previously developed codon usage statistics were tested to determine if these could explain the measured mRNA levels . Of the tested methods , CAI ( Carbone et al . , 2005 ) and tAI ( Tuller et al . , 2010 ) provided the best correlation but the predictive capacity was limited with Spearman’s rank correlation coefficients of ρ = 0 . 16 and ρ = 0 . 13 respectively ( Figure 3—figure supplement 1 ) . These measures may be unsuitable for this type of analysis as each assumes the effectiveness of the 'best' codon for each amino acid is equivalent , for example that the most efficient codon for glycine will have the same effect as the most efficient codon for alanine . There is evidence that this is not the case ( Gardin et al . , 2014 ) . The geCAI values for each codon in Table 1 are derived from the mRNA levels measured in the transcriptome analysis used in this study . There is variation in the measurements of mRNA levels in transcriptome analyses from different studies that probably has a range of origins: identity of the cell line , growth conditions , methodology of RNA preparation , RNA-Seq and data analysis and similar variation has been found in studies in yeast ( Harigaya and Parker , 2016 ) . The consequence is that any table of geCAI codon values contains an element reflecting how mRNA levels were measured , but the underlying principle that codon choice is a major determinant of mRNA levels is unaffected . To illustrate this point , transcriptome data from three other studies were analysed to determine geCAI values as above . The Pairwise Pearson correlation ( r ) between log2 ( mRNA abundance ) measures from this study and three others ranged from 0 . 68 to 0 . 86 ( Supplementary file 5A ) . The calculation of geCAI values for each codon showed variation derived from the differences in transcriptome measurements ( Supplementary file 5B and C ) . Each set of geCAI codon values was then used to derive geCAI scores for each of the 5136 mRNAs and these were plotted against the mRNA expression level determined in the cognate study ( Figure 3—figure supplement 2 ) . In each case , there was still a relationship between geCAI scores for ORFs and expression levels with a range of Spearman’s rank correlation coefficients from ρ = 0 . 35 to ρ = 0 . 50 . Twenty-two different GFP ORFs ( Supplementary file 3 ) were used to systematically test the predictions of expression levels made by geCAI scores . For GFP protein , expression level was measured by flow cytometry of three independent clones for each GFP transgene , in every case there was less than 5% variation in GFP protein abundance between the three clones ( Supplementary file 6A and B ) . The GFP protein abundance was expressed relative to eGFP and plotted against geCAI score ( Figure 4A ) ; the geCAI score was able to predict GFP protein abundance with 84% accuracy ( Pearson’s correlation coefficient , r2 = 0 . 84 ) . This compared favourably with the correlation coefficients between GFP protein abundance and either CAI ( r2 = 0 . 25 ) or tAI ( r2 = 0 . 61 ) ( Supplementary file 6B ) . Next , the mRNA abundance for four GFP transgenes with different coding sequences with a range of geCAI scores was determined by RNA-Seq analysis . Equal numbers of cells of the four cell lines were mixed , RNA extracted and expression levels for each GFP mRNA measured in transcripts per million transcripts ( TPM ) . Three separate analyses were performed using three independent clones for each GFP transgene . In analysing the results , the measured expression level of each GFP mRNA was adjusted to compensate for the effect of gene copy number and the dilution effect from the mixing of four different cell lines . Trypanosomes are diploid and any endogenous single copy gene will be present in two copies per cell , each of the four cell lines contained a haploid copy of a different GFP transgene so in the mixed population used to prepare RNA there are eight copies of an endogenous gene for one copy of each of the individual GFP transgenes , effectively an 8-fold lower copy number . After adjusting for the effective difference in copy number within the sequenced pool , GFP mRNA abundance was plotted against geCAI values ( Figure 4B ) . The Pearson’s correlation coefficient between GFP mRNA levels and geCAI was r2 = 0 . 92 ( Figure 4B and Supplementary file 7 ) . Therefore , as the genomic integration site , UTRs and amino acid sequence were all identical , codon use was the major determinant of steady state levels for the transgene derived GFP mRNA and protein . The ~25 fold range of GFP mRNA levels , 2 . 5 to 60 mRNAs/diploid gene pair equivalent/cell ( Figure 4B ) is similar to the range present in mRNAs from single copy genes in the transcriptome . The measurements for the GFP mRNAs are in the higher end of the endogenous transcriptome , probably because the range of GFP geCAI scores included values higher than those present in most endogenous single copy genes . Thus , altering codon use can account for the range of steady state level of most mRNAs and the geCAI score is a good predictor of mRNA level for single copy genes not subject to developmental regulation . Furthermore , discrepancy between geCAI score for a gene and the observed mRNA or protein abundance for that gene is an indicator that other regulatory mechanisms , such as cis-acting RNA binding proteins may be acting on that gene to modulate mRNA abundance . mRNAs for cytosolic ribosomal proteins have significantly higher geCAI scores than those encoding mitochondrial ribosomal proteins The mRNA abundance measurements used to obtain geCAI codon weights did not include mRNAs for 67 of the 75 cytosolic ribosomal proteins as these are encoded by gene families with two or three members . Comparison of the geCAI scores for mRNAs encoding cytoplasmic and mitochondrial ribosomal proteins provided a test of geCAI scores for endogenous genes with orthologous function but different expression levels , there are many more ribosomes in the cytoplasm than in the mitochondrion . Lists of ribosomal protein genes were derived from the structures of both types of ribosome ( Hashem et al . , 2013; Zíková et al . , 2008 ) and geCAI scores calculated for each mRNA encoding a ribosomal protein . The more abundant cytosolic ribosomal protein mRNAs had higher geCAI scores than the less abundant mitochondrial ribosome protein mRNAs ( Figure 5 and Supplementary file 8 ) . The mean geCAI scores ( ±SD ) were 0 . 439 ( ±0 . 034 ) for cytoplasmic and 0 . 378 ( ±0 . 033 ) for mitochondrial ribosomal protein mRNAs respectively . The probability of this difference arising by chance is <0 . 00001 ( unpaired two-sample t test with equal variance ) . These measurements are consistent with more abundant proteins are encoded by mRNAs with higher geCAI scores . The origin of the differences in GFP mRNA levels from transgenes with differently coded ORFs was investigated by measuring turnover of five different GFP mRNAs by quantitative northern blotting after the inhibition of trans-splicing , and thus mRNA maturation , with sinefungin ( Pugh et al . , 1978 ) ( Figure 6A ) . The two GFPs with the highest geCAI scores ( eGFP: 0 . 547 and GFP P3: 0 . 520 ) had no detectable decay over the time course of 60 min , these both have geCAI scores greater than nearly all endogenous mRNAs . The three other GFPs with lower geCAI scores ( GFP P2: 0 . 493 , GFP 226: 0 . 432 and GFP 102: 0 . 375 ) had a rate of decay inversely proportional to the geCAI value ( Figure 6B and Supplementary file 9 ) . For these three GFP mRNAs decay to 50% occurred between 10 and ~80 min , similar to the range of half-lives reported for endogenous mRNAs ( Fadda et al . , 2014 ) . One study has estimated the half-lives of most mRNAs in procyclic form trypanosomes ( Fadda et al . , 2014 ) , and these measurements were plotted against geCAI scores determined from the cognate transcriptome and this study ( Figure 6—figure supplement 1 ) . The Spearman’s rank correlation coefficient was ρ = 0 . 22 for the cognate geCAI values and ρ = 0 . 33 for the geCAI values used in this study . Thus , geCAI scores calculated using either study can explain a significant proportion of variance in mRNA turnover rates at a transcriptome wide level . The observations above provide evidence that codon use is a major determinant of mRNA half-life , which in turn implies translation is involved . This was tested directly by blocking or reducing translation through the inclusion of secondary structures in the 5’UTR . Five different GFP transgene constructs with a range of geCAI scores were modified by insertion of a 48 base hairpin ( 24 base pairs ) in the 5’UTR located 59 nucleotides from the cap and 88 nucleotides from the initiation codon ( Supplementary file 2B ) . The same approach has been used previously in trypanosomes to block translation of a specific mRNA ( Webb et al . , 2005b ) . Measurements of GFP protein and mRNA expression were made in three independent clones for each transgene . As expected , there was no detectable expression of GFP by flow cytometry in any of the cell lines containing a transgene with a hairpin ( Supplementary file 6C and 6D ) , therefore the hairpin-containing mRNAs were not translated . Measurements of the mRNA levels by quantitative northern blotting revealed that the effect of inclusion of the hairpin in the 5’UTR was to stabilise the GFP mRNAs ( Figure 7A and Supplementary file 6E ) and the lower the geCAI score the greater the relative increase in abundance of the hairpin to control mRNAs from 1 . 4-fold for eGFP ( geCAI = 0 . 547 ) to 14-fold for GFP 102 ( geCAI = 0 . 375 ) ( Figure 7B ) . Could this observation be explained by the persistence of decapped mRNA stabilised by the hairpin blocking 5’ to 3’ decay ? This is very unlikely as an mRNA with a similar size and GC-content hairpin in the 5’UTR was readily degraded in procyclic trypanosomes ( Webb et al . , 2005b ) . These observations provide evidence that: ( i ) the translation of an ORF is necessary for the variation in mRNA half-life conferred by the geCAI , ( ii ) the different GFP mRNA levels result from active destabilisation , and ( iii ) none of the mRNAs with differently coded GFPs are inherently unstable due to sequence alone . The experiments above indicated that translation is necessary for geCAI score-mediated mRNA turnover . The 5’UTR hairpin approach above was modified by decreasing the length of double stranded RNA inserted into the 5’UTR until eGFP protein was expressed . Reduction of the length of the hairpin from 24 bp to 18 bp still blocked translation and no GFP fluorescence was observed . In contrast , a 12 bp stem loop resulted in eGFP protein expression that was 11% of that from the parental transgene without a hairpin in the 5’UTR ( Supplementary file 6F ) . The mRNA levels from three independent clones expressing GFP transgenes with the wild type 5’UTR , 18 bp hairpin containing 5’UTR and 12 bp stem loop containing 5’UTR were measured ( Figure 8 and Supplementary file 6G ) , the presence of the 12 bp stem loop resulted in a modest reduction to 76% in the steady state GFP mRNA levels ( p=0 . 13 for a difference between the mRNAs with and without a 12 bp stem loop; unpaired two-sample t test ) . This observation that a nine-fold reduction in the frequency of translation has only a small effect on the GFP mRNA levels favours a model in which geCAI score is interpreted by the translating ribosome by a mechanism mostly independent of ribosome density/frequency of translation , possibly the rate of ribosome progression . Two sets of ribosomal profiling data are available that contain values for ribosome density for >5000 of the 5136 non-developmentally regulated mRNAs expressed from single copy genes that were used to derive geCAI values ( Vasquez et al . , 2014; Jensen et al . , 2014 ) . There is a correlation between ribosome density and mRNA levels but there is also great variation in the ribosome density on different mRNAs with similar expression levels ( Jensen et al . , 2014 ) indicating frequency of translation/ribosome density alone is not sufficient to determine translationally regulated mRNA turnover . The geCAI scores of mRNAs were plotted against ribosomes density for the two ribosome profiling datasets ( Supplementary file 10 ) . The Pearson’s correlation coefficient between ribosome footprint levels and geCAI was R2 = 0 . 141 and 0 . 244 for the two sets of ribosome density values . Thus occupancy/frequency of translation can explain a significant component of the geCAI measure , but suggests a model in which geCAI mediated mRNA turnover is also dependent on other factors . The geCAI values for each codon were calculated using a set of values for mRNA expression levels that excluded developmentally regulated mRNAs , defined in the case as having a twofold or greater difference in abundance when PCF cells were compared with BSF cells . Analysing only single copy genes , the geCAI scores for 372 mRNAs upregulated in PCF cells , and 176 mRNAs upregulated in BSF cells were calculated and plotted against expression levels in PCF cells ( Figure 9 ) . mRNAs upregulated in PCFs are largely expressed at higher levels than would be predicted by geCAI alone . In contrast , mRNAs downregulated in PCFs ( upregulated in BSFs ) are expressed at levels lower than predicted by geCAI . Taken together , these measurements provide evidence that developmental regulation results from both stabilisation and destabilisation pathways that act in concert to modulate the geCAI determined mRNA levels .
Initial attempts at a numerical analysis of codon choice and mRNA levels using existing codon metrics were not particularly successful especially for the expression of reporter genes containing codons found infrequently in abundant mRNAs . For example , the most commonly used codon metric is the codon adaptation index ( CAI ) ( Sharp et al . , 1987 ) and previous work had found only a weak correlation between CAI and protein levels in both T . brucei ( Horn , 2008 ) and the related species Leishmania mexicana and L . major ( Subramanian and Sarkar , 2015 ) . However , this association between codon use and mRNA abundance suggested that perhaps an alternative metric may have increased explanatory power . The aim in developing geCAI was to account for the full range of mRNA levels present in the procyclic form of T . brucei . The measured mRNA levels for the set of single copy , non-developmentally regulated genes were used to compute a Spearman’s rank correlation coefficient between transcript abundance and the geCAI score . The final Spearman’s rank correlation coefficient between geCAI score and mRNA levels was ρ = 0 . 55 , indicating that codon usage is responsible for more than 50% of the variation in the expression of mRNAs from different genes . The main difference between the geCAI and other CAI statistics is that the codon weight matrix is global rather than local . Specifically , for a conventional CAI matrix one codon for each amino acid must be assigned a value of 1 , other synonymous codons for each amino acid must have values ≤ 1 and>0 . This makes the assumption that the translation efficiency of the optimal synonymous codon for each amino acid is equivalent . This local constraint ( one codon for each amino acid has a value = 1 ) is not enforced when calculating geCAI but a single global constraint that stipulates that at least one from the 61 codons must have the maximum value of 1 and that all other codons must have values ≤ 1 and>0 . Thus , unlike other codon weight matrices , such as CAI ( Sharp et al . , 1987 ) and tAI ( dos Reis et al . , 2004 ) , the optimal synonymous codon for one amino acid does not necessarily have the same value as the optimal synonymous codon for any other amino acid . This more accurately represents the real relationship between codons and thus enables greater explanatory power . In the set of geCAI values for all amino acids , only six codons have a value of 1 and there is also an unexpected variation in values between codons encoding similar amino acids . For example , all six codons for serine have geCAI values below 0 . 25 whereas threonine codons range from 0 . 75 to 1 ( Figure 3B ) . The selection for these contrasting geCAI values for similar amino acids is not obvious but may be related to maintaining most mRNAs at low copy number . The geCAI scores provided a prediction of a range mRNA abundance , for example the vast majority of mRNAs encoding ORFs with a geCAI score of 0 . 35 vary between 1 and 5 mRNA molecules per cell . This range of mRNA levels suggests that there are determinants of the final mRNA levels that lie outside codon use: first , through the action of cis-elements and trans-acting factors the alter mRNA stability directly . Second , it is not known whether all the polycistronic transcription units are transcribed at the same rate , so some of the variation might arise from differential transcription rates . Third , the cohort of single copy genes will include some that are developmentally regulated in one of the several other developmental forms not considered here . However , the predictive ability of geCAI scores for mRNA expression level is remarkable and , along with the experiments altering translation of the reporter mRNA , provides good evidence that the half-life of most mRNAs is set by a translation associated process . Developmentally regulated mRNAs provide an example of cohorts of mRNAs that alter in response to external signals . In the PCFs used in experiments here , the expression levels of the sets of developmentally regulated mRNAs do not correlate well with geCAI scores . Many mRNAs down regulated in PCFs ( when compared to BSFs ) were expressed at levels lower than predicted by geCAI scores . This might be expected if the intrinsic geCAI expression level of these mRNAs is modulated by cis-elements in the UTRs to destabilise the mRNA ( Webb et al . , 2005a; Clayton , 2013 ) . Conversely , many mRNAs upregulated in PCFs relative to BSFs were expressed at levels greater that would be predicted by the geCAI score . This indicates that these mRNAs are actively stabilised in PCFs and that developmental regulation comes from a combination of active stabilisation and destabilisation that modulates geCAI score mediated stability . Two comparative analyses of BSFs and PCFs by ribosome profiling showed that regulated translation was common ( Vasquez et al . , 2014; Jensen et al . , 2014 ) . Thus , could inhibition or stimulation of translation directly cause alteration to mRNA levels ? The evidence here where reducing the frequency of translation had little effect on mRNA levels indicates the two are not necessarily linked . Further , the weak correlation between geCAI score for an ORF and the ribosome density indicates that abundant mRNAs ( high geCAI score ) are not necessarily translated frequently . However , should a trans-acting factor act in a similar manner to Fragile X mental retardation protein ( Darnell et al . , 2011 ) and cause ribosome slowing or stalling during translation then an effect on mRNA stability could follow . The differential speed of progression of a ribosome on different codons is generally held to be dependent on the supply of cognate charged tRNAs ( Fluitt et al . , 2007; Chu et al . , 2011; Chu et al . , 2014 ) and this forms the basis of the tRNA adaptation index ( tAI ) ( dos Reis et al . , 2004 ) . The availability of charged tRNAs in vivo is not easily determined and approximations using the copy number of genes for each tRNA ( the larger the copy number , the more tRNA ) do not take account of post-transcriptional modifications to the tRNA that alter codon binding ( Novoa and Ribas de Pouplana , 2012 ) ; the latter has been used to improve the interpretation of codon bias ( Novoa et al . , 2012 ) . In yeast there are >270 tRNA genes , whilst in the T . brucei genome there are only 66 tRNA genes including two for selenocysteine ( Tan et al . , 2002; Padilla-Mejía et al . , 2009 ) . The partially characterised modifications of tRNAs in trypanosomes ( for example [Rubio et al . , 2007; Rubio et al . , 2006; Gaston et al . , 2007; Bruske et al . , 2009; Ragone et al . , 2011; Krog et al . , 2011] ) and the uncertain effect of base modifications and wobble on efficiency of decoding remains unclear and means it is not trivial to relate the geCAI value for each codon to the cognate tRNA abundance . Codon optimality has also been shown to be a major determinant of mRNA stability in yeast ( [Presnyak et al . , 2015] and reviewed in [Heck and Wilusz , 2018] ) and operates through the detection of slower ribosome progression by the RNA helicase DHH1 ( Radhakrishnan et al . , 2016 ) which in turn can activate mRNA decay ( Coller et al . , 2001; Sweet et al . , 2012 ) . The findings here provide evidence that a similar process that links speed of ribosome progression to mRNA decay operates in trypanosomes . An orthologue of DHH1 is present in trypanosomes ( Schwede et al . , 2008; Kramer et al . , 2010 ) and co-precipitates with the NOT deadenylase complex ( Schwede et al . , 2008; Färber et al . , 2013 ) . Moderate over expression of DHH1 resulted in altered levels of developmentally regulated mRNAs and a slowing of growth ( Kramer et al . , 2010 ) but little else is known and unlike yeast , DHH1 is essential in trypanosomes . In several animals , maternal mRNAs are marked by codon use for degradation at the maternal-to-zygotic transition and this may reflect changes in tRNA availability or re-activation of the degradation pathway ( Mishima et al . , 2016; Bazzini et al . , 2016 ) . Changes in tRNA pools in response to external stimuli is an attractive model for regulating gene expression in developmental transitions and deletion of individual tRNA genes can alter mRNA levels in yeast ( Bloom-Ackermann et al . , 2014 ) . Direct measurements of the tRNA pools in trypanosomes would inform whether a similar change underpins any of developmental transitions in trypanosomes . In summary , we have developed and validated a new codon use metric , the gene expression codon adaptation index ( geCAI ) and shown that it can be used to predict the expression level of many mRNAs in PCF trypanosomes . The lack of selective regulation of transcription of most protein coding genes in trypanosomes means that the effect of codon use on mRNA levels is particularly apparent . Mechanistically , the evidence supports a model in which the speed or processivity of ribosome progression determines mRNA half-life . This is similar to the process identified in yeast and , if there is a common mechanism in yeast and trypanosomes , it is likely that it has a very early evolutionary origin possibly dating from the last eukaryotic common ancestor , and that it may be present in many diverse eukaryotes .
A script to enable estimation of geCAI codon weights is available for download under the GPL V3 . 0 licence at https://github . com/SteveKellyLab/geCAI ( Kelly , 2017; copy archived at https://github . com/elifesciences-publications/geCAI ) . A codon weight matrix was generated with each codon randomly assigned a weight in the interval [0 , 1] with a constraint that at least one codon had a value of 1 . These codon values were used to calculate a score each ORF as the geometric mean of the constituent codon values . The Spearman’s rank correlation coefficient between the transcript abundance estimates and the ORF scores was then computed . The Spearman’s rank correlation coefficient was then maximised by using a Markov chain Monte Carlo algorithm to introduce stochastic changes into the codon weight matrix and selecting matrices that led to increased Spearman’s rank correlation coefficients . 1000 chains were run for 5000 generations and the final codon geCAI value was calculated as the median value of the 1000 chains . Three biological replicates of T . brucei TREU 927 procyclic form cells were grown in SDM-79 ( Brun and Jenni , 1977 ) and harvested at 6 × 106/ ml and RNA prepared using the Qiagen RNAeasy kit . Cells were washed once with ice cold PBS before the lysis step in RNA preparation procedure . This lysis step was performed within 5 min of removal from the growth incubator . The cDNA libraries were prepared and sequenced at the Beijing Genomics Institute ( Shenzhen , China ) ( Fiebig et al . , 2015 ) . In brief , polyadenylated RNA was purified from total RNA , converted to cDNA using random hexamer primers sheared and size selected for fragments ~ 200 bp in length using the Illumina TruSeq RNA Sample Preparation Kit v2 . RNAseq of the resulting libraries was used for the determination of transcript abundances . Sequencing was performed on an Illumina Hiseq 2000 ( Illumina , CA ) platform . Paired end reads were subject to quality trimming and adaptor filtering using Trimmomatic ( Bolger et al . , 2014 ) using the settings ‘LEADING:10 TRAILING:10 SLIDINGWINDOW:5:15 MINLEN:50’ . Trimmed reads were then quantified against the T . brucei TREU927 v6 transcripts using RSEM ( Li and Dewey , 2011 ) using the default settings for RSEM . Bowtie 2 ( Langmead and Salzberg , 2012 ) is used as part of the RSEM protocol using setting described in the paper ( Li and Dewey , 2011 ) . For computing of geCAI scores all transcript abundance estimates ( Transcripts per million transcripts ) were averaged across the three biological replicates . The transcript reads are in EBI ArrayExpress E-MTAB-3335 and developmental regulation of mRNA expression was derived from a comparison of data in EBI ArrayExpress E-MTAB-3335 . eGFP was purchased from Clontech ( Takara , Kyoto , Japan ) . GFPs 71 , 183 , 188 , 194 , 226 , 102 , 163 , 205 , 211 and 224 were kindly provided by Grzegorz Kudla ( Kudla et al . , 2009 ) . All others GFP sequences were commercially synthetized ( Eurofins MWG Operon ) . For all GFP sequences used see Supplementary file 3 . All plasmids and oligonucleotides used in this work are described in Supplementary file 11 . All GFP sequences were cloned into the vector p3827 using HindIII and BamHI restriction sites . Plasmid p3827 is designed to integrate transgenes into the T . brucei tubulin locus which contains a long tandem array of alternating alpha and beta tubulin genes ( Figure 2 ) . Digestion with PacI released a fragment that has the following components in order: alpha to beta tubulin inter-ORF sequence , transgene ORF; actin inter-ORF sequence , blasticidin resistance ORF , alpha to beta tubulin inter-ORF . After integration , the transgene is transcribed from the distant endogenous RNA polymerase II promoter . In order to generate a hairpin in the 5’-UTR of a GFP transgene , p4432 ( Supplementary file 11 ) was first modified by the addition of an EcoRV restriction site using BglII and HindIII sites and phosphorylated oligonucleotides D799 , D780 , D781 and D782 , making the plasmid p4699 . After , p4699 was linearized with EcoRV and ligated with the phosphorylated oligonucleotides D797 and D798 , generating p4724 . GFP ORFs for GFP 102 , GFP 226 , GFP P2 and GFP 065 were cloned into p4724 using HindIII and BamHI restriction sites . Similarly , in order to generate shorter hairpins on the 5’-UTR of the eGFP , p4699 was linearized with EcoRV and ligated with the phosphorylated oligonucleotides E417 and E418 generating p4841 , or E419 and E420 , generating p4842 . Standard procedures were used for generating all plasmids , the sequences of the GFP transcript , modifications to the 5’UTR and p4432 are in Supplementary file 2 . All experiments were performed using the procyclic developmental form of Trypanosoma brucei Lister 427 KG ( a kind gift of Keith Gull ) . Cells were cultured in SDM-79 at 27°C and 5% CO2 . Electroporations were as described in ( McCulloch et al . , 2004 ) using 5–10 μg of PacI digested plasmid . 10 μg/ml of blasticidin was use to select the transfectants . Independent clones were picked on the tenth day after transfection . Three independent clones of each cell line , growing in mid-logarithmic phase , were analysed by flow cytometry . Data was acquired using FACScan ( Becton Dickinson , Franklin Lakes , NJ ) and analysed using Cell Quest V3 . 3 software . Flow Check Fluorospheres ( Beckman Coulter , Pasadena , CA ) were used to normalise the readings from experiments performed on different days . Protein samples were loaded in 17 . 5% SDS-PAGE and transferred to Immobilon membrane ( Millipore , Burlington , MA ) . GFP Rabbit IgG Polyclonal Antibody Fraction ( Life Technologies , Waltham , MA ) was used to detect eGFP and a cross-reacting band was used as loading control . Second incubation was with peroxidase conjugated donkey anti-rabbit IgG . Detection was carried by enhanced chemiluminescence ( ECL ) using Fuji Medical X-Ray Film . Films were digitalised and the signal was quantified using ImageJ 1 . 48 v . RNA was prepared from cultures with cell densities between 4 and 6 × 106/ ml using the RNeasy Mini Kit ( Qiagen , Hilden , Germany ) and quantified using Nanodrop and ethidium bromide stained agarose gels . All experiment used three biological replicates with independent clones for each cell line . Cells were washed once with room temperature serum-free culture medium before the lysis step in RNA preparation procedure . This lysis step was performed within 4 min of removal from the growth incubator . For the measurement of GFP mRNA levels , equal numbers of cells expressing eGFP , GFP 226 , GFP 102 and GFP S5 were mixed and RNA prepared as above . RNAseq and quantitation was as above . For the quantification of the half-lives of different GFP , cells expressing eGFP , GFP P2 , GFP P3 , GFP 226 , GFP 102 were treated with 2 μg/ml of sinefungin and RNA prepared as above at 0 , 15 , 30 and 60 min . For northern blots , 3 μg of each sample was loaded in 1 . 2% agarose gel after denaturation with glyoxal ( Carrington et al . , 1987 ) . Quantitative analyses were done using phosphoimager and ImageQuant TL 1D v7 . 0 ( GE Healthcare ) . All probes used were the complete ORFs: each GFP was probed specifically with the cognate ORF . 18S ribosomal RNA was used as loading control . | Genes are made up of DNA and contain the instructions to make molecules called proteins . This information is stored as a genetic code consisting of four bases: adenine ( A ) , cytosine ( C ) , guanine ( G ) and thymine ( T ) . The order of these bases and their different combinations serves as a blueprint for making thousands of different proteins and to assemble living cells . Converting the information in the genes into proteins requires several steps . First , the code from the DNA needs to be transcribed into RNA and then processed to make messenger RNA , or mRNA for short , which in turn is translated into proteins . Cells decode mRNAs by reading the bases as groups of three , also called codons . Most codons specify an amino acid – the building blocks of proteins – but certain codons also mark the start and end point of a protein . This ensures that the mRNA is read in the correct ‘frame’ and the desired proteins are made . Any cell contains thousands of different proteins and each protein has its own unique level . The mechanisms used to set the number of each different type of protein can operate at every point in the process . In many organisms , the number of times a gene is transcribed to make an mRNA , underpins differences in protein levels . Trypanosomes , for example , are parasites that cause a range of devastating diseases in humans and livestock . They lack the ability to set individual mRNA levels by regulating how often the gene is transcribed . This suggests that the expression of thousands of mRNAs is regulated by a common control mechanism later in the process ending in protein synthesis . However , until now , it was unclear what these mechanisms are . Most amino acids are encoded by more than one codon . The different codons for one amino acid are not equivalent and using a different codon can lead the mRNA to yield more or less protein . Evolution acts on these differences between codons , and the ‘codon choice’ in any one mRNA represents the outcome of natural selection . Now , Nascimento , Kelly et al . found that codon choice directs both the levels of mRNAs and the level of translation . For the experiments , a new metric that enables a prediction of the level of expression for each mRNA was created . This metric ( known as the ‘gene expression codon adaptation index’ or geCAI for short ) could relate the codon choice to mRNA levels . For example , mRNAs with a low index score had shorter half-lives , i . e . , how long that mRNA remained in the cell before being broken down . Nascimento , Kelly et al . confirmed this by measuring mRNA levels in specific genes tagged with distinguishable markers and revealed that the codon choice indeed dictated the rate at which an mRNA would be broken down . A separate study by Jeacock , Faria and Horn looked more closely at how codon choice contributes to the control of the copy number of proteins . However , genes and mRNAs involved in development could deviate from the levels predicted by the geCAI metric , which suggests that other mechanisms may be in place to control the stability these mRNAs . The importance of codon choice in setting mRNA levels has now been demonstrated in several organisms , including yeast and trypanosomes , which suggests that this process is more widespread than previously realised . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"microbiology",
"and",
"infectious",
"disease"
] | 2018 | Codon choice directs constitutive mRNA levels in trypanosomes |
Context-dependent changes in genetic interactions are an important feature of cellular pathways and their varying responses under different environmental conditions . However , methodological frameworks to investigate the plasticity of genetic interaction networks over time or in response to external stresses are largely lacking . To analyze the plasticity of genetic interactions , we performed a combinatorial RNAi screen in Drosophila cells at multiple time points and after pharmacological inhibition of Ras signaling activity . Using an image-based morphology assay to capture a broad range of phenotypes , we assessed the effect of 12768 pairwise RNAi perturbations in six different conditions . We found that genetic interactions form in different trajectories and developed an algorithm , termed MODIFI , to analyze how genetic interactions rewire over time . Using this framework , we identified more statistically significant interactions compared to end-point assays and further observed several examples of context-dependent crosstalk between signaling pathways such as an interaction between Ras and Rel which is dependent on MEK activity . Editorial note: This article has been through an editorial process in which the authors decide how to respond to the issues raised during peer review . The Reviewing Editor's assessment is that all the issues have been addressed ( see decision letter ) .
Gene-gene interactions – the epistatic influences of one gene’s effect on the function of another gene – have widespread effects on cellular and organismal phenotypes , ranging from fitness defects in unicellular organisms to interactions between germline and somatic variants in cancer ( Baryshnikova et al . , 2013; Billmann and Boutros , 2017; Boone et al . , 2007; Burgess , 2016; Carter et al . , 2017; Ideker and Krogan , 2012; Mani et al . , 2008; Phillips , 2008; Taylor and Ehrenreich , 2015 ) . In past studies , statistical genetic interactions ( also simply referred to as genetic interactions ) have been defined as an unexpected phenotypic outcome observed upon simultaneous perturbations ( or knock-outs ) of two genes that cannot be explained from the genes’ individual effects ( Beltrao et al . , 2010; Fisher , 1930; Mani et al . , 2008 ) . Genetic interactions can be discovered using pairwise perturbations of genes , a strategy which has been experimentally used at large scale in yeast ( Collins et al . , 2007; Costanzo et al . , 2010; Fiedler et al . , 2009; Tong et al . , 2001 ) , C . elegans ( Lehner et al . , 2006 ) , Drosophila ( Fischer et al . , 2015; Horn et al . , 2011 ) , E . coli ( Babu et al . , 2011 ) and human cells ( Kampmann et al . , 2013; Laufer et al . , 2013; Roguev et al . , 2013; Shen et al . , 2017 ) . To create genetic interaction maps , these studies systematically identified alleviating ( e . g . better fitness than expected ) or aggravating ( e . g . worse fitness than expected ) genetic interactions , which can then be used to generate ‘genetic interaction profiles’ for each gene . Several studies have shown that genes involved in the same cellular processes have highly similar genetic interaction profiles , which therefore can be used to create maps of cellular processes at a genome-wide scale ( Costanzo et al . , 2010; Costanzo et al . , 2016; Fischer et al . , 2015; Pan et al . , 2018; Rauscher et al . , 2018; Tsherniak et al . , 2017; Wang et al . , 2017; Yu et al . , 2016 ) . In addition to univariate phenotypes , such as fitness and growth phenotypes of cells or organisms , genetic interactions can be measured for a broader spectrum of phenotypes by microscopy and image-analysis ( Horn et al . , 2011; Laufer et al . , 2013; Roguev et al . , 2013 ) . Importantly , by allowing to infer the direction of specific genetic interactions , multivariate phenotypes further opened the possibility to predict a hierarchy of epistatic relationships of components in genetic networks ( Fischer et al . , 2015 ) . To date , most studies of genetic interactions focused on ‘static’ environmental conditions ( e . g . under optimal culture conditions ) , ignoring the impact of context-dependent changes . Recently , several studies have more specifically analyzed the influence of environmental changes on genetic interactions ( Bandyopadhyay et al . , 2010; Billmann and Boutros , 2017; Díaz-Mejía et al . , 2018; Guénolé et al . , 2013; Martin et al . , 2015; St Onge et al . , 2007; Wong et al . , 2015 ) . For example , Bandyopadhyay et al . ( 2010 ) defined static , positive and negative differential interactions that vary under changing environmental conditions . ( Billmann and Boutros , 2017 ) used extrinsic and intrinsic changes of Wnt signaling in cultured Drosophila cells to map differential genetic interactions using a pathway-centric functional readout . These studies demonstrated that widespread changes in genetic interactions occur upon changes in environmental conditions . RNA interference ( RNAi ) can be used to perturb gene function with high efficiency and specificity to study gene function and map genetic interactions in Drosophila tissue cell culture ( Heigwer et al . , 2018 ) . Upon treatment , for example , with small molecules , genetic interactions change over time due to time-dependent inhibition of components or other changes in the underlying composition of its molecular constituents . To date , little is known about the trajectories genetic interaction networks ‘rewire’ over time and models for their analysis as well as proof-of-principle data sets are missing . In this study , we devised an experimental and analytical approach to gain insights into higher order ( e . g . gene-gene-drug ) interactions . To analyze how genetic interactions manifest over time , we used a high-throughput , image-based , multivariate phenotypic readout . By combining combinatorial RNAi with a MEK inhibitor or control treatment , we measured higher order chemo-genetic interactions in Drosophila S2 cells to gain new insights into the wiring diagram of the Ras signaling cascade . Ras signaling is an important oncogenic pathway and Ras and EGFR family proteins are frequently mutated in cancer ( Rodriguez-Viciana et al . , 2005 ) . MEK1/2 ( the ortholog of Drosophila Dsor1 ) acts downstream of Ras and phosphorylates ERK1/2 ( the ortholog of Drosophila rl ) , which phosphorylates many other proteins ( e . g . ETS-family transcription factors [Friedman et al . , 2011] ) . The topology of the Ras signaling pathway and its key components are widely conserved between human and Drosophila ( Kolch , 2005; Perrimon , 1994; Wassarman et al . , 1995 ) . In Drosophila , the Ras-pathway has been implicated in early embryonic patterning , growth of wing imaginal discs , differentiation of photoreceptors and blood cell proliferation ( Asha et al . , 2003; Prober and Edgar , 2000; Wassarman et al . , 1995 ) . In this study , we first performed a series of high-throughput image-based genome-wide RNAi screens to identify a set of 168 genes with phenotypic profiles sensitive to MEK inhibition . To construct the differential genetic interaction network , we then created a 168 × 76 double-perturbation matrix and measured the effect of 12 , 768 gene-gene perturbations under differential time and treatment conditions . These perturbations were characterized by 16 reproducible and non-redundant phenotypic features . Notably , we assessed how each treatment-sensitive interaction changes over time and used this information to construct maps of context-dependent biological modules . Context-dependent interactions mapped the plasticity of Ras signaling and cross-talk to other signaling pathways , such as Rel and Stat signaling . Our analyses help to better understand the principles of interaction changes in higher order combinations of genetic perturbations .
Previous studies defined positive differential , negative differential and stable interactions between two genes associated with changes in environmental conditions such as DNA-damage inducing agents ( Bandyopadhyay et al . , 2010; St Onge et al . , 2007 ) . Positive differential interactions are newly forming under stress conditions and mark resistance or other mechanisms counter-acting the noxious stimulus ( e . g . drug treatment ) . Negative differential interactions , on the contrary , mark connections that are required for homeostasis under normal , unperturbed conditions but are either obsolete or harmful under stress conditions . Within these studies , the wiring diagrams of genetic interaction networks were studied at steady state conditions between two endpoints . The information gained from observations of isolated gene-gene-drug interactions thus missed dynamic responses of differential interactions ( Bandyopadhyay et al . , 2010; Ideker and Krogan , 2012; Mani et al . , 2008; Martin et al . , 2015 ) . Based on the observation that the formation of measurable genetic interactions appears to be time dependent ( Figure 1A ) , our study aims to extend the previously established framework of differential genetic interactions by adding a time component . Often , when genetic interactions such as a synthetic sick or lethal interaction between two genes are quantified , different interactions-scores ( π ) are found at different time points ( Figure 1B ) . This indicates that , next to a perturbation by external stresses ( e . g . chemicals ) , also time influences the experimental outcome of genetic interaction measurements systematically . We thus extended the theoretical concept of context-dependent interactions by adding a temporal component and distinguished time-dependent from time-independent interactions , treatment sensitive versus treatment insensitive and alleviating ( rescuing ) from aggravating interactions ( Figure 1C ) . By a systematic exploration of the time’s influence on stress-sensitive genetic interactions , we can gain an understanding on the mechanisms that change genetic interactions over time , and thus the possibility to map stress responsive interactions in greater depth and the chance to assess the time dependence of stress response of specific biological processes upon chemical perturbation of MEK . Thus , we asked: ( i ) What is the behavior of genetic interactions over time and how can we describe it ? ( ii ) What do we learn about the genetic interaction network in response to a compound treatment when observed over time ? ( iii ) What specific biological processes underlie time-dependent and treatment-sensitive genetic interactions . ( iv ) Can we in turn reveal new characteristics of the biological pathways under study , for example regulatory feedback loops in Ras signaling in response to MEK inhibition ? To recover a broad spectrum of cellular phenotypes upon MEK-inhibition , we used a cell morphology assay and automated image analysis in Drosophila cells ( Breinig et al . , 2015; Fischer et al . , 2015; Horn et al . , 2011 ) . Willoughby et al . ( 2013 ) previously compared the effect of multiple small molecule MEK inhibitors in vivo and in S2 cell culture and showed that all but one inhibitor significantly reduced the levels of phosphorylated rl . In this assay , we perturbed cells by small molecule treatment and genetic perturbagens before we arrested cellular morphology by fixation and stained for DNA ( visualizing the nucleus ) , actin ( visualizing cell adhesion and cytoskeleton organization ) and α-tubulin ( visualizing cell morphology and spindle apparatus ) . Using automated high-throughput microscopy combined with a real-time image analysis framework we then recorded morphological phenotypes on a single-cell level . The resulting multivariate phenotypic feature vectors describe the quantitative phenotype resulting from the perturbations ( Figure 2 , Figure 2—figure supplement 1A , Materials and methods ) . As combinatorial gene perturbation screens scale poorly with the number of genes , we first sought to identify genes which phenotypes change in a MEK-inhibitor-sensitive manner . Previous studies have found that genes involved in gene-gene interactions are enriched for genes that themselves display a phenotype distinguishable from the wild type ( Deshpande et al . , 2017; Koch et al . , 2017 ) . Hence , the identification of genes showing a phenotype as a single knockdown will likely enrich combinatorial screens for genes that form higher order interactions . To this end , we performed multiple genome-wide RNAi screens under different environmental conditions ( Figure 2—figure supplement 1 , Materials and methods , Appendix 1 ) . For the following gene-gene interaction analysis , we selected a set of 168 genes from the genome-wide screens that showed: ( i ) high reproducibility between biological replicates , ( ii ) high correlation between sequence-independent dsRNA reagents ( Pearson’s correlation coefficient [PCC]> 0 . 5 ) , ( iii ) measurable effects that deviate from the negative controls , ( iv ) differential phenotypes upon Dsor1 inhibition , and ( v ) are expressed in S2 cells ( log normalized read count > 0 , see Supplementary file 1 ) . We also prioritized genes that were largely uncharacterized ( Materials and methods , Appendix 1 ) . The resulting gene list for gene-gene interaction screening includes 168 target genes that also cover a number of signaling pathways including Ras signaling , innate immunity , Wnt signaling , mRNA splicing , protein translation , cell cycle regulation , Jak/STAT and Tor signaling ( see Supplementary file 2 ) . The query gene set , a subset of the 168 target genes , contained 76 well-described genes to aid biological interpretability . To quantitatively analyze treatment-sensitive genetic interactions in a time-dependent manner , we set up an experimental design based on co-RNAi treatment and high-throughput microscopy ( Figure 2A ) . A combinatorial gene-gene matrix covering 168 target genes and 76 query genes was used to measure 12768 genetic interactions under the different conditions . The library was screened under MEK ( Dsor1 ) inhibitor and control conditions at 48 , 72 and 96 hr after compound addition . The screen was performed using two sequence-independent dsRNA design replicates and in two biological replicates for each condition . In total 4 . 4 Mio . fluorescent images were captured , and 155 image features measured the perturbation effects for every single cell in the experiment ( Appendix 1 ) . Following automated image analysis , we transformed the phenotypic features using the generalized logarithm , normalized , centered and scaled them ( Materials and methods , Appendix 1 ) . Plates failing technical quality control ( Z’-factor between RasGAP1 RNAi and Diap1 RNAi <0 . 3 and biological correlation <0 . 6 PCC for cell number ) were masked in further analysis . Overall , <3% of all plates were excluded according to these criteria . Most of the 155 features showed a high reproducibility ( 80% having a PCC greater than 0 . 6 , Figure 2—figure supplement 2A ) . The two features cell count ( relative cellular fitness ) and actin eccentricity ( morphology of cells ) were among the features with the highest replicate correlation ( Figure 2—figure supplement 3A , B ) and are highlighted as exemplary features in some of the following visualizations . All features that failed to meet a replicate correlation of PCC >0 . 6 were removed , leaving 114 features for further analyses . In addition , 90% of sequence-independent dsRNA pairs correlate with a PCC >0 . 6 with an average correlation of PCC = 0 . 77 ( Figure 2—figure supplement 3C ) . Since many of the remaining 114 features provide redundant information ( Figure 2—figure supplement 2B ) , overlap was reduced by first clustering all features according to the pairwise PCC of the genetic interactions . Second , we fixed the first feature ( cell number ) and removed all remaining features that correlated with PCC >0 . 7 . Third , we selected the next most reproducible and biologically interpretable feature and removed all highly correlated features; this scheme was iterated until all features were passed . The remaining 16 features ( see Supplementary file 3 ) were selected for further analysis . As a confirmation , we verified that cell number and actin eccentricity show a weak correlation ( PCC = 0 . 48 ) and thus provide independent information ( Figure 2—figure supplement 2C ) . An unbiased ‘information gain’ analysis by stability selection , as carried out in an earlier study ( Fischer et al . , 2015 ) , validated this approach showing that each of the chosen features also delivers independent but reproducible information ( Figure 2—figure supplement 2D ) . As they enrich biologically interpretable and reproducibly measurable features , we however kept the features selected by correlation-based analyses . An analysis of the multivariate Z’-factors between RasGAP1 , a negative regulator of Ras signaling and Pvr , a positive regulator of Ras signaling ( Zhang et al . , 1999 ) showed a multi-variate Z’ of 0 . 814 , indicating high assay quality ( Figure 2—figure supplement 3D ) . In a first quality control step , we systematically analyzed whether: ( i ) π-score analysis recapitulates earlier studies using a cell morphology readout in Drosophila , ( ii ) π-scores were reproducible between biological replicates , ( iii ) the interaction profile changed considerably when target and query genes switch roles and ( iv ) interaction profiles were independent for different features . To this end , we compared gene-gene interactions that overlapped between this and previous studies of genetic interactions in Drosophila S2 cell culture ( Figure 2—figure supplement 4 ) . We found significant agreement between π-scores measured in various features in the different studies ( FDR << 0 . 1 , for linear dependence between π-scores measured in different studies ) . We found , for example , that the DNA texture feature we used , could also explain the phospho-histone H3 staining used in Fischer et al . ( 2015 ) . Next , we confirmed a high correlation of interactions between biological replicates , as illustrated on the phenotypic features ‘DNA eccentricity’ and ‘cell number’ ( Figure 2B , Figure 2—figure supplement 5A , A’ ) . As the combinatorial matrix contained all query genes also in the target gene set , we tested whether interaction phenotypes were in accordance regardless of the assignment of target and query . In theory , all interactions should be symmetric , and it should not matter which gene was assigned as target and which as query . However , in practice target and query RNAi reagents were added independently during the experiment which could skew symmetry . Our analysis demonstrated that both combinatorial conditions highly correlate ( Figure 2—figure supplement 5B , B’ , PCC = 0 . 76 for cell number; PCC = 0 . 75 for actin eccentricity ) . We furthermore confirmed that different features provide independent information about genetic interactions as indicated by low correlation ( PCC = −0 . 21 and 0 . 04 , Figure 2—figure supplement 5C , C’ ) . We also confirmed the suitability of our cell-based assay to score compound induced phenotypes without the need to measure its biochemical effect , and determined the ED50 of the MEK-inhibitor PD-0325901 on S2 cells ( Figure 2—figure supplement 6 ) . These experiments demonstrated that S2 cells show a sustained phenotypic response toward PD-0325901 . A high correlation ( PCC = 0 . 81 ) between small molecule and RNAi perturbation of MEK indicates high compound specificity . The most drastic phenotypic changes among a number of features occurred in a concentration window around the drug’s ED50 ( 1 . 5 nM ) . Thus , we selected a concentration of 1 . 5 nM PD-0325901 as an optimal condition for the co-RNAi screening experiments . This ranges within an order of magnitude of the ED50 known for treatment of mammalian tissue cells cultures ( Ciuffreda et al . , 2009; Hatzivassiliou et al . , 2013 ) . Under control conditions , phenotype vectors also reliably separated control RNAi treatments ( RasGap1 [RASAL3] vs . drk [GRB2] , Figure 2—figure supplement 7 ) . In addition , the multi-variate Z’ factor is significantly higher than univariate Z’ using cell count only ( Zhang et al . , 1999 ) . We also found that the knockdown phenotypes of known Ras pathway components Dsor1 ( MEK1/2 ) and drk showed a high correlation ( PCC = 0 . 91 , Figure 2—figure supplement 8 ) . Accordingly , knockdown phenotypes of genes with antagonizing function like the negative regulator of Ras signaling RasGAP1 and Dsor1 inversely correlate ( PCC = −0 . 78 ) . dsRNA targeting the same gene were also highly reproducibly producing similar phenotypic vectors ( e . g . PCCRasGAP1 = 0 . 88 ) . Hierarchical clustering of phenotypic profiles recapitulated known functional relationships of Ras pathway components , whereas regulators of translation show distinct phenotypes . These experiments demonstrated that the morphological assay captures meaningful phenotypes for MEK inhibition , robustly distinguishes controls and groups functionally related genes into clusters of phenotypic similarity . Following quality control , we calculated genetic interaction scores ( π-scores ) for each feature under each condition using a multiplicative model as described previously by Horn et al . ( Horn et al . , 2011 , Materials and methods ) . Overall , we analyzed over 1 . 3 million gene-gene interactions in two conditions , three time points and 16 cellular features . 72922 interactions showed a significant deviation from the expected combinatorial phenotype . Only 9090 ( 12% ) genetic interactions are measured significantly ( moderated t-test [limma] , FDR < 0 . 1 ) for the cell number phenotype underlining the value of the multiparametric analysis . Figure 2C shows an excerpt of the genetic interaction matrices obtained for each treatment and time condition . We found that our analyses recapitulated known genetic interactions . For example , Ras signaling components showed negative interactions with the Jak/STAT pathway ( e . g . Pvr , dos and Sos show negative genetic interactions with dome and Stat92E ( STAT5B ) , Baeg et al . , 2005; Li et al . , 2003; Xu et al . , 2011 ) . The observed interactions become stronger over the three time points measured , and interactions such as a negative interaction between Ras signaling components and Rho1 are stronger upon MEK inhibitor treatment . Next , we sought a suitable statistical framework to score significant context-dependent interactions . Previous studies employed different statistical tests that score the significance of interaction differences between endpoint measurements ( B-Score , dS-Score , limma-based moderated t-test , Bandyopadhyay et al . , 2010; Bean and Ideker , 2012; Billmann and Boutros , 2017; Guénolé et al . , 2013 ) . In a pooled genetic interaction screen in human cells , Shen et al . used the time dependence of fitness defects to improve statistical power ( Shen et al . , 2017 ) . Thus , we tested whether we can also leverage a time- and treatment-dependent model ( Materials and methods , Figure 3A ) to identify differential genetic interactions more sensitively than time-independent statistical models ( Figure 3B ) . We found that a robust linear model of serial measurements ( MODIFI , Figure 3A ) identifies the most differential interactions ( 4723 in total , 2 . 31% of all possible interactions , FDR < 0 . 1 ) . When using only end-point measurements , the robust statistic ( rlm ) is more sensitive than the moderated t-test ( limma , Billmann and Boutros , 2017; Fischer et al . , 2015; Laufer et al . , 2013 ) used in previous studies to score treatment-sensitive interactions and the two-tailed t-tests ( Bandyopadhyay et al . , 2010; Guénolé et al . , 2013 ) of each interaction between conditions ( 1907 vs 874 vs 21 interactions , respectively ) . We further found that MODIFI increased statistical power , identifying 147% more differential interactions across all features when compared to the best endpoint measurements ( 4723 vs 1907; 96 hr/rlm ) . We conclude that by employing robust statistics MODIFI outperforms conventional models and more accurately estimates the parameters treatment sensitivity δ and the time dependence σ . MODIFI between to genes i an j , is described by the following equation:πij=cij+σij*time+δij*treatment+εij . There σ estimates the rate by which interactions change and δ estimates the amplitude of interaction change between treatment conditions ( Materials and methods ) . We next sought to test if the linear models are practical to describe time and treatment dependent genetic interactions . To this end , we compared the unweighted residuals of each fit with the actual experimental variance measured at each time point . If the model fails to fit the data appropriately ( e . g . the comparison does not behave monotonic , or the form of the input-curve is not linear ) one would expect that the residuals are unexpectedly greater than the variance . However , analyses of all interactions for each phenotypic feature reveals that this is rarely the case ( Figure 3—figure supplement 1 ) . In most models , remaining residuals of fit are explained by the variance between biological replicates ( avg . PCC = 0 . 96 , R2 = 0 . 92 ) . Interestingly , this is true for all features we assessed . By initial feature transformation , centering and scaling , systematic differences between features were removed . During subsequent interaction calling , where only the residual of activity not explainable by each single gene knockdown is kept as a phenotype , specific time-dependent behaviors of features affecting each gene are removed as well ( Diss and Lehner , 2018 ) . We thus concluded that the π-score dependence on time and treatment for each phenotypic feature can be reliably quantified using linear model statistics . Next , we quantified to what extent gene-gene interactions changed due to MEK inhibitor treatment ( δ ) . δ serves as a surrogate for the integrated area between the trajectories of the two treatments . If δ is close to zero , only little changes occur upon treatment and high δ marks highly treatment-sensitive interactions . We found that treatment-sensitive interactions were equally likely to be positively or negatively shifted over all analyzed genes ( Figure 3C , grey distribution ) . Of note , especially treatment-sensitive interactions of Rel ( NFKB1 , a downstream effector of the Drosophila Imd signaling pathway , Myllymäki et al . , 2014 ) , or Ras/Map and Jak/STAT related genes enriched as either negatively shifted π-score or positively shifted π-scores because of MEK inhibition , respectively . This implies that pathways which are positively regulated by MEK tend to form interactions that are less aggravating under MEK inhibition . Interactions formed by Rel are negatively enhanced by MEK inhibition . We further found no or little significant difference between housekeeping modules ( such as proteasome , translation machinery ) and all genes ( p>0 . 1 , two-sided KS-test ) . Taken together , these data suggest that components of the same pathway share differential interaction sensitivity and directionality in response to Ras pathway inhibition . Assessing all interactions for which MODIFI identified statistically significant hits ( FDR < 0 . 1 ) , we identified four main types of time and treatment-dependent interactions that we expected would be recovered by MODIFI ( Materials and methods , Figure 1C III-VI ) . Interactions that are neither time-dependent nor treatment-sensitive were not covered by MODIFI ( see also Figure 1C I , II ) . Among the time-dependent interactions , we observed alleviating treatment-insensitive interactions where the π-score raised over time ( Figure 4A ) . These interactions often involve core essential genes whose influence on the phenotype ( e . g . cell count ) is not altered by MEK inhibition . This is for example the case for mts knockdown ( PP2CA , lethal by itself; Snaith et al . , 1996 ) where the simultaneous loss of the proteasomal subunit Prosbeta4 ( PSMB2 , Wójcik and DeMartino , 2002 ) dominates the combinatorial phenotype that do not change further regardless of the treatment . In this case , a positive interaction that strengthens over time was measured ( Figure 4A ) . Accordingly , we termed interactions aggravating , treatment insensitive when the π-score declined over time and its trajectories were indifferent between treatments ( Figure 4B ) . Aggravating treatment-insensitive interactions on cell count often include signaling transducers where the loss of one only has a mild phenotype while the double perturbation disturbed homeostasis which cannot be buffered buffer and a synthetic sick or lethal interaction is observed . For example , ksr ( KSR1 ) and rl ( ERK1/2 ) , two core members of the Ras signaling cascade ( Morrison , 2001; Wassarman et al . , 1995 ) , interact significantly ( p=0 . 0017 ) . This synthetic sick interaction is stable upon MEK inhibition and thus appears independent of phospho-rl levels which hints toward a kinase-independent function of rl in combination with its scaffolding protein ksr ( Figure 4B ) . We defined interactions as treatment sensitive when trajectories differed significantly between treatments ( FDR < 0 . 1 , Materials and methods ) . If the π-score is lower under control than under treatment conditions , we termed it a positive treatment-sensitive interaction ( MEK inhibition lifts the phenotype , Figure 4C ) and negative treatment-sensitive interaction ( MEK inhibition dampens the interaction , Figure 4D ) in the opposite situation . For instance , skd ( MED13 , an integral component of the mediator complex; Janody et al . , 2003 ) showed a positive differential interaction with Stat92E ( Drosophila ortholog of human STAT receptor; Bina and Zeidler , 2009 ) ( Figure 4C ) . Under control conditions skd knockdown aggravated the fitness loss induced by Stat92E knockdown to a lethal phenotype . This aggravation was attenuated under MEK inhibition . Our data suggest that a synthetic lethal relationship connects both genes when they are otherwise unperturbed . Only little is known about the cooperative function of mediator and STAT or crosstalk toward Ras signaling ( Bina and Zeidler , 2009 ) . Interestingly , under control conditions , the loss of fitness phenotypes of Stat92E and skd single knock down are not time dependent , while the interaction is strongly time and treatment dependent . This is indicative of a longer-term transcriptional response when cooperative action of skd and Stat92E is disturbed ( Figure 4—figure supplement 1 ) . In contrast , a negative differential interaction occurred between Rel and pnt . While Rel knockdown rescued the fitness-defect induced by pnt knockdown under normal conditions , it aggravated the pnt knockdown phenotype after MEK inhibition ( Figure 4D ) . Thus , we hypothesize that both , the aggravating interaction between skd and Stat92E and the alleviating interaction of Rel and pnt depend on the proper function of Dsor1 . Mixed forms , such as interactions that deviate strongly in the beginning experiment and converge later or interactions that were almost time independent but treatment sensitive , were also observed . To assess whether different features ( which we grouped into meta features , such as cell shape or nuclear texture ) or pathways show enrichments in one or the other interaction type , we analyzed enrichment of interaction counts over a random distribution . We found considerably more treatment-insensitive than treatment-sensitive interactions for all feature ( 18468 vs . 4723 , 16 phenotypic features , Figure 4E ) . While , as expected , the distribution of negative and positive interactions over all features was symmetric , specific phenotypic features capture surprisingly high numbers of alleviating ( nuclear shape ) or aggravating ( nuclear texture ) treatment-insensitive interactions . This indicates that different phenotypic features identify specific biological reactions of cells toward double gene perturbations . A possible explanation would be that different biological processes influence different cellular features , for example perturbations of the cytoskeleton organization mostly influence shape features while perturbation of nuclear factor alters mostly nuclear texture . The direction of interactions then follows the genes that are involved and so do the different features enrich distinct interaction types . Core essential housekeeping genes , for example , show exceptionally high numbers of alleviating interactions on cell shape but simultaneously display mostly aggravated phenotypes on their nuclear texture . These observations indicate a complex interdependence between specific genes under investigation and the phenotypic features that are used to assess them . Additionally , we found that treatment-sensitive interactions , compared to treatment-insensitive interactions , enriched in specific signaling pathways related to MEK inhibition . While , for example , ribosome or spliceosome-related genes formed mostly alleviating and treatment-insensitive interactions ( Figure 5A ) , the JNK pathway was enriched for alleviating treatment-insensitive and negative treatment-sensitive interactions ( Figure 5B ) . Other pathways , such as Ras signaling , Rel , Mediator signaling or Jak/STAT signaling were equally overrepresented in treatment-sensitive and treatment-insensitive interactions . Among the pathways tested , the enrichment of treatment-insensitive interactions highlights pathways with large impact on the interaction network controlling cell viability . The enrichment of differential interactions highlights mainly signaling pathways that are sensitive to MEK inhibition . Differential genetic interactions are not equally distributed over all genes that were tested . Jak/STAT signaling components ( Stat92E , dome , upd3 ) alongside Ras signaling members ( drk , rl , dos , Sos , pnt ) and , interestingly , Imd signaling ( Rel ) showed specific enrichment of differential interactions ( cell count feature , Figure 5C ) . Specifically , pnt forms many positive differential interactions ( alleviated upon MEK inhibition ) while Pvr is involved in many negative differential interactions ( aggravated by MEK inhibition ) . This could be attributed to pnt acting as a terminal transcriptional effector of the signal triggered by the activated receptor Pvr . We also found that genes , which form more treatment-insensitive genetic interactions also enrich treatment-sensitive interactions ( compare linear trendline , Figure 5D ) . However , some particular genes are involved in unexpectedly many differential interactions . This indicates that a rather specific response to the treatment is reflected in the differential interactions . These data demonstrate that time-dependent modeling of interaction scores sensitively detects treatment differential interactions which enrich in and thus highlight Ras-sensitive biological processes . MODIFI estimates the time dependence ( σ ) of each treatment-sensitive interaction . This term can be interpreted as the slope by which an interaction changes ( e . g . strengthens or weakens ) over time . Depending on the initial difference ( compare Figure 3A ) , π-scores increase or decrease over time , diverge or converge . The most abundant interaction in this study describes a treatment-insensitive interaction that could not be measured initially but forms over the course of the experiment ( 78% of all significant interactions , FDR < 0 . 1 ) . In the following analyses , we use genetic interactions based on cell count as an example to test whether genes or pathways react at different specific rates . For example , from 48 hr to 96 hr after compound addition , genetic interactions with Rel remained stable , whereas interactions of Jak/STAT or Ras signaling-related genes changed significantly over time . Interactions with housekeeping-related genes ( proteasomal or ribosomal subunits ) show phenotypes of an exceptionally high time dependence ( Figure 6A ) . These data indicate that interactions of the different biological processes rewire at different rates after perturbation . We also hypothesized that the difference of interaction scores and their time dependence could inform about the influence of MEK inhibition on different biological modules or phenotypic features . Cell fitness-based interactions formed by proteasome related genes show the strongest phenotypic differences between treatments at the initial and last measured time point ( Figure 6B ) . This suggests that proteasome-related genes are involved in particularly strong treatment-sensitive interactions upon MEK inhibition . These interactions interfere with cell proliferation early on during our experiment and also become stronger over time . This supports reports of synergistic effects between proteasome and MEK inhibition on perturbing cell viability ( Chang-Yew Leow et al . , 2013 ) . Next , we hypothesized that phenotypic features measure different initial interaction differences and analyzed initial π-score differences between phenotypic features . Especially , cell morphology features ( nucleus/cell eccentricity ) and their variance within the population of cells show initial differences that are significantly higher than those measured by cell count ( p<0 . 0001 , Figure 6C ) . Of note , nuclear eccentricity and its variance among the population of cells ( nucleus eccentricity sd ) are also the only initially different features that are masked later on . All other phenotypes show an increased interaction difference over time ( Figure 6D ) . Surprisingly , cell count as the traditional readout for fitness after gene-gene perturbation shows the smallest interaction differences between the treatments in general , irrespective of the time point . Together , these analyses demonstrate that the time dependence of genetic interactions is specific to certain biological process . It further highlights that phenotypes beyond cell viability excel to capture early treatment-sensitive interactions . Next , we analyzed whether interaction networks formed by different biological modules or core signaling pathways change systematically over time and treatment . In the following examples , we used cell eccentricity as an exemplary feature which we found to capture early cellular responses . Figure 7A shows how an interaction sub-network including Jak/STAT signaling , Ras/Map signaling components and spliceosome related genes rewires over time in reaction to MEK inhibition . Core housekeeping modules ( ribosome , spliceosome or proteasome ) were highly interconnected by alleviating treatment-insensitive interactions . In contrast , components of the Ras signaling , Jak/STAT signaling or Tor signaling cascade showed aggravating interactions with housekeeping modules . We observed that ( i ) alleviating interactions ( π > 0 ) dominate early time points , ( ii ) many initially alleviating interactions reverse over time ( π > 0 → π < 0 ) , ( iii ) differences attributed to the compound treatment become more profound over time . Lastly , we noted that genes in proximity tend to have similar interaction patterns coherently changing over time and treatment ( Figure 7—source data 1 ) . Previous studies implied that similarities of treatment sensitive genetic interaction profiles can identify functionally related genes ( Bean and Ideker , 2012 ) . Thus , interactions of related genes change coherently upon network perturbation . Hence , we defined treatment-insensitive interaction profiles for each target gene . We used the modeled interaction difference between treatments over time ( δ ) to quantify interaction change due to Dsor1 inhibition . For every target , we calculated δ with every query gene in a vector comprising 76 measurements for cell eccentricity . Correlations between profiles ( Figure 7B ) confirmed known functional relationships of genes , as for example the profiles of the genes Stat92E and dome , members of the Drosophila Jak/STAT pathway , were similar ( PCC 0 . 73 ) confirming that both genes share biological function upon perturbation of Ras signaling ( Xu et al . , 2011 ) . Furthermore , our analysis showed a correlation of treatment-sensitive genetic interactions for all features between Stat92E , dome and Ras signaling . Interestingly , the profile of Rel was similar to negative regulators of Ras signaling ( RasGAP1 , PCC 0 . 38 ) , but was anti correlated with positive regulators ( pnt , PCC −0 . 37 ) indicating a potential crosstalk between the two pathways . We expected that a correlation-based network drawn from treatment-sensitive interaction profiles across all phenotypic features reveals modules of functionally related genes . Thus , we calculated the pairwise correlation coefficients ( PCC ) of treatment-sensitive interaction profiles ( interactions with 76 query genes ) including all 16 cellular features of all 176 target genes . We visualized resulting positive correlations in a network graph highlighting biological processes and candidate genes ( Figure 7C , Supplementary file 2 ) . This revealed that correlations of treatment-sensitive interaction profiles clustered genes into known pathway modules . Of note , Rel and Fur1 ( FURIN ) and swm ( RBM26 ) showed unexpected correlations with members of the Ras signaling cascade ( Figure 7B ) . It is expected that genes with similar functions irrespective of the treatment show similar interaction profiles between and within conditions . In contrast , genes with a treatment-dependent function should lose or gain correlations to other genes when compared between treatments ( Billmann and Boutros , 2017 ) . To test this , we defined profiles of all interactions across all cellular features and time points and correlated them between genes and between conditions . Most interaction profile correlations did not differ significantly between conditions , compared to within conditions ( Figure 7—figure supplement 1 ) . Specifically affected gene pairs were mostly Ras signaling components . Interestingly , also profile correlations of Jak/STAT signaling components ( Stat92E , dome ) as well as of the two genes Fur1 and swm differed between and within conditions . This provides further clues that Fur1 and swm are implicated Ras signaling . Only few , weak interaction profile correlations were higher between than within conditions . We have shown that the treatment-sensitive interaction profiles of Rel and pnt were negatively correlated , whereas Rel profiles were positively correlated with RasGAP1 , a negative regulator of Ras ( Figure 7B ) . This suggested that Rel itself might function as a negative regulator of Ras signaling . We observed that Rel depletion alone had little impact on cell growth , as compared to pnt , but showed a cell length ( major axis ) phenotype ( Figure 8A , Figure 8—figure supplement 1 ) . Co-depletion of pnt and Rel altered both cell number and cell length . Under control conditions , depletion of Rel alleviated the loss of viability and cell length phenotypes after pnt knockdown ( Figure 8A ) . This interaction was attenuated under MEK inhibition ( Figure 8B ) when co-depletion of Rel and pnt led to a synthetic lethal phenotype ( FDR < 0 . 1 , Figure 8C , C’ ) . These interactions were observed for both dsRNA designs ( PCC = 0 . 88 and 0 . 96 for Rel and pnt , Figure 8—figure supplement 2 ) . Pvf2 ( orthologue of human VEGF ) is upregulated in the absence of Rel ( log2fold-change = 1 . 5 ) ( Boutros et al . , 2002 ) . The data presented here indicate that a knockdown of Rel induced a re-activation of the Ras pathway which is dependent on Dsor1 activity ( Figure 8A–C ) . We hypothesized that Rel negatively regulates Ras signaling by repressing the expression of Pvf2 , the ligand activating the Pvr-Ras-phl-Dsor1-rl-pnt signaling cascade after binding to Pvr ( PDGFR ) . To test this hypothesis , we performed qPCR analysis of pnt , Rel , Pvf2 , sty ( SPRY2 ) and RasGAP1 expression levels ( Figure 8D , E ) . We first confirmed the upregulation of Rel after depletion of Ras ( Figure 8D ) and showed that upregulation of Rel was suppressed by pnt co-RNAi . Pvr knockdown , as a control for loss-of-Ras signaling activity , led to a downregulation of pnt and RasGAP1 . Pvr knockdown also induced a strong upregulation of Rel expression . Finally , co-RNAi of Rel and pnt induced a significant increase in Pvf2 expression , not observed by depletion of either gene alone ( Figure 8E ) . The Rel/pnt co-RNAi also induced upregulation of negative regulators of Ras signaling sprouty ( sty ) ( Casci et al . , 1999 ) and RasGAP1 ( Feldmann et al . , 1999 ) ( Figure 8E ) , thereby providing a mechanistic explanation how Rel could negatively regulate Ras signaling . We hypothesized that this regulatory loop is mediated by the transcriptional regulation of Pvf2 and requires Dsor1-mediated Ras signaling activity , as summarized in Figure 8F . These changes were observed both at 48 hr and 96 hr time-points ( Figure 8—figure supplements 3 and 4 ) . Interestingly , protein levels of rl were down regulated by pnt-or rl-RNAi and rescued by Rel co-RNAi ( Figure 8—figure supplement 4F , G ) . Overall , these experiments provide a mechanistic basis how Rel acts as a negative regulator of Ras signaling in a context-dependent manner .
To better understand context-dependent differences in genetic networks upon changes in environmental conditions is a current frontier in genetics ( Rancati et al . , 2018 ) . Many biological processes rely on context-dependent changes in genetic requirements , from robustness of cell differentiation during development to responses of cancer cells to chemotherapeutic treatments . However , only few studies on selected phenotypes have systematically analyzed how environmental changes impact genetic interaction networks . Previous studies have analyzed genetic networks after activation of the DNA damage response signaling in yeast or changes in Wnt signaling activity in Drosophila cells ( Bandyopadhyay et al . , 2010; Billmann and Boutros , 2017; Díaz-Mejía et al . , 2018 ) . In these studies , positive and negative treatment-sensitive , and treatment-insensitive interactions have been determined based on fitness phenotypes or pathway reporter activity in static end-point assays . Aim of the present study was to analyze changes in genetic networks that impact a broad spectrum of phenotypes by imaging and multiparametric image analysis and to determine how treatment-sensitive interactions change over time after small molecule perturbation of the Ras signaling pathway . In this study , we established a high-throughput image-based assay which enabled us to reproducibly measure many phenotypes including cell proliferation and cell morphology which are influenced by many cellular processes ( Breinig et al . , 2015; Fuchs et al . , 2010; Horn et al . , 2011 ) . We used this assay to measure genetic interactions between differential treatment conditions over the course of three time points . To this end , we assessed the phenotypes of 76608 di-genic interactions in Drosophila hemocyte-like cells . Each interaction was characterized by a vector of 16 non-redundant and quantitatively reproducible phenotypic features . Further , we developed MODIFI , a two-factor robust linear model to quantitatively describe the time and treatment-dependent changes of genetic interactions . MODIFI also allowed us to describe whether an interaction is treatment sensitive ( treatment could predict π-score ) or time dependent ( time predicted the π-score ) . Using MODIFI we found , for example , treatment-insensitive interactions within the Ras signaling cascade ( rl-ksr interaction , Figure 4B ) as well as treatment-sensitive crosstalk between Mediator , STAT and Ras signaling ( Figure 4C ) . Discovery of such interactions can lead to new treatment options in cases where the pharmacologic inhibition of MEK had no effect and an inhibition of the ERK ( rl ) -KSR ( ksr ) interaction becomes an interesting target ( Roy et al . , 2002; Yu et al . , 1998 ) . Regarding the example of a signaling axis between Ras , STAT and Mediator signaling , some evidence indicates that mediator and STAT signaling engage in cooperative transcriptional regulation dependent on the phosphorylation status of Stat92E ( Kuuluvainen et al . , 2014; Wienerroither et al . , 2015 ) . Additionally , evidence exists on mutual crosstalk between phosphorylation-dependent Ras signaling and Stat92E ( Li et al . , 2002 ) . Hence , our data suggests that these pathways could be interconnected and Stat92E and Mediator only show cooperative action when Ras signaling is active and Stat92E phosphorylation is not impaired . Our analysis showed that we detected treatment sensitive interactions more sensitively as compared to endpoint measurements or single time point replicates ( see examples in Figure 4C , D & Figure 8C , C’ ) . Enrichment of treatment-sensitive interactions among stress-responsive pathways and genes underlines their biological relevance . Using this approach , we also analyzed the treatment ( δ ) and time ( σ ) dependency of interactions of specific genes and pathways . Overall , measuring phenotypes resulting from genetic interactions increased our ability to detect treatment-sensitive interactions . Furthermore , the measurement of multiple phenotypic features simultaneously enabled more detailed characterization of the observed treatment-sensitive interaction . We also tested whether the establishment of phenotypes is dependent on a gene’s expression level but found no correlation of high gene expression and high time dependency ( data not shown ) . Our data further suggests that σ is influenced by the general resilience of a pathway or signaling module to perturbations . For example , housekeeping genes , widely believed to form extremely time and condition stable regulatory networks took the most time to rewire their interactions . This makes it unlikely that the stability or turnover of a single gene product is a major driver of time-dependent establishment of genetic interactions . We found , for example , that genetic interactions of ‘core’ ( or housekeeping ) modules such as the translation machinery , proteasome and others induce phenotypes that are much stronger at later time-points . In contrast , other cellular modules such as signaling and innate immunity ‘rewire’ early in the experiment . We could also show that while single perturbation phenotypes in some instances do not change over time interactions still do , hinting toward a time-dependent combinatorial effect . Our analysis classified genes into categories of genetic interactions that are ( i ) signaling modules central to the cells’ physiological role , ( ii ) signaling modules required for maintaining homeostasis and ( iii ) resilient ‘core’ modules whose network hubs form interactions on a longer timescale . We also found that measuring different phenotypes provided more information about the development of interaction differences over varying time scales and demonstrate a number of example treatment-sensitive interactions that could not have been found in end-point assays . While the cell count ( comparable to yeast colony size ) as a phenotype captures cellular reactions rather late in the experiment , other phenotypes , such as nuclear morphology or cytoskeleton texture , enabled to measure immediate cellular reactions . In the gene-drug interaction experiments , we found that pathways interacting with Ras signaling reacted strongest to the Dsor1/MEK inhibition . This observation was translated to map signaling modules that react similarly toward Ras signal perturbation and we correlated δ-profiles along all features between all target genes . By this means , genes whose interactions change coherently upon Dsor1 inhibition are grouped into highly interconnected modules . Consequently , this correlation network clusters genes of similar functions in proximity with each other . Each module is also characterized by a coherent reaction towards Dsor1 perturbation . Interestingly , rolled ( rl ) , Dsor1 and pole hole ( phl ) ( ERK1/2 , MEK1/2 and Raf ) were not connected to the rest of Ras-signaling-related genes in the correlation network . In contrast , they correlated with Ras when using interaction profiles of the control treatment . This indicates that the chemico-genetic analysis identified ‘responsive’ factors that can be uncoupled upon environmental modulation of specific signaling modules . Our analysis also revealed three genes that unexpectedly connected to Ras signaling: Fur1 , a serine-type endopeptidase ( Kim et al . , 2015 ) , swm , involved in mitotic checkpoint regulation and hedgehog signaling ( Casso et al . , 2008; Dong et al . , 1997 ) and Rel ( Foley and O'Farrell , 2004 ) . The correlation of Fur1 and swm with positive regulators of Ras signaling indicates that they respond similarly towards Dsor1 inhibition as Ras pathway members . In addition , we identified Rel ( NF-κB ) as a strong treatment-sensitive genetic interactor , suggesting that mitogenic Ras signaling and innate immune pathways depend on each other . Once Rel is lost , cells become more dependent on Ras signaling; a phenotype that can be blocked by perturbing Dsor1 activity chemically or genetically . Already at a low dose , both perturbations result in a synthetic lethal phenotype that kills Drosophila hemocyte-like cells . Conversely , it was previously shown that Ras signaling influences Rel activity by regulation of its negative transcriptional regulator pirk ( Ragab et al . , 2011 ) . We hypothesize that this mutual negative feedback regulation could be the basis for a ‘fight’ or ‘flight’ response of the immune cells; balancing an immune and proliferative response in the same cell . Large-scale studies on gene essentiality have challenged the concept of a static repertoire of essential genes . In contrast , loss-of-function screens in different genetic background of cancer cells identified ‘core’ and ‘genotype’-dependent sets of essential genes . This indicates that essentiality is modulated in a context-dependent manner ( Hart et al . , 2015; McDonald et al . , 2017; Rauscher et al . , 2018 ) . At this point , our study is the largest exploration of gene-gene-drug interactions based on multiparametric , non-essential phenotypes . We demonstrate how different vulnerabilities for a diverse set of automatically scored phenotypes change upon time and environmental conditions . Our modeling approach increases the confidence to call treatment sensitive interactions upon changes of environmental conditions . This allows to map a correlation network of cellular modules that react coherently toward the external stimulus . We expect that , when further studies of context-dependent genetic interactions will become available , a comparative analysis will provide fundamental insights into how different cellular networks react to environmental stimuli with implications for therapy resistance and timing of drug treatments . In future studies , MODIFI could be further expanded to include terms assessing the influence of the treatment on the behavior over time ( rate of interaction change ) . This will , for example , aid to understand the qualitative relationships between different treatment trajectories beyond the current analysis and sets the basis for further experiments . This study introduced an experimental and analysis framework to explore time-dependent rewiring of genetic networks which can be used to dissect the complexity of biological networks in model organisms and human cells .
All cells used in this project were from the same culture of the , serum-free medium adapted , Drosophila melanogaster S2 cell line ( S2 ) and will be referred to as S2 cells ( Schneider's Drosophila Line 2 [D . Mel . ( 2 ) , SL2] ( ATCC CRL1963 ) from ThermoFisher ( Waltham , MA , Billmann et al . , 2018; Horn et al . , 2011; Fischer et al . , 2015 ) . We used a genome-wide D . melanogaster dsRNA library ( HD3-dsRNA library ) in this study , as previously described ( Billmann and Boutros , 2017; Horn et al . , 2011 ) . The library contains 28941 dsRNA reagents targeting 14242 unique gene IDs in the D . melanogaster genome and contains two sequence independent reagents targeting 13617 IDs twice and the remaining genes once . The reagents were optimized for the BDGP5 mRNA annotations in D . melanogaster by for example avoiding CAN repeats and non-unique sequences ( off-targets ) . 250 ng dsRNA , synthesized as described previously , were aliquoted to 384 Greiner µClear plates prior to the image-based assay at a mass of 250 ng/well . A table containing all sequences that were used in the genome-wide RNAi screen can be found in Supplementary file 4 . Another table containing sequence IDs ( HD3 ) that were used in the combinatorial RNAi screen can be found in Supplementary file 5 . dsRNA reagents dissolved in water were spotted into barcoded 384-well microscopy plates ( Greiner µClear , black , flat-transparent-bottom , Ref: 781092 , Greiner Bio One International GmbH , Frickenhausen , Germany ) to reach a final mass of 250 ng dsRNA per well ( 5 µl of a 50 ng/µl solution ) . Express V medium ( Gibco , Ref: 10486–025 , Life Technologies GmbH , Darmstadt , Germany ) with 10% Glutamax ( Gibco , Ref: 35050–061 ) was pre-warmed to 25°C and 30 µl were dispensed on top of the spotted dsRNA using a MultiDrop Combi dispenser and standard cassette ( Thermo Fisher Scientific , Ref: 5840400 , Life Technologies GmbH , Darmstadt , Germany ) . 10 µl of pre-diluted S2 cell suspension were seeded to a final concentration of 9000 cells/well into the prepared assay plates using MultiDrop Combi dispensing under constant stirring of the suspension in a sterile spinner flask ( Corning , Ref: CLS4500500 , Kaiserslautern , Germany ) . After cell addition , the assay plates were heat sealed using a PlateLoc ( peelable seal , 2 . 3 s at 180°C , Agilent Technologies Deutschland GmbH and Co . KG , Waldbronn , Germany ) and centrifuged at 140x g for 60 s . Cells were incubated for 24 hr at 25°C without CO2 adjustment . After 24 hr incubation , plates with growing cells were opened and small molecule treatment was performed . The concentration of applied compound is outlined with the separate experiments in the following paragraphs . Per well 5 µl of a solution containing 5% DMSO ( Sigma Aldrich , Ref: 41644–1 l , Merck KGaA , Darmstadt , Germany ) in medium , or the MEK-inhibitor PD-0325901 ( Cayman chemical , Ref: CAY-13034–5 , Biomol GmbH , Germany ) dissolved in 5% DMSO in medium , were added to achieve a final assay concentration of 0 . 5% DMSO and varying small molecule concentrations . After compound addition , plates were sealed again and incubated at 25°C without CO2 adjustment for 48 hr , 72 hr or 96 hr depending on the experiment . Assays were stopped after the second incubation period by fixation using a robotics procedure on a CyBioWell vario ( 384-well pipetting head , Analytic Jena AG , Jena , Germany ) . The supernatant was removed , and cells were washed with 50 µl PBS ( Sigma Aldrich , Ref: P3813-10PAK ) per well . After addition of 40 µl Fix-Perm solution ( 4% Para-formaldehyde ( Roth , Ref: 0335 . 3 , Karlsruhe , Germany ) ; 0 , 3% Triton X-100 ( Sigma Aldrich , Ref: T8787-250ml ) ; 0 , 1% Tween20 ( Sigma Aldrich , Ref: P1379-100ML ) ; 1% BSA ( GERBU Biotechnik GmbH , Ref: 1507 . 0100 , Heidelberg , Germany ) ) , plates were incubated for 60 min at RT and then washed twice with 50 µl of PBS . 50 µl of PBS were added again and plates were stored at 4°C before staining . For staining , remaining PBS was removed and fixed cells were first blocked by adding 30 µl of blocking solution ( 4% BSA; 0 , 1% Triton X-100 , 0 , 1% Tween20 ) and incubated for 30 min at RT . Next , the blocking buffer was removed and 10 µl of staining solution ( 1:4000 Hoechst ( Thermo Scientific , Ref: H1399 , Life Technologies GmbH , Darmstadt , Germany ) , 1:1500 primary FITC labelled anti α-tubulin antibody ( Sigma Aldrich , P1951 ) , 1:6000 Phalloidin-TRITC conjugate ( Sigma Aldrich , F2168- . 5ml ) in 1x blocking buffer ) were added . After addition of the staining solution plates were incubated for 60 min at RT in the dark . After staining , 30 µl of PBS were added and the staining solution was removed . After two additional washing steps with 50 µl PBS another 50 µl of fresh PBS were added per well and stored at 4°C until imaging . We performed genome-wide RNAi screens in combination with drug and solvent control treatment to verify dsRNA reagent efficiency , identify candidate genes for combinatorial screening and to find which genes react most differentially to the Dsor1 inhibitor ( PD-0325901 ) treatment . Four sets of 88 × 384 well Greiner µClear plates were spotted with the HD3 library , 5 µl of 50 ng/µl dsRNA in each well . The HD3 library is arrayed to target one gene with one dsRNA design per well . Two additional plates , containing only controls were added to control assay reproducibility , robustness and effect size . Controls were chosen to spread over the complete dynamic range of cell fitness . dsRNAs against RLUC and GFP expressing plasmids serve as non-targeting negative controls , such that we could control for unspecific dsRNA induced phenotypes . dsRNA containing plates were thawed , seeded with cells and left for 24 hr at 25°C without CO2 adjustment for incubation prior to drug treatment . Plates were opened and 5 µl of 15 nM PD-0325901 in 5% DMSO were added resulting in a final assay concentration of 1 . 5 nM PD-0325901 in 0 . 5% DMSO in medium . Cells were left to incubate for another 72 hr at 25°C without CO2 adjustment prior to fixation , staining and imaging . Images were acquired using the standard protocol described below with low illumination timings ( DAPI: 100 ms , Cy3: 200 ms , FITC: 300 ms ) . The resulting images were analyzed in line with the acquisition using the standard image analysis pipeline and progress was monitored using our automated analysis pipeline as described below . The design of the library for combinatorial screening is described in a separate paragraph . 168 genes were chosen for design of a combinatorial RNAi library . The dsRNA sequences that were used in the combinatorial library can be found in Supplementary file 4 and 5 . All used labware and reagents , which are not further detailed here have been the same as in previous experiments . The library contained 12 batches for screening , each comprising 80 × 384 well Greiner µClear plates spotted with 250 ng dsRNA/well dissolved in 5 µl of DNase , RNase-free water . dsRNAs were obtained from the HD3-library templates and synthesized accordingly . To avoid contaminations , all dsRNAs were sterile filtered using Steriflips-0 . 22µm ( Merck Millipore , Ref: SCGP00525 , Darmstadt , Germany ) for the query dsRNAs and MultiScreenHTS-GV 0 . 22 µm filter plates ( Merck Millipore , Ref: MSGVN2250 ) for the target dsRNAs . Genes were divided into target and query genes based on prior knowledge on key pathway components and screened a matrix of 76 genes combined with 168 genes . All query genes were included in the list of target genes . We screened each target gene in two sequence independent designs and each query gene in one design . This way , we screened 25536 dsRNA combinations ( 12768 gene pairs ) in each batch . Combinatorial dsRNA spotting was achieved with combining the query and target master plates such that 2 . 5 µl of each query dsRNA were spotted onto 2 . 5 µl target dsRNA using a Beckman FX robotic liquid handling station ( Beckman Coulter , Krefeld , Germany ) . In order to control for RNAi induced phenotypes and per-plate batch effects control dsRNAs against Dsor1 , drk , Diap1 , RasGAP1 , Pten , pnt , Pvr , Rho1 and RLUC expressing plasmid were spotted on each plate and not paired with a second query dsRNA perturbation . Two control plates containing only the target gene dsRNA reagents with 250 ng dsRNA per well complete one screening batch of 80 plates and controlled for screening batch effects . To achieve differential treatment and time resolution , 12 screening batches were prepared . They were divided into two groups of six batches , which then were treated under different conditions in duplicate . six library batches are needed to screen two conditions ( 1 . 5 nM PD-0325901% and 0 . 5% DMSO ) at three time points ( fixation 48 hr , 72 hr , 96 hr after small molecule addition ) , all in all comprising 480 screened plates . The entire experiment was repeated twice . This way we screened 960 × 384 well plates . The assay workflow followed the same procedures as outlined above . Briefly , 9000 cells per well were seeded onto 384-well Greiner µClear plates for microscopy , which were pre-spotted with a combinatorial dsRNA library . After centrifugation , plates were sealed and left to incubate for 24 hr at 25°C prior to compound addition . Therefore , a PD-0325901 dilution ( 15 nM in medium with 5% DMSO ) , and a 5% DMSO-only dilution in medium were prepared and added to the opened plates . This resulted in either 1 . 5 nM PD-0325901 or 0 . 5% DMSO in-assay concentrations . Plates were sealed again using a heat sealer and left to incubate until the experiment was stopped by fixation after 48 hr , 72 hr and 96 hr , respectively . Stained plates were imaged using an InCell Analyzer 2200 automated fluorescence microscope according to the protocol described above with 20x magnification , three channels per field and four fields per well . Resulting images were analyzed using the R/EBImage based pipeline described below . All plates were imaged using the same protocol . There , an InCell-Analyzer 2200 automated fluorescence microscope ( GE Healthcare GmbH , Solingen , Germany ) with a Nikon SAC 20x objective ( NA = 0 . 45 ) was used . The microscope was adjusted to scan Greiner µClear plates by setting the bottom height to 2850 µm and the bottom thickness to 200 µm and the laser autofocus function was applied to identify the well bottoms with attached cells . This Z-position was used for image acquisition in three fluorescence channels: Hoechst ( excitation: 390 ± 18 nm , emission: 435 ± 48 nm ) at 400 ms exposure ( 100 ms in dose response experiments and genome wide screens ) , Cy3 ( excitation: 475 ± 28 nm , emission: 511 ± 2 nm 3 ) at 300 ms exposure ( 200 ms in dose response experiments and genome wide screens ) and FITC ( excitation: 542 ± 27 nm , emission: 597 ± 45 nm ) at 300 ms exposure ( 300 ms in dose response experiments and genome wide screens ) . Four fields of view were imaged per well at 20x magnification each representing a 665 . 60 µm x 665 . 60 µm area covered ( approximately 20% of total well area ) by 2048 × 2048 pixels . For plate handling , the microscope was equipped with a KinedX robotic arm ( PAA Scara , Peak Analysis and Automation Ltd , Hampshire , UK ) allowing a fully automated image acquisition . Plates were imaged and analyzed in batches of 40 plates and a custom pipeline allowed parallel image processing and analysis by bundling images of fields and channels of several wells . During imaging an automated pipeline scheduled the processing of image files for each field of view through the following analysis workflow , here described representatively for one field . Raw images of three channels with a size of 8 . 4 MB ( 16-Bit grey scale , 2048 × 2048 pixels ) per image were captured with the InCell Analyzer 2200 software and saved as TIFF files on a server cluster for image processing and analysis . The image processing- and analysis pipeline covered two main blocks , first a sequence of pre-processing steps which was followed by extraction of phenotypic features from single cells . First , the images were read in and each channel was assigned to the subcellular structure that was selectively stained with the above described assay ( Hoechst: DNA , Phalloidin-Cy3: F-actin , α-tubulin-FITC: tubulin ) . To identify cell and nuclei boundaries , a duplicate image of each channel is ln transformed , scaled between 0 to 1 and smoothened by a Gaussian filter using a sigma of one . This reduced optical noise , improved the dynamic range and smoothened the image gradients for further segmentation by thresholding . For segmentation , the normalized actin and tubulin images were binarized by global thresholding . Second , the cell nuclei were identified by applying a local adaptive average threshold to binarize the DNA channel nuclei image and assigning objects . The resulting binary image was then subjected to morphological operations of opening and hull filling such that filled objects with smoothly roundish outlines result . Offsets for segmentation were varied if the channels surpassed certain thresholds . If more than 30 nuclei were counted per field , the objects were subjected to further propagation of nuclei objects into the an á priori defined cell body mask . Starting from the nucleus objects as seed regions , the cell bodies are segmented by propagating the nuclei objects into foreground area ( Carpenter et al . , 2006 ) . This strategy allowed to identify single cells and corresponding nuclei as objects . Using the segmented object outlines as masks , features on each object and channel were calculated on the original image using the R/Bioconductor package EBImage ( Pau et al . , 2010 ) . Specifically , numeric descriptors for five feature classes are defined in the computeFeatures function from EBImage ( Supplementary fle 6 ) : ( i ) shape features which quantify the shape of cells and nuclei , ( ii ) basic features that describe the summary statistics , such as 5% quantiles , of pixel intensity within the borders of the object , ( iii ) moment features that describe the spatial orientation of the objects , ( iv ) Haralick features derived from a pixel intensity co-occurrence matrix as texture descriptors ( Haralick et al . , 1973 ) and ( v ) social features such as distance to the first 20 nearest neighboring cells . Social features are derived by a k-nearest-neighbor search based on the geometric center points of single cells . Single-cell data were stored and aggregated to well averaged data by calculating the trimmed mean ( q = 0 . 01 ) of all cells belonging to all fields of one well and its standard deviation . For the analysis of the genome-wide drug-gene interaction screens the following analysis strategy was pursued: feature data was collected in a data frame containing per-well aggregated values as trimmed mean and standard deviation . This data frame was then reformatted to a 4-dimensional data cube featuring the dimensions: feature , plate , well , screen . Per feature , the feature’s minimum value was added to each value prior to logn ( x + 1 ) transformation to approach the features’ histogram to normal distribution . Following transformation , each plate in each screen was normalized separately for each feature by B-Score normalization ( Ljosa et al . , 2013; Mpindi et al . , 2015 ) . The B-Score normalization centers and scales the data to be the residuals of the median polish divided by the median absolute deviation ( mad ) across all values of the plate and thus be symmetrically centered around zero and scaled in units of the mad . Here , 38 representative features were chosen based on their biological significance ( our ability to refer them back to cellular phenotypes ) and their biological reproducibility between the two mock ( DMSO ) treated replicate screens and their information content , as measured by added variance ( Supplementary file 6 ) . For the combinatorial screens , the obtained data frame containing rows for each well and columns for each feature . In addition , we added well , plate and batch identifiers as annotation columns . Data acquired from single cells was aggregated by calculating the trimmed mean ( q = 0 . 01 ) for each feature extracted in the respective well together with its standard deviation . This way , outliers , produced by over or under segmentation of cells , were mostly excluded from further analysis . Data was normalized by dividing each feature in each plate by the median of the non-targeting control wells ( if that was not zero ) . Further , the values of each feature were transformed on a logarithmic scale using the generalized logarithm with c being the 3% quantile of the features value distribution over all values ( Caicedo et al . , 2017; Fischer et al . , 2015 ) . For each feature , data was subsequently scaled and centered around 0 by using the robust Z-transformation , where the feature median is subtracted from each value and the result was divided by the median absolute deviation ( x’= ( x-median ( x ) ) /mad ( x ) ) . After that , all features were in normalized units of median absolute deviation from the median of that feature and normalized per plate . The normalized feature vector provided the basis to all further analyses . The metrics used for judging the quality of dsRNA reagents and to assess the gene’s suitability for the combinatorial screen are summarized in Supplementary file 1 . For this purpose , several metrics have been deployed . Summarized , the applied metrics were used to assesses for each individual gene in the genome-wide HD3 library ( i ) the quality of the RNAi reagent , ( ii ) the effect size of the induced phenotypes under solvent control treatment as well as the differential effect size of the treatment-sensitive phenotype between small molecule treatment and control conditions , ( iii ) the quality of the target gene as a candidate for gene-gene-drug combinatorial screening . Effect size was quantified using the Euclidean distance ( ∑i=1n ( xi-yi ) 2 ) between sample and control measurements under different conditions . Quality of dsRNAs was assessed by calculating Pearson correlation coefficients between phenotypic profiles of biological and dsRNA design replicates . The quality of genes as screening candidates was assessed by gene expression analysis and literature analysis . The Q1 metric shows the strength of a knockdown induced profile when compared to the non-targeting control knockdown ( here: GFP ) . This was calculated as the Z-Score normalized Euclidean distance of the genes profile to the control profile and can be used to inform if a phenotype of a gene is exceptionally strong or weak . In general , strong phenotypes ( Q2 ) were preferred since they were more robust to experimental noise and are likely to engage in many genetic interactions ( Costanzo et al . , 2010 ) . Q3 gives to what extent the phenotypic profile of those genes’ knockdown changes upon drug treatment . An ideal candidate for drug gene interaction screening shows a high value in this metric . Q4 and Q5 allow inferring the reproducibility of the measured phenotype by comparing the correlation of two sequence-independent dsRNA designs targeting one gene and the correlation of one design across screen replicates , respectively . There , 7957 genes were targeted by designs whose feature vectors correlate with PCC >0 . 5 while 17263 designs were reproducible between screens ( PCC >0 . 5 ) . Q6 was used to infer if the respective gene is expressed under the screened conditions ( S2 cells , 4 days in culture in Express-V medium ) . 12567 genes ( 88% of all genes screened ) had a log2 normalized read count greater than 0 . In contrast , the knowledge sum in Q7 was used to avoid over enrichment of well-characterized genes in the final combinatorial library . The ‘unknown’ was defined by means of assigning each gene a score describing how well it has been studied and characterized . Therefor the Gene Ontology terms associated to each Drosophila gene were downloaded from Flybase . In Flybase , each ontology term is annotated with evidence codes as provided by the gene ontology consortium ( Ashburner et al . , 2000 ) . Each of these codes was then used to assign weights to the ontology terms for each gene ( Supplementary file 7 ) . Ontology terms derived from experimental evidence , such as genetic interactions , direct assays or physical interactions were assigned the highest weight while computational annotations were weighted the lowest . For each gene , the sum of ontology terms was computed and used as a proxy for the current state of its functional characterization . For example , the cell fate determining receptor Notch is the most well studied gene with a score of 973 , while all genes have an average score of 34 . 7 and the third quartile ends at 41 . This means that only a minor fraction of genes is as well studied as Notch and most genes can be accounted as uncharacterized if their score is beneath 100 ( 90 % quantile ) . An example for such a gene is tzn with a knowledge sum of 14 . Only known fact about tzn is its function as Hydroxyacylglutathione hydrolase in response to hypoxia ( Neely et al . , 2010; Jha et al . , 2016 ) . For screening , genes with a low knowledge sum were preferentially chosen . The data frame with normalized feature data per well was reformatted into a five-dimensional data cube representing the experimental design . The dimensions are: target dsRNA ( two entries for each gene ) , query gene , time , treatment and feature . The data cube was further subjected to genetic interaction analysis following the protocol established by Bernd Fischer ( Fischer et al . , 2015; Horn et al . , 2011; Laufer et al . , 2013 ) . There , genetic interactions are defined as the residuals of a modified median polish over the double perturbation matrix of one replicate , feature , treatment and time point . The median polish presents a robust linear fit ( Mij=mi+nj+πij+ε ) that lifts the main effects ( m , n ) of each query such that it resembles the value of a single gene knockdown . The residuals of this fit scaled by their median absolute deviation are defined as π-scores . π-scores further provide us with a quantitative measure of genetic interaction following the multiplicative model plus some error term ( ε ) estimating the experimental noise . There , the interaction of two genes is defined as the deviation of the measured combined phenotype ( Mij ) from the expected phenotype for a target-query gene pair i and j . The expected phenotype is defined as the product of the two independent single knockdown phenotypes . The resulting π-scores are then collected for all replicates ( dsRNA and experimental , each interaction is measured four times ) . The significance of their mean over all measured scores is estimated by a moderated students t-test as is implemented in the R-package limma . There , the t-test is adapted for situations where a small amount of observations is tested in many tests , normally causing large test variability , using an empirical Bayes variance estimator . p-values were adjusted using the methods of Benjamini Hochberg ( Benjamini and Hochberg , 1995 ) . From there on , adjusted p-values can be treated as false discovery rates . The FDR estimates the chance that the finding was observed by random chance given the entire dataset . This described procedure was applied to quantitatively calculate genetic interactions for each phenotypic feature . To identify a hit-list of condition-sensitive gene-gene interactions , we tested whether the changes of genetic interactions over time and between different conditions could be quantitatively described by a multi factorial linear model . This would provide the possibility to ( i ) quantify the time dependence of an interaction and ( ii ) to measure the phenotypic difference between treatment conditions with high confidence . For every gene-gene combination [i , j] screened across time and chemical treatment , we used a two-factor robust linear model , which we termed model of differential interactions ( MODIFI ) , to estimate the predictive strength and influence of time and differential compound treatment on the π-score ( πij=cij+ σij*time+δij*treatment+εij ) . Therein , the coefficient σij models the time dependence , δijmodels the quantitative offset between treatments , c estimates the intercept and the residual εij , estimates the error of fit for each combination of the target gene i and the query gene j . σ and δare thus parameter estimates that uniquely describe the behavior of each gene-gene interaction . A separate model was fitted for every feature and every gene-gene combination using the “rlm” function of the R/MASS package . A p-value denoting the predictive power of each covariate ( time , treatment ) on the π-score was estimated by a robust F-test as implemented in the function f . robftest function from the R/sfsmisc package . For statistical assessment the difference in interaction strength is used , as opposed to the interaction in a single condition . This way MODIFI identifies a great number of treatment sensitive interactions where the interaction score in each isolated condition is small , but the difference between conditions is significant . This resulted in a data frame that , for each gene-gene combination and each feature , contains a p-value for each covariate and its estimate . The p-value was multiple testing corrected by FDR analysis ( Benjamini and Hochberg , 1995 ) . Interactions with an FDR<0 . 1 in either term ( time , treatment or both ) were called significant . The FDR threshold also served as the basis for classifying context-dependent interactions into the different classes ( Figure 1C ) . Interactions are time dependent if the adjusted p-value for the time term is below 0 . 1 , treatment sensitive when the adjusted p-value for the treatment term is below 0 . 1 , and context-independent else . Interactions are aggravating when the π-score is negative and alleviating if it is positive . Quantitative real-time PCR ( qPCR ) was used to analyze the transcriptional response following Rel/pnt co-RNAi . To this end , as 5*105 cells / well were seeded in 630 µl ExpressFive ( Gibco ) culture medium and reverse transfected with 14 µg dsRNA . All dsRNAs denoted with #2 were used in three biological replicates and combinatorial RNAi was achieved by mixing 7 µg of dsRNA targeting each gene ( Supplementary file 8 ) . After 72 hr incubation ( 25°C , no CO2 adjustment ) , cells were washed once in 750 µl PBS ( Gibco ) and lysed in 350 µl RLT buffer shipped with the RNAeasy-mini Kit ( Qiagen ) . RNA was then purified from all samples according to manufacturer's standard instructions for spin column purification . An optional DNase digestion step was performed using the RNase-Free DNase Set ( Qiagen ) . Samples were prepared for qPCR by reverse transcription of 1 µg of RNA using RevertAid H minus First strand cDNA Synthesis kit ( Thermo scientific ) according to the manufacturer's standard protocol . A qPCR reaction was prepared using PrimaQuant 2x qPCR-Mastermix ( Steinbrenner ) by mixing 5 µl of sample ( 1:10 diluted cDNA ) with 5 µl of Mastermix ( including 0 , 3 µM of forward and reverse primer , Supplementary file 9 ) on a 384-well qPCR plate ( LightCycler 480 Multiwell Plate 384 , white , Roche ) . The plate was then centrifuged ( 2 min , 2000 rpm ) and processed for qPCR in a Roche 480 LightCycler using the following PCR program: ( i ) 10 min at 95°C , ( ii ) 15 s at 95°C , ( iii ) 60 s at 60°C , repeat step ii ) and iii ) 40 times and measure fluorescence at 494 nm-521 nm during step iii ) . Melting curve analysis of each sample was performed to assess reaction quality . Relative expression of each gene in each sample ( normalized to rps7 expression ) was analysis as log2-foldchange over RLUC dsRNA-treated samples ( Nolan et al . , 2006; Schmittgen and Livak , 2008 ) . qPCR primers were designed using the GETprime web service ( Gubelmann et al . , 2011 ) . For analysis , all genes in the combinatorial library were annotated manually using FlyBase and literature annotations ( Marygold et al . , 2013 ) . A more detailed description of all methods including those for supplementary materials can be found in Appendix 1 . All code used for the analysis presented in this study is available for download at: https://github . com/boutroslab/Supplemental-Material/tree/master/Heigwer_2018 ( Heigwer , 2018; also forked at https://github . com/elifesciences-publications/Supplemental-Material/tree/master/Heigwer_2018 ) . All raw data is available at: https://doi . org/10 . 6084/m9 . figshare . 6819557 | Within a cell , communication routes that involve many different genes work to control how the cell responds to the environment . Although different communication routes – so called signaling pathways – control different cell activities , they do not work in isolation . Instead , they form part of larger regulatory networks that maintain the cell in an appropriate state . As such , changing the activity of one pathway may in turn affect another seemingly unrelated pathway . The Ras signaling pathway helps to control when cells divide . When this signaling is not regulated correctly , cells can start to divide uncontrollably , leading to cancer . Drugs that suppress the activity of overactive Ras pathways could help to treat cancer . But how do the wider regulatory networks in the cell rewire themselves over time in response to this treatment ? To investigate this question , Heigwer et al . used a method called RNA interference to alter the activity of different pairs of 168 genes in fruit fly cells that had been grown in the laboratory . This meant 12 , 768 gene interactions were examined in total . Some of the cells had been treated with a drug that suppresses Ras signaling . By developing a new cell imaging and analysis system , Heigwer et al . could examine how the cell’s regulatory networks were affected by the drug at three different time points after treatment . The results show that housekeeping genes , which handle basic cell duties , take more time to rewire their interactions than signaling pathways . Heigwer et al . also developed a computational method – called MODIFI – to analyze how environment and time affect how genes interact . This highlighted a number of signaling pathways that are strongly affected by the suppression of Ras signaling , including an unexpected immune signaling pathway . In the future , more research will be needed to study the context-dependency of interactions between genetic networks in different cell types and in living organisms . A better understanding of this context-dependency will be important to understand how cancerous cells develop drug resistance . The data collected by Heigwer et al . could also be used by other researchers to explain any unexpected gene interactions that affect the signaling pathways they are studying . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"research",
"communication",
"computational",
"and",
"systems",
"biology",
"genetics",
"and",
"genomics"
] | 2018 | Time-resolved mapping of genetic interactions to model rewiring of signaling pathways |
A central problem in infection biology is understanding why two individuals exposed to identical infections have different outcomes . We have developed an experimental model where genetically identical , co-housed Drosophila given identical systemic infections experience different outcomes , with some individuals succumbing to acute infection while others control the pathogen as an asymptomatic persistent infection . We found that differences in bacterial burden at the time of death did not explain the two outcomes of infection . Inter-individual variation in survival stems from variation in within-host bacterial growth , which is determined by the immune response . We developed a model that captures bacterial growth dynamics and identifies key factors that predict the infection outcome: the rate of bacterial proliferation and the time required for the host to establish an effective immunological control . Our results provide a framework for studying the individual host-pathogen parameters governing the progression of infection and lead ultimately to life or death .
Two hosts exposed to apparently identical infections may have dramatically different outcomes . In one case , the host may experience only mild symptoms and recover easily , while in another the host may suffer devastating illness or even death . In many pathogenic infections , only a fraction of the infected hosts die from the infection or carry a high pathogen burden , while other hosts may strongly limit parasite proliferation . For example , mosquitoes infected with malaria experience high and low pathogen load ( Molina-Cruz et al . , 2008 ) , rats infected with Enterococcus faecalis experience bimodal pathogen burden ( Frank et al . , 2015 ) , and even the abundance of gut microbes exhibits bimodal variation ( Lahti et al . , 2014 ) . Genetically identical hosts may still experience drastically different infection outcomes ( e . g . clonal population of Daphnia magna infected by one strain of bacteria [Duneau et al . , 2012] ) . These observations suggest that the process of infection could itself be intrinsically variable . While such differences can sometimes stem from genetic or environmental variation among hosts , in other cases the difference in outcome appears to be more arbitrary or stochastic , and the relative contributions of genetic , environmental and random individual variation remain unclear . Variation in within-host disease processes can affect pathogen evolution by modulating the within-host evolutionary dynamic during the course of the infection ( Alizon et al . , 2011; Mideo et al . , 2011 ) . Theory predicts that variation in traits such as bacterial growth rate or host response time to an infection can affect the outcome of the infection and thus the epidemiology of the disease ( Alizon et al . , 2011; Fenton et al . , 2006 ) . To better understand the selection pressures that shape host-parasite interactions , it is important to empirically open the ‘within-host black box’ and characterize the different phases of an infection and to define the evolutionarily relevant parameters in the host and parasite ( Alizon et al . , 2013 ) . Drosophila is a prime model to study host-microbe interactions ( Buchon et al . , 2014 ) . Drosophila relies on both humoral and cellular responses to fight infection ( Lemaitre and Hoffmann , 2007 ) . The humoral response consists of the secretion of antimicrobial peptides ( AMPs ) into the hemolymph ( insect blood ) and the activation of the melanization cascade , both of which limit microbial proliferation . The cellular responses include encapsulation of foreign bodies and phagocytosis by dedicated immune cells . Antimicrobial peptides are constitutively expressed in a restricted set of barrier tissues at low levels but are massively upregulated in the fat body in response to infection . Two major signaling pathways , the Toll and Imd pathways , drive the induction of AMPs in the fat body . Importantly , we and other groups have observed strong inter-individual variation in the outcome of Drosophila infected with a wide range of bacteria and viruses ( Chambers et al . , 2014; Clemmons et al . , 2015; Ferreira et al . , 2014; Howick and Lazzaro , 2014; Kutzer and Armitage , 2016 ) , which suggests that Drosophila is a good model to study the stochastic nature of infection . In the present work , we aim to identify key parameters of the intra-host dynamic that underlie the inter-individual variation in the outcome of infection . We first demonstrate that infection of Drosophila melanogaster by a series of bacterial pathogens leads to highly variable infection outcomes . Notably , infections with the Gram-negative Providencia rettgeri and the Gram-positive Enterococcus faecalis result in one of two binary outcomes: some individuals die with a high pathogen burden , whereas others survive indefinitely with fairly asymptomatic and low-burden persistent infections . Different individuals of the same age and genotype may experience either outcome , even when co-housed in the same rearing vial . We find that bacterial burden upon host death did not correlate with the time post-injection at which death occurs , and that the lethal burden varies across bacterial species and strains . Surviving flies sustain a much lower pathogen burden , which also varies across bacterial species and can depend on the initial inoculum and on the immune response . Using a variety of experimental manipulations , we determined that inter-individual variation occurs as a consequence of the interplay between bacterial growth rate and the activation of the host immune response . We developed a theoretical model that describes the probabilistic nature of bacterial growth and thus host survival based on three key factors: the net rate of bacterial proliferation ( µ ) , the timing of effective immunological control ( Tc ) , and an inferred threshold pathogen density ( ntip ) that must be reached to enable the bacteria to switch to a lethally acute infection instead of a chronically persistent one . Our model predicts infection dynamics accurately and suggests that inter-host variation in survival must originate in the ability of the pathogen to reach the ntip threshold before effective immune control is established . We observe empirically that the quantitative activation of the immune response varies among individual hosts , suggesting that inter-individual stochasticity early in infection might lead to differences in the probability of ultimate survival . Altogether , our results illustrate the mechanisms underlying the variable nature of infection outcome and provide a framework for studying the distinct parameters that govern the progression of infection and lead ultimately to life or death .
Infection with different pathogens produces dramatically different outcomes that range from the certain death of all host individuals to the nearly certain survival of all host individuals . In order to define the range of outcomes , we infected male Drosophila melanogaster with a panel of seven bacteria by injection of a controlled dose into the abdominal cavity ( Figure 1A ) . Infection with Providencia alcalifaciens and Serratia marcescens Db11 was lethal to the flies in a short timeframe , killing all individuals within one day of infection . Conversely , nearly all individuals survived infection with E . coli and Erwinia carotovora Ecc15 ( formally known as Pectinobacterium carotovorum carotovorum 15 ) ( Figure 1A ) . Infections with different natural pathogens of Drosophila including P . rettgeri , E . faecalis , P . burhodogranariea resulted in the death of only a fraction of the infected population ( Figure 1A ) . It is commonly assumed that inter-individual differences can stem from genetic differences or environmental variation , including nutrition , gut microbes or rearing conditions . We therefore controlled for these sources of variation by infecting same-aged , genetically identical hosts ( isogenic lines of the Drosophila Genome Resource Panel and w1118 ) , reared in a common environment with a controlled microbiota and given experimentally identical infections ( Figure 1B ) . Even under experimentally identical conditions , we continued to see dual outcomes of infection . In addition , the relative probability of the infection achieving either outcome varied among genotypes . Inter-individual variation could also stem from small variation in fly handling at the time of infection . For instance , differences in exposure to the CO2 used for anesthesia during the experiments could influence the probability of death upon infection ( Helenius et al . , 2009 ) . However , we found that even 15 min of exposure to CO2 prior to or after injection did not affect the chance of survival relative to 2 min of exposure ( Figure 1—figure supplement 1 ) . These results demonstrate stochastic variation in the outcome of infection ( life or death ) even among genetically identical individuals , although host and bacterial genotype combine to determine the probability of surviving infection . One could reasonably predict that individuals that survive their infections may do so by having cleared the infecting pathogen . To test this , we quantified bacterial loads in individual flies that were still alive seven days after injection with E . coli , E . carotovora , E . faecalis , and P . rettgeri . At this point , mortality curves had plateaued and little to no further death was observed ( Figure 1A ) . We found that none of the bacteria had been entirely cleared from the host and nearly all individual hosts still had measurable loads of bacteria ( Figure 1C ) . In the case of the most benign pathogens , which did not lead to the death of any host individuals , a few host individuals either cleared the infection or sustained loads below our detection threshold of 10 bacteria per fly ( ~22% for E . coli and ~8% for E . carotovora ) . However , most surviving hosts developed persistent , chronic infections . In addition , the bacterial load sustained during the persistent phase of infection was constant but distinct for each bacterium ( e . g . , Figure 1—figure supplement 2A ) . We termed this parameter the ‘Set Point Bacterial Load’ ( SPBL ) , in analogy to the commonly used ‘Set Point Viral Load’ for chronic viral infections ( Fraser et al . , 2014 ) . Interestingly the SPBL for E . coli and E . carotovora were significantly lower than those from E . faecalis and P . rettgeri infections , which also differed between each other ( Figure 1C ) . The SPBL for E . coli and E . carotovora , our two least virulent bacteria , was a few hundred bacteria per infected host . In contrast , the SPBL for pathogens of intermediate virulence , such as P . rettgeri and E . faecalis , centered around a few thousand bacteria , and no single fly ever cleared the infection . In addition , we found the SPBL for P . rettgeri to be significantly different among three D . melanogaster host genotypes ( Figure 1C ) , increasing with the size of the initial inoculum ( Figure 1—figure supplement 2B ) . We next examined the individual hosts that died from their infections . In the case of P . rettgeri , only a subset of infected hosts die . Mortality begins as soon as 16 hr post-injection and continues over the course of a few days before survivorship plateaus ( Figure 1A , B ) . We quantified the Bacterial Load Upon Death ( BLUD ) by sampling individual flies within 30 min after their death , before the confounding effect of post-mortem bacterial proliferation . We compared those loads to pathogen burdens of still-living individuals sampled from the population all along the course of infection ( Figure 2Ai ) or at a unique time point where 5% of the population died at once ( Figure 2Aii ) . The P . rettgeri loads in recently-dead individuals were 223 . 4 ± 21 . 6 ( mean ±sd ) bacteria/fly , regardless of the time post-injection at which the host died ( Figure 2B and Figure 2—figure supplement 1A ) . These loads were consistently 100-fold higher than those of living individuals ( 215 . 8 ± 23 . 8 bacteria/fly; Figure 2A ) . The narrow range of bacterial load upon death and the fact that live flies carry fewer bacteria than recently dead flies ( Figure 2Aii ) suggests that the BLUD is a lethal load and cannot be sustained in living individuals . We found that there is a BLUD associated with every bacterium and that the BLUD differed significantly between bacteria ( Figure 2C ) , indicating that this threshold is a property of the pathogen . Surprisingly , the BLUD did not reflect the time it took various pathogens to kill the host . For example , the BLUD for S . marcescens , which killed wild-type flies within 8 hr of injection , was 220 . 2 ± 20 . 7 bacteria/fly . This is lower than that of P . alcalifaciens ( 224 . 3 ± 20 . 5 ) , which killed within 24 hr , but higher than for S . aureus ( 218 . 1 ± 20 . 7 ) , which killed within 28 hr ( Figure 2C ) . Because the BLUD varies across bacteria but is constant for a given bacterial strain in a given environment , we propose that it provides a novel measure of bacterial pathogenicity or toxicity to the host . Reciprocally , we reasoned that the BLUD also represents a measure of host disease tolerance corresponding to the maximal bacterial load that can be sustained before the individual dies from the infection . As a measure of disease tolerance , the BLUD might vary among hosts . We therefore next measured both inter- and intra-specific variation in the BLUD after infection with P . rettgeri . We first assessed the BLUD of P . rettgeri-infected flies belonging to four different Drosophila species ( D . sechellia , D . erecta , D . mojavensis and D . ananassae ) and to four different D . melanogaster wild-derived isogenic lines ( RAL-714 , RAL-359 , RAL-138 and RAL-882 ) . We found that the BLUD varied significantly across Drosophila species ( Figure 2D ) and across different D . melanogaster genotypes ( Figure 2E ) . To our surprise , however , the BLUD did not depend on the initial inoculum starting the infection ( Figure 2—figure supplement 1B ) . We tested this by challenging wild-type D . melanogaster lines with P . rettgeri infection at five different doses ranging from ~300 ( OD600 = 0 . 01 ) to ~150 , 000 ( OD600 = 5 ) bacteria injected . Flies injected with higher doses died earlier than flies injected with lower doses ( Figure 2—figure supplement 1C–D ) but BLUD did not correlate with the dose injected ( Figure 2—figure supplement 1B ) . These data affirm that hosts die at a lethal burden that does not depend on their initial infection dose or on their time to death . Having described the BLUD as a measure of disease tolerance , we sought to determine whether the immune response , the main mechanism of bacterial elimination ( i . e . resistance ) , affected the BLUD . We did not detect significant differences in the BLUD between any of the immune-deficient hosts and their corresponding wild-type genotypes ( Figure 2F ) . This result supports the idea that the BLUD is an exclusive measure of tolerance and not resistance . Collectively , our results indicate that bacterial infections of Drosophila result in binary outcomes . Either the host dies from acute infection with a stereotypical bacterial burden that is independent of dose , the BLUD , or the host survives but sustains a chronic infection at low and also stereotypical bacterial burden that does depend on initial dose , the SPBL . These binary outcomes are general to D . melanogaster and are observed in all wild-type host genotypes and species that we tested , although the relative proportion of hosts entering either state may vary with pathogen identity and host genotype ( Figure 1A and B ) . Although the BLUD is a key parameter to describe bacterial virulence and host disease tolerance , it does not explain the observed inter-individual variation in the outcome of infection ( death versus chronic persistence ) . We have shown that individual hosts may either survive or die from their infections , and we hypothesized that among-individual variation in survival might result from differences in bacterial growth dynamics at early stages of infection . In order to test this hypothesis , we initiated an in-depth study of bacterial growth in response to three categories of pathogens: those that quickly kill all infected hosts , those that kill none , and those that kill only a fraction of the infected population . We measured the bacterial load in individual flies at regular one-hour intervals for the first 16 hr after injection ( Figure 3 ) . Upon infection with two avirulent bacteria , E . coli and E . carotovora , bacterial burdens inside individual flies decreased soon after injection ( Figure 3A ) , suggesting that these bacteria are controlled within a couple of hours by the host , even though they are not completely cleared . After injection with bacteria that kill all infected hosts in few hours ( P . alcalifaciens or S . marcescens ) , the pathogen burdens increased monotonically until all individuals were dead ( Figure 3B ) , suggesting that the host fails to control the infection . The pattern was very different for pathogens killing only a fraction of the infected host population: E . faecalis , P . burhodogranariea , P . rettgeri ( Figure 3C–E ) . For these , we found that bacteria proliferated homogenously in all individuals during the early phase of infection ( up to around 6–8 hr ) , but that subsequently two groups of hosts could be distinguished . Hosts of one group carried high and increasing bacterial loads , and we hypothesized that these individuals are fated to die . Hosts in the other group exhibited a smaller bacterial load that stabilized over time , and we hypothesize that these are the individuals who will sustain chronic infections . Remarkably , the in vivo proliferation of P . rettgeri in the early phase of infection is equivalent to the rate of in vitro growth in LB medium ( Figure 3—figure supplement 1 ) , suggesting that the bacteria are growing largely without constraint in this early phase . The same is true of the always-lethal P . alcalifaciens and S . marcescens ( Figure 3—figure supplement 1 ) . Proliferation continues at this same rate in the host group with the higher bacterial load up until the BLUD is reached , at which point the host dies . These results are consistent with the hypothesis that host variation in the outcome of infection might stem from differences in the ability of the fly to control infection in the early stage . Although perhaps unlikely , one possible explanation for the observed data would be that a subset of hosts infected with P . rettgeri , E . faecalis , P . burhodogranariea do not survive their infections because a mutation that provides resistance to the host immune system occurs in the bacterial population , allowing these bacteria to continue to grow instead of being controlled . If this were true , the bacteria that kill the fly would have evolved hyper-virulence and be able to outcompete their wild-type counterparts . Conversely , the flies that survive their infections might conceivably do so because the bacteria infecting them sustained loss-of-virulence mutations . To test these hypotheses , we sampled bacteria from flies bearing either low ( i . e . close to the SPBL ) or high bacterial load ( i . e . close to the BLUD; Figure 4Ai ) at 12 hr post-injection and used them to infect new cohorts of flies . In both cases , we found that bacteria isolated from infected hosts were as likely or more likely than the original bacterial clone to kill new hosts , although we see no difference in mortality upon infection with bacteria from high-load versus low-load flies ( Figure 4Aii ) . These data collectively provide no evidence that the probability of a host controlling or failing to control bacterial infection results from heritable changes in the bacterial population during infection . We next hypothesized that the host immune response might be primarily responsible for regulating pathogen burden , and that variation among individual hosts in immune activity might determine whether infections proceed to acute lethality or settle into chronic persistence . To test the hypothesis that the existence of a binary outcome to infection depends on the immune system , we studied P . rettgeri growth during an in vivo infection and the probability of death in hosts deficient for either the cellular or humoral immune responses . Engulfment by macrophage-like cells , called phagocytes , is considered the first line of immune defense once the parasite is inside an insect . We thus tested whether phagocyte activity predicts inter-individual variation in bacterial growth . We infected hosts in which phagocytes were genetically ablated with P . rettgeri and compared them to infected control hosts ( Figure 5A ) . Individual hosts separated into high-load and low-load groups over the first 16 hr of infection both in the presence and in the absence of phagocytes . We therefore concluded that phagocytosis is not a determinant of the switch between lethal and persistent P . rettgeri infection under our experimental conditions . We next investigated the contribution of the humoral response to inter-individual variation in bacterial control . Gram-negative bacteria are mainly controlled by the Imd pathway , a major regulator of AMP production ( Buchon et al . , 2014 ) . We tested whether flies deficient for the Imd pathway ( DreddEP142 mutant , Figure 5B ) show inter-individual variation in P . rettgeri proliferation . No individual DreddEP142 mutant fly could control infection and all mutant hosts died with a high pathogen burden ( Figure 5B ) . This demonstrates that the Imd pathway is required to control P . rettgeri infection and shift the infection to a persistent state in a subset of hosts . The Toll pathway provides the primary response against Gram-positive bacteria but can also be activated by host damage and pathogen-derived virulence factors ( El Chamy et al . , 2008 ) . We infected Toll-deficient ( spzrm7 mutants ) and wild-type hosts with P . rettgeri and observed the emergence of discrete high-load and low-load groups of flies of both genotypes during the first 16 hr of infection ( Figure 5C ) . This demonstrates that Toll-deficient flies are sometimes able to control P . rettgeri infection . We next tested the role of the Toll pathway upon infection with a Gram-positive bacterium , E . faecalis , which also results in either acute lethality or chronic persistence in wild-type flies ( Figure 3C ) . Without exception , flies deficient for the Toll pathway ( spzrm7 mutants ) failed to control E . faecalis infection and all hosts died with high bacterial loads ( Figure 5D ) . These data collectively demonstrate that the specific and appropriate arm of the humoral immune response is critical for the control of bacterial infection and establishment of persistent infection . Altogether , our data indicate that the capacity to control infection versus allowing it to proceed to the lethal state is determined by the action of the specific and appropriate arm of the immune response , but that the lethal pathogen burden is independent of immunity . This reinforces the idea that the BLUD is a readout of disease tolerance , but that the probability of the ultimate outcome of infection is mechanistically distinct and determined by host immunological control . We have empirically demonstrated that even genetically identical hosts differ in their ability to control infection , which ultimately leads to binary outcomes of death at high pathogen burden or survival with chronically persistent infection . In order to quantitatively define how the kinetics of bacterial proliferation and host response determine ultimate infection outcome , we developed a model to estimate bacterial growth within each individual host ( illustrated in Figure 6A ) . Our model assumes that in the early phase of infection , the bacteria grow unchecked within the host at rate μ . Empirical support for this component of the model comes from the observation that bacteria proliferate at the same rate as in in vitro culture over the first few hours of infection in a wild-type host and until death in an immune deficient host ( as shown in Figure 3—figure supplement 1 ) . We assume that there is a time post-injection where the immune response of an individual fly becomes sufficiently active to restrict bacterial proliferation , which we term the ‘time to control’ ( Tc ) , and we define t̄c as the average time to control of a given population . This is based on the empirical observation that an intact immune response is required for establishment of a persistent infection ( see Figure 5B ) and that the timing of our observed divergence in host trajectories corresponds roughly to the time at which the D . melanogaster humoral immune response begins to become active ( Lemaitre and Hoffmann , 2007 ) . We also observed that infections that kill only a fraction of the hosts reach a state where flies either have a high bacterial load , in which case they die , or a low bacterial load , in which case they survive . We model this divergence between the two types of bacterial kinetics by hypothesizing a threshold bacterial number ( ntip ) that , if reached prior to control , results deterministically in ultimate host death at high bacterial load ( the BLUD ) . By applying our model to the empirical data , we can obtain for each individual host a probability corresponding to the likelihood that control is effective ( see Figures 3 and 5 , where white dots are infections that are most likely uncontrolled while red dots are most likely to be controlled ) . This probability is computed as the probability that bacterial load does reach ntip before control becomes effective . Using this computation , we estimated the overall probability that an individual host controls an infection by a given bacterium . We found that the estimated probability of control in our model , which is obtained from the measurements of bacterial load in different flies at different times along the infection ( i . e . there was no repeated measurement on the host ) , correlates well with the empirically observed proportion of hosts that are killed ( Spearman rank correlation test: S = 3 . 02 , r = 0 . 97 , p-value=8 . 17e-6; Figure 6B ) . This is both an indication that the model adequately describes our data and a confirmation that the probability of death is determined by the early phase of infection as described by the model . One parameter that our model allows to estimate from experimental data on early growth is the average time to control ( t̄c; Figure 6C ) . We hypothesized that ‘priming’ the immune system prior to infection should dramatically reduce this parameter and therefore yield much higher probability of controlling the infection . To test that , we infected transgenic flies that had constitutive activation of the Imd pathway with P . rettgeri and monitored bacterial growth . We found that bacterial loads in hosts with an immune system that was genetically activated prior to infection began to decrease immediately after injection , demonstrating effective control at the moment of infection ( t̄cc0 ) ( Figure 7A ) . Surprisingly , however , the bacteria grew to acute lethality in a small proportion of individuals with constitutive Imd pathway activation ( Figure 7A ) , suggesting that the Imd pathway alone is not always sufficient to control P . rettgeri . We compared the average time to control ( t̄c ) estimated from the model with experimental measurements of the time at which host immunity impacts bacterial load after infection with the same bacteria . We quantified proliferation of E . carotovora and P . rettgeri in wild-type and Imd pathway mutant hosts ( Figure 7B ) and the growth of E . faecalis in wild-type and Toll pathway mutant hosts ( Figure 5D ) , using the logic that bacterial loads should become measurably different between wild-type and immune deficient flies soon after control by the immune response occurs in the wild-type host . Using this method , we found that the effect of the immune response is detectable starting at 2 hr post-injection for E . carotovora , at 8 hr post-injection for P . rettgeri , and at 2 hr post-injection for E . faecalis ( Figures 7B and 5D ) . These empirical measurements of t̄c ranges overlap with the confidence intervals of t̄c that our model predicts for these infections ( Figure 6C ) . We found variation in t̄c across bacterial infections , and we next hypothesized that this variation , because it impacts the probability of control ( Pc ) , could partly determine the outcome of infection . In our model , the probability of control also depends on parameters that determine early bacterial growth . To analyze the contribution of t̄c and bacterial growth rate to the outcome of infection , we compared the likelihood ( i . e . how well the model fits the data ) of a complete model fitting the data through the estimation of 11 parameter values for each type of infections to the likelihood of models where some of the parameters are held constant for all infection types . For example , to estimate the role of the parameter t̄c , we compared likelihood of our full model , where each type of infection has its own value of t̄c , with a model fitting our data by estimating a common t̄c value for all infection types . A significantly better fit of the full model reveals a significant role of the parameter which was held constant in the second model . We performed this analysis for all the parameters involved in modeling the control of infection ( Control model , i . e . t̄c , Vc , ntip ) or the parameters involved in modeling early bacterial growth ( Growth model , i . e . tlag , µ , σb ) to test their relative impact on how accurately the model predicts survivorship ( Table 2 ) . We found that both sets of parameters involved in modeling the control of infection and early bacterial growth both have a strong impact on adequately predicting bacterial load ( Table 2 ) . We found that variation in survivorship is best explained by variation in t̄c when only bacteria displaying intermediate survival are considered ( i . e . when excluding E . coli , E . carotovora which have very low growth and mortality rates , and S . marcescens and P . alcalifaciens which have high growth and mortality rates , see Figure 6B ) . Conversely , when all bacteria were included in the analysis , differences in survivorship are best explained by differences in early bacterial growth ( Table 2 ) . These results suggest that the probability of surviving infection is affected by the average time at which control occurs ( t̄c ) , but that this effect can be masked by large differences in growth rates ( µ ) . Our analyses so far indicate that variation in infection outcome is strongly dependent on t̄c , which should be determined by both the kinetics of the host immune response and the bacterial sensitivity to host immunity . For infection by a given bacterium , with a known and constant t̄c , our model predicts that inter-individual variation in infection outcome must stem from variance in time to control ( Tc ) , quantified by Vc . Setting this variance to zero for all infections indeed produces a probability of control which is either zero , with all individuals dying from infection , or one , with all infections being controlled . Since we see empirical variability in control of some infections , we hypothesized this might arise from biological variance in Tc . We empirically tested the hypothesis that the speed and magnitude of Imd pathway induction might vary among individual flies , potentially resulting in variation in Tc that leads to variation in ultimate infection outcome . To test this , we first determined the kinetics of Imd pathway activation after infection with P . rettgeri by quantifying mRNA levels of the Diptericin gene , which encodes an antimicrobial peptide and provides an excellent read-out of Imd pathway activity ( Buchon et al . , 2014; Lemaitre et al . , 1997 ) . We began by measuring Diptericin expression in pools of 20 flies Canton S . Diptericin expression was extremely low at the time of injection ( Figure 7C , left ) , but a measurable induction was detected at 2 hr post-injection and its level continuously increased during the first 8 hr post-injection ( up to ~100 fold induction compared to the start of the infection ) . Thus , the Imd pathway is activated transcriptionally very early upon infection , well before we detected any impact on bacterial proliferation . To test whether inducibility is variable among individual hosts , we quantified Diptericin expression level in single flies at 4 hr post-injection ( Figure 7C , right ) . Time points equal or earlier than 8 hr were chosen because they occur before t̄c and before we begin to see inflated variance in the bacterial load of individual flies . At later times of infection , higher bacterial number leads to increased immune system activation in a positive feedback loop , so variation in AMP gene expression level becomes an indicator of the differences in bacterial loads instead of an estimation of the variance of induction in response to a unique bacterial load . At 4 hr post-injection , however , immune induction can be detected ( Figure 7C ) but immune control is not yet effective ( Figure 7B ) and no divergence in bacterial load is yet observed ( e . g . Figure 3D and E ) . At 4 hr post-injection , we detected strong among-individual variation in Diptericin expression , ranging from little induction to strong immune response , with approximately 5-fold difference between extremes ( Figure 7C ) , which may result from experimentally undetectable differences in handling , infection , or developmental history . We then did paired measurements of AMP expression and bacterial loads from single flies at 8 hr post inoculation , which is just before the predicted t̄c . The Imd pathway is activated and Diptericin is expressed primarily in the D . melanogaster fat body , most of which is located in the abdomen , and we determined that the bacterial burden in the head of a fly is tightly correlated with the load estimated from the thorax and abdomen: ( r = 0 . 84 , p=8 . 8e-9; Figure 7—figure supplement 1 ) , so we tested whether pathogen burden estimated from the heads of individual flies could be predicted by the Diptericin expression measured from abdominal RNA extractions . We found no significant correlation between bacterial load and the level of Diptericin transcripts 8 hr after injection ( Spearman correlation: r = 0 . 06 , S = 4225 . 88 , p=0 . 75 ) . This indicates that empirical variability in Imd activation is independent of variation in bacterial load at the early stage of infection . The disconnect may allow the infections of some hosts to grow through the critical threshold ntip prior to t̄c , ultimately leading to host death , while the seemingly identical infections of other hosts are effectively controlled .
We evaluated infection of Drosophila by multiple bacteria that cover a range of pathogenicity , ranging from completely lethal infections that kill all hosts to fairly benign infections where most or all hosts survive . Most interestingly , we demonstrate binary outcome of infection with pathogens such as P . rettgeri or E . faecalis . Even among hosts that are identical in genotype , sex and age and are raised in a common environment , a fraction of hosts die with high bacterial burden while the remainder survive indefinitely with persistent low pathogen load . We identify a set of key parameters that determine which of these outcomes occurs and we propose a general model framework to describe the dynamic interplay between host and pathogen . We find that host death is associated with a bacterial load , the BLUD , which varies with host genotype and pathogen strain but does not vary with inoculum dose . The BLUD does not correlate with the time a host will die from infection . Instead , we suggest that inter-individual variation in the probability of death stems from variation in the time it takes for each individual fly to establish effective immunological control of the infection ( Tc ) , with the understanding that this must occur before the pathogen reaches a critical density ( ntip ) if control is to be achieved . Hosts that survive their infections do so carrying a fixed bacterial load , the SPBL . Based on the quantification of bacterial load in the host , we have built a model that describes inter-individual variability in bacterial growth ( Figure 6A ) . In an early phase of the infection by pathogens , bacterial proliferation occurs at a rate similar to that of in vitro growth , suggesting that growth is essentially unconstrained in the first hours of infection . After a brief delay , the host immune response becomes sufficiently active so as to limit or perhaps even reverse bacterial growth . This time , which we parameterize as t̄c in our model , will certainly depend on both the kinetics of immune induction by the host and the sensitivity of the pathogen to the host immune response . In a following phase of our model , the infection resolves in one of two ways . The bacteria may continue to proliferate at the unconstrained rate until the host dies at a lethal load , the BLUD , or the bacteria may cease proliferation and settle into a phase of long-term persistent infection with a burden defined as the SPBL . In principle , the SPBL could be complete elimination of the bacteria , although we very rarely see that with D . melanogaster hosts . We infer that the probability of an infection entering into the lethal versus persistent state is defined by the probability that control is effective before bacteria grow above the critical threshold density , ntip . Thus , the probability of lethal infection is most influenced and therefore predicted by the rate of bacterial proliferation ( µ ) and the average time required to mount an effective immune response , t̄c . The simplicity of our statistical model yields some crucial insights and implications . First , a complex and apparently stochastic organism-level phenotype ( survival versus death ) can be satisfactorily predicted by modeling the probability of transitions between a few discrete states of infection ( early growth , uncontrollable growth , and decrease/stabilization after control ) . In the course of infection , each may occur in discrete temporal phases ( termed ‘early phase’ , ‘resolution phase’ , ‘terminal phase’ or ‘chronic phase’ in Figure 6A ) that have distinct and possibly deterministic outcomes , as those described in ( Duneau et al . , 2011; Ebert et al . , 2016; Hall et al . , 2017; Levin and Antia , 2001; Schmid-Hempel and Ebert , 2003 ) . Furthermore , each of these states and the parameters that determine them has a distinct mechanistic underpinning and may be differentially subjected to natural selection . For example , host evolution of shortened time to control through increased inducibility of the immune response would have a different genetic basis than one that involved reduced bacterial proliferation through host sequestration of nutrients ( e . g . metal ions or carbon ) . Both mechanisms might result in apparent ‘resistance’ to infection , both would contribute to increased probability of survival from infection , and each would carry distinct costs or tradeoffs . From the pathogen’s perspective , changes in its growth rate or sensitivity to host immune responses would also alter the probability that they become lethal versus persistent , also with distinct mechanistic bases and costs or tradeoffs . Each predictive parameter in the model can be experimentally manipulated , and the entire model can be theoretically evaluated to identify evolutionary optima in different environments and under different transmission requirements . We note that this model can be easily extended to infection by other classes of pathogen , to specific tissues such as the gut , or to hosts that display secondary immune responses ( e . g . , antibody-mediated acquired immune responses ) . Interestingly , the two endpoint parameters of infection , bacterial load in survivors ( SPBL ) and in hosts that succumb ( BLUD ) , were remarkably invariant among individual hosts of a same genotype even though they varied substantially across infections by different pathogens , host species and genotypes . The mechanisms underlying the death of a fly upon infection have been curiously overlooked , and the nature of damage to D . melanogaster from bacterial infection remains elusive ( Dionne and Schneider , 2008 ) . We find that flies die at a specific bacterial load that does not depend on initial infection dose or eventual time to death . This implies that death is unlikely to be a direct consequence of accumulated infectious damage , since longer term exposure to low pathogen burdens might trigger as much damage as short-term exposure to high bacterial burden . Instead , our results are more compatible with a model for death occurring as a consequence of sepsis or multiple organ failure driven by pathogen load ( Baue , 1975 ) . Of course , the lethal load of different pathogens would vary depending on the toxicity of the particular pathogen . In that sense , we can argue that the BLUD is a measure of pathogenicity of the bacterium , as it represents the minimal density of bacteria that kills a host . We hesitate to use the word ‘virulence’ though , because the BLUD does not correlate with the speed at which pathogens kill or with the probability of host death . Reciprocally , the BLUD also represents a measure of host disease tolerance ( Soares et al . , 2017 ) as it corresponds to the maximal bacterial load that can be sustained in the host without dying from the infection . Our results are somewhat inconsistent with those of Haine et al . , 2008 , which documented rapid cellular elimination of bacteria injected into Tenebrio molitor and posited an only minor role for the humoral response . In contrast , our study indicates that the timing and intensity of AMP production is the most crucial factor for controlling bacterial infection in Drosophila . One intriguing hypothesis for reconciling the two observations is that adult Drosophila have only a small number of hemocytes due to loss of hemocytes upon aging ( Guillou et al . , 2016 ) , while this might not be the case in Tenebrio . Thus , the relative importance of cellular versus humoral immunity may differ between the two insects . Additionally , different bacterial pathogens were employed in the two studies , and different microbes may be differentially sensitive to alternative arms of the immune response so generalizations should be made with caution . We found that complete bacterial clearance rarely happens in Drosophila . With all seven pathogens in this study , in no case were the bacteria fully cleared from all hosts even a week after the initial infection , even in cases where 100% of hosts survived the infection . Persistent infections stabilized at a fixed load for each pathogen , the SPBL . On rare occasion , individual flies in our study would carry persistent infections for several days before suddenly dying with a pathogen burden that had reached the BLUD . This suggests that there are some conditions under which a persistent infection can re-emerge as an acute infection . This is reminiscent of chronic infections by multiple human pathogens , including HIV , eukaryotic parasites such as Toxoplasma gondii , and bacteria such as Salmonella enterica , Mycobacterium tuberculosis or Streptococcus pneumoniae . These parasites are considered to be ‘specialists’ at chronic infection , yet our results suggest that qualitatively similar phenomena can be obtained from infection with a broad range of generalist bacteria . As the SPBL varies with both host and pathogen genotype , it is presumably subject to selection in both . In that context , it is worth noting that forcing an infection into the persistent state is effectively a disease tolerance strategy from the host perspective , yet the pathogen burden borne may still carry lifelong fitness cost . From the pathogen’s perspective , either persistence or acute lethality may be favored depending on the conditions of transmission . Moreover , selection for high growth rate due to competition between bacterial genotypes within a host ( Alizon et al . , 2009; 2013 ) may trade off against the capacity to establish persistent infection . Levin and Antia ( 2001 ) suggested that ‘us and other living organisms are little more than soft , thin-walled flasks of culture media’ . With the present work , we argue that hosts are slightly more complex and that potentially minute variations in host or pathogen physiology during the early phase of infection can have dramatic effects on the ultimate outcome . We have defined three parameters ( t̄c , μ , nTip ) that are sufficient to predict ultimate infection outcome , although we still do not know whether variation in those parameters is due to micro-environmental variation , uncontrolled plasticity in developmental history , somatic mutations , or other factors that are uncontrolled or uncontrollable in experiments ( e . g . depth of penetration of the needle during injection , site at which bacteria accumulate etc . ) and in nature ( e . g . time since last meal or mating , psychological status of the fly etc . ) . Nevertheless , what is becoming clear is that small differences in pathogen infectivity and host immunological control , especially in the early stages of infection , may manifest as large differences in the outcome of infection .
Drosophila melanogaster were reared on glucose-yeast medium ( 82 g/L yeast , 82 g/L glucose , 1% Drosophila agar , supplemented with 2 . 5 mg/L methylparaben and 10 mL of a solution of phosphoric and propionic acid: 41 . 5 ml Phosphoric Acid + 418 mL Propionic Acid + 540 . 5 mL distilled water ) . At day two after eclosion , adults were isolated in groups of five males and five females . All experiments were conducted with mated males 5 to 8 day post-eclosion . Rearing and experiments were conducted at 25°C ( ±1°C ) with a 12 hr light/dark cycle . Canton S , Oregon R and w1118 were used as wild-type laboratory strains . Eight fully isogenic lines from the Drosophila Genome Reference Panel ( DGRP: RAL-85 , RAL-138 , RAL-359 , RAL-882 , RAL-594 , RAL-707 , RAL-712 , RAL-714 ) were randomly chosen and used to assay the role of inter-individual genetic variation on several infection parameters ( Mackay et al . , 2012 ) . All wild-type stocks and isogenic lines were made axenic then re-associated with a controlled mixture of 5 bacterial species that compose the core microbiota of Drosophila ( Wong et al . , 2011 ) to ensure homogeneity of associated microbiological communities . Briefly , eggs of these stocks were sterilized by bleaching and reassociated with a mixture of 5 common gut microbes of Drosophila ( Acetobacter pomorum , Acetobacter tropicalis , Lactobacillus plantarum , Lactobacillus brevis , Lactobacillus fructivorans ) as previously published . Immune mutant stocks have been described previously and include mutants of the Imd pathway ( PGRP-LE112 , PGRP-LCΔE double mutants ( Takehana et al . , 2004 ) , DreddEP1412 ( RRID:BDSC_10456 , Leulier et al . , 2000 ) and RelishE20 mutants ( RRID:BDSC_55714 , Hedengren et al . , 1999 ) , mutants of the Toll pathway ( spzrm7 , [Lemaitre et al . , 1996] ) , and mutants of the melanization cascade ( double mutant PPO1Δ , PPO2Δ , [Binggeli et al . , 2014] ) . We used the Gal80TS; daughterless-Gal4 driver in combination with UAS-Imd to ubiquitously induce the Imd pathway at the adult stage . To generate phagoless adult flies , we ablated phagocytes by inducing the pro-apoptotic gene Bax in hemocytes using the hemocyte-specific driver Hml-Gal4 ( Hml-Gal4 >UAS GFP [RRID:BDSC_30140]; UAS-Bax , Gal80TS ) in adult flies at 2 days post-eclosion ( Defaye et al . , 2009 ) . The bacterial strains used in this study include the Gram-negative Providencia rettgeri ( strain Dmel ) ( Juneja and Lazzaro , 2009 ) , P . burhodogranariea ( Juneja and Lazzaro , 2009 ) , Serratia marcescens ( strain Db11 ) ( Kurz et al . , 2003 ) , E . coli ( Type strain ) , Pectinobacterium carotovorum carotovorum 15 ( strain Ecc15-GFP; formerly genus Erwinia , [Basset et al . , 2000] ) , Pseudomonas entomophila ( Vodovar et al . , 2005 ) , and Providencia alcalifaciens ( Juneja and Lazzaro , 2009 ) , as well as the Gram-positive Enterococcus faecalis ( isolated from wild-caught D . melanogaster by B . Lazzaro ) , and Staphylococcus aureus ( strain PIG1 [Liu et al . , 2005] ) . Cultures were grown to saturation overnight at 37°C ( 29°C for E . carotovora ) in LB liquid medium ( LB broth , Miller , VWR ) . Saturated cultures were suspended and diluted in Phosphate Buffered Saline ( PBS , pH 7 . 4 ) to the desired optical density ( OD600 = 1 or as otherwise indicated ) . We injected 23 nL of bacterial suspension ( c . a . 30 , 000 bacteria for OD600 = 1 ) into each fly abdomen using a Nanoject II ( Drummond ) ( Khalil et al . , 2015 ) . Flies were anesthetized with light CO2 for about five minutes during the injection procedure and were observed shortly after injection to confirm recovery from manipulations . We measured host survivorship post-injection in groups of 20 or 50 males kept with ad libitum access to food . Differences in survivorship were tested with Cox regression models ( R package Survival ) . To characterize the dynamics of within-host bacterial loads , after being dipped in Ethanol 70% and washed in PBS to limit contamination by external bacteria , individual flies were homogenized in 500 µl of sterile PBS with an HT homogenizer ( OPS Diagnostics ) at each timepoint post-injection . The homogenate was then diluted to 1:100 or 1:1000 in PBS to ensure that plate counts remained within the limits of resolution of the plating system . We plated 70 µl onto LB agar using a WASP II Autoplate spiral plater ( Microbiology International ) . Plates were incubated overnight at 37°C ( 29°C for E . carotovora ) and bacterial colonies were counted using an EZ-Count Automated Colony Counter ( Microbiology International ) to estimate the number of viable bacteria per fly . To estimate the Bacterial Load Upon Death ( BLUD ) , infected hosts were checked every 30 min and newly dead flies ( flies not moving and on their side or back ) were collected and homogenized , with bacterial load quantified as described above . Total RNA was extracted from either 20 flies or single flies with TRIzol reagent ( Invitrogen ) . Template RNA ( 1 µg ) was used to generate cDNA by reverse transcription using the SuperScript II cDNA synthesis kit ( Promega ) and then analyzed by quantitative polymerase chain reaction ( qPCR ) using the PerfeCTa SYBR Green SuperMix ( Quantabio ) . Expression values were normalized to RpL32 . Primer sequences are available in method supplementary 1 . At least three independent repeats were done . We used non-parametric tests to compare bacterial loads between groups of flies: Wilcoxon for two groups comparison and Kruskal-Wallis for more than two groups . The only exception was the comparison of loads between immune deficient mutants and wild-type controls , where , to take into account that we used more than one genotype to test the same biological hypothesis , we used a linear regression with genotype as a random effect . We used Spearman correlations to assess the relationship between bacterial loads and several quantitative traits ( initial inoculum , time to death , and Dpt expression ) . All data analysis was performed using R version 3 . 3 . 1 ( R Core Team , 2017 ) . Our results demonstrate that experimentally identical hosts exposed to the same pathogenic infections can either control the infection or succumb to acute bacterial proliferation . In addition , we see that bacterial growth in the absence of immune control has a similar rate to that of in vitro LB cultures ( see Results ) . We therefore built a model that consists of a mixture of two demographic models , one describing within-host bacterial growth and one modeling bacterial elimination ( see Figure 6A for a representation of the conceptual model this statistical model is based upon ) . More precisely , we considered that each bacterial load estimated on a single fly ( n ) was a log-normal random variable with the mean computed from a Baranyi model ( Baranyi and Roberts , 1994 ) when bacteria grow in the host uncontrolled by the fly's immune system withnt=nmaxeμt-1+eμtlageμt-1+eμtlagnmaxn0 or from an exponential decrease model , when bacteria are controlled by the host , withnt=1+nc-1e-δt With increasing time , t , n begins at n0 and approaches nmax in the Baranyi model , while it decreases from nc to asymptotically reach zero in the exponential decrease model . Actual bacterial load was assumed to follow a log-normal distribution , with average given by either of the two previous equations and variance fixed at σ2b for the Baranyi model and σ2d for the exponential decrease . To reflect the probabilistic nature of the infection process , we developed a model to compute the probability that bacteria are either controlled ( and thus have an exponential decrease ) or are not ( and thus have a Baranyi growth dynamic ) . We considered that the time required for the host immune defense to efficiently control bacterial infection follows a Gamma distribution with a fixed average tlag + t̄c and a variance Vc . t̄c is the average time host defenses take to control bacteria after they have started to grow ( Tc ) , and Vc quantifies inter-individual variation in Tc . We further assumed that the host cannot control infections after the bacterial load reaches a fixed threshold , which we termed the tipping point ( ntip ) . The eleven parameters of this model are all described in Table 1 . Under the model , the probability that a host survives infection is the probability that the control of bacterial growth occurs before the tipping point is reached . More precisely , if Pc is the probability of survival , ttip is the time at which the tipping point is reached , and Tc is a random variable corresponding to the time of control observed for a given host , we can writePc=Pr ( Tc < ttip ) The probability of survival Pc therefore depends on parameters that determine how fast bacteria grow initially ( n0 , nmax , tlag , µ ) , on the tipping point itself ( ntip ) , and on parameters , t̄c and Vc , that govern the distribution of time to control Tc . All these parameters can be estimated by adjusting the model on experimental data consisting of bacterial loads measured over time in different individuals . This was performed by computing the likelihood of the model for each bacterial load x observed at time t asLlog2x , t=Lblog2x , t1-pt+Ldlog2x , tpt where Lblog2x , t is the likelihood under the Baranyi model and Ldlog2x , t the likelihood under the exponential decrease model . The global log-likelihood of the model is then obtained by summing log-likelihoods computed for each point of the datasets . This was possible because each point of our data set consists of a single individual host , and all measurements at all sampling times are therefore somewhat independent from each other ( i . e . the measurements are not repeated on the same fly ) . We illustrated simulated outcomes of this model of within-host dynamics in Figure 6—figure supplement 1 . We estimated the 11 parameters of this mixture model by maximizing log-likelihood , using the Optim procedure in R . Confidence intervals for parameter estimates were then obtained by bootstrapping data for each experimental time point . This model , once adjusted , allows estimation of the probability that a host controls an infection by computing the probability that effective control occurs before ntip is reached . Probability of control therefore depends on the parameters of the Baranyi model ( n0 , nmax , tlag , µ ) and of those describing the distribution of time to control ( t̄c , Vc , ntip ) . | Sick individuals do not all respond to an infection in the same way . One individual may experience mild symptoms and recover easily , while another may suffer devastating illness or even death . A number of factors are often assumed to account for these differences , including the sex , age and genes of the individuals , and differences in the environments the individuals have been exposed to . However , random variations in how an individual’s immune system interacts with the infection could also play an important role in recovery . Duneau et al . have now studied how genetically identical fruit flies who were raised in the same environment respond to different bacterial infections . This enabled them to develop a mathematical model that describes how a bacterial infection develops in an individual . In an initial phase , the bacteria proliferate freely . If the immune defenses activate in time to control the infection , the number of bacteria in the fly decreases to a constant level and the infection enters a long-term , or chronic , phase . The sooner this happens the more likely it is that the fly will survive . If the immune control happens too late , the infection enters a terminal phase and the fly will die once the number of bacteria increases to a certain level . The model therefore reveals that the precise time at which the immune system takes control of the bacterial population – termed the “Time to Control” – determines the outcome of the infection . Duneau et al . confirmed this by injecting bacteria into identical flies . A small variation in the Time to Control was sometimes the difference between a fly living or dying . Understanding what controls this apparently random variation is key to understanding infection and potentially developing more efficient treatments for a wide range of diseases – not just those caused by bacteria , but also those caused by viruses and parasites , like HIV and malaria . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2017 | Stochastic variation in the initial phase of bacterial infection predicts the probability of survival in D. melanogaster |
CRISPR-based homing gene drives have sparked both enthusiasm and deep concerns due to their potential for genetically altering entire species . This raises the question about our ability to prevent the unintended spread of such drives from the laboratory into a natural population . Here , we experimentally demonstrate the suitability of synthetic target site drives as well as split drives as flexible safeguarding strategies for gene drive experiments by showing that their performance closely resembles that of standard homing drives in Drosophila melanogaster . Using our split drive system , we further find that maternal deposition of both Cas9 and gRNA is required to form resistance alleles in the early embryo and that maternally-deposited Cas9 alone can power germline drive conversion in individuals that lack a genomic source of Cas9 .
Homing gene drives have the potential to rapidly spread through a population by converting wild-type alleles to drive alleles in the germline of heterozygous individuals , thereby enabling super-Mendelian inheritance of the drive allele ( Esvelt et al . , 2014; Champer et al . , 2016; Burt , 2014; Unckless et al . , 2015; Alphey , 2014; Noble et al . , 2017; Deredec et al . , 2011 ) . Such systems could be a tool for the eradication of vector-borne diseases such as malaria or dengue by propagating transgenes through mosquito populations that prevent disease transmission ( Esvelt et al . , 2014; Champer et al . , 2016; Burt , 2014; Alphey , 2014 ) . Other proposed applications include the direct suppression of vector populations , invasive species , or agricultural pests ( Esvelt et al . , 2014; Champer et al . , 2016; Burt , 2014; Alphey , 2014 ) . Proof-of-principle studies using CRISPR-based homing drive constructs have now been demonstrated in a variety of potential target systems . The first experiments to achieve successful drive conversion were conducted in yeast ( DiCarlo et al . , 2015; Roggenkamp et al . , 2018; Basgall et al . , 2018; Shapiro et al . , 2018 ) , flies ( Champer et al . , 2018a; Oberhofer et al . , 2018; KaramiNejadRanjbar et al . , 2018; Gantz and Bier , 2015; Champer et al . , 2017 ) , and mosquitoes ( Hammond et al . , 2017; Hammond et al . , 2016; Gantz et al . , 2015 ) . These experiments revealed highly variable conversion efficiencies , ranging from close to 100% in Saccharomyces cerevisiae to between 19–62% in Drosophila melanogaster and 87–99% in Anopheles . Such variability could be due to several factors , including differences in the level and timing of Cas9 expression , the genomic targets , and organism-specific factors such as recombination rate . In Anopheles , for instance , conversion rates generally tend to be higher than in D . melanogaster , especially in males , consistent with the fact that there is no recombination in male flies . A recent study has further demonstrated successful drive conversion in mice ( Grunwald et al . , 2019 ) , although with a lower efficiency than most of the other systems . It has also become clear that homing gene drives face a significant obstacle due to the frequent formation of resistance alleles when cleavage is repaired by end-joining , which typically generates mutations at the target site ( Champer et al . , 2017 ) . This process has been observed to take place in the germline during failed drive conversion , but also in the embryo due to the persistence of maternally-deposited Cas9 ( Champer et al . , 2017 ) . Similar to conversion rates , resistance rates too are highly variable between drive systems and organisms ( DiCarlo et al . , 2015; Champer et al . , 2017; Gantz et al . , 2015; Grunwald et al . , 2019; Champer et al . , 2018b; Hammond et al . , 2018 ) . However , strategies for improving conversion efficiency and lowering resistance rates have already been successfully tested , including gRNA multiplexing ( Champer et al . , 2018b ) , improved promoters ( Champer et al . , 2018b; Hammond et al . , 2018 ) , and careful selection of target sites to render resistance alleles non-viable ( Kyrou et al . , 2018 ) . In fact , total population elimination with a CRISPR gene drive was recently achieved in laboratory cages of Anopheles gambiae for the first time ( Kyrou et al . , 2018 ) . While some have touted CRISPR homing drives as a potential game-changer in the fight against vector-borne diseases , key questions loom large about our ability to predict the outcome of releasing such a drive into a natural population . Unintended effects or even an accidental release could result in severe societal backlash . These concerns may seem hypothetical at present , given that most drives are still prone to rapid evolution of resistance ( KaramiNejadRanjbar et al . , 2018; Champer et al . , 2017; Unckless et al . , 2017 ) . Yet even an inefficient drive that reaches only a modest fraction of the population may spread resistance alleles to the entire population ( Noble et al . , 2018 ) . Furthermore , the first examples of effective drives are already on the horizon ( Kyrou et al . , 2018 ) , and even more powerful drives will likely be developed in the near future . Regardless of the likelihood of their escape from a lab or a field trial , it is imperative that we safeguard laboratory gene drives so that they cannot accidentally spread into a natural population . Current strategies typically rely on physical confinement of drive-containing organisms . However , it is doubtful whether this sufficiently reduces the likelihood of any accidental escape into the wild given the possibility of human error . Since very few escapees can establish an effective drive in a population ( Unckless et al . , 2015; Noble et al . , 2018; Marshall and Hay , 2012; Marshall , 2009 ) , additional safety measures should be employed in any experiments with drives potentially capable of spreading indefinitely . Two molecular safeguarding strategies have recently been proposed that go beyond physical or ecological confinement ( Akbari et al . , 2015 ) . The first is synthetic target site drive , which homes into engineered genomic sites that are absent in the wild . The second is split drive , where the drive construct lacks its own endonuclease , relying on one engineered into an unlinked site instead . Both strategies should thereby reliably prevent efficient drive outside of their respective laboratory lines . One potential drawback of these strategies is that each requires an additional transgenesis step compared to a standard drive . For a split drive , the line containing the Cas9 gene needs to be engineered , although one such line could be used for multiple split drive systems , and the transformation of the two individual elements may be easier since each is smaller than a standard drive . For the synthetic target site drive , the line containing the synthetic target needs to be separately engineered . However , such a system can also provide distinct advantages over standard drives in addition to confinement . For example , moving a target gene from a pest species into a model organism would permit researchers to test some aspects of the drive system in the model organism prior to release in the pest population . Additionally , the flexibility of synthetic target site drives allows targeting a dominant marker such as a fluorescent gene , facilitating the measurement of drive performance parameters while preventing the need to target a natural marker gene that may have significant fitness effects . Here , we provide the first experimental demonstration of synthetic target site drives and split drives in an insect system and show that their behavior closely resembles that of standard drives , with similar rates of drive conversion efficiency and resistance allele formation . This suggests that these strategies can serve as appropriate molecular safeguards in the development and testing of CRISPR homing gene drives .
We designed and tested three synthetic target site drives in D . melanogaster , each targeting an enhanced green fluorescent protein ( EGFP ) gene introduced at two autosomal sites and an X-linked site adjacent to the yellow gene ( Figure 1a ) . To determine conversion efficiencies of these drives ( the percentage of EGFP alleles converted to drive alleles in the germline ) , we scored dsRed phenotype in the progeny of crosses between EGFP homozygotes and drive/EGFP heterozygotes . We found drive conversion efficiencies of approximately 52–54% in females and 32–46% in males in these drive/EGFP heterozygotes ( Table 1 , Supplementary file 2-Datasets S1-S3 ) , which were similar to our previous homing drives targeting natural sites ( Champer et al . , 2017; Champer et al . , 2018b ) . We next measured the rate at which ‘r2’ resistance alleles ( those that disrupt the target gene ) were formed in the embryo by scoring the progeny of female heterozygotes for EGFP phenotype . This rate was high in all three drives , ranging from 80 to 91% ( Table 1 , Supplementary file 2-Datasets S1-S3 ) . It is thus likely that nearly all EGFP target alleles were converted to resistance alleles . These rates are again similar to what we found for previous drive constructs targeting the autosomal cinnabar and X-linked white loci ( Champer et al . , 2018b ) , but significantly less than a construct targeting the X-linked yellow ( Champer et al . , 2017 ) gene ( p < 0 . 001 , Fisher’s exact test ) . This difference is likely due to location-specific variation in expression levels of Cas9 ( and possibly also the gRNA ) between constructs inserted into the yellow gene and other sites . For our split drive system , we designed a drive construct targeting the X-linked yellow gene , similar to the one used in a previous study ( Champer et al . , 2017 ) , but lacking Cas9 ( Figure 1b ) . We then designed a second construct containing Cas9 driven by a nanos promoter for germline-restricted expression , which was inserted into chromosome 2R . We assessed drive performance of this system by first crossing males that had the drive element but no Cas9 to females that were homozygous for Cas9 but lacked the drive element . Similarly , we also crossed females homozygous for the drive but lacking Cas9 to males homozygous for Cas9 but lacking the drive . The progeny of these crosses followed Mendelian inheritance rules , indicating that both Cas9 and gRNA must be maternally deposited for resistance alleles to form in the early embryo . The progeny of w1118 males and drive/wild-type heterozygous females containing one copy of Cas9 were then scored for dsRed and yellow phenotype to assess drive conversion efficiency and resistance rates ( Table 1 , Supplementary file 2-Datasets S4 ) . Compared to our previous results with a standard gene drive targeting yellow ( Champer et al . , 2017 ) , where drive conversion efficiency was 62% ( and which shared the same genomic location , gRNA , dsRed marker , and nanos-Cas9 element , albeit at a different location ) , we measured a significantly higher drive conversion efficiency of 74% for the split drive using the same experimental parameters ( p<0 . 0001 , Fisher’s exact test ) . This improvement may be due to increased efficiency of homology-directed repair for the split drive element compared to the larger standard drive . However , we also observed that early embryo r2 resistance allele formation was much higher in the split drive at 74% compared to the 20% for the standard drive ( p<0 . 001 , Fisher’s exact test ) . This is likely because Cas9 , rather than the gRNA , is the main limiting factor in determining the cleavage rate and that Cas9 at its new site had higher expression than the Cas9 in the standard drive at yellow ( it was located only 277 nucleotides away from synthetic target site B , and the drive at this site had 88% embryo resistance ) . One concern regarding the use of split drives as a surrogate for standard drives is that every genome in the experimental split drive population would contain Cas9 , so maternally-deposited Cas9 would likely be present in each embryo , even if the mother did not have a drive element . In combination with the zygotically expressed gRNA from a paternal allele , this might then result in a higher rate of embryo resistance allele formation . However , our finding that both Cas9 and gRNA must be maternally deposited to form such embryo resistance alleles suggests that a split drive in a laboratory population should behave similarly to a standard drive . A hypothetical split drive where Cas9 is encoded in the driving element and the gRNA forms the supporting element ( the reverse of our split drive ) would presumably have nearly identical behavior to a standard drive . This is because in such a drive the Cas9 gene would always be present in the same copy number per individual as in a standard drive , and it would be located at the same genomic position , eliminating the possibility of position-based differences in Cas9 expression levels between the two drives . It would also be much closer to the standard drive in total size , minimizing potential differences in the efficiency of homology-directed repair . However , such a strategy would be experimentally less flexible because both elements would have to be redesigned for every new target site , rather than just the drive element when Cas9 is in the supporting element . Nevertheless , such a strategy may be advantageous for testing standard homing drives , integral homing drives ( Nash et al . , 2019 ) with gRNAs driving in a separate synthetic target site , and other future types of CRISPR-based gene drives . The flexibility of the split drive system , facilitated by its genomic separation of Cas9 and gRNA , allowed us to further refine our understanding of the general mechanisms by which homing drives operate . Previous studies have indicated that germline resistance alleles can form in pre-gonial germline cells ( KaramiNejadRanjbar et al . , 2018; Champer et al . , 2018b ) , but it remained unclear whether this could also occur at other stages . The fact that we observed a higher drive conversion efficiency of the split drive compared with a standard drive strongly implies that not all resistance alleles form prior to drive conversion , since resistance allele formation alone should not be affected by the reduced size of the split drive . This raises the possibility that drive conversion could potentially take place as an alternative to resistance allele formation in pre-gonial germline cells , where resistance alleles are known to form . However , a perhaps more likely explanation would be that only a portion of resistance alleles form in pre-gonial germline cells and the remainder form either in gametocytes as an alternative to drive conversion or afterward in late meiosis , when a template for homology-directed repair is no longer available . We further found evidence that even in individuals lacking a genomic source of Cas9 , maternally-deposited Cas9 can persist through to gametocytes in the germline , where it can then facilitate successful drive conversion . This was demonstrated by crosses of w1118 males with females that were mosaic yellow and had inherited a drive allele with dsRed from their mother but not the Cas9 allele itself , as evidenced by the absence of a EGFP phenotype . These females still received maternally-deposited Cas9 from their heterozygous mothers . Despite lacking a Cas9 gene , fifteen out of sixteen of these flies showed an average drive conversion efficiency of 54% , with a single fly showing no drive conversion ( Supplementary file 2-Datasets S4 ) . By contrast , nine out of eleven females from the same cross that were fully yellow , rather than mosaic , and two wild-type females showed no detectable drive conversion , while two fully yellow females showed successful drive conversion . In the wild-type females , ‘r1’ resistance alleles that preserve the function of the target gene had presumably formed at the embryo stage . It is unlikely that any drive conversion had occurred in these females with full yellow phenotype at the early embryo stage , because in that case their progeny should have consistently displayed biased inheritance of the drive allele . Thus , it appears that when Cas9 and gRNA are maternally-deposited , they can fail to cleave the target site in the early embryo and induce end-joining repair , while nonetheless showing significant cleavage activity in later stages when homology-directed repair is possible . This creates yellow phenotype over most of their body while also usually enabling drive conversion in the germline . Such Cas9 did not persist to embryos of the subsequent generation , as indicated by the lack of yellow phenotype in female progeny . We also found that maternally-deposited gRNA was not necessary to achieve drive conversion in conjunction with maternally-deposited Cas9 when a genomic source of gRNA is provided . For example , we found that the progeny of drive-heterozygous females receiving a paternal drive allele without genomic Cas9 ( but receiving maternally-deposited Cas9 from a mother with a single copy of Cas9 ) showed 38% germline drive conversion efficiency ( Supplementary file 2-Datasets S4 ) . An additional 12% of wild-type alleles were converted to r2 resistance alleles , and the remainder most likely remained wild-type . These rates were lower than those of the full split drive , likely because of reduced Cas9 and gRNA activity due to the fact that only maternally-deposited Cas9 could be utilized . Taken together , our results suggest that in the absence of early embryo resistance alleles , germline drive rates in an individual may be affected by the level of maternally-deposited Cas9 and gRNAs . Individuals inheriting a drive allele from their mother , as opposed to their father , will also receive maternally-deposited Cas9 , which could increase the level of cleavage during the window for homology-directed repair , thereby increasing the rate of drive conversion . On the other hand , cleavage by the maternally-deposited Cas9 prior to this stage in pre-gonial germline cells , in addition to the early embryo , should usually form additional resistance alleles compared to individuals that have no such persistent Cas9 . Our results demonstrate that CRISPR homing gene drives with synthetic target sites such as EGFP will show highly similar behavior to standard drives and can thus be used for most testing in lieu of these drives . Split drives also show similar performance , while allowing for the targeting of natural sequences in situations where the use of synthetic targets is difficult , such as for certain resistance reduction strategies and population suppression drives that require the targeting of wild-type genes . We therefore suggest that gene drive research should consistently adopt these molecular safeguarding strategies in the development and testing of new drives . This will be particularly important for large-scale cage experiments aimed at gaining a better understanding of the expected population dynamics of candidate drives , which will be integral for any informed discussion about their feasibility and risks .
We constructed three synthetic target site drives at different genomic sites ( B , E , and Y ) into which the EGFP target was inserted . Sites B and E were on chromosomes 2R and 3R , respectively , located 3’ of two protein- coding genes . Site Y was on the X chromosome immediately downstream of yellow . All of our synthetic target site drives used a slightly recoded 3xP3 promoter ( 3xP3v2 ) to drive the dsRed marker and also used a P10 3’UTR . This was to reduce potential misalignment with the 3xP3 promoter and SV40 3’UTR in the homology arms ( see Figure 1a ) , which we found to result in poor drive efficiency in initial tests of drive constructs that used the same 3xP3 promoter and SV40 3’UTR for the EGFP and dsRed markers . Since all our synthetic target site drives home into EGFP , successful insertion of the drives will disrupt this marker ( Figure 1a ) . For the split drive , the driving element disrupts yellow , causing a recessive yellow body phenotype ( Figure 1b ) . If cleavage is repaired by end-joining , rather than homology-directed repair , this will typically result in a mutated target site , creating a resistance allele . Most such resistance alleles will render the target gene nonfunctional due to a frameshift or otherwise sufficient change in the amino acid sequence . We term these alleles ‘r2’ . Resistance alleles that preserve the function of the target gene are termed ‘r1’ . In some cases , we observed mosaicism for EGFP in the eyes of heterozygotes for the drive and the synthetic target site . This indicates that the germline nanos promoter may still drive a low level of expression in somatic cells . However , since no mosaicism was observed in the body of the split drive flies , this mosaicism may be due to proximity of the nanos promoter to the nearby 3xP3 promoter that drives expression in the eyes ( the promoters were only eight nucleotides apart in the synthetic target site flies but 68 nucleotides apart in the split Cas9 construct ) . The different phenotypes and genotypes of our drive systems are summarized in Supplementary file 2-Datasets S1-S3 , as are calculations for determining drive performance parameters based on phenotype counts . One line in the study was transformed at GenetiVision by injecting the donor plasmid ( ATSabG ) into a w1118 D . melanogaster line , and seven lines were transformed at Rainbow Transgenic Flies by injecting the donor plasmid ( ATSaeG , ATSxyG , BHDgN1bv2 , BHDgN1e , BHDgN1y , BHDaaN , IHDyi2 ) into the same w1118 line . Cas9 from plasmid pHsp70-Cas9 ( Gratz et al . , 2013 ) ( provided by Melissa Harrison and Kate O'Connor-Giles and Jill Wildonger , Addgene plasmid #45945 ) and gRNA from plasmids BHDaag1 , BHDabg1 , BHDaeg1 , or BHDxyg1 were included in the injection , depending on the target site . For Genetivision injections , concentrations of donor , Cas9 , and gRNA plasmids were 102 , 88 , and 60 ng/µL , respectively in 10 mM Tris-HCl , 23 µM EDTA , pH 8 . 1 solution . For Rainbow Transgenic Flies injections , concentrations of donor , Cas9 , and gRNA plasmids were approximately 350–500 , 250–500 , and 50–100 ng/µL , respectively in 10 mM Tris-HCl , 100 µM EDTA , pH 8 . 5 solution . Note that the synthetic target site drives were transformed into the w1118 line in parallel with the synthetic targets themselves , including elements of the target on either side of the drive . This avoids the need to transform the drive into lines already possessing the synthetic target site . To obtain homozygous lines , the injected embryos were reared and crossed with w1118 flies . The progeny with dsRed or EGFP fluorescent protein in the eyes , which usually indicated successful insertion of the donor plasmid , were selected and crossed with each other for several generations . The stock was considered homozygous at the drive locus after sequencing confirmed lack of wild-type or resistance alleles . All flies were reared at 25 ˚C with a 14/10 hr day/night cycle . Bloomington Standard medium was provided as food every 2–3 weeks . During phenotyping , flies were anesthetized with CO2 and examined with a stereo dissecting microscope . Flies were considered ‘mosaic’ if any discernible mixture of green fluorescence was observed in either eye . However , for the synthetic target site drives , flies that carried a drive allele were only considered mosaic if either eye had less than 50% EGFP phenotype coverage , to avoid identifying flies with possible somatic expression of Cas9 as mosaic for EGFP . This definition was stringent enough that no mosaic insects without the drive were found that would have avoided mosaic classification based on this definition . Fluorescent phenotypes were scored using the NIGHTSEA system only in the eyes ( SFA-GR for dsRed and SFA-RB-GO for EGFP ) . Even though dsRed did bleed through into the EGFP channel , both types of fluorescence could still be easily distinguished . All experiments involving live gene drive flies were carried out using Arthropod Containment Level two protocols at the Sarkaria Arthropod Research Laboratory at Cornell University , a quarantine facility constructed to comply with containment standards developed by USDA APHIS . Additional safety protocols regarding insect handling approved by the Institutional Biosafety Committee at Cornell University were strictly obeyed throughout the study , further minimizing the risk of accidental release of transgenic flies . To obtain the DNA sequences of gRNA target sites , individual flies were first frozen and then ground in 30 µL of 10 mM Tris-HCl pH 8 , 1 mM EDTA , 25 mM NaCl , and 200 µg/mL recombinant proteinase K ( Thermo Scientific ) . The homogenized mixture was incubated at 37 ˚C for 30 min and then 95 ˚C for 5 min . 1 µL of the supernatant was used as the template for PCR to amplify the gRNA target site . DNA was further purified by gel extraction and Sanger sequenced . Sequences were analyzed using the ApE software , available at: http://biologylabs . utah . edu/jorgensen/wayned/ape . The starting plasmid pCFD3-dU6:3gRNA ( Port et al . , 2014 ) ( Addgene plasmid #49410 ) was kindly supplied by Simon Bullock , starting plasmid pJFRC81-10XUAS-IVS-Syn21-GFP-p10 ( Pfeiffer et al . , 2012 ) was a gift from Gerald Rubin ( Addgene plasmid # 36432 ) , and starting plasmid IHDyi2 was constructed in our previous study ( Champer et al . , 2017 ) . All plasmids were digested with restriction enzymes from New England Biolabs ( HF versions , when available ) . PCR was conducted with Q5 Hot Start DNA Polymerase ( New England Biolabs ) using DNA oligos and gBlocks from Integrated DNA Technologies . Gibson assembly of plasmids was conducted with Assembly Master Mix ( New England Biolabs ) and plasmids were transformed into JM109 competent cells ( Zymo Research ) . Plasmids used for injection into eggs were purified with ZymoPure Midiprep kit ( Zymo Research ) . Cas9 gRNA target sequences were identified using CRISPR Optimal Target Finder ( Gratz et al . , 2014 ) . Tables of the DNA fragments used for Gibson Assembly of each plasmid , the PCR products with the oligonucleotide primer pair used , and plasmid digests with the restriction enzymes are shown in the Supporting Information . | Gene drives are a new genome editing technology where artificial gene packages are designed to create a mutation that will quickly spread within a population . These packages target a specific sequence in a genome , where they could potentially add , remove or deactivate a gene . They also trigger a process known as drive conversion , which ensures the mutation will be inherited at a higher rate than normal . Within several generations , nearly every organism in the population will carry this genetic change . This technology could , for example , help us eradicate disease-carrying mosquitoes , crop pests or invasive species . However , it could also have unforeseen and dangerous consequences . It is therefore crucial to keep gene drives within laboratory walls before they are ready to be released . Even if a small numbers of genetically modified animals were to escape , they could rapidly spread the packages within a wild population . To prevent this , scientists have devised two safeguarding strategies . One , called synthetic target site gene drive , uses target sequences that have been introduced on purpose in research organisms , but which are absent in wild populations . If the gene drive were to escape , it could not spread in the genomes of wild creatures because they lack the synthetic target site . Alternatively , split drive systems can also limit risk . There , the different components required for a gene drive are not packaged together , but in separate locations in the genome . Some of these elements are inherited at a normal rate , so the gene drive fizzles out after a few generations . However , it was still unclear whether synthetic gene drives and split drive systems could be used instead of the classic approach and yield the same results in research . Champer et al . compared traditional gene drives , synthetic target site gene drives , and split drive systems in fruit flies raised in the laboratory . The experiments show that the three approaches lead to similar results , with the genetic package spreading and creating resistance in a similar way . They also confirm that , in split drive systems , both components of the drive must be genetically inherited to create the intended mutation . Synthetic gene drives and split drive systems could therefore be used in experiments on gene drives , especially in studies with large numbers of organisms . Ultimately , adopting these measures could help to keep gene drive research safe , which may encourage more scientific teams to work on this technology and exploit its potential . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"genetics",
"and",
"genomics"
] | 2019 | Molecular safeguarding of CRISPR gene drive experiments |
Motoneurons ( MNs ) control muscle contractions , and their recruitment by premotor circuits is tuned to produce accurate motor behaviours . To understand how these circuits coordinate movement across and between joints , it is necessary to understand whether spinal neurons pre-synaptic to motor pools have divergent projections to more than one MN population . Here , we used modified rabies virus tracing in mice to investigate premotor interneurons projecting to synergist flexor or extensor MNs , as well as those projecting to antagonist pairs of muscles controlling the ankle joint . We show that similar proportions of premotor neurons diverge to synergist and antagonist motor pools . Divergent premotor neurons were seen throughout the spinal cord , with decreasing numbers but increasing proportion with distance from the hindlimb enlargement . In the cervical cord , divergent long descending propriospinal neurons were found in contralateral lamina VIII , had large somata , were neither glycinergic , nor cholinergic , and projected to both lumbar and cervical MNs . We conclude that distributed spinal premotor neurons coordinate activity across multiple motor pools and that there are spinal neurons mediating co-contraction of antagonist muscles .
The spinal cord is ultimately responsible for organizing movement by controlling the activation pattern of motoneurons ( MNs ) , which in turn produce appropriate patterns of muscle contractions to produce limb movement . Across any single limb joint , there are fundamentally three types of control – or three ‘syllables of movement’ – possible . The three basic syllables are: ( 1 ) changing a joint angle , ( 2 ) stiffening a joint , and ( 3 ) relaxing a joint . The concatenation of these syllables across joints within and between limbs ultimately produces behaviour ( Brownstone , 2020; Wiltschko et al . , 2015 ) . To change a joint angle , MNs innervating synergist muscle fibres are activated whilst those that innervate antagonist muscle fibres are inhibited . This ‘reciprocal inhibition’ ( Eccles , 1969; Eccles et al . , 1956 ) is mediated locally by spinal interneurons ( INs ) throughout the spinal cord; this syllable has been fairly well characterized , with responsible neurons identified and classified ( Alvarez et al . , 2005; Benito-Gonzalez and Alvarez , 2012; Sapir et al . , 2004; Zhang et al . , 2014 ) . The other two syllables are less well studied , but it is clear that behavioural joint stiffening requires co-activation of MNs innervating antagonist muscle groups , while joint relaxation would require co-inhibition of these MNs . Co-contraction has largely been thought to result from brain activity ( Humphrey and Reed , 1983 ) , whereas circuits mediating co-inhibition remain elusive . Since the spinal cord controls movement not only across single joints but throughout the body , it is natural to consider whether it contains the circuits necessary to produce these different syllables . To identify whether these syllables are produced by spinal circuits , several questions can be asked: Does the spinal cord contain circuits that lead to co-activation or co-inhibition of different pools of MNs – either synergists or antagonists ? Does each motor pool have its own dedicated population of premotor INs , and are these INs interconnected in such a way that they can produce contraction of different muscle groups ? Or are there populations of INs that project to multiple motor pools in order to effect contraction ( or relaxation ) of multiple muscles ? Indeed , INs that have activity in keeping with innervation of multiple synergists , leading to motor ‘primitives’ or synergies ( Bizzi and Cheung , 2013; Giszter , 2015; Hart and Giszter , 2010; Takei et al . , 2017; Tresch and Jarc , 2009 ) have been identified , but knowledge of their locations and identities remains scant . Normal behaviours in quadrupeds as well as bipeds require coordination of syllables across joints between forelimbs and hindlimbs . This coordination relies on populations of propriospinal neurons projecting in either direction between the lumbar and cervical enlargements ( Eidelberg et al . , 1980; Giovanelli Barilari and Kuypers , 1969; Miller and van der Meché , 1976; Ruder et al . , 2016 ) . Long descending propriospinal neurons ( LDPNs ) were first proposed in cats and dogs more than a century ago ( Sherrington and Laslett , 1903 ) , and their existence has been confirmed in several other species including humans ( Alstermark et al . , 1987a; Alstermark et al . , 1987b; Ballion et al . , 2001; Brockett et al . , 2013; Flynn et al . , 2017; Giovanelli Barilari and Kuypers , 1969; Jankowska et al . , 1974; Mitchell et al . , 2016; Nathan et al . , 1996; Ni et al . , 2014; Reed et al . , 2009; Ruder et al . , 2016; Skinner et al . , 1979 ) . While LDPNs that establish disynaptic connections to lumbar MNs have been identified , it was initially suggested that at least some cervical LDPNs could establish monosynaptic inputs to lumbar MNs ( Jankowska et al . , 1974 ) . This connectivity was later confirmed using monosynaptic modified rabies virus ( RabV ) tracing ( Ni et al . , 2014 ) . More recently , descending and ascending spinal neurons and their involvement in the control of stability and interlimb coordination have been characterized , but these studies did not directly focus on monosynaptic premotor circuits ( Pocratsky et al . , 2017; Ruder et al . , 2016 ) . It is likely that LDPNs function to ensure coordination between forelimbs and hindlimbs , and they could be an important source of premotor input to MNs , providing a substrate for coordination between distant joints . In the present study , we examine circuits underlying co-activation and co-inhibition in the spinal cord by assessing premotor neurons through the use of RabV tracing techniques ( Ronzano et al . , 2021; Ugolini , 1995; Wickersham et al . , 2007 ) . We used glycoprotein ( G ) -deleted RabV ( ΔG-Rab ) , and supplied G to MNs through crossing ChAT-Cre mice with RΦGT mice ( Ronzano et al . , 2021; Takatoh et al . , 2013 ) . We injected ΔG-RabV tagged with two different fluorescent proteins into hindlimb muscle pairs of ChAT-Cre mice to retrogradely trace premotor circuits throughout the spinal cord . At the lumbar level , this method revealed apparent low rates of INs projecting to both MN pools targeted . As the distance from targeted MN pool to premotor INs increased , the density of infected premotor INs decreased . But the apparent rate of divergence to multiple pools was higher in thoracic and cervical regions than in the lumbar spinal cord . Interestingly , the extent of divergence throughout the spinal cord was similar whether injections were performed in flexor or extensor pairs , or in synergist or antagonist pairs of muscles . In addition , a population of premotor LDPNs was identified in the cervical spinal cord . These neurons had a high rate of divergence and large somata , projected contralaterally , were neither glycinergic nor cholinergic , located in lamina VIII , and projected to cervical MNs as well as lumbar MNs . Together , these data show that the spinal cord contains premotor INs that project to multiple motor pools ( including antagonists ) , and could thus form substrates for the fundamental syllables of movement .
Given evidence that INs are involved in motor synergies ( Hart and Giszter , 2010; Levine et al . , 2014; Takei et al . , 2017; Takei and Seki , 2010 ) , we would expect that there would be INs in the lumbar spinal cord that project to synergist motor pools . We thus first investigated whether such premotor INs could be infected with two RabVs expressing two different fluorescent proteins injected in pairs of muscles . We injected ΔG-Rab expressing eGFP or mCherry into synergist ankle extensors ( lateral gastrocnemius [LG] and medial gastrocnemius [MG] ) or synergist ankle flexors ( tibialis anterior [TA] and peroneus longus [PL] ) in ChAT-Cre;RΦGT P1-P3 mice ( Ronzano et al . , 2021 ) . These mice selectively express rabies G in cholinergic neurons ( including MNs ) , providing the necessary glycoprotein for retrograde trans-synaptic transfer from infected MNs to premotor INs ( Figure 1A ) . After 9 days , we visualized the distribution of premotor INs that expressed one or both fluorescent proteins , specifying the premotor INs that make synaptic contact with two motor pools as ‘divergent’ premotor INs ( Figure 1—figure supplement 1A , B , Figure 1—figure supplement 2A , B ) . We found divergent premotor INs distributed across the lumbar spinal cord ( Figure 1—figure supplement 4A , B ) bilaterally in the ventral quadrants and ipsilaterally in the dorsal quadrant of the spinal cord ( Figure 1C , D ) , consistently across experiments ( Figure 1—figure supplement 5A , Supplementary file 1 ) . Across the lumbar spinal cord , we quantified infected MNs and found that 380 MNs were labelled from synergist injections ( n = 4 , two extensor and two flexor pairs ) . Notably , five MNs were double labelled , most likely due to secondary infection of synaptically connected MNs ( Supplementary file 1 , Bhumbra and Beato , 2018 ) . We then quantified premotor INs on one of every three sections and found that 4 . 0% ± 0 . 3% ( 276/7043 , n = 4 , two extensor and two flexor pairs , Figure 1—figure supplement 1C , D , and Figure 1—figure supplement 2C , D ) of labelled premotor INs were double labelled , confirming that INs can be infected by more than one RabV . We would expect this to be an underestimate of the number of INs that have divergent projections since RabV is not expected to label 100 % of presynaptic neurons and as there is a reduced efficiency of double infections compared to single infections ( Ohara et al . , 2009; see Discussion ) . We next sought to determine whether this divergence was restricted to synergist motor pools or whether there are also premotor INs that diverge to antagonist pools and could thus be involved in co-contraction or joint stiffening . Following injections into flexor ( TA ) and extensor ( LG ) muscles , 260 MNs were labelled ( n = 3 antagonist pairs ) , one of them being double labelled ( Supplementary file 1 ) . Following these injections , we also found divergent INs ( Figure 1B , Figure 1—figure supplement 3A , B ) . We found a similar rate of divergence to antagonist pools as to synergist muscles , with 4 . 7% ± 0 . 5 % ( 206/4341 , n = 3 antagonist pairs , Figure 1E , Figure 1—figure supplement 3C–E ) double labelled . The mapping of all divergent INs in every section revealed that , whether injections were in synergist ( n = 4 , two extensor and two flexor synergist pairs ) or antagonist ( n = 3 pairs ) pairs of muscles , double-labelled premotor INs were distributed similarly ( Figure 1F , G , Figure 1—figure supplement 4 , Figure 1—figure supplement 5A , and Supplementary file 1 for summary of individual experiments ) . The proportion of divergent cells was calculated from the ratio of double and single infected cells in 1/3 sections , in order to avoid double counting cells present in consecutive sections ( see Methods ) . Equal proportions of divergent premotor INs were found in the ventral ipsilateral quadrant synergists: 74/1913 ( 3 . 9% ) vs antagonists: ( 46/1046 [4 . 4%] ) , ventral contralateral quadrant ( 46/1020 [3 . 8%] vs 21/502 [4 . 2%] ) , and dorsal ipsilateral quadrant ( 153/3874 [3 . 9%] vs 134/2651 [5 . 1%] ) . There were few labelled neurons in the dorsal contralateral quadrant following either synergist or antagonist injections and a similarly low proportion were double labelled ( in 1/3 sections: 3/236 [1 . 3%] and 5/142 [3 . 5%] , respectively ) . Divergence in premotor circuits is thus common , with at least 1/25 ( see Discussion ) premotor INs diverging to two MN pools , whether synergists or antagonists . Since motor synergies can span across more than a single joint , it is possible that divergent premotor INs could project to motor pools other than those injected . Indeed , following injection of ΔG-Rab-mCherry into the TA muscle , we could visualize mCherry-positive excitatory ( vGluT2+ ) boutons in apposition to L1 ( Figure 1—figure supplement 6A , D–E ) , as well as to thoracic ( as rostral as at least T10 ) MNs ( Figure 1—figure supplement 6B , C ) , that is , three to seven segments rostral to the infected motor pool . mCherry-positive excitatory boutons on MNs were consistently observed in all upper lumbar and thoracic sections taken from three injected mice ( three to four sections in each region ) . This observation , in agreement with a previous study that described premotor INs coordinating the activity of multiple lumbar motor groups from L2 to L5 ( Levine et al . , 2014 ) , supports the possibility that thoraco-lumbar premotor circuits comprise a substrate for multi-joint synergies . In order to maintain posture and stability , trunk muscles are coordinated with hindlimb movements . Neurons in the thoracic cord that are premotor to lumbar MNs have previously been described ( Ni et al . , 2014 ) ; we thus next examined the projections of thoracic premotor neurons to lumbar motor pools . These premotor neurons were found with decreasing density from T11 through T3 whether the injections were in extensor ( LG and MG; Figure 2—figure supplement 1A , B ) or flexor ( TA and PL , Figure 2A , Figure 2—figure supplement 2A-B ) pairs of muscles ( Figure 2—figure supplement 4A , B ) . The distributions of single labelled as well as divergent premotor neurons were similar whether injections were performed in flexor , extensor , or antagonist pairs of muscles ( Figure 2B–H , Figure 2—figure supplement 4A–C , Figure 1—figure supplement 5B ) . Divergence rates calculated from the whole thoracic spinal cords were similar between synergist and antagonist injections with 16 . 2% ± 5 . 7% ( 77/497 , n = 4 , 2 extensor and two flexor pairs , Figure 2B , C , Figure 2—figure supplement 1C , D , Figure 2—figure supplement 2C , D ) and 9 . 0% ± 0 . 7% ( 59/401 , n = 3 antagonist pairs , Figure 2D , Figure 2—figure supplement 3C , E ) , respectively . In all animals ( 7/7 ) , the overall proportion of double-labelled neurons in the thoracic spinal cord was higher than in the lumbar cord ( 13 . 1% ± 5 . 6% , Figure 7B ) . In all animals ( 7/7 ) , most divergent premotor neurons in the thoracic cord were located in the ipsilateral dorsal quadrant ( 46/77 , n = 4 synergist and 42/59 , n = 3 antagonist pairs , Figure 2E–H ) , and within this quadrant 22 . 1% ± 8 . 6% ( 46/188 synergists and 42/211 antagonists , Figure 2E , F ) of premotor neurons were double labelled . The divergence rates in the two ventral quadrants were lower: in the ventral cord , double-labelled neurons were observed in 5/7 animals ( 3/4 synergist; 2/3 antagonist in both quadrants ) ipsilaterally ( 6 . 7% ± 4 . 8%; 10/118 synergist and 5/70 antagonist pairs ) , as well as contralaterally ( 11 . 1% ± 10 . 7%; 21/167 synergist and 12/104 antagonist pairs ) to the injection ( Figure 2E , F ) . Thus , there are premotor neurons throughout the thoracic cord that project directly to more than one motor pool , including antagonist pairs , in the lumbar spinal cord , with most of these located in the ipsilateral dorsal quadrant . Cervical long descending propriospinal neurons ( LDPNs ) have been shown to modulate interlimb coordination to provide stability ( Eidelberg et al . , 1980; Miller and van der Meché , 1976; Pocratsky et al . , 2017; Ruder et al . , 2016 ) . Given that cervical premotor LDPNs projecting to TA MNs have previously been demonstrated ( Ni et al . , 2014 ) , we asked whether these neurons could be premotor to hindlimb and/or hindlimb–forelimbs MN pairs . We found that premotor LDPNs projecting to flexor ( TA and PL ) and extensor ( LG and MG ) MNs were localized throughout the rostro-caudal extent of the ventral cervical cord with an enrichment between C6 and T1 ( Figure 3—figure supplement 4 ) . Of 92 premotor LDPNs , 88 were localized in the ventral quadrants , 68 of which were in contralateral lamina VIII ( n = 7 , four synergist and three antagonist pairs , Figure 3A–F , Figure 3—figure supplement 4 ) . A substantial proportion of premotor LDPNs was double labelled , with the proportion and location of double labelling similar across experiments ( Figure 1—figure supplement 5C and Supplementary file 1 ) whether injections were into synergist or antagonist pairs ( 42 . 4% ± 22 . 1 % per animal , total of 19/55 neurons , n = 4 synergist pairs and 47 . 9% ± 7 . 1 % per animal , total of 19/37 neurons , n = 3 antagonist pairs , Figure 3E , F , Figure 3—figure supplement 1 , Figure 3—figure supplement 2 , Figure 3—figure supplement 3 ) . This apparent divergence rate of LDPNs in the cervical cord was higher than in the lumbar and thoracic cords in all animals ( 7/7 , Figure 7B ) . These divergent premotor LDPNs exhibited a stereotypical morphology with an unusually large soma ( 774 ± 231 µm2 , n = 38 premotor LDPNs ) compared to the double-labelled premotor neurons in the thoracic and lumbar cords ( respectively , 359 ± 144 and 320 ± 114 µm2 , n = 135 premotor neurons [thoracic] , n = 61 premotor INs [lumbar] , p < 0 . 0001 Kruskal–Wallis test , p < 0 . 0001 [lumbar vs cervical] , and p < 0 . 0001 [thoracic vs cervical] , Dunn’s multiple comparisons test ) . On average , the cross-sectional area of divergent cervical LDPNs was comparable to that of cervical MNs ( 661 ± 86 µm2 , n = 17 MNs , Figure 3H ) . Their location and size suggest that these divergent , commissural cervical premotor LDPNs may constitute a somewhat homogenous population . To determine the neurotransmitter phenotype of the premotor LDPNs , we used single ΔG-Rab-mCherry injections in ChAT-Cre;RΦGT mice crossed with mice expressing eGFP under the control of the promoter for the neuronal glycine transporter GlyT2 ( Figure 4A , Zeilhofer et al . , 2005 ) . GlyT2 is expressed in the vast majority of spinal inhibitory INs ( Todd et al . , 1996; Todd and Sullivan , 1990 ) , making GlyT2-eGFP mice a suitable tool to determine whether premotor LDPNs are inhibitory . Given that at least 40 % of the labelled INs in the cervical region are divergent ( see above ) , many of the neurons labelled following even single RabV injections would be expected to be divergent . Following injection into LG ( Figure 4A ) , we found that only 1/21 infected cervical commissural premotor LDPNs was eGFP positive ( n = 3 LG injections , Figure 4B , C and F ) . Since none of the labelled neurons expressed ChAT , the majority of cervical premotor LDPNs are likely to be glutamatergic by exclusion . However , in agreement with previous results from TA injections ( Ni et al . , 2014 ) , single-labelled thoracic premotor neurons comprised a mixed population of inhibitory and non-inhibitory neurons ( 34 . 4% ± 5 . 9% , 96/273 , mCherry+ eGFP + premotor neurons , n = 3 LG injections , Figure 4D–F ) . We cannot determine whether the thoracic or lumbar GFP+ or GFP− premotor INs are divergent , as these data were obtained following single injections . However , in the lumbar cord , as expected , we observed that some divergent INs were cholinergic ( Figure 4—figure supplement 1 ) . We next sought to determine the genetic provenance of divergent cervical LDPNs . Among the classes of ventral INs defined by the early expression of transcription factors ( Lee and Pfaff , 2001 ) , the V0 and V3 cardinal classes are known to project to contralateral MNs . These classes can be further subdivided , with all V3 subclasses being glutamatergic ( Zhang et al . , 2008 ) , and V0 INs being neuromodulatory V0C , cholinergic ( Miles et al . , 2007 ) , inhibitory V0D , dorsal ( Talpalar et al . , 2013 ) , or excitatory V0V , ventral ( Talpalar et al . , 2013 ) , or V0G , medial glutamatergic neurons that project to dorsal and intermediate lamina but not to MNs ( Zagoraiou et al . , 2009 ) . Since previous studies showed that none of the LDPNs with soma in the cervical cord belong to the V3 population ( Flynn et al . , 2017 ) , we sought to determine whether these LDPNs were of the V0 class . V0 INs are defined by their embryonic expression of the transcription factor Dbx1 ( Pierani et al . , 2001 ) and Evx1 ( Moran-Rivard et al . , 2001 ) . However , neither of these two transcription factors can reliably be detected at the postnatal ages of our mice . On the other hand , Lhx1 is expressed throughout the V0 and V1 populations ( as well as dI2 , dI4 , and dILA populations ) and may be detectable at this early postnatal stage ( Skarlatou et al . , 2020 ) . However , V1 and V0D INs are glycinergic ( Alvarez et al . , 2005; Talpalar et al . , 2013 ) , V0C are cholinergic ( Miles et al . , 2007 ) . Since we have shown that LDPNs are negative for GlyT2 and ChAT and dI4 and dILA INs are dorsal neurons ( Glasgow et al . , 2005; Pillai et al . , 2007 ) expression of Lhx1 would point to cervical LDPNs belonging to either the V0V or dI2 class . In fact , it has very recently been shown that dorsally derived excitatory dI2 INs migrate to this region in the chick spinal cord and have divergent axons along the length of the cord and to the cerebellum ( Haimson et al . , 2021 ) . While these neurons are not premotor in the chick ( Haimson et al . , 2021 ) , it is possible that they are in the mouse . Following injection of gastrocnemius ( GS , n = 4 , Figure 5A ) , we detected 33 premotor LDPNs . Of these infected cervical premotor LDPNs , 8 ( ~24% ) were clearly Lhx1 positive ( Figure 5B , C ) . Given that there is a decrease of Lhx1 expression along the course of postnatal development ( Figure 5—figure supplement 1 ) , it is possible that the proportion of premotor LDPNs that were positive for Lhx1 was underestimated . Nevertheless , although we cannot conclude that the identified LDPNs arise from a homogenous population , it is likely that at least a portion of them arise from V0v neurons and/or dI2 neurons . Given that propriospinal neurons are involved in interlimb coordination , we next sought to determine whether the divergent cervical premotor LDPNs also project to cervical MNs . We therefore performed a series of experiments in which we injected forearm muscles ( FMs ) with ΔG-Rab-mCherry , and extensor hindlimb GS with ΔG-Rab-eGFP . Since it has been suggested that LDPNs participate in ipsilateral control of forelimb and hindlimb ( Miller and van der Meché , 1976 ) , we sought to determine if premotor LDPNs project to homolateral lumbar and cervical motor pools ( Figure 6A ) . When homolateral limbs were targeted , we found that some premotor LDPNs infected from ankle extensor injections were also infected from homolateral FMs injection ( in 5/6 animals , 16/80 premotor LDPNs were also infected from FMs injection 18 . 7% ± 12 . 9% , Figure 6B–D and Figure 7 ) . These divergent premotor LDPNs that projected to lumbar and cervical MNs were all located in the ventral quadrants with 11/16 located in contralateral lamina VIII , and were distributed throughout the rostro-caudal extent of the cervical cord , including segments rostral ( C4 ) to the MN pools innervating the injected forelimb muscles . Furthermore , they had a soma size similar to the premotor LDPNs double labelled by dual hindlimb injections ( 632 ± 236 µm2 , p = 0 . 056 , n1 = 16 premotor LDPNs infected from both homolateral forelimb and hindlimb injections vs n2 = 38 divergent premotor LDPNs infected from dual hindlimb injections ( see above ) , Mann–Whitney test; Figure 6E ) . Given the involvement of LPDNs in the diagonal synchronization of forelimb and hindlimb during locomotion ( Bellardita and Kiehn , 2015; Ruder et al . , 2016; Sherrington et al . , 1906 ) , we also injected contralateral FMs and GS ( Figure 6—figure supplement 1A ) . We found that 2/26 cervical premotor LDPNs were also infected from the FMs injection with one divergent LDPNs in the lamina VIII contralateral to the hindlimb injection in each of two of the three injected animals ( Figure 6—figure supplement 1B ) . Thus , at least a few cervical premotor LDPNs monosynaptically project to diagonal lumbar and cervical MNs . However , given the paucity of these cervical premotor LDPNs projecting to local cervical MNs , we could not reliably determine whether this subpopulation shared the same morphology as described above . While sharing similar features with the LDPNs infected from dual hindlimb injections , it remains to be determined whether these neurons premotor to hindlimb and forelimb muscles form a homogenous population with the divergent LDPNs . Having identified a population of divergent premotor LDPNs with projections from the cervical to the lumbar region , we next investigated whether ascending propriospinal neurons projecting from the lumbar or thoracic segments to cervical MNs could be identified . Following FMs injections , ascending premotor INs were observed throughout the cord ( thoracic to sacral ) . There were very few ( <1% ) bifurcating ( ascending/descending ) premotor neurons in the thoracic cord after injections in homolateral GS and FMs ( 4/523 double-labelled premotor neurons between T2 and T11 , n = 3 , Figure 6—figure supplement 2A , B ) . We identified premotor long ascending propriospinal neurons ( LAPNs ) in the lumbar cord , about half of which were localized in the dorsal ipsilateral quadrant ( 56/117 , n = 6 forelimb–hindlimb injections ) . This distribution of lumbar premotor LAPNs is different from that of cervical premotor LDPNs , which were almost exclusively ventral ( 164/172 , n = 13 pair of injections , see above ) . Of the 117 lumbar premotor LAPNs identified , 10 were also labelled from GS injections , indicating that some neurons projected both to local lumbar MNs as well as to cervical MNs ( n = 6 ipsilateral forelimb–hindlimb injections , Figure 6—figure supplement 2C , D ) . However , the position of these particular divergent premotor LAPNs was different from that of the premotor LDPNs , in that they were not localized within one quadrant of the cord ( Figure 6—figure supplement 2D ) . Finally , we turned our attention to the sacral spinal cord , where we found few premotor LAPNs ( 12 neurons in four of six mice ) . Of these , however , 10/12 were in the ventral contralateral quadrant ( n = 6 ipsilateral forelimb–hindlimb injections , Figure 6—figure supplement 2E , F ) , similar to the location of the cervical premotor LDPNs . Like these cervical neurons , the sacral LAPNs had strikingly large somata ( 710 ± 310 µm2 , n = 11 premotor LAPNs , Figure 6—figure supplement 2G ) . Of 12 labelled neurons , 3 were also infected from the hindlimb ( LG ) injections ( Figure 6—figure supplement 2E–G ) . Given that the size and location of these sacral premotor LAPNs were similar to the population of cervical divergent premotor LDPNs , they may represent a ‘reverse counterpart’ of this descending system .
The control of MNs across motor pools through spinal premotor circuits is required for the performance of all motor tasks involving limb movements . Previous studies showed the importance of motor synergies in the production of complex movements ( Giszter , 2015; Takei et al . , 2017 ) , with the spinal cord identified as a potential site for muscle synergy organization ( Bizzi and Cheung , 2013; Levine et al . , 2014 ) . In this regard , it might be expected that a significant proportion of local spinal premotor INs innervate multiple motor pools , in particular those corresponding to synergist muscles . Perhaps surprisingly , we found similar rate of divergence throughout the spinal cord be the targeted MN pools synergist or antagonist . In the lumbar region , at least 4 % of the local premotor INs project to two motor pools . More remotely , in thoracic as well as cervical premotor circuits , the apparent rate of divergence was higher but with a decreased density of labelled premotor neurons . Regardless of the proportion of divergent premotor neurons amongst the total premotor population , it is possible that these neurons effectively modulate the synchrony of MN activation and participate in co-activation or co-inhibition of different MN populations . What proportion of premotor neurons project to more than one motor pool ? To investigate the presence of premotor neurons projecting to multiple motor pools in the spinal cord , we used RabV tracing , injecting ΔG-RabV expressing eGFP or mCherry into different pairs of muscles . Although this technique allowed for visualization of divergent premotor neurons throughout the spinal cord , the proportion of divergent premotor neurons has undoubtedly been underestimated . A divergent neuron will be double labelled only if each virus has been efficiently transmitted across its synapses with MNs from both motor pools . Therefore , due to the stochastic nature of the process of crossing a synapse , any given transfer efficiency lower than 100 % will inevitably give rise to an underestimate of the real number of divergent neurons . The efficiency of trans-synaptic jumps for the SADB19 RabV that we used is unknown , and may depend in part on the type of synapse , with stronger connections facilitating transmission of the virus ( Ugolini , 2011 ) . The only indirect indication of efficiency comes from the direct comparison of the SADB19 and the more efficient CVS-N2c strains , for which there was at least a fourfold increase in the ratio of local secondary to primary infected premotor INs ( Reardon et al . , 2016 ) . This result suggests that the trans-synaptic efficiency of SADB19 is no higher than 25 % . While there is no evidence for a bias towards stronger or weaker synapses ( i . e . , the actual number of physical contacts ) between proximal and distal premotor INs , such a bias could affect efficiency of viral transmission , and could thus also have potentially skewed our relative estimate of divergence . With the simplifying assumption that the efficiencies of viral transfer are equal and independent from each other across spinal cord regions , we simulated a double injection experiment , extracting a binomial distribution , and calculated the relation between the observed and true rate of divergence . With a jump efficiency of 25% , the 4 % divergence rate we observed in the lumbar spinal cord would correspond to an actual rate of divergence of 18 % ( Figure 8 ) . And this calculated rate is almost certainly an underestimate because of the phenomenon of viral interference , whereby there is a reduced probability of subsequent infection with a second RabV after a window of a few hours after the first infection ( Ohara et al . , 2009 ) . It is therefore likely that the actual rate of divergence of premotor circuit throughout the cord is substantially higher than we observed . Specifically , it is possible that the vast majority of , if not all , premotor LDPNs innervate more than one motor pool . In our experimental model , the rabies glycoprotein is expressed only in neurons expressing ChAT , such as MNs . By restricting primary infection to specific MNs via intramuscular injection of RabV , trans-synaptic viral spread was thus restricted to neurons presynaptic to the infected MN population . It is therefore theoretically possible that there might be double jumps via other presynaptic cholinergic neurons such as medial partition neurons V0C neurons ( Zagoraiou et al . , 2009 ) . MNs also form synapses with other MNs ( Bhumbra and Beato , 2018 ) , so it could also be possible that specificity is lost due to second-order jumps via these cells . We consider double jumps unlikely for two main reasons: ( 1 ) following muscle injections , the first trans-synaptic labelling occurs after 5–6 days . Since the tissue was fixed 9 days after injections , it is unlikely that many secondary jumps could have occurred in such a brief time window . And ( 2 ) most presynaptic partners of V0C INs are located in the superficial dorsal laminae ( Zampieri et al . , 2014 ) , a region in which we did not observe any labelled INs . We are thus confident that the labelled neurons are premotor . We also acknowledge the possibility that some of the labelled premotor cells might originate from tertiary infection originating from secondary infection of synaptically connected MNs ( Bhumbra and Beato , 2018 ) . Such events might be rare ( Ronzano et al . , 2021 ) and would not alter our findings on the organization of divergent premotor neurons , since we have shown that their distributions are similar , regardless of the particular pair of injected muscles . The similar rate of divergence between synergist and antagonist pairs might be surprising . But divergence to agonist and antagonist motor pools has been shown in adult mice ( Gu et al . , 2017 ) , indicating that these circuits are not limited to an early developmental stage . Apart from the cervical divergent premotor LDPNs that are likely to represent a rather homogenous group of excitatory neurons , the divergent premotor neurons in the thoracic and lumbar regions could be comprised of different neural populations , with a mixed population of excitatory , inhibitory , and , in lower proportion , cholinergic neurons ( Figure 1—figure supplement 6 and Figure 4—figure supplement 1 ) . These INs that project to antagonist motor pools could thus be involved in modulating either joint stiffening ( excitatory ) or relaxation ( inhibitory ) . For example , during postural adjustment and skilled movements , divergent excitatory premotor INs would lead to co-contraction of antagonist muscles to facilitate an increase in joint stiffness and to promote stability ( Hansen et al . , 2002; Nielsen and Kagamihara , 1993; Nielsen and Kagamihara , 1992 ) . In invertebrates , co-contraction of antagonist muscles has also been described preceding jumping ( Pearson and Robertson , 1981 ) : co-contraction could thus also be important for the initiation of movement . On the other hand , divergent inhibitory premotor neurons would lead to joint relaxation . This phenomenon is less well studied ( Leis et al . , 2000; Manconi et al . , 1998 ) . One example could be their involvement in the loss of muscle tone that accompanies rapid eye movement sleep ( Uchida et al . , 2021; Valencia Garcia et al . , 2018 ) . In the cat , long descending fibres originating in the cervical cord have been shown to innervate lumbar MNs ( Giovanelli Barilari and Kuypers , 1969 ) and trigger monosynaptic potentials ( Jankowska et al . , 1974 ) . The existence of LDPNs has been confirmed anatomically in neonatal mice ( Ni et al . , 2014 ) and functionally in adult cats ( Alstermark et al . , 1987a; Alstermark et al . , 1987b ) , where they are thought to play a role in posture and stability . Our study confirms the existence of premotor LDPNs , and also indicates that they have a high rate of divergence ( up to ~40 % compared to ~13 % for thoracic neurons ) . Most cervical LDPNs are clustered in contralateral lamina VIII , are virtually all excitatory , and have a distinct morphology with somal size ~twofold larger than other local cells ( and similar to MNs ) . These findings contrast with the divergent premotor neurons found in the thoracic spinal cord: these are distributed in ipsilateral lamina VI and VII as well as in contralateral lamina VIII and thus clearly comprise multiple neuronal populations . In contrast to thoracic divergent premotor neurons , cervical LDPNs may thus have a more unifying function . Given their apparent widespread divergence , it is possible that these LDPNs are involved in producing widespread increases in muscle tone . One step towards being able to further assess the function of this population of INs would be through understanding their lineage . Given the poor detection of the Lhx1 transcription factor in postnatal mice ( Figure 5—figure supplement 1 ) , we could not conclude that the labelled cervical LDPNs are a population that derive from the V0v or dI2 class . Although in the chick , dI2 INs do not project to MNs ( Haimson et al . , 2021 ) , it is possible that they could in the mouse: these are large neurons located in the ventromedial spinal cord ( Haimson et al . , 2021 ) , and express Lhx1 ( Avraham et al . , 2009 ) . Further experiments using a Dbx1-IRES-GFP mouse line ( Bouvier et al . , 2010 ) , for example , could help to determine the identity of these divergent cervical LDPNs . Genetic access to this particular set of INs would also allow the design of experiments aimed at acute and specific activation or inactivation of divergent LDPNs , and could unravel their anatomy and function in behaviour . The completion of movements requires well-controlled muscle contractions across multiple joints within and between limbs . The control of any one joint is analogous to the production of syllables of speech , with the three most fundamental syllables of movement being a change in joint angle ( requiring reciprocal inhibition of flexors and extensor MNs ) , a stiffening of a joint ( requiring co-activation of flexors and extensor MNs ) , and a relaxation of a joint ( requiring co-inhibition of flexor and extensor MNs ) . While neural circuits for reciprocal inhibition have been well studied over many decades ( Eccles , 1969; Eccles et al . , 1956 ) , circuits for stiffening or relaxation have not been . Our anatomical data identify neurons that could be potentially implicated in these circuits and show that they are present within and distributed throughout the spinal cord . Thus , the mechanisms that lead to the production of the fundamental syllables of movement could be contained within the spinal cord itself .
All experiments ( n = 27 ) were performed according to the Animals ( Scientific Procedures ) Act UK ( 1986 ) and certified by the UCL AWERB committee , under project licence number 70/7621 . Homozygous ChAT-IRES-Cre mice ( which have an IRES-Cre sequence downstream of the ChAT stop codon , such that Cre expression is controlled by the endogenous ChAT gene promoter without affecting ChAT expression; Rossi et al . , 2011 , Jackson lab , stock #006410 ) crossed with homozygous RΦGT mice ( Takatoh et al . , 2013 , Jackson lab , stock #024708 ) , that have Cre dependent expression of the rabies glycoprotein and the avian viral receptor TVA , whose expression is not employed in this study were used for double injections ( see the detail of animal use for each type of injection ) . For single injections , homozygous ChAT-IRES-Cre mice ( termed ChAT-Cre here ) were crossed with hemizygous GlyT2-eGFP mice ( BAC transgene insertion in exon 2 of Slc6a5 gene allowing specific eGFP expression in GlyT2-positive cells , MGI:3835459 , Zeilhofer et al . , 2005 ) and their eGFP-positive offspring was mated with homozygous RΦGT ( see Supplementary file 3 ) . We used the glycoprotein G-deleted variant of the SAD-B19 vaccine strain rabies virus ( a kind gift from Dr M . Tripodi ) . Modified RabV ( ΔG-Rab ) with the glycoprotein G sequence replaced by mCherry or eGFP ( ΔG-Rab-eGFP/mCherry ) was produced at a high concentration with minor modifications to the original protocol ( Osakada et al . , 2011 ) . BHK cells expressing the rabies glycoprotein G ( BHK-G cells ) were plated in standard Dulbecco modified medium with 10 % foetal bovine serum ( FBS ) and split after 6–7 hr incubating at 37°C and 5 % CO2 . They were inoculated at a multiplicity of infection of 0 . 2–0 . 3 with either ΔG-Rab-eGFP or mCherry virus in 2 % FBS , and incubated at 35°C and 3 % CO2 . Plates were then split in 10 % FBS at 37°C and 5 % CO2 . After 24 hr the medium was replaced by 2 % FBS medium and incubated at 35°C and 3 % CO2 for 3 days ( virus production ) . The supernatant was collected and medium was added for another cycle ( three cycles maximum ) , after which the supernatant was filtered ( 0 . 45 µm filter ) and centrifuged 2 hr at 19 , 400 rpm ( SW28 Beckman rotor ) . The pellets were re-suspended in phosphate-buffered saline ( PBS ) and centrifuged together at 21 , 000 rpm , 4°C , 4 hr in a 20 % sucrose gradient . Pellets of each collection were then re-suspended and stored in 5–10 µl aliquots at –80°C . Virus titration was performed on BHK cells plated in 10 % FBS medium at 1 . 5 × 105 cells/ml and incubated overnight at 37°C and 10 % CO2 ( growth ) . The virus was prepared for two serial dilutions with two different aliquots and added in the well after an equal volume of medium had been removed ( serial dilution from 10−3 to 10−10 ) and incubated 48 hr at 35°C and 3 % CO2 . The titre was determined from the count of cells in the higher dilution well and was between 109 and 1010 infectious units ( IU ) /ml . A subcutaneous injection of analgesic ( carprofen , 1 µl , 10% wt/vol ) was given to the neonatal pups ( P1–P3 ) prior to surgery and all procedures were carried out under general isoflurane anaesthesia . After a skin incision to expose the targeted muscle , the virus ( 1 µl ) was injected intramuscularly using a Hamilton injector ( model 7652-01 ) mounted with a bevelled glass pipette ( inner diameter 50–70 µm ) . The mice were injected in TA and PL ( ankle flexor pair ) , LG and MG ( ankle extensor pair ) for synergist pairs and TA and LG for antagonist pairs . In hindlimb/forelimb double injections , the LG and MG were both injected with 1 µl of one RabV to increase the number of long projecting cells infected . In addition , 1 µl of the second RabV was injected in FMs ( see Supplementary file 2 ) without selecting a specific muscle . The injected viruses were used at a titre between 109 and 1010 IU/ml . The incisions were closed with vicryl suture , and the mice were closely monitored for 24 hr post-surgery . Mice were perfused 9 days after the injections . Due to the proximity of synergist pairs of muscles , prior to spinal tissue processing , we dissected the injected leg and confirmed that there was no contamination of virus across the injected muscles or in adjacent muscles below or above the knee . When injecting FMs , we could not target a single muscle . To visualize which muscles had been infected , we carefully dissected each FM and assess for the presence of fluorescent signal ( see Supplementary file 2 ) . Three heterozygous RΦGT mice were also injected ( LG muscle ) with an EnvA pseudotyped RabV in order to test simultaneously for ectopic expression of G or of the TVA receptors . In three control animals we observed one to three labelled MNs , but no IN labelling . This indicates the presence of minimal ectopic TVA expression , but not of G ( Ronzano et al . , 2021 ) . The mice were perfused with PBS ( 0 . 1 M ) followed by PBS 4 % paraformaldehyde under terminal ketamine/xylazine anaesthesia ( i . p . 80 and 10 mg/kg , respectively ) . The spinal cords were then collected through a ventral laminectomy and post-fixed for 2 hr . The cords were divided into the different parts of the spinal cord ( cervical [C1–T1] , thoracic [T2–T11] , lumbar [L1–L6] , and sacral [S1–S4] ) , cryoprotected overnight in 30 % sucrose PBS , embedded in optimal cutting temperature compound ( Tissue-Tek ) and sliced transversally ( 30 µm thickness ) with a cryostat ( Bright Instruments , UK ) . Sections were incubated with primary antibodies for 36 hr at 4°C and with secondary antibodies overnight at 4°C in PBS double salt , 0 . 2 % Triton 100-X ( Sigma ) , 7 % donkey normal serum ( Sigma ) . The primary antibodies used were: goat anti-choline acetyltransferase ( ChAT , 1:100 , Millipore , AB144P ) , chicken anti-mCherry ( 1:2500 , Abcam , Ab205402 ) , rabbit anti-GFP ( 1:2500 , Abcam , Ab290 ) , guinea pig anti-vGluT2 ( 1:2500 , Millipore , AB2251-I ) , and rabbit anti-Lhx1 ( 1:5000 , from Dr . T Jessell , Columbia University , New York ) ; and the secondary antibodies: donkey anti-rabbit Alexa 647 ( 1:1000 , Abcam , Ab150079 ) , donkey anti-goat preadsorbed Alexa 405 ( 1:200 , Abcam , Ab175665 ) , donkey anti-rabbit Alexa 488 ( 1:1000 , Thermo Fisher , A21206 ) , and donkey anti-chicken Cy3 ( 1:1000 , Jackson ImmunoResearch , #703-165-155 ) . The slides were mounted in Mowiol ( Sigma , 81381-250 G ) and coverslipped ( VWR , #631-0147 ) for imaging . Images of the entire sections were obtained using a Zeiss LSM800 confocal microscope with a ×20 air objective ( 0 . 8 NA ) and tile advanced set up function ( ZEN Blue 2 . 3 software ) . A ×63 oil objective was used for Airy scan imaging of somata and excitatory boutons . Tiles were stitched using Zen Blue and analyses were performed using Zen Blue and Imaris ( Bitplane , version 9 . 1 ) software packages . Location maps were plotted setting the central canal as ( 0 , 0 ) in the ( x , y ) Cartesian system and using the ‘Spots’ function of Imaris . The y-axis was set to the dorso-ventral axis . Positive values were assigned for dorsal neurons in the y-axis and ipsilateral ( to the hindlimb injection ) neurons in the x-axis . Coordinates were collected on every section and normalized through the cervical , thoracic , lumbar , and sacral parts separately using grey matter borders and fixing the width and the height of the transverse hemisections . To calculate divergence rates , given the high density of premotor INs infected in the lumbar cord all infected premotor INs ( eGFP+ , mCherry+ and eGFP+ mCherry+ ) were quantified in one of every three sections which further allowed to avoid counting the same cells twice on consecutive sections . In the cervical , thoracic , and sacral regions , all cells were quantified , as their low density allowed for manually excluding premotor neurons found in consecutive sections . Since MNs are big cells localized as a restricted column of the ventral spinal cord , we quantified them on every other sections , to avoid counting the same cell twice on consecutive sections . All statistical analyses and plots were made using R ( R Foundation for Statistical Computing , Vienna , Austria , 2005 , http://www . r-project . org , version 3 . 6 . 2 ) and GraphPad PRISM ( version 7 . 0 ) . To compare cell sectional areas , non-parametric rank tests were used as specified in each related result . The numbers of animals/cells in each experiment and statistical tests used are reported in the figure legends or directly in the text . Results and graphs illustrate the mean ± standard deviation . Statistical significance levels are represented as follows: *p < 0 . 05 , **p < 0 . 01 , ***p < 0 . 001 , ****p < 0 . 0001 , and ns: not significant . | We are able to walk , run and move our bodies in other ways thanks to circuits of neurons in the spinal cord that control how and when our muscles contract and relax . Neurons known as premotor neurons receive information from other parts of the central nervous system and control the activities of groups ( known as pools ) of motor neurons that directly activate individual muscles . To bend a joint or move our limbs , the movement of different muscles needs to be coordinated . Previous studies have focused on how premotor neurons activate a pool of motor neurons to contract a single muscle , but it remains unclear if and how some of these premotor neurons can co-activate different pools of motor neurons to control more than one muscle at the same time . Here , Ronzano , Lancelin et al . injected mice with modified rabies viruses labelled with different fluorescent markers to build a map of the premotor neurons that connect to motor neurons controlling the leg muscles . The experiments revealed that many of the individual premotor neurons in the spinal cords of mice connected to different pools of motor neurons . In the upper region of the spinal cord – which is primarily responsible for controlling the front legs – some large premotor neurons activated motor neurons in this region as well as other motor neurons in a lower region of the spinal cord that controls the back legs . This suggests that these large premotor neurons may be important for coordinating muscles contraction within and between limbs . Many neurological diseases are associated with difficulties in contracting or relaxing muscles . For example , individuals with a condition called dystonia experience disorganized and excessive muscle contractions that prevent them from being able to bend and straighten their joints properly . By helping us to understand how the body coordinates the activities of multiple limbs at the same time , the findings of Ronzano , Lancelin et al . may lead to new lines of research that ultimately improve the quality of life of patients with dystonia and other similar neurological diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2021 | Proximal and distal spinal neurons innervating multiple synergist and antagonist motor pools |
Visual pigments can be spontaneously activated by internal thermal energy , generating noise that interferes with real-light detection . Recently , we developed a physicochemical theory that successfully predicts the rate of spontaneous activity of representative rod and cone pigments from their peak-absorption wavelength ( λmax ) , with pigments having longer λmax being noisier . Interestingly , cone pigments may generally be ~25 fold noisier than rod pigments of the same λmax , possibly ascribed to an ‘open’ chromophore-binding pocket in cone pigments defined by the capability of chromophore-exchange in darkness . Here , we show in mice that the λmax-dependence of pigment noise could be extended even to a mutant pigment , E122Q-rhodopsin . Moreover , although E122Q-rhodopsin shows some cone-pigment-like characteristics , its noise remained quantitatively predictable by the ‘non-open’ nature of its chromophore-binding pocket as in wild-type rhodopsin . The openness/closedness of the chromophore-binding pocket is potentially a useful indicator of whether a pigment is intended for detecting dim or bright light .
Retinal rod and cone photoreceptors , although having similar phototransduction mechanisms , elaborate different morphological and molecular features for functioning in dim and bright light , respectively . At the pigment level , rod pigments have a low rate of spontaneous activation in darkness ( Baylor et al . , 1980 ) , thus offering a good signal-to-noise ratio for dim-light vision . Spontaneous activation originates from internal thermal energy of the pigment molecule , generating an electrical event indistinguishable from that triggered by an absorbed photon ( Baylor et al . , 1980 ) , thus interfering with real-light detection . Recently , Luo et al . ( 2011 ) have developed a macroscopic physicochemical theory about pigment noise based on the notion that a pigment’s spontaneous activity originates from thermal isomerization , with an energy barrier closely related to the pigment’s λmax . By using multi-vibrational-mode statistical mechanics ( Ala-Laurila et al . , 2004; Hinshelwood , 1940; St George , 1952 ) , the theory was able to explain quantitatively the λmax-dependence of pigment noise , with the noise increasing by 107-fold from blue ( short-wavelength-sensitive , or SWS ) cone pigment to red ( long-wavelength-sensitive , or LWS ) cone pigment ( Fu et al . , 2008; Kefalov et al . , 2003; Luo et al . , 2011 ) . This theory clarifies the decades-long uncertainty about whether the spontaneous pigment activity arises from canonical isomerization of the pigment’s chromophore ( as in photoisomerization ) or from some different , unknown chemical reaction . Very interestingly , noise measurements in conjunction with the theory indicate that , for a given λmax , a cone pigment may be generally ~25 fold more spontaneously active than a rod pigment ( Luo et al . , 2011 ) . The simplest interpretation is that a cone pigment has a higher molecular frequency of attempting to cross the isomerization barrier ( Luo et al . , 2011 ) . Concurrently , unlike rod pigment , a number of cone pigments show observable dark chromophore-exchange without isomerization when exposed to another chromophore ( Kefalov et al . , 2005; Matsumoto et al . , 1975 ) , suggesting a tendency of spontaneous dissociation between apo-cone-opsin and 11-cis-retinal by Schiff-base hydrolysis , in turn implicating the binding pocket being accessible – or ‘open’ – to external water . It was hypothesized that this ‘openness’ of cone pigments’ chromophore-binding pocket – defined by the property of dark chromophore-exchange – imposes less constraint on the chromophore’s attempts to isomerize spontaneously , resulting in a higher thermal noise compared to rod pigments for a given λmax ( Luo et al . , 2011 ) . Considering the fundamental success of the above theory in explaining the spontaneous activities of several representative rod and cone pigments , it is important to test the theory’s overall predictive power more generally . However , this test is non-trivial , requiring in each case a separate genetic mouse line expressing a test pigment for stringent interrogation in vivo . As such , the already-available RhoE122Q/E122Q knock-in mouse ( Imai et al . , 2007 ) offers an unusual opportunity . Its rods express a mutant rhodopsin with its Glu122 residue ( conserved in rhodopsin ) in the chromophore-binding pocket replaced by Gln , which is common in cone pigments . This E122Q mutation causes a blue-shift in λmax to ~480 nm from ~500 nm in wild-type ( WT ) rhodopsin ( Imai et al . , 2007 ) , substantial enough for validating the quantitative connection between pigment noise and λmax . Equally interestingly , this mutant rhodopsin has acquired some cone-pigment-like properties such as faster decays of the meta-II and meta-III states as well as a shift of the meta-I/meta-II equilibrium ( Imai et al . , 2007 ) , although retaining the indication of a closed chromophore-binding pocket as gleaned from in vitro experiments ( Sakurai et al . , 2007 ) . Thus , we can also check in this ‘hybrid’ pigment the correlation between pigment noise and openness/closedness of the chromophore-binding pocket as we hypothesized .
A paper just appeared ( Tian et al . , 2017 ) reporting that rhodopsin purified from bovine rod outer segments dissociates into opsin and 11-cis-retinal in darkness , with a half life for holo-rhodopsin of the order of days . This rhodopsin behavior is not incompatible with our findings here because it is clearly still very different in time scale from the chromophore exchange in cone pigments , which occurs within hours .
All animal experiments were carried out according to protocols approved by the Institutional Animal Care and Use Committee at Johns Hopkins University ( MO14M199 for mouse ) and Boston University ( AN15427 for both mouse and zebrafish ) . Animals used in this study include RhoWT/WT;Gcaps-/- ( RRID:MGI:3586516 ) and RhoE122Q/E122Q;Gcaps-/- mice as well as zebrafish ( AB Danio rerio; RRID:ZIRC_ZL1 ) . An eyeball of an acutely-euthanized animal was fixed in an alcohol-based zinc-formalin solution ( Z-fix , Anatech , Battle Creek , MI ) at room temperature overnight . The eyeball was then sent to the Johns Hopkins Medical Laboratories , where it was dehydrated through a series of increasing concentrations of ethanol , embedded in paraffin , and sectioned at a thickness of 5–8 µm . Sections close to the plane of the optic disc were collected , then de-paraffinized and rehydrated by passing through Xylene and a series of ethanol solutions of decreasing concentrations . After rinsing with water , the sections were stained with haematoxylin for 3 min . Following a wash with water , the sections were cleared , rinsed and blued . The sections were then rinsed again and stained with eosin for 1 min . Finally , the slides were rinsed , dehydrated through graded alcohols , cleared by Xylene and mounted . Retinas were isolated from euthanized mice into RIPA lysis buffer ( 140 mM NaCl , 0 . 1% Na-deoxycholate , 10 mM Tris-HCl , pH 8 . 0 , 1 mM EDTA , 0 . 5 mM EGTA , 1% Triton X-100 and 0 . 1% SDS ) . Proteins were extracted by grinding the tissues with plastic pestles and vortexing every 5 min over a total of 30 min of incubation . Protein concentrations were determined using the bicinchoninic acid ( BCA ) Protein Assay Kit ( Thermo Fisher Scientific , Waltham , MA ) . Subsequently , protein extracts ( 20 μg ) were separated on 3–20% continuous SDS-PAGE gels ( Bio-Rad , Hercules , CA ) and transferred to polyvinylidene difluoride ( PVDF ) membrane . The membranes were blocked with 5% normal non-fat milk in TBST ( 500 mM NaCl , 20 mM Tris-HCl , pH 7 . 4 , 0 . 1% Tween-20 ) for 1 hr and then incubated with different primary antibodies at 4°C overnight . Primary antibodies included a mouse anti-bovine rhodopsin ( RHO ) monoclonal antibody ( 1D4; 1:50; gift from Dr . Robert Molday , University of British Columbia ) , a rabbit anti-human transducin ( Gtα ) polyclonal antibody ( RRID:AB_2294749; 1:500; Santa Cruz , Dallas , TX ) , a mouse anti-bovine phosphodiesterase-6 ( PDE6 ) monoclonal antibody ( 1: 1000; gift from Dr . Theodore Wensel , Baylor College of Medicine ) , a mouse anti-bovine cyclic-nucleotide channel subunit A1 ( CNGA1 ) monoclonal antibody ( PMc1D1; 1:100; gift from Dr . Robert Molday ) , a mouse anti-bovine CNG channel subunit B1 ( CNGB1 ) monoclonal antibody ( GARP4B1; 1:1000; gift from Dr . Robert Molday ) , a rabbit anti-mouse arrestin-1 ( ARR1 ) polyclonal antibody ( 1:2500; gift from Dr . Jason C . -K . Chen ) , a rabbit anti-mouse regulator of G protein signaling isoform 9 ( RGS9 ) polyclonal antibody ( 1:1000; gift from Dr . Jason C . -K . Chen ) , and a chicken anti-human glyceraldehyde 3-phosphate dehydrogenase ( GAPDH ) polyclonal antibody ( RRID:AB_10615768; 1:500; Millipore , Germany ) . After being washed with TBST , the blots were incubated with the appropriate HRP-conjugated secondary antibodies ( 1:10 , 000; Bio-Rad ) at room temperature for 1 hr . Finally , the proteins on the membranes were detected by using the Enhanced Chemiluminescence ( ECL ) system ( Thermo Fisher Scientific ) . One- to three-month-old mice were dark-adapted overnight , euthanized and their eyes removed under dim red light . The eyes were hemisected and the retinas were removed under infrared light in Locke’s solution [112 . 5 mM NaCl , 3 . 6 mM KCl , 2 . 4 mM MgCl2 , 1 . 2 mM CaCl2 , 3 mM Na2-succinate , 0 . 5 mM Na-glutamate , 0 . 02 mM EDTA , 10 mM glucose , 0 . 1% vitamins ( Sigma-Aldrich , St . Louis , MO ) , 0 . 1% amino-acid supplement ( Sigma-Aldrich ) , 10 mM HEPES , pH 7 . 4 and 20 mM NaHCO3] . Retinas were stored in Locke’s solution bubbled with 95% O2/5% CO2 at room temperature until use over not longer than 6 hr . When needed , a fraction of the retina was chopped into small pieces with a razor blade in the presence of DNase I ( ~20 U/ml ) and was transferred to the recording chamber perfused with bubbled Locke’s solution at 37 . 5°C ± 0 . 5°C . Temperature was monitored by a thermistor situated close to the recorded cell . Single-cell recordings were made under infrared light by drawing the outer segment of a rod projecting from a fragment of retina into a tight-fitting glass pipette containing the following pipette solution: 140 mM NaCl , 3 . 6 mM KCl , 2 . 4 mM MgCl2 , 1 . 2 mM CaCl2 , 0 . 02 mM EDTA , 10 mM glucose and 3 mM HEPES , pH 7 . 4 . In most experiments with light stimulation , 10- to 30-msec monochromatic flashes were used . Signals were sampled at 1 kHz through an Axopatch 200B amplifier and low-pass filtered at 20 Hz ( RC filter , Krohn-Hite , Brockton , MA ) , unless specified otherwise . The average single-photon response function [f ( t ) ] was computed by first obtaining the average response profile of a rod to 80–100 identical dim flashes and then scaling it to the amplitude of the single-photon response , which was calculated as the ensemble variance-to-mean amplitude ratio at the transient peak of these dim-flash responses . Continuous 10 min recordings were obtained from rods in complete darkness , and the rate constant of spontaneous activation were measured by two methods . In the direct counting method , traces were usually low-pass filtered at 3 Hz for identifying and counting quantal events . Two criteria were imposed during identification: ( 1 ) the amplitude of the event should be >30% of the single-photon response amplitude of the same cell , and ( 2 ) the integration time of the event should be within 50–200% of that of the average dim-flash response . The cellular rate constant of thermal activation was given by the total number of spontaneous events divided by the total recording time for each cell . Alternatively , the dark-recording traces were divided into 100 s epochs . The frequency of observing no event , one event , two events , etc . in an epoch was plotted and fitted with the Poisson distribution p ( u ) =wue−w/u ! , where p ( u ) was the probability of observing u events in each epoch and w was the average number of spontaneous activation event per 100 s epoch . In the second method , power spectra were computed from the entire dark-recording trace and from a segment of it containing no obvious spontaneous events ( based on visual inspection ) for each cell by using Clampfit 9 ( Molecular Devices , Sunnyvale , CA ) in 8 . 192 s segments with 50% overlap . The difference spectrum between these two spectra constituted the spectrum for the spontaneous events . This difference spectrum was fitted with a scaled power spectrum of the average single-photon response function [f ( t ) ; see above] of the same cell . The rate of spontaneous isomerization is given by the scaling factor divided by the acquisition time ( 8 . 192 s ) . The molecular rate constant was obtained by dividing the measured cellular rate by the number of pigment molecules ( 6 . 5 × 107 ) per rod . The expression of rhodopsin appeared normal in RhoE122Q/E122Q;Gcaps-/- retinas based on Western blotting ( Figure 1B ) . Another way to assess pigment content in a rod outer segment is to measure the probability ( ps ) of successfully eliciting electrical responses in an experiment using repeated dim-flash trials of known intensity . This probability is related to the rod outer segment’s effective collecting area ( Ae ) and the flash intensity ( I ) by ps = 1 − e−AeI . In turn , Ae is directly proportional to the pigment content . As such , we found Ae to be 0 . 44 ± 0 . 09 µm2 ( mean ± SD , n = 10 ) for RhoWT/WT;Gcaps-/- rods and 0 . 35 ± 0 . 10 µm2 ( mean ± SD , n = 18 ) for RhoE122Q/E122Q;Gcaps-/- rods . Thus , the E122Q/WT pigment-content ratio is 1/1 . 26 . Meanwhile , microspectrophotometry ( see below ) measured an average relative peak optical density of 0 . 37 ± 0 . 07 unit ( mean ± SD , n = 44 recordings from seven experiments ) for RhoE122Q/E122Q;Gcaps-/- rods and 0 . 30 ± 0 . 09 unit ( n = 84 recordings from 33 experiments ) for RhoWT/WT;Gcaps-/- rods , giving a E122Q/WT pigment-content ratio of 1 . 23/1 . The mild discrepancy between methods may reflect measurement uncertainties . Taken together , the pigment levels appear similar between RhoWT/WT;Gcaps-/- and RhoE122Q/E122Q;Gcaps-/- rods . The rate of spontaneous activation ( k ) is given by: ( 1 ) k=Ae−0 . 84hcRTλmax∑1m1 ( m−1 ) ! ( 0 . 84hcRTλmax ) m−1 , where A is the pre-exponential factor , h is Planck’s constant ( 1 . 58 × 10−37 kcal sec ) , c is speed of light ( 3 . 00 × 1017 nm sec−1 ) , R is universal gas constant ( 1 . 99 × 10−3 kcal oK−1 mol−1 ) , T is absolute temperature ( 310 . 5 oK ) and m is the nominal number of vibrational modes contributing thermal energy to pigment activation . Based on previous work ( Luo et al . , 2011 ) , m is 45 for rhodopsin and is taken to be the same for cone pigments , given the same chromophore . The average A-values were empirically determined to be 7 . 19 × 10−6 s−1 for rod pigments and 1 . 88 × 10−4 s−1 for cone pigments ( Luo et al . , 2011 ) . Predictions were made by substituting these parameters and λmax = 481 nm ( for E122Q-rhodopsin ) into Equation 1 . For experiments on mouse rods , mice were dark-adapted for 12 hr before experiment . After euthanization , eyes were removed under dim red light . Under infrared illumination , the eyes were hemisected and the retinas were isolated in HEPES ( 10 mM , pH 7 . 4 ) -buffered Ames medium ( Sigma-Aldrich ) . Each retina was divided in half , yielding altogether four pieces of tissues to be subjected to different treatments . Two pieces of retina were kept dark-adapted and incubated for 3 hr in darkness in HEPES-buffered Ames medium containing 1% fatty-acid-free bovine serum albumin ( BSA ) with or without 15 μM 9-cis-retinal; the other two pieces of retina were subjected to a 99%-bleach ( see below ) and then incubated in the same HEPES-buffered , BSA-supplemented Ames medium as above with or without 15 μM 9-cis-retinal . After their respective treatments , the absorbance spectra of the retinal pieces were measured using a custom-built microspectrophotometer . A retinal piece was gently flattened by forceps and a slice anchor ( Warner Instruments , Hamden , CT ) on a quartz cover-slip window in the bottom of a 2 mm-deep Plexiglass recording chamber with the photoreceptors facing up . The recording chamber was placed on a microscope stage located in the beam path of the microspectrophotometer . The retinal tissue was superfused at a rate of 4 ml/min with Ames medium ( Sigma-Aldrich ) buffered with sodium bicarbonate and equilibrated with 95% O2/5% CO2 . Temperature was maintained at 35–37°C . Absorption spectra were obtained from a region of the retina along its edge where outer segments could be seen protruding perpendicular to the light beam , with tens of outer segments in the light path . The measured area contained predominantly rod instead of cone photoreceptors , as evinced by the λmax . Measurements were made over the wavelength range of 300–700 nm with 2 nm resolution , with the polarization of the incident beam parallel to the plane of the intracellular disks ( T-polarization ) . The absorbance spectrum was calculated from Beers’ Law OD = log ( Ii/It ) , where OD is the optical density or absorbance , Ii is the light transmitted through a cell-free space adjacent to the outer segments , and It is the light transmitted through the tissue . Generally , 10 complete sample scans and 10 baseline scans were averaged to increase the signal-to-noise ratio . All absorbance spectra were baseline-corrected . For experiments on zebrafish photoreceptors , wild-type ( AB ) zebrafish ( Danio rerio ) , obtained from the colony held by the Animal Science Department at Boston University School of Medicine , was dark-adapted for 12 hr prior to experimentation . Euthanasia , dissection and tissue manipulation were performed in darkness with the aid of infrared image converters . Fishes were euthanized by exposure to cold ( 0°C ) water followed by decapitation . The eyes were removed and hemisected in recording solution containing 104 mM NaCl , 2 . 5 mM KCl , 1 . 2 mM MgCl2 , 1 . 6 mM CaCl2 , 0 . 1 mM NaHCO3 , 1 mM-NaH2PO4 , 1 mM sodium pyruvate , 15 mM glucose , 15 mM HEPES ( acid ) , 5 mM HEPES ( base , Na-salt ) , 0 . 5 µg/ml insulin , 5 µg/ml d-biotin , 70 µl/ml fetal bovine serum , 10 µl/ml penicillin streptomycin , 150 µg/ml L-glutamine , 10 µL/ml 50× MEM amino acids , 5 µl/ml 100× MEM vitamins , pH = 7 . 8 . The retinas were then isolated from the eyecups and the retinal pigment epithelium . Retinal tissues not immediately used for experiment were stored in the above solution in a dark container on ice . For experiments in Figure 3B , a retina was treated off-stage in one of the following four ways: ( 1 ) directly used for microspectrophotometric measurements to obtain dark spectra , ( 2 ) incubated in recording solution with additional 1% bovine serum albumin ( fatty-acid-free ) and 15 µM 9-cis-retinal in darkness for 3 hr , ( 3 ) incubated with 1% bovine serum albumin and 15 µM 9-cis-retinal in the same way for 6 hr , and ( 4 ) bleached ( see below ) and incubated with 1% bovine serum albumin and 15 µM 9-cis-retinal in the same way for 3 hr . After treatment , the retina was cut into small ( ~50 µm×50 µm ) pieces and then triturated in solution , producing isolated photoreceptors . Cells were transferred to a recording chamber containing recording solution maintained at 20–22°C . Different types of photoreceptors were identified by their morphology and confirmed by spectral absorbance , which was recorded similarly as in mouse experiments except from single zebrafish photoreceptors . To measure the time course of chromophore-exchange in zebrafish LWS cones ( Figure 3C ) , dark-adapted photoreceptors were dissociated as above directly after retina isolation and were transferred to the MSP recording chamber . A LWS cone was identified and its dark spectrum measured . The solution in the recording chamber was then replaced by recording solution containing 1% bovine serum albumin and 15 µM 9-cis-retinal . Measurements of spectral absorbance were made periodically over 3 hr , at 20–22°C . The light for probing the spectrum at a given time point was at an intensity that would bleach less than 0 . 1% of the pigment content per scan . To quantify the degree of chromophore-exchange , we used the spectrum of dark-adapted LWS cones ( Figure 3B , black ) and that of bleached LWS cones regenerated with 9-cis-retinal ( Figure 3B , green ) as the spectra for 11-cis- and 9-cis-pigment , respectively . A polynomial with degree 10 was fitted into each of these spectra . For each LWS cone , the spectrum acquired at each time point during 9-cis incubation was fitted in the 510–750 nm range with a linear combination of the two polynomials to obtain the percentage of 9-cis-conjugated pigment . Data from all cells were then averaged . For mouse rods , bleaching was performed off-stage on a portable optical bench consisting of a tungsten/halogen lamp , a set of neutral density filters , a 500 nm interference filter and a small aperture ( 3 . 25 mm ) . The retinal tissue was placed in Ames medium in a 35 mm petri dish under the focused circular light spot . The onset of light was controlled by a manual shutter . The bleached fraction , F , was estimated from the relation F = 1− e−IPt , where I was the bleaching light intensity ( 1 . 33 × 106 photons µm−2 s−1 ) , P was the photosensitivity [5 . 7 × 10−9 µm2; see ( Woodruff et al . , 2004 ) ] of mouse rhodopsin measured in situ at its λmax and t was the duration of light exposure; the retinal tissue was typically light-exposed for 16 min to achieve a > 99% bleach . For zebrafish photoreceptors , bleaching was done as above except for using recording solution instead of Ames solution , and using white light of the same source intensity ( i . e . , not including the 500 nm interference filter ) . 9-cis-retinal was handled in dim red light . A stock solution of 30 mM 9-cis-retinal was prepared by dissolving 9-cis-retinal in ethanol . The peak absorbance ( OD ) of retinoid in the stock solution was measured using a conventional spectrophotometer , and its concentration was calculated as c = ( OD373l ) /ε373 , where l was a 1 cm path length and ε373 = 36 , 100 M−1 cm−1 was the extinction coefficient of 9-cis-retinal in ethanol . Working solutions containing 9-cis-retinal were prepared by first adding 1 µl of stock solution to a conical vial . HEPES-buffered Ames medium containing 1% delipidated BSA was then added in multiple times in increasing amounts ( 9 × 5 μl , 1 × 50 μl , 2 × 450 μl , 1 × 1000 μl ) until the final volume was 2 ml; the concentration of 9-cis-retinal in the working solution was 15 µM . | At the back of our eyes is a thin layer of cells that contain light-absorbing pigment molecules . These cells convert light energy into electrical signals that the brain then interprets to allow us to see . In this cell layer , the so-called cone cells work in bright light and provide us with the sense of color , whereas rod cells are for vision in dim light . Each visual pigment consists of a protein with a pocket-like space that holds a compound called a chromophore . Light causes the chromophore to change shape inside the pocket , which in turn activates the pigment . However , the pigments can also become activated at random , even in darkness . These false signals , nicknamed “dark light” , are caused by heat instead of light and essentially create a kind of visual noise that can interfere with vision . In 2011 , researchers found that pigments that are most sensitive to the longer wavelengths of light ( that is , light redder in color ) tend to be noisier . The researchers also found that cone pigments are noisier than rod pigments even if they are most sensitive to the same wavelengths of light . To understand what causes this difference between cone and rod pigments , Yue , Frederiksen et al . – who include many of the researchers involved in the 2011 study – made use of mice with a mutated pigment in their rod cells . The mutant pigment was more sensitive to light of shorter wavelengths and , importantly , it behaved like a cone pigment in some ways but kept the closed pocket that is found in rod pigments . Indeed , Yue , Frederiksen et al . showed that the noise level of this mutant pigment could be accurately predicted from the wavelength it was most sensitive to and how closed its pocket was ( in other words , the pocket's “closedness” ) . Further analyses revealed that an open pocket seems to be common to cone pigments from different species . So , it appears that cone pigments are noisier because they have a more open pocket , and the extra space might allow the chromophore to move around and change shape more easily . Going forward , more visual pigments need to be tested to confirm the relationship between the openness of the chromophore-binding pocket and spontaneous activity . If confirmed , it might be possible to one day predict whether a pigment is intended for dim- or bright-light vision simply by knowing whether its chromophore-binding pocket is more open or closed . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"neuroscience"
] | 2017 | Spontaneous activation of visual pigments in relation to openness/closedness of chromophore-binding pocket |
Cytoplasmic dynein powers intracellular movement of cargo toward the microtubule minus end . The first step in a variety of dynein transport events is the targeting of dynein to the dynamic microtubule plus end , but the molecular mechanism underlying this spatial regulation is not understood . Here , we reconstitute dynein plus-end transport using purified proteins from S . cerevisiae and dissect the mechanism using single-molecule microscopy . We find that two proteins–homologs of Lis1 and Clip170–are sufficient to couple dynein to Kip2 , a plus-end-directed kinesin . Dynein is transported to the plus end by Kip2 , but is not a passive passenger , resisting its own plus-end-directed motion . Two microtubule-associated proteins , homologs of Clip170 and EB1 , act as processivity factors for Kip2 , helping it overcome dynein's intrinsic minus-end-directed motility . This reveals how a minimal system of proteins transports a molecular motor to the start of its track .
Cytoskeletal motor proteins transport and position a variety of macromolecules , organelles and mRNAs in the cell interior ( Vale , 2003 ) , but how these motors are themselves targeted to specific locations within the cell is an important unsolved question . Cytoplasmic dynein , a large and complex AAA+ motor protein ( Carter , 2013 ) , uses the energy from ATP hydrolysis to move cargoes toward the minus end of microtubules ( typically toward the cell center ) . In addition , dyneins can act while anchored at the cell cortex , where they pull the microtubule network toward them ( Moore et al . , 2009 ) . Live-cell imaging in diverse organisms reveals that , surprisingly for a minus-end-directed motor , dynein accumulates at the plus ends of microtubules that grow and shrink near the cell periphery ( Vaughan et al . , 1999; Han et al . , 2001; Ma and Chisholm , 2002; Lee et al . , 2003; Sheeman et al . , 2003; Lenz et al . , 2006; Kobayashi and Murayama , 2009 ) . By localizing to dynamic microtubule plus ends ( Howard and Hyman , 2009 ) , dynein is thought to ‘search-and-capture’ ( Kirschner and Mitchison , 1986 ) cargo molecules , before transporting them toward the minus end ( Wu et al . , 2006 ) . The targeting of the dynein machinery to the microtubule plus end is crucial for proper dynein function in budding yeast ( Moore et al . , 2009 ) , filamentous fungi ( Wu et al . , 2006 ) , and a subset of metazoan cells , including mammalian neurons ( Lomakin et al . , 2009; Lloyd et al . , 2012; Moughamian and Holzbaur , 2012; Moughamian et al . , 2013 ) . However , the molecular mechanisms that target dynein to the microtubule plus end are poorly understood . Microtubule plus ends are dynamic binding platforms for a variety of proteins , collectively referred to as +TIPs ( plus-end tracking proteins ) ( Akhmanova and Steinmetz , 2008 ) . Previous studies have illuminated two general mechanisms by which proteins can be targeted to microtubule plus ends . First , a subset of +TIPs preferentially bind to unique structural features present at the plus end vs the body of the microtubule , and can recruit additional proteins to these sites . The prototypical example of this class of +TIP are EBs ( end-binding proteins ) , which are involved in recruiting binding partners such as Clip170 and the dynein regulator dynactin to plus ends in metazoans ( Watson and Stephens , 2006; Bieling et al . , 2007; Dixit et al . , 2009; Zanic et al . , 2009; Maurer et al . , 2012; Moughamian et al . , 2013 ) . Second , kinesin motor proteins can power the vectorial transport of proteins along the microtubule body to the plus end ( Bieling et al . , 2007; Subramanian et al . , 2013 ) . These two mechanisms ( direct recruitment and vectorial transport ) are not mutually exclusive , and evidence suggests that both may contribute to the plus-end targeting of dynein ( Wu et al . , 2006 ) . In the yeast Saccharomyces cerevisiae , genetic studies implicate at least three proteins in targeting dynein to the microtubule plus end: Lis1 , a regulator of dynein motility ( Lee et al . , 2003; Sheeman et al . , 2003 ) ; Bik1 , a homolog of Clip170 ( Sheeman et al . , 2003; Markus et al . , 2009 ) ; and Kip2 , a plus-end-directed kinesin motor ( Carvalho et al . , 2004; Caudron et al . , 2008; Markus et al . , 2009 ) . Deletion of the EB homolog ( Bim1 ) in S . cerevisiae appears to have little effect on the plus-end targeting of dynein and its co-factors ( Carvalho et al . , 2004; Caudron et al . , 2008; Markus et al . , 2011 ) , despite ( 1 ) Bim1 binding directly to Bik1 , an essential factor for dynein plus-end targeting ( Sheeman et al . , 2003; Blake-Hodek et al . , 2010 ) ; and ( 2 ) The homologs of Kip2 , Bik1 and Bim1 functioning together as a plus-end tracking system in fission yeast ( Bieling et al . , 2007 ) . While dynein plus-end targeting can persist in the absence of Bim1 , it is unknown if Bim1 is involved in the dynein pathway in the native situation , because +TIPs often interact in an interconnected and , in some cases , redundant manner ( Caudron et al . , 2008 ) . More generally , how a system of molecules can function together to target dynein to the plus end is not clear in any system . This problem is particularly pressing for S . cerevisiae dynein: a forceful minus-end-directed motor with constitutive activity in motility assays ( Reck-Peterson et al . , 2006; Gennerich et al . , 2007 ) . In vitro reconstitution can provide powerful mechanistic insights into cellular processes , complementary to those derived from the complex in vivo environment ( Liu and Fletcher , 2009 ) . Here , we have reconstituted kinesin-driven transport of dynein to the microtubule plus end using purified proteins , allowing us to dissect the mechanism using single-molecule imaging , protein engineering and DNA origami .
We began by purifying the S . cerevisiae proteins Lis1 , Bik1 , Bim1 and Kip2 , in addition to a well-characterized dynein motor construct ( GST-dynein331 kDa; referred to herein as ‘dynein’ ) ( Reck-Peterson et al . , 2006 ) . This dynein construct lacks the cargo-binding tail , which is dispensable for plus-end targeting in vivo ( Markus et al . , 2009 ) , and is dimerized by GST to yield a motor with highly-similar motile properties to full-length dynein ( Reck-Peterson et al . , 2006 ) . Each purified protein migrated as a single band by SDS-PAGE ( Figure 1A ) with the exception of Kip2 , which migrated as a doublet . The Kip2 bands correspond to differently phosphorylated forms of the protein; after treatment with λ phosphatase , the Kip2 doublet collapsed into a single band ( Figure 1A , inset ) . 10 . 7554/eLife . 02641 . 003Figure 1 . Purification and analysis of the putative dynein plus-end transport machinery . ( A ) SDS-PAGE and diagrams of purified proteins . Kip2 from S . cerevisiae migrates as a doublet of bands , corresponding to differently phosphorylated isoforms as verified by phosphatase treatment ( right panel ) . ( B ) Analysis of sub-complex formation by size-exclusion chromatography . Elution volumes of the individual proteins are shown on the top axis . V0: void volume . Dynein and Lis1 co-elute as a complex ( green trace ) , as do Bik1 , Bim1 and Kip2 ( magenta trace ) . A mixture of all five proteins elutes as dynein-Lis1 and Bik1-Bim1-Kip2 sub-complexes , rather than one stable assembly ( blue trace ) . SDS-PAGE analysis of the fractions is shown below , pseudo-colored to match the traces . ( C ) TIRF microscopy microtubule recruitment assay . Fluorescently labeled wild-type ( WT ) dynein or a mutant with weak microtubule affinity ( K3116A , K3117A , E3122A , R3124A: Weak ) were incubated with microtubules in the presence of the indicated proteins . Representative microtubule images and the corresponding dynein channel are shown in each condition . The width of each panel corresponds to 5 . 4 μm . Mean dynein fluorescence intensity ± SEM is shown below ( N = 30–50 microtubules ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02641 . 00310 . 7554/eLife . 02641 . 004Figure 1—figure supplement 1 . Analysis of interactions between Bik1 , Bim1 and Kip2 by size-exclusion chromatography . The void volume ( V0 ) and elution volumes of the individual proteins are shown on the top axis . SDS-PAGE analysis of the corresponding fractions is shown below , pseudo-colored to match the traces . Bik1 and Bim1 co-elute as a complex that is upshifted relative to the individual proteins , in agreement with Blake-Hodek et al . ( 2010 ) . There is a partial upshift and co-elution of Kip2 with Bik1 , whereas Bim1 and Kip2 do not co-elute as a stable complex . However , all three proteins ( Bik1 , Bim1 and Kip2 ) together co-elute as an upshifted ternary complex . Removal of Bik1's C-terminal zinc knuckle domain ( Bik1ΔC ) does not prevent its ability to form a complex with Bim1 and Kip2 . DOI: http://dx . doi . org/10 . 7554/eLife . 02641 . 004 To explore the interactions within this putative dynein plus-end transport machinery , we mixed the proteins in different combinations and analyzed their behavior by size-exclusion chromatography . Dynein and Lis1 co-elute in a complex , as verified by SDS-PAGE of the fractions ( Figure 1B; Huang et al . , 2012 ) . We also found that Bik1 , Bim1 and Kip2 co-elute in a ternary complex ( Figure 1B ) , resembling the behavior of related proteins from Schizosaccharomyces pombe ( Bieling et al . , 2007; Table 1 ) . A pair-wise mixture of Kip2 with Bim1 did not elute as a complex under the same conditions ( Figure 1—figure supplement 1 ) , suggesting that Bim1 binds with higher affinity to the Kip2-Bik1 complex . In summary , these first experiments with all five purified components demonstrate that dynein-Lis1 and Bik1-Bim1-Kip2 each form complexes that are sufficiently stable to co-elute by size-exclusion chromatography . 10 . 7554/eLife . 02641 . 005Table 1 . Protein homolog namesDOI: http://dx . doi . org/10 . 7554/eLife . 02641 . 005GenericS . cerevisiaeS . pombeKinesinKip2Tea2Clip170Bik1Tip1EB1Bim1Mal3Lis1Pac1– Given that the dynein-Lis1 and Bik1-Bim1-Kip2 sub-complexes did not stably associate in solution ( Figure 1B ) , we hypothesized that these two assemblies might interact dynamically on the microtubule . To test this idea , we devised a microtubule recruitment assay . When wild-type dynein is mixed with microtubules in the absence of nucleotide , it binds strongly to the microtubule , as visualized by total internal reflection fluorescence ( TIRF ) microscopy in Figure 1C ( lane 1 ) . In order to monitor dynein recruitment to the microtubule by other proteins , we introduced four mutations into its microtubule-binding domain ( Koonce and Tikhonenko , 2000; Carter et al . , 2008; Redwine et al . , 2012 ) , which severely weakened its association with microtubules ( Figure 1C , lane 2; see also Figure 2—figure supplement 2 ) . However , the weak-binding dynein could be recruited efficiently to the microtubule by the combined presence of Lis1 , Bik1 , Kip2 and Bim1 ( Figure 1C , lane 3 ) . This dynein recruitment depended strictly on the presence of Lis1 and Bik1 ( Figure 1C , lanes 4 and 5 ) , consistent with the idea that Lis1 and Bik1/Clip170 interact directly ( Coquelle et al . , 2002; Sheeman et al . , 2003; Lansbergen et al . , 2004 ) . To determine if the reported interaction between Lis1 and the C-terminal zinc-knuckle of Bik1/Clip170 ( Coquelle et al . , 2002; Sheeman et al . , 2003; Lansbergen et al . , 2004 ) is involved in connecting the dynein-Lis1 and Bik1-Bim1-Kip2 complexes , we purified a Bik1ΔC construct lacking the C-terminal region . The Bik1ΔC construct retained the ability to bind Kip2 and Bim1 ( Figure 1—figure supplement 1 ) . However , its ability to recruit the weak-dynein to the microtubule in the presence of Lis1 , Kip2 and Bim1 was abolished ( Figure 1C , lane 6 ) . Together , these results suggest that dynein-Lis1 and Bik1-Bim1-Kip2 each form sub-complexes , which in turn interact in a manner dependent on Lis1 and the C-terminal region of Bik1 . Having established a potential chain of connection between dynein and Kip2 with purified proteins , we next sought to visualize the emergent motile behavior of these opposite polarity motors on dynamic microtubules . We first monitored the motility of dynein and Kip2 together , but in the absence of the other proteins , on dynamic microtubules grown from stabilized seeds using three-color TIRF microscopy ( Figure 2A ) . Dynein and Kip2 were labeled with different-colored fluorophores ( tetramethylrhodamine and Atto647 , respectively ) , while microtubules contained a fraction of tubulin labeled with a third color ( Alexa488 ) . In the absence of Lis1 , Bik1 , and Bim1 , dynein moved toward the minus end of the microtubule , and Kip2 moved toward the plus end , as expected ( Figure 2B ) . Notably , upon reaching the minus end of the microtubule , dynein accumulated at , and moved with , the slowly growing end of the polymer ( Figure 2B ) . Dynein showed the same behavior in the absence of Kip2 , indicating that dynein has the intrinsic ability to track dynamic microtubule minus ends . 10 . 7554/eLife . 02641 . 006Figure 2 . Reconstitution of dynein transport to the microtubule plus end . ( A ) Diagram of dynamic microtubule assay , after Bieling et al . ( 2007 ) . A bright , biotinylated , GMP-CPP-stabilized microtubule seed is attached to the coverslip via streptavidin and biotin-PEG . Dimly labeled microtubule extensions grow from this seed in the presence of tubulin and GTP . Tubulin , dynein and Kip2 are visualized using TIRF microscopy , and motor protein movement on the dynamic extensions is analyzed . ( B ) Dynein and Kip2 motility on dynamic microtubules . Kymographs of the dynein and microtubule channels are overlaid , and the Kip2 channel is shown separately for clarity . Plus ( + ) and minus ( − ) denote microtubule polarity . Dynein moves to the minus end and Kip2 moves to the plus end . Scale bar , 5 μm . ( C ) In the presence of Lis1 , Bik1 , Bim1 and Kip2 , dynein moves to the plus end of the microtubule ( lane 1 ) . Weakening dynein’s microtubule affinity via mutagenesis ( Figure 2—figure supplement 2 ) increases its plus-end velocity and accumulation ( lane 2 ) . In the absence of Lis1 , Bik1 or Kip2 , dynein resumes minus-end-directed motion ( lanes 3–5 ) . Arrows mark microtubule catastrophe events seen in the absence of Kip2 ( lane 5 ) . When Bim1 is omitted , dynein displays both plus- and minus-end-directed movements ( lane 6 ) . Mean velocities ± SD for dynein and free Kip2 ( without dynein bound ) are shown at bottom ( N = 36–122 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02641 . 00610 . 7554/eLife . 02641 . 007Figure 2—figure supplement 1 . Examples of plus-end-directed motion of GST-Dynein331 kDa and full-length dynein . Dynein at the growing microtubule plus end is marked with arrowheads . Plus ( + ) and minus ( − ) denote microtubule polarity . DOI: http://dx . doi . org/10 . 7554/eLife . 02641 . 00710 . 7554/eLife . 02641 . 008Figure 2—figure supplement 2 . Weakening dynein's microtubule affinity increases the velocity of its plus-end-directed transport . ( A ) Four charged amino acids were mutated to weaken dynein's microtubule affinity ( K3116A , K3117A , E3122A and R3124A ) . These residues are depicted in red sphere representation using the cryo-EM derived model of dynein's microtubule-binding domain bound to an α/β-tubulin dimer ( PDB 3J1T ) ( Redwine et al . , 2012 ) . The inset shows the location of dynein's microtubule-binding domain ( boxed ) in relation to the rest of the molecule . ( B ) These mutations substantially weaken , but do not abolish , the binding of S . cerevisiae dynein to the microtubule in the absence of nucleotide , as shown by a microtubule co-sedimentation assay . Under conditions in which wild-type ( WT ) dynein is fully bound to the microtubule ( [α/β-tubulin]: 600 nM ) , the weak binding mutant only partially co-sediments with the microtubule in the pellet ( P ) and partially remains in the supernatant ( S ) . See also Figure 1C . ( C ) In the presence of Lis1 , Bik1 , Bim1 and Kip2 , both the WT and weak binding dynein constructs are transported by Kip2 to the plus-end of the microtubule ( Figure 2C ) . However , the plus-end velocity of the weak-binding construct is increased relative to the wild-type dynein ( p<0 . 0001 , Student's t test ) , becoming close to that of free Kip2 . This suggests that wild-type dynein can resist its own transport to the plus end through its microtubule-binding domain . DOI: http://dx . doi . org/10 . 7554/eLife . 02641 . 00810 . 7554/eLife . 02641 . 009Figure 2—figure supplement 3 . Kymographs showing colocalized , plus-end-directed runs of weak dynein and Kip2 in the presence of Bik1 and Lis1 . Plus ( + ) and minus ( − ) denote microtubule polarity . DOI: http://dx . doi . org/10 . 7554/eLife . 02641 . 009 Next , we explored the effect of adding the remaining factors . Strikingly , in the presence of Lis1 , Bik1 , Bim1 and Kip2 , the large majority of dynein movements along the body of the microtubule were directed to the plus end ( Figure 2C , lane 1 ) ( 90 ± 3%; proportion ± SE , N = 122 ) . Moreover , on 63 ± 7% of microtubules , dynein was present at the growing plus end itself ( Figure 2—figure supplement 1 ) , reminiscent of dynein localization in living yeast cells ( Lee et al . , 2003; Sheeman et al . , 2003; Markus et al . , 2009 ) . Minus-end-directed dynein movements were now rare ( 10 ± 3% of events ) . We observed similar behavior for the full-length dynein complex purified from S . cerevisiae ( Reck-Peterson et al . , 2006; Figure 2—figure supplement 1 ) . Thus , we conclude that a system of four proteins ( Lis1 , Bik1 , Bim1 , and Kip2 ) is sufficient to strongly bias dynein's movement toward the microtubule plus end . Dynein's plus-end-directed motion depended strictly on Kip2 ( Figure 2C , lane 5 ) , but was slower than that of free Kip2 ( Figure 2—figure supplement 2 ) , leading us to hypothesize that dynein might have the capacity to resist its own transport to the plus end . To test this model , we repeated the reconstitution using the dynein mutant engineered to have weak affinity for the microtubule . Strikingly , the plus-end velocity of the weak-binding dynein was increased relative to the wild-type protein , becoming close to that of Kip2 ( Figure 2C , Figure 2—figure supplement 2 ) . Moreover , the weak-binding dynein also showed strong accumulation at the plus end of the microtubule ( Figure 2C , lane 2 ) . These results indicate that wild-type dynein is not a passive passenger during its plus-end-directed transport . Instead , the data support a model in which wild-type dynein and Kip2 have the potential to engage simultaneously with the microtubule in a ‘tug-of-war’-like mechanism . If dynein's plus-end-directed movement is driven directly by Kip2 , these two species should colocalize and move together on the microtubule . As expected from our biochemical data ( Figure 1B ) , binding events between the dynein and Kip2 subcomplexes were extremely rare when both species were diluted to the sub-nanomolar concentrations required for single-molecule imaging . However , by exploiting the weak-binding dynein construct that does not bind appreciably to the microtubule at low nanomolar concentrations ( Figure 1C , lane 2 ) , we were able to observe clear co-localized movements of dynein with Kip2 in the presence of Bik1 and Lis1 ( Figure 2—figure supplement 3 ) . Notably , the dynein construct frequently localized with Kip2 midway through a run , and dissociated before its end , consistent with transient binding as indicated by our biochemical experiments ( Figure 1B ) . These observations suggest that the movement of dynein toward the plus end is driven directly by Kip2 . To probe the roles played by individual components in the plus-end-directed transport of dynein , we performed dropout experiments in which proteins were selectively omitted from the reconstitutions ( Figure 2C ) . Lis1 and Bik1 were critical for dynein's plus-end-directed movement ( Figure 2C , lanes 3 and 4 ) . In their absence , dynein moved to the minus end of the microtubule , despite the continued movement of Kip2 to the plus end . These results indicate that Lis1 and Bik1 are required to connect dynein to Kip2 . Omitting Kip2 abolished dynein's plus-end-directed movement , showing that the pathway we have reconstituted is kinesin dependent ( Figure 2C , lane 5 ) . Interestingly , removal of Kip2 also increased the frequency of microtubule catastrophes ( the conversion from growth to shrinkage; Figure 2C , arrows ) , consistent with the short microtubule phenotype caused by Kip2 deletion in vivo ( Cottingham and Hoyt , 1997 ) . Finally , we observed an unexpected influence of Bim1 ( Figure 2C , lane 6 ) . In the absence of Bim1 , dynein could still be transported to the plus end , but the fraction of plus- vs minus-end-directed dynein movements was roughly balanced ( 47 vs 53 ± 5% , respectively ) . Thus , Lis1 , Bik1 and Kip2 constitute the minimal machinery for transporting dynein to the plus end , and Bim1 can regulate this machinery to increase the fraction of dynein movements that are directed to the plus end . Because Bim1 forms a ternary complex with Bik1 and Kip2 ( Figure 1B ) , we suspected that the influence of Bim1 on dynein's plus-end transport might be exerted through these proteins . Furthermore , analysis of the Kip2 amino acid sequence revealed that there is an SxIP motif within an extension N-terminal to the kinesin motor domain ( Figure 3A ) . The SxIP motif is a signature sequence that interacts with EB-family proteins such as Bim1 ( Honnappa et al . , 2009 ) , suggesting that Bim1 may interact with Kip2 through this motif . Bim1 also binds to the Cap-Gly domain of Bik1 ( Blake-Hodek et al . , 2010 ) and the microtubule lattice ( Zimniak et al . , 2009 ) , while Bik1's coiled coil is thought to interact with Kip2 ( Newman et al . , 2000; Carvalho et al . , 2004 ) . These observations suggest that a network of interactions exists within the Bik1-Bim1-Kip2 complex ( Figure 3A ) , consistent with our pairwise binding analysis ( Figure 1—figure supplement 1 ) . Therefore , we wanted to test if Bik1 and Bim1 could promote Kip2's microtubule interactions by providing additional , transient microtubule-binding sites ( Figure 3A ) . 10 . 7554/eLife . 02641 . 010Figure 3 . Bik1 and Bim1 are Kip2 processivity factors . ( A ) Top: diagram of Kip2 , Bik1 and Bim1 domain structure . Arrows indicate reported and putative interactions ( see main text for details ) . Bottom: Kip2 contains an SxIP motif ( orange ) within an N-terminal extension to its kinesin motor domain . The SxIP motif is flanked by serine residues ( blue ) that can be phosphorylated ( Holt et al . , 2009; Bodenmiller et al . , 2010 ) . ( B ) Top: kymographs showing the movement of Kip2-Atto647 along Taxol-stabilized microtubules in the absence and presence of Bik1 and Bim1 . Scale bars , 1 min and 5 μm . Bottom: histograms showing the distribution of Kip2 run lengths in each condition . Orange curves show single exponential fits , based on the cumulative distribution function of each dataset . Average run lengths were determined from the decay constant . Standard errors were determined by bootstrapping , with each dataset resampled 200 times . ( C ) Bar chart of run length ± SE of dynein and Kip2 toward the minus and plus end of the microtubule , respectively , in the indicated conditions ( N = 233–331 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02641 . 01010 . 7554/eLife . 02641 . 011Figure 3—figure supplement 1 . While increasing Kip2 processivity , Bik1 and Bim1 confer a small reduction in Kip2 velocity and do not change Kip2 copy number . ( A ) Impact of Bik1/Bim1 on dynein and Kip2 velocity on Taxol-stabilized microtubules . Graphs show mean velocity ± SEM ( N = 175–331 events per condition ) . Bik1 and Bim1 together confer a small reduction in the velocity of Kip2’s plus-end-directed movement ( p < 0 . 0001 , Student’s t test ) . ( B ) Bik1 and Bim1 do not change Kip2 copy number , as judged by average fluorescence intensity per Kip2 spot . Graphs show mean intensity ± SEM ( N = 29–40 spots per condition ) . Fluorescence intensity per Kip2 spot does not change upon addition of Bik1 and Bim1 ( P = 0 . 1529 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02641 . 01110 . 7554/eLife . 02641 . 012Figure 3—figure supplement 2 . Bik1 and Bim1 increase Kip2’s microtubule on-rate and decrease its off-rate . ( A ) Left: kymographs showing Kip2 at a fixed concentration in the absence or presence of Bik1 and Bim1 . Right: quantification of Kip2 landing rate on the microtubule , expressed as events per micron of microtubule per nM Kip2 per minute . Bik1 and Bim1 increase Kip2's microtubule landing rate . Orange lines show mean ± SEM . ( B ) Histograms showing the duration of Kip2 runs in the absence or presence of Bik1 and Bim1 . Orange curves show single exponential fits , based on the cumulative distribution function of each dataset . The average run duration and off-rate were determined from the decay constant . Standard errors were determined by bootstrapping , with each dataset resampled 200 times . Bik1 and Bim1 reduce Kip2's microtubule off-rate . DOI: http://dx . doi . org/10 . 7554/eLife . 02641 . 012 To directly visualize the impact of Bik1 and Bim1 on Kip2's motility , we fluorescently labeled Kip2 with Atto647 and tracked the movement of single molecules along Taxol-stabilized microtubules ( Figure 3B ) . In the absence of regulators , the average distance traveled by Kip2 per microtubule encounter was 1 . 2 ± 0 . 1 μm ( Figure 3C ) . The addition of Bik1 or Bim1 individually had minor effects on Kip2 run length . However , in the presence of both Bik1 and Bim1 , the run length of Kip2 increased fourfold , to 4 . 8 ± 0 . 3 μm ( Figure 3C ) . Concomitantly , there was a small decrease in Kip2 velocity , while the Kip2 copy number ( judged by fluorescence intensity per spot ) was unchanged ( Figure 3—figure supplement 1 ) . These effects were specific to Kip2: Bik1 and Bim1 had minimal effects on dynein motility ( Figure 3—figure supplement 1 ) . Analysis of Kip2 landing events and run durations revealed that Bik1 and Bim1 increase the on-rate and reduce the off-rate of Kip2's microtubule encounters ( Figure 3—figure supplement 2 ) , indicating that these factors augment Kip2's microtubule affinity . In summary , we conclude that Bik1 and Bim1 together act as processivity factors for Kip2 . Finally , we sought to investigate if regulation of Kip2‘s processivity could affect its ability to overcome dynein in a tug-of-war . Our reconstitutions show that Bik1 is required to connect dynein to Kip2: thus , any additional role of Bik1 in regulating the tug-of-war was obscured . Therefore , we used DNA origami ( Shih and Lin , 2010 ) to couple dynein and Kip2 directly , independent of any other proteins ( Figure 4A ) . Dynein and Kip2 were attached at opposite ends of a 225 nm long DNA structure comprising a 12-helix bundle , which we call the ‘chassis’ ( Derr et al . , 2012 ) . Under this regime , the large majority ( 80 ± 2%; mean ± SEM ) of dynein/Kip2-chassis movements on the microtubule were in the minus-end ( dynein ) direction , as visualized by the incorporation of fluorescently labeled DNA strands ( Figure 4B , C ) . Velocity analysis indicated that the chassis-coupled motors were in a tug-of-war: both minus- and plus-end-directed runs exhibited reduced velocity compared to single-motor chassis ( Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 02641 . 013Figure 4 . Bik1 and Bim1 regulate the outcome of a tug-of-war between dynein and Kip2 . ( A ) Coupling dynein and Kip2 to a DNA origami ‘chassis’ allows them to be pitted directly against each other in a tug-of-war . Single-stranded DNA oligonucleotides were attached to dynein and Kip2 via SNAP tags at their N- or C-terminus , respectively ( Derr et al . , 2012 ) . These DNA oligonucleotides base pair with complementary sequences extending from the chassis . ( B ) Top: kymographs of dynein/Kip2-chassis motility on Taxol-stabilized microtubules in the presence and absence of Bik1 and Bim1 . Bottom: Color-coded schematics , highlighting runs that are minus-end-directed ( green ) , plus-end-directed ( purple ) , bi-directed ( orange ) or non-mobile ( black ) . ( C ) Quantification of the different types of dynein/Kip2-chassis behavior in the indicated conditions , expressed as the average percentage ± SEM ( N = 3 separate dynein/Kip2-chassis assembly reactions , with 306–353 runs analyzed in each case ) . The addition of Bik1 and Bim1 causes the fraction of plus-end-directed runs to increase relative to dynein/Kip2-chassis alone ( p < 0 . 05; Student’s t test ) . Scale bar , 20% . DOI: http://dx . doi . org/10 . 7554/eLife . 02641 . 01310 . 7554/eLife . 02641 . 014Figure 4—figure supplement 1 . Dynein and Kip2 engage in a tug-of-war when coupled via the DNA origami chassis . Graph shows the average velocity ± SD of chassis movements toward the microtubule minus end ( gray bars ) or plus end ( purple bars ) . 1D denotes a chassis with one dynein-attachment site . 1K denotes a chassis with one Kip2-attachment site . 1D1K denotes a chassis with one dynein and one Kip2 attachment site . The velocity of the 1D1K chassis in the minus-end ( dynein ) direction is reduced compared to the 1D chassis , suggesting that Kip2 resists transport to the minus end ( p < 0 . 0001 , Student’s t test ) . The converse is also true for the velocity in the plus-end direction ( p < 0 . 0001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02641 . 014 The DNA chassis allowed us to directly determine if Bik1 and Bim1 influence the outcome of the tug-of-war . The addition of either Bik1 or Bim1 alone had a modest effect on the motility of the dynein/Kip2-chassis complex: the fraction of minus- and plus-end-directed runs was changed little ( Figure 4C ) . In contrast , inclusion of Bik1 and Bim1 together caused the fraction of plus-end-directed runs to markedly increase ( Figure 4C ) . Under these conditions , the dynein/Kip2-chassis also underwent switches in direction during a run ( Figure 4B; orange trace; 10 ± 1% of events ) . These results reveal that regulatory proteins can tune the outcome of a tug-of-war between dynein and the kinesin Kip2 .
Through in vitro reconstitution , protein engineering and DNA origami , we have elucidated a mechanism that spatially targets dynein toward the microtubule plus end—the start of its track ( Figure 5 ) . The essence of the mechanism involves coupling dynein's motor domain to a plus-end-directed kinesin , Kip2 . By reconstituting the functional coupling between these opposite-polarity motors for the first time , we show that two proteins , Lis1 and Bik1 , are sufficient to connect dynein and Kip2 . This connection is capable of bearing load on the microtubule , but is not stable enough to be observed by size-exclusion chromatography . As is the case for interactions found in other self-organizing systems ( Kirschner et al . , 2000 ) , a high off-rate between dynein and Kip2 might allow for dynamic regulation , for example enabling dynein to readily detach from Kip2 following targeting . Indeed , in vivo , the timescale of dynein ‘offloading’ from the microtubule plus end onto its receptor on the cell cortex is on the order of seconds ( Markus and Lee , 2011 ) . 10 . 7554/eLife . 02641 . 015Figure 5 . Model of the minimal machinery for transporting dynein to the microtubule plus end . Based on results from this study and earlier work . Dynein is connected to the plus-end-directed motor Kip2 by Lis1 and Bik1 . Specifically , Lis1's ß-propeller domain binds to the AAA4 subdomain of dynein's motor domain ( Huang et al . , 2012 ) . The dynein/Lis1 complex interacts dynamically with the C-terminal zinc-knuckle domain of Bik1/Clip170 ( Coquelle et al . , 2002; Sheeman et al . , 2003; Lansbergen et al . , 2004; Markus et al . , 2011 ) as indicated by the double arrow . Bik1's coiled coil is posited to interact with Kip2 ( Newman et al . , 2000; Carvalho et al . , 2004 ) . Bim1 is not strictly required for dynein's plus-end transport , but together with Bik1 , it can enhance Kip2's processivity and help Kip2 overcome dynein's intrinsic minus-end-directed motility . As depicted with dotted lines , it is likely that the C-terminal tail of Bim1 interacts with the N-terminal Cap-Gly domain of Bik1 ( Weisbrich et al . , 2007 ) , and the cargo-binding domain of Bim1 interacts with the SxIP motif of Kip2 ( Honnappa et al . , 2009 ) . This arrangement would leave the calponin-homolgy domain of Bim1 ( Slep and Vale , 2007 ) free to interact transiently with the microtubule lattice , and thus promote Kip2's microtubule association . A single copy of each molecule is shown for clarity; interplay with the numerous other + TIPs in S . cerevisiae ( not shown ) is expected . DOI: http://dx . doi . org/10 . 7554/eLife . 02641 . 015 Our experiments show that dynein has the potential to resist its plus-end transport through its microtubule-binding domain . We also find that regulatory factors can tune the efficiency with which Kip2 ‘wins’ a tug-of-war against dynein . Bik1 and Bim1 can promote Kip2's microtubule interactions and help this kinesin to overcome dynein's intrinsic minus-end-directed motility . The effect of these factors is recapitulated when dynein and Kip2 are connected synthetically using DNA origami , suggesting that it is mechanical in its basis and not contingent on a specific coupling geometry . Moreover , our biochemical demonstration of Kip2 phosphorylation suggests that additional layers of regulation may exist . Specifically , the addition and removal of phosphate groups on the N-terminal extension of Kip2's motor domain has the potential to influence its interaction with Bim1 ( Figure 3A; Honnappa et al . , 2009 ) , as well as the negatively charged microtubule surface , both of which could impact Kip2 motility . It remains to be seen if additional factors mitigate resistance from dynein during its plus-end transport in vivo , akin to the mutations in dynein's microtubule-binding domain that we found to enhance plus-end targeting in vitro . For example , dynein's tail domain is crucial for events downstream of plus-end targeting ( Markus et al . , 2009 ) , but it is conceivable that the tail and its associated factors regulate the transport process as well . Once dynein reaches the plus end , the high local concentration of Lis1 at plus ends in S . cerevisiae ( Markus et al . , 2011 ) may help retain dynein by prolonging its microtubule attachments ( Huang et al . , 2012; McKenney et al . , 2010; Yamada et al . , 2008 ) . Nudel , a Lis1 and dynein binding partner , may also promote dynein plus-end targeting by tethering Lis1 to dynein ( Li et al . , 2005; McKenney et al . , 2010; Wang and Zheng , 2011; Zylkiewicz et al . , 2011; Huang et al . , 2012 ) . Notably , genetic studies indicate there is a second pathway for targeting dynein directly from the cytoplasm to the plus end that does not require Kip2 ( Caudron et al . , 2008; Markus et al . , 2009 ) . Overlap between these pathways in vivo may explain why Bim1's capacity to regulate dynein plus-end transport was previously hidden . In summary , these results provide a vivid example of how the interplay between dynein , kinesin and microtubule-associated proteins can give rise to a coherent behavior: the recycling of a molecular motor to the start of its track .
Proteins were analyzed by SDS-PAGE on 4–12% Tris-Bis gels with Sypro Red staining ( Invitrogen; Carlsbad , CA ) , and imaged using an ImageQuant 300 gel imaging system ( Bio-Rad; Hercules , CA ) . Protein concentrations were determined by comparisons with standards using Bradford protein assays . All protein concentrations are expressed for the monomer , with the exception of α/β-tubulin , for which the dimer concentration is given . To verify Kip2 phosphorylation , purified Kip2 ( 2 μg ) was incubated ±280 units of λ phosphatase ( NEB ) with 1 mM MnCl2 in PMP NEBuffer for 30 min at 25°C . With the exception of Figure 1A ( inset ) , the Kip2 used in all experiments was not treated with phosphatase . For size-exclusion chromatography , indicated combinations of purified protein ( 100 picomoles each ) were pre-incubated at a concentration of 1 . 1 μM for 10 min at 4°C . Samples were fractionated on a Superose 6 PC 3 . 2/30 column using an ÄKTAmicro system ( GE Healthcare; Piscataway , NJ ) that had been equilibrated with gel filtration buffer ( 50 mM Tris–HCl [pH 8 . 0] , 150 mM potassium acetate , 2 mM magnesium acetate , 1 mM EGTA , 5% glycerol , and 1 mM DTT ) . Fractions ( 50 μl ) were analyzed by SDS-PAGE . Taxol-stabilized microtubules containing 10% Alexa488-tubulin and biotin-tubulin were prepared using standard methods ( http://mitchison . med . harvard . edu/protocols . html ) . Flow chambers for total internal reflection fluorescence ( TIRF ) microscopy were assembled using biotin-PEG cover glasses from MicroSurfaces , Inc . ( Englewood , NJ ) or prepared as described ( Bieling et al . , 2010 ) . Chambers were incubated sequentially with the following solutions , interspersed with two washes with assay buffer ( BRB80 [80 mM PIPES-KOH pH 6 . 8 , 1 mM MgCl2 , 1 mM EGTA] , 0 . 5 mg/ml casein and 1 mM DTT ) with 20 μM taxol: ( 1 ) 0 . 5 mg/ml streptavidin in BRB80 ( 4 min incubation ) ; ( 2 ) a 1:100 dilution of microtubule solution ( 2 min incubation ) . Finally , TMR-labeled dynein ( 2 . 5 nM ) was added with Lis1 ( 50 nM ) , Bik1 ( 50 nM ) , Bim1 ( 5 nM ) and Kip2 ( 2 . 5 nM ) as indicated . The final reaction solution contained no nucleotide , 20 μM taxol and an oxygen scavenging system ( Yildiz et al . , 2003 ) in assay buffer . Alexa488-microtubules and dynein-TMR were visualized using an Olympus IX-81 TIRF microscope with a 100x 1 . 45 N . A . oil immersion TIRF objective and CW 491 nm and 561 nm lasers , controlled by Metamorph software ( Qiu et al . , 2012 ) . Images were recorded with a 100 ms exposure on a back-thinned electron multiplier CCD camera , giving 159 nm/pixel at the specimen level . All conditions were imaged with identical microscope settings ( 491 nm laser power: 0 . 65 mW; 561 nm laser power: 0 . 35 mW; EMCCD gain: 150 ) . Dynein fluorescence intensity on the microtubule was quantified using ImageJ . Background signal was subtracted using a rolling ball radius of 5 pixels . Intensities were determined over a 20-pixel wide line drawn perpendicular to the long axis of the microtubule , and averaged for n = 30–50 microtubules in each condition . Dynamic microtubule assays and visualization by TIRF microscopy were performed essentially as described ( Bieling et al . , 2010 ) . Brightly-labeled , biotinylated microtubule seeds were polymerized by mixing Alexa488-tubulin ( 10 μM ) , biotin-tubulin ( 10 μM ) and unlabeled tubulin ( 10 μM ) with 0 . 5 mM GMP-CPP ( Jena Bioscience ) in BRB80 and incubating for 30 min at 37°C . Following the addition of 10 vol of BRB80 , polymerized seeds were pelleted in a benchtop centrifuge ( 15 min at 16 , 100×g ) and resuspended in a volume of BRB80 equal to the original polymerization volume . Flow chambers for TIRF microscopy were assembled using biotin-PEG cover glasses . Chambers were incubated sequentially with the following solutions , interspersed with two washes with assay buffer: ( 1 ) 1% plurionic F-127 and 5 mg/ml casein in BRB80 ( 8 min incubation ) ; ( 2 ) 0 . 5 mg/ml streptavidin in BRB80 ( 4 min incubation ) ; ( 3 ) a fresh 1:2000 dilution of microtubule seed solution ( 2 min incubation ) . The final reaction solution contained 1 mM Mg-ATP , 1 mM GTP , 0 . 1% methylcellulose , an oxygen scavenging system , and 15 μM tubulin ( 7 . 5% Alexa488 labeled , 92 . 5% unlabeled ) in assay buffer . Dynein-TMR ( 0 . 04–0 . 3 nM ) , Lis1 ( 50 nM ) , Bim1 ( 5 nM ) , Bik1 ( 50 nM ) and Kip2-Atto647 ( 1 nM ) were added as indicated . After sealing with vacuum grease , flow chambers were imaged immediately . Three-color TIRF movies ( capturing Alexa488 , TMR and Atto647 fluorescence ) were recorded with 3 s intervals per channel for 10 min using the microscopy setup described above and in Qiu et al . ( 2012 ) . Microtubule plus ends were assigned based on two criteria , which were consistent: ( 1 ) the direction of Kip2 movement and ( 2 ) the microtubule end with faster growth rate . Only motility on the dimly labeled plus-end extension of the microtubule was analyzed . Dynein velocities and directionalities were calculated from kymographs using an ImageJ macro ( provided in Supplementary file 1 ) . The density of Kip2 on the microtubule precluded measuring the velocity of every run in each kymograph . Hence the velocity of a subset of discrete runs was determined as an estimate of the population velocity . These velocities match closely those of Kip2 at single-molecule concentrations in equivalent conditions ( Figure 3—figure supplement 1 ) . In some cases , minor image drift was corrected using the ImageJ plugin Turboreg . Movies showing significant drift were discarded . Flow chambers were assembled using cover glasses ( Corning no . 1½ ) and Taxol-stabilized microtubules containing 10% Alexa488- and biotin-tubulin were immobilized on the glass surface with Biotin-BSA and streptavidin as described ( Derr et al . , 2012; Huang et al . , 2012 ) . Atto647-labeled Kip2 ( 0 . 01–0 . 05 nM ) was added with Bik1 ( 100 nM ) and Bim1 ( 5 nM ) as indicated , in BRB80 supplemented with 1 mM Mg-ATP , 1 mM DTT , 20 μM taxol , 2 . 5 mg/ml casein , and an oxygen scavenging system . Motility assays were imaged as described ( Derr et al . , 2012; Huang et al . , 2012 ) . Motor velocities and run lengths were calculated from kymographs . For landing rate quantification , Kip2 was analyzed at 0 . 01 and 0 . 02 nM in the presence or absence of Bik1 ( 100 nM ) and Bim1 ( 5 nM ) . Landing events were determined from kymographs and the rate was calculated as the number of events per micron of microtubule per nM Kip2 per minute . For visualizing the colocalization of weak dynein with Kip2 , assay chambers were prepared as for dynamic microtubule assays except taxol-stabilized microtubules were attached to the biotin-PEG surface instead of GMP-CPP stabilized seeds , and all buffers contained 20 μM taxol . The final reaction contained the weak dynein construct ( 2 nM ) , Lis1 ( 2 nM ) , Bik1 ( 2 nM ) and Kip2-Atto647 ( 0 . 2 nM ) . The 12 helix-bundle DNA origami ‘chassis’ was made as described ( Derr et al . , 2012 ) , by rapidly heating and slowly cooling an 8064-nucleotide , single-strand DNA ( ssDNA ) ‘scaffold’ in the presence of short ssDNA ‘staples’ that base-pair with the scaffold to fold it into a desired shape . The folded chassis was purified by glycerol gradient centrifugation as described ( Derr et al . , 2012 ) . The chassis contained five TAMRA-labeled ssDNAs for fluorescent visualization . The chassis also had two projecting ‘handle’ sequences of ssDNA at opposite ends , termed ‘A’ and ‘B’ , which were used to attach DNA-labeled dynein and Kip2 respectively . The sequences for the scaffold and all oligonucleotides are listed in Derr et al . ( 2012 ) . Motor-chassis complexes were assembled by incubating dynein labeled with oligo A′ ( complementary to handle A ) and Kip2 labeled with oligo B′ ( complementary to handle B ) with the chassis for 30 min on ice . Complex assembly was verified by agarose gel shift ( Derr et al . , 2012 ) . Chassis motility was visualized in the presence of Bik1 ( 100 nM ) and Bim1 ( 5 nM ) on taxol-stabilized microtubules as indicated . To determine microtubule polarity , a low concentration ( ∼0 . 1 nM ) of a highly processive dynein mutant ( E3197K ) labeled with a different fluorophore ( Atto647 ) was included . Chassis velocities and directionalities were determined from kymographs using ImageJ . To determine the fraction of motile events , the number of observations for a given event ( e . g . , the number of plus end runs ) was tallied and divided by the total number of observations . Fractions are expressed as the average percentage ± SEM for three separate dynein/Kip2-chassis assembly reactions , with 306–353 runs analyzed in each case . | Eukaryotic cells use transport systems to efficiently move materials from one location to another . Much transport in the cell interior is achieved using molecular motors , which carry cargoes along tracks called microtubules . Unlike roads of human construction , microtubules are very dynamic . One of their ends ( the ‘plus’ end ) explores the outskirts of the cell , growing and shrinking through the addition and loss of protein building blocks . The other microtubule end ( the ‘minus’ end ) typically lies in a hub near the center of the cell . There are two types of molecular motor that move on microtubules . Kinesin motors move toward the plus end of the microtubule , and dynein motors move in the opposite direction , toward the minus end . But if dynein only moves to the minus end of the microtubule , a problem arises: how would dynein initially reach the plus end of the microtubule and the outskirts of the cell , where it collects cargoes ? Using purified yeast proteins , Roberts et al . reveal that a group of three proteins can solve this problem by transporting dynein to the plus end of the microtubule . The proteins comprise a kinesin motor , and two additional proteins that connect the dynein motor to the kinesin . Imaging the transport process shows that the dynein motor is not a passive passenger: it is able to resist against the kinesin . However , an additional microtubule-associated protein can help the kinesin motor to win this ‘tug of war’ , and so the protein complex—including the dynein motor—moves toward the plus end of the microtubule . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2014 | Reconstitution of dynein transport to the microtubule plus end by kinesin |
The cross-talk between dynamic microtubules and integrin-based adhesions to the extracellular matrix plays a crucial role in cell polarity and migration . Microtubules regulate the turnover of adhesion sites , and , in turn , focal adhesions promote the cortical microtubule capture and stabilization in their vicinity , but the underlying mechanism is unknown . Here , we show that cortical microtubule stabilization sites containing CLASPs , KIF21A , LL5β and liprins are recruited to focal adhesions by the adaptor protein KANK1 , which directly interacts with the major adhesion component , talin . Structural studies showed that the conserved KN domain in KANK1 binds to the talin rod domain R7 . Perturbation of this interaction , including a single point mutation in talin , which disrupts KANK1 binding but not the talin function in adhesion , abrogates the association of microtubule-stabilizing complexes with focal adhesions . We propose that the talin-KANK1 interaction links the two macromolecular assemblies that control cortical attachment of actin fibers and microtubules .
Cell adhesions to the extracellular matrix support epithelial integrity and cell migration , and also provide signaling hubs that coordinate cell proliferation and survival ( Hynes , 1992 ) . Integrin-based adhesions ( focal adhesions , FAs ) are large macromolecular assemblies , in which the cytoplasmic tails of integrins are connected to the actin cytoskeleton . One of the major components of FAs is talin , a ~2500 amino acid dimeric protein , which plays a key role in adhesion formation by activating integrins ( Anthis et al . , 2009 ) , coupling them to cytoskeletal actin ( Atherton et al . , 2015 ) , regulating adhesion dynamics and recruiting different structural and signaling molecules ( Calderwood et al . , 2013; Gardel et al . , 2010; Wehrle-Haller , 2012 ) . While the major cytoskeletal element associated with FAs is actin , microtubules also play an important role in adhesion by regulating the FA turnover ( Akhmanova et al . , 2009; Byron et al . , 2015; Kaverina et al . , 1999 , 1998; Krylyshkina et al . , 2003; Small and Kaverina , 2003; Stehbens and Wittmann , 2012; Yue et al . , 2014 ) . The recent proteomics work showed that microtubule-FA cross-talk strongly depends on the activation state of the integrins ( Byron et al . , 2015 ) . Microtubules can affect adhesions by serving as tracks for delivery of exocytotic carriers ( Stehbens et al . , 2014 ) , by controlling endocytosis required for adhesion disassembly ( Ezratty et al . , 2005; Theisen et al . , 2012 ) and by regulating the local activity of signaling molecules such as Rho GTPases ( for review , see [Kaverina and Straube , 2011; Stehbens and Wittmann , 2012] ) . In contrast to actin , which is directly coupled to FAs , microtubules interact with the plasma membrane sites that surround FAs . A number of proteins have been implicated in microtubule attachment and stabilization in the vicinity of FAs . Among them are the microtubule plus end tracking proteins ( +TIPs ) CLASP1/2 and the spectraplakin MACF1/ACF7 , which are targeted to microtubule tips by EB1 , and a homologue of EB1 , EB2 , which binds to mitogen-activated protein kinase kinase kinase kinase 4 ( MAP4K4 ) ( Drabek et al . , 2006; Honnappa et al . , 2009; Kodama et al . , 2003; Mimori-Kiyosue et al . , 2005 ) . The interaction of CLASPs with the cell cortex depends on the phosphatidylinositol 3 , 4 , 5-trisphosphate ( PIP3 ) -interacting protein LL5β , to which CLASPs bind directly , and is partly regulated by PI-3 kinase activity ( Lansbergen et al . , 2006 ) . Other components of the same cortical assembly are the scaffolding proteins liprin-α1 and β1 , a coiled-coil adaptor ELKS/ERC1 , and the kinesin-4 KIF21A ( Lansbergen et al . , 2006; van der Vaart et al . , 2013 ) . Both liprins and ELKS are best known for their role in organizing presynaptic secretory sites ( Hida and Ohtsuka , 2010; Spangler and Hoogenraad , 2007 ) ; in agreement with this function , ELKS is required for efficient constitutive exocytosis in HeLa cells ( Grigoriev et al . , 2007 , 2011 ) . LL5β , liprins and ELKS form micrometer-sized cortical patch-like structures , which will be termed here cortical microtubule stabilization complexes , or CMSCs . The CMSCs are strongly enriched at the leading cell edges , where they localize in close proximity of FAs but do not overlap with them ( [Lansbergen et al . , 2006; van der Vaart et al . , 2013] , reviewed in [Astro and de Curtis , 2015] ) . They represent a subclass of the previously defined plasma membrane-associated platforms ( PMAPs ) ( Astro and de Curtis , 2015 ) , which have overlapping components such as liprins , but may not be necessarily involved in microtubule regulation , as is the case for liprin-ELKS complexes in neurons , where they are part of cytomatrix at the active zone ( Gundelfinger and Fejtova , 2012 ) . Several lines of evidence support the importance of the CMSC-FA cross-talk . In migrating keratinocytes , LL5β and CLASPs accumulate around FAs and promote their disassembly by targeting the exocytosis of matrix metalloproteases to FA vicinity ( Stehbens et al . , 2014 ) . Furthermore , liprin-α1 , LL5α/β and ELKS localize to protrusions of human breast cancer cells and are required for efficient cell migration and FA turnover ( Astro et al . , 2014 ) . In polarized epithelial cells , LL5β and CLASPs are found in the proximity of the basal membrane , and this localization is controlled by the integrin activation state ( Hotta et al . , 2010 ) . CLASP and LL5-mediated anchoring of MTs to the basal cortex also plays a role during chicken embryonic development , where it prevents the epithelial-mesenchymal transition of epiblast cells ( Nakaya et al . , 2013 ) . LL5β , CLASPs and ELKS were also shown to concentrate at podosomes , actin-rich structures , which can remodel the extracellular matrix ( Proszynski and Sanes , 2013 ) . Interestingly , LL5β-containing podosome-like structures are also formed at neuromuscular junctions ( Kishi et al . , 2005; Proszynski et al . , 2009; Proszynski and Sanes , 2013 ) , and the complexes of LL5β and CLASPs were shown to capture microtubule plus ends and promote delivery of acetylcholine receptors ( Basu et al . , 2015 , 2014; Schmidt et al . , 2012 ) . While the roles of CMSCs in migrating cells and in tissues are becoming increasingly clear , the mechanism underlying their specific targeting to integrin adhesion sites remains elusive . Recently , we found that liprin-β1 interacts with KANK1 ( van der Vaart et al . , 2013 ) , one of the four members of the KANK family of proteins , which were proposed to act as tumor suppressors and regulators of cell polarity and migration through Rho GTPase signaling ( Gee et al . , 2015; Kakinuma et al . , 2008 , 2009; Li et al . , 2011; Roy et al . , 2009 ) . KANK1 recruits the kinesin-4 KIF21A to CMSCs , which inhibits microtubule polymerization and prevents microtubule overgrowth at the cell edge ( Kakinuma and Kiyama , 2009; van der Vaart et al . , 2013 ) . Furthermore , KANK1 participates in clustering of the other CMSC components ( van der Vaart et al . , 2013 ) . Here , we found that KANK1 is required for the association of the CMSCs with FAs . The association of KANK1 with FAs depends on the KN domain , a conserved 30 amino acid polypeptide sequence present in the N-termini of all KANK proteins . Biochemical and structural analysis showed that the KN domain interacts with the R7 region of the talin rod . Perturbation of this interaction both from the KANK1 and the talin1 side prevented the accumulation of CMSC complexes around focal adhesions and affected microtubule organization around FAs . We propose that KANK1 molecules , recruited by talin to the outer rims of FA , serve as 'seeds' for organizing other CMSC components in the FA vicinity through multivalent interactions between these components . This leads to co-organization of two distinct cortical assemblies , FAs and CMSCs , responsible for the attachment of actin and microtubules , respectively , and ensures effective cross-talk between the two types of cytoskeletal elements .
Our previous work showed that the endogenous KANK1 colocalizes with LL5β , liprins and KIF21A in cortical patches that are closely apposed to , but do not overlap with FAs ( van der Vaart et al . , 2013 ) . We confirmed these results both in HeLa cells and the HaCaT immortal keratinocyte cell line , in which CMSC components CLASPs and LL5β were previously shown to strongly cluster around FAs and regulate their turnover during cell migration ( Stehbens et al . , 2014 ) ( Figure 1—figure supplement 1A , B ) . Inhibition of myosin-II with blebbistatin , which reduces tension on the actin fibers and affects the activation state of FA molecules , such as integrins and talin ( Parsons et al . , 2010 ) , caused not only FA disassembly but also a strong reduction in clustering of CMSC components at the cell periphery ( Figure 1—figure supplement 2A , B ) , as described previously ( Stehbens et al . , 2014 ) . To investigate this effect in more detail , we partially inhibited contractility using a Rho-associated protein kinase 1 ( ROCK1 ) inhibitor , Y-27632 ( Oakes et al . , 2012 ) . In these conditions , the number of FAs was not affected although their size was reduced ( Figure 1—figure supplement 2C–E ) . This treatment was sufficient to diminish CMSC clustering at the cell edge ( Figure 1—figure supplement 2C , F ) . Interestingly , at the same time we observed a very significant increase in the overlap of KANK1 with FA adhesion markers ( Figure 1—figure supplement 2C , G ) . Live imaging of KANK1 together with the FA marker paxillin showed a gradual redistribution of KANK1 into the areas occupied by FAs upon ROCK1 inhibitor-induced attenuation of contractility ( Figure 1—figure supplement 2H , Video 1 ) . These data indicate that the organization of CMSCs at the cell cortex might be controlled by interactions with tension-sensitive components of FAs . 10 . 7554/eLife . 18124 . 003Figure 1 . The KN motif of KANK1 interacts with the R7 domain of talin1 . ( A ) Schematic representation of KANK1 and the deletion mutants used in this study , and the summary of their interactions and localization . N . d . , not determined in this study . ( B ) TIRFM images of live HeLa cells transiently expressing the indicated GFP-tagged KANK1 deletion mutants together with the focal adhesion marker mCherry-paxillin . In these experiments , endogenous KANK1 and KANK2 were also expressed . ( C ) Identification of the binding partners of Bio-GFP-tagged KANK1 and its indicated deletion mutants by using streptavidin pull down assays from HEK293T cells combined with mass spectrometry . ( D ) Streptavidin pull down assays with the BioGFP-tagged KANK1 or the indicated KANK1 mutants , co-expressed with GFP-talin1 in HEK293T cells , analyzed by Western blotting with the indicated antibodies . ( E ) Sequence alignment of KANK1 and KANK2 with the known talin-binding sites of DLC1 , RIAM and Paxillin . The LD-motif and the interacting hydrophobic residues are highlighted green and blue respectively . ( F ) Schematic representation of talin1 and the deletion mutants used in this study , and their interaction with KANK1 . ( G ) Streptavidin pull down assays with the BioGFP-tagged talin1 or the indicated talin1 mutants , co-expressed with HA-KANK1 in HEK293T cells , analyzed by Western blotting with the indicated antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 18124 . 00310 . 7554/eLife . 18124 . 004Figure 1—figure supplement 1 . KANK1 colocalizes with CMSC components around FAs . ( A ) Widefield fluorescence images of HeLa cells stained for endogenous proteins as indicated . In the two bottom panels , cells were transfected with GFP-KANK1 . pY , phospho-tyrosine , a FA marker . ( B ) Widefield fluorescence images of HaCaT cells stained for endogenous proteins as indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 18124 . 00410 . 7554/eLife . 18124 . 005Figure 1—figure supplement 2 . Role of myosin II activity in KANK1 localization to FA . ( A ) Widefield fluorescence images of serum-starved HeLa cells stimulated with fetal calf serum with or without blebbistatin and stained as indicated . ( B ) Quantification of peripheral clustering of LL5β in cells treated as in panel ( A ) ( n=25–30 , 6 cells per condition ) . ( C ) Widefield fluorescence images of HeLa treated for 45 min with the ROCK1 inhibitor Y-27632 at indicated concentrations and stained as indicated . ( D–E ) Average FA number ( D ) and individual area ( E ) in cells treated as in ( C ) ( 373–642 FAs/ condition averaged per cell; n=7 ) . ( F ) Quantification of peripheral clustering of KANK1 in cells treated as in panel ( C ) ( n=12 , 5 cells per condition ) . ( G ) Quantification of KANK1 and talin colocalization . Pearson R value , n=30 in 10–12 cells per conditions . ( H ) Dual fluorescence time-lapse images acquired using TIRFM in HeLa cells stably expressing TagRFP-paxillin and GFP-KANK1 and treated as indicated . Red line , FA rim obtained by using a threshold-based mask . In all plots: error bar , SEM; ns , non-significant , ***p<0 . 001 , Mann- Whitney U test . DOI: http://dx . doi . org/10 . 7554/eLife . 18124 . 00510 . 7554/eLife . 18124 . 006Figure 1—figure supplement 2—source data 1 . An Excel sheet with numerical data on the quantification of peripheral clustering of different markers , FA number and area and colocalization of KANK1 with talin represented as plots in Figure 1—figure supplement 2B , D–G . DOI: http://dx . doi . org/10 . 7554/eLife . 18124 . 00610 . 7554/eLife . 18124 . 007Figure 1—figure supplement 3 . FA localization of KN-bearing proteins . GFP-KN-LacZ fusion localizes to the inner part of FAs . TIRFM images of live HeLa cells transfected with TagRFP-paxillin and GFP-tagged KN peptide , ΔANKR KANK1 mutant or KN-LacZ . DOI: http://dx . doi . org/10 . 7554/eLife . 18124 . 00710 . 7554/eLife . 18124 . 008Video 1 . Effect of myosin II inhibition on KANK1 localization to FA . TIRFM-based time-lapse imaging of HeLa cells stably expressing GFP-KANK1 and TagRFP-paxillin and treated when indicated with 10 μM ROCK1 inhibitor Y-27632 . Both red and green fluorescence images were acquired at 1 min interval and displayed at 15 frames/second ( accelerated 900 times ) . DOI: http://dx . doi . org/10 . 7554/eLife . 18124 . 008 To identify the domains of KANK1 required for cortical localization , we performed deletion mapping . KANK1 comprises an N-terminal KANK family-specific domain of unknown function , the KN domain ( residues 30–68 ) ( Kakinuma et al . , 2009 ) , a coiled coil region , the N-terminal part of which interacts with liprin-β1 , and a C-terminal ankyrin repeat domain , which binds to KIF21A ( van der Vaart et al . , 2013 ) , while the rest of the protein is predicted to be unstructured ( Figure 1A ) . Surprisingly , the KN domain alone strongly and specifically accumulated within FAs ( Figure 1B ) . A similar localization was also seen with a somewhat larger N-terminal fragment of KANK1 , Nter , as well as the Nter-CC1 deletion mutant , which contained the first , liprin-β1-binding coiled coil region of KANK1 ( Figure 1A , B ) . However , an even larger N-terminal part of KANK1 , encompassing the whole coiled coil domain ( Nter-CC ) showed a pronounced enrichment at the FA rim ( Figure 1A , B ) . The KANK1 deletion mutant missing only the C-terminal ankyrin repeat domain ( △ANKR ) was completely excluded from FAs but accumulated in their immediate vicinity , similar to the full-length KANK1 ( Figure 1A , B ) . A tight ring-like localization at the outer rim of FAs was also observed with a KANK1 mutant , which completely missed the coiled coil region but contained the ankyrin repeat domain ( △CC ) , while the mutant which missed just the KN domain showed no accumulation around FAs ( Figure 1A , B ) . To test whether the exclusion of larger KANK1 fragments from the FA core was simply due to the protein size , we fused GFP-tagged KN domain to the bacterial protein β-D-galactosidase ( LacZ ) , but found that this fusion accumulated inside and not around FAs ( Figure 1—figure supplement 3 ) . Since GFP-KN-LacZ and GFP-KANK1-△ANKRD have a similar size ( 1336 and 1400 amino acids , respectively ) , but one accumulates inside FAs , while the other is excluded to their periphery , this result suggests that features other than the mere protein size determine the specific localization of KANK1 to the FA rim . We conclude that the KN domain of KANK1 has an affinity for FAs , but the presence of additional KANK1 sequences prevents the accumulation of the protein inside FAs and instead leads to the accumulation of KANK1 at the FA periphery . To identify the potential FA-associated partners of KANK1 , we co-expressed either full-length KANK1 or its N-terminal and C-terminal fragments fused to GFP and a biotinylation ( Bio ) tag together with biotin ligase BirA in HEK293T cells and performed streptavidin pull down assays combined with mass spectrometry . In addition to the already known binding partners of KANK1 , such as KIF21A , liprins and LL5β , we identified talin1 among the strongest hits ( Figure 1C ) . Talin2 was also detected in a pull down with the KANK1 N-terminus though not with the full-length protein ( Figure 1C ) . The interaction between KANK1 and talin1 was confirmed by Western blotting , and subsequent deletion mapping showed that the talin1-binding region of KANK1 encompasses the KN domain ( Figure 1A , D ) , while liprin-β1 binds to the N-terminal part of the coiled coil domain , as shown previously ( van der Vaart et al . , 2013 ) . Sequence analysis of the KN domain showed that it is predicted to form a helix and contains a completely conserved leucine aspartic acid-motif ( LD-motif ) ( Alam et al . , 2014; Zacharchenko et al . , 2016 ) . The LD-motifs in RIAM ( Goult et al . , 2013 ) , DLC1 and Paxillin ( Zacharchenko et al . , 2016 ) have recently been identified as talin-binding sites that all interact with talin via a helix addition mechanism . Alignment of the KN domain of KANK with the LD-motif of DLC1 , RIAM and Paxillin ( Zacharchenko et al . , 2016 ) revealed that the hydrophobic residues that mediate interaction with talin are present in the KN domain ( Figure 1E ) . Using deletion analysis , we mapped the KANK1-binding site of talin1 to the central region of the talin rod , comprising the R7-R8 domains ( Figure 1F ) . This R7-R8 region of talin is unique ( Gingras et al . , 2010 ) , as the 4-helix bundle R8 is inserted into a loop of the 5-helix bundle R7 , and thus protrudes from the linear chain of 5-helix bundles of the talin rod ( Figures 1F , 2A ) . This R8 domain serves as a binding hub for numerous proteins including vinculin , synemin and actin ( Calderwood et al . , 2013 ) . R8 also contains a prototypic recognition site for LD-motif proteins , including DLC1 ( Figure 2B ) , Paxillin and RIAM ( Zacharchenko et al . , 2016 ) . Based on the presence of the LD-binding site , we anticipated that KANK1 would also interact with R8 . However , deletion mapping revealed that KANK1 in fact binds to the talin1 rod domain R7 ( Figure 1F , G ) , suggesting that KANK1 interacts with a novel binding site on talin1 . 10 . 7554/eLife . 18124 . 009Figure 2 . Biochemical and structural characterization of the Talin-KANK interaction . ( A ) Schematic representation of Talin1 , with F-actin , β-integrin and vinculin binding sites highlighted . The KANK1 binding site on R7 is also shown . ( B ) The structure of the complex between talin1 R7-R8 ( white ) and the LD-motif of DLC1 ( yellow ) bound on the R8 subdomain ( PDB ID . 5FZT , [Zacharchenko et al . , 2016] ) . Residues W1630 and Y1389 ( blue ) and S1641 ( magenta ) are highlighted . ( C–D ) The KANK KN domain binds to a novel site on talin R7 . 1H , 15N HSQC spectra of 150 μM 15N-labelled talin1 R7 ( residues 1357–1659 Δ1454–1586 ) in the absence ( black ) or presence of KANK1 ( 30–68 ) C peptide ( red ) ( top panel ) or KANK1-4A ( green ) ( bottom panel ) at a ratio of 1:3 . ( D ) Mapping of the KANK1 binding site on R7 as detected by NMR using weighted chemical shift differences ( red ) – mapped onto the R7-R8 structure in ( B ) . Residues W1630 and Y1389 ( blue ) and G1404 and S1641 ( magenta ) are highlighted . ( E ) Structural model of the talin1:KANK1 interface . Ribbon representation of the KANK1 binding site , comprised of the hydrophobic groove between helices 29 and 36 of R7 . Two bulky conserved residues , W1630 and Y1389 ( blue ) hold these two helices apart forming the binding interface . A small glycine side chain ( G1404 ) creates a pocket between the helices . S1641 ( magenta ) has been shown previously to be a phosphorylation site ( Ratnikov et al . , 2005 ) . The KN peptide ( green ) docked onto the KANK binding surface . ( F–G ) Biochemical characterization of the talin:KANK interaction . ( F ) Binding of BODIPY-labeled KANK1 ( 30–60 ) C , KANK2 ( 31–61 ) C and KANK1-4A peptides to Talin1 R7-R8 ( 1357–1659 ) was measured using a Fluorescence Polarization assay . ( G ) Binding of BODIPY-labeled KANK1 ( 30–60 ) C to wild type R7-R8 , R7-R8 S1641E , R7-R8 G1404L and R7-R8 W1630A . Dissociation constants ± SE ( μM ) for the interactions are indicated in the legend . All measurements were performed in triplicate . ND , not determined . DOI: http://dx . doi . org/10 . 7554/eLife . 18124 . 00910 . 7554/eLife . 18124 . 010Figure 2—figure supplement 1 . NMR validation of the Talin1 R7-R8 mutants . 1H , 15N HSQC spectra of; ( A ) Talin1 R7-R8 domain , ( B ) G1404L Talin1 R7-R8 mutant , and ( C ) W1630A Talin1 R7-R8 Mutant . The mutant spectra show good peak distribution with uniform intensity similar to the wildtype suggesting that they are correctly folded . DOI: http://dx . doi . org/10 . 7554/eLife . 18124 . 01010 . 7554/eLife . 18124 . 011Figure 2—figure supplement 2 . Biochemical characterization of the Talin2:KANK interaction . Binding of BODIPY-labeled KANK1 ( 30–60 ) C , KANK2 ( 31–61 ) C and KANK1-4A peptides to Talin2 R7-R8 measured using a Fluorescence Polarization assay . Dissociation constants ± SE ( μM ) for the interactions are indicated in the legend . ND , not determined . DOI: http://dx . doi . org/10 . 7554/eLife . 18124 . 011 To explore the interaction between talin1 and KANK1 in more detail , we used NMR chemical shift mapping using 15N-labeled talin1 R7-R8 ( residues 1357–1653 ) and a synthetic KANK1 peptide of the KN domain , KANK1 ( 30–60 ) . The addition of the KANK1 ( 30–60 ) peptide resulted in large spectral changes ( Figure 2C ) , most of which were in the slow exchange regime on the NMR timescale indicative of a tight interaction . In agreement with the pull down data , the signals that shifted in slow exchange upon the addition of KANK1 ( 30–60 ) mapped largely onto the R7 domain with only small progressive shift changes evident on R8 . To validate R7 as the major KANK1-binding site on talin , we repeated the NMR experiments using the individual domains , R8 ( residues 1461–1580 ) and R7 ( residues 1359–1659 Δ1454–1586 ) . Addition of KANK1 ( 30–60 ) induced small chemical shift changes on the R8 domain indicative of a weak interaction ( most likely due to the interaction of LD with the LD-recognition box on R8 ) . However , the addition of a 0 . 5 molar ratio of KANK1 ( 30–60 ) to R7 induced large spectral changes with many of the peaks displaying two locations , corresponding to the free peak position and the bound peak position . This is indicative of slow-exchange and confirms a high affinity interaction between R7 and KANK1 . The KN peptide is the first identified ligand for the R7 domain . NMR chemical shifts also provide information on the residues involved in the interaction , as the peaks in the 15N-HSQC spectrum pertain to individual residues in the protein . To map these chemical shift changes onto the structure of R7-R8 , it was first necessary to complete the backbone chemical shift assignments of the R7 domain . This was achieved using conventional triple resonance experiments as described previously ( Banno et al . , 2012 ) , using 13C , 15N labeled R7 . The chemical shift changes mapped predominantly onto one face of R7 , comprised of helices 2 and 5 of the 5-helix bundle ( talin rod helices 29 and 36 ) , as shown in Figure 2D–E . Our recent elucidation of the interaction between the LD-motif of DLC1 and talin R8 has generated insight into how LD-motifs are recognized by helical bundles ( PDB ID . 5FZT , [Zacharchenko et al . , 2016] ) . In the DLC1:talin R8 complex the DLC1 peptide adopts a helical conformation that packs against two helices on the side of the helical bundle . It is becoming increasingly clear that other LD-motif proteins bind to talin through a similar interaction mode ( Zacharchenko et al . , 2016 ) . The surface of α2 and α5 on R7 forms a hydrophobic groove that the KANK1 helix docks into . A striking feature of this KANK1 binding surface is that the two helices are held apart by the conserved aromatic residues , W1630 at the end of α5 and Y1389 at the end of α2 ( Figure 2B , E ) . W1630 and Y1389 thus essentially serve as molecular rulers , separating helices α2 and α5 by ~8Å ( compared with ~5–6Å for the other bundles in R7-R8 ) . The spacing between the two helices is enhanced further as the residues on the inner helical faces , S1400 , G1404 , S1411 on α2 and S1637 and S1641 on α5 , have small side chains which have the effect of creating two conserved pockets midway along the hydrophobic groove of the KANK1-binding site ( Figure 2E ) . The talin-binding site on KANK1 is unusual as it contains a double LD-motif , LDLD . The structure of R7 revealed a potential LD-recognition box with the positive charges , K1401 and R1652 positioned on either side to engage either one , or both , of the aspartic residues . Using the docking program HADDOCK ( van Zundert et al . , 2016 ) , we sought to generate a structural model of the KANK1/R7 complex , using the chemical shift mapping on R7 and a model of KANK1 ( 30–60 ) as a helix ( created by threading the KANK1 sequence onto the DLC1 LD-motif helix ) . This analysis indicated that the KANK-LD helix can indeed pack against the sides of α2 and α5 ( Figure 2E ) . Interestingly , all of the models , irrespective of which way the KANK1 helix ran along the surface , positioned the bulky aromatic residue , Y48 in KANK1 , in the hydrophobic pocket created by G1404 . This raised the possibility that mutation of G1404 to a bulky hydrophobic residue might block KANK1 binding by preventing Y48 engagement . We also noticed that S1641 , one of the small residues that create the pocket , has been shown to be phosphorylated in vivo ( Ratnikov et al . , 2005 ) and might have a regulatory function in the KANK1-talin1 interaction . To test these hypotheses , we generated a series of point mutants in talin R7 and also in the KANK1 KN-domain , designed to disrupt the talinR7/KANK1 interaction . On the KANK1 side , we mutated the LDLD motif to AAAA , ( the KANK1-4A mutant ) , while on the talin1 side , we generated a series of R7 mutants . These included G1404L , in which a bulky hydrophobic residue was introduced instead of glycine to occlude the hydrophobic pocket in R7 , S1641E , a phosphomimetic mutant aimed to test the role of talin phosphorylation in regulating KANK1 binding , and W1630A , a substitution that would remove one of the molecular rulers holding α2 and α5 helices apart at a fixed distance . These mutants were introduced into talin1 R7-R8 and the structural integrity of the mutated proteins confirmed using NMR ( Figure 2—figure supplement 1 ) . The relative binding affinities were measured using an in vitro fluorescence polarization assay . In this assay , the KANK1 ( 30–60 ) peptide is fluorescently labeled with BODIPY and titrated with an increasing concentration of talin R7-R8 , and the binding between the two polypeptides results in an increase in the fluorescence polarization signal ( Figure 2F ) . The KANK1-4A mutant abolished binding to talin ( Figure 2C , F ) . The S1641E mutant had only a small effect on binding ( Figure 2G ) , suggesting that either talin1 phosphorylation does not play a major role in modulating the interaction with KANK1 or that the S-E mutation is not a good phosphomimetic , possibly because phosphorylation might also affect helix formation integrity , an effect not mimicked by a single aspartate residue . However , strikingly , both the W1630A and the G1404L mutants abolished binding of KANK1 to talin R7 ( Figure 2G ) , confirming the validity of our model . Finally , we also tested whether the KN-R7 interaction is conserved in talin2 and KANK2 , and found that this was indeed the case ( Figure 2—figure supplement 2 ) . We conclude that the conserved KN domain of KANKs is a talin-binding site . Next , we set out to test the importance of the identified interactions in a cellular context by using the KANK1-4A and the talin G1404L mutants . We chose the G1404L talin mutant over W1630A for our cellular studies , because removing the bulky tryptophan from the hydrophobic core of the R7 might have the off target effect of perturbing the mechanical stability of R7 , and our recent studies showed that the mechanostability of R7 is important for protecting R8 from force-induced talin extension ( Yao et al . , 2016 ) . As could be expected based on the binding experiments with purified protein fragments , the 4A mutation reduced the interaction of the full-length KANK1 with talin1 in a pull-down assay ( Figure 3A ) . An even stronger reduction was observed when KANK-△CC or the KN alone were tested ( Figure 3A ) . Furthermore , the introduction of the G1404L mutation abrogated the interaction of full-length talin1 or its R7 fragment with full-length KANK1 ( Figure 3B ) . 10 . 7554/eLife . 18124 . 012Figure 3 . KANK1-talin interaction is required for recruiting KANK1 to FAs . ( A ) Streptavidin pull-down assays with the BioGFP-tagged KANK1 or the indicated KANK1 mutants , co-expressed with GFP-talin1 in HEK293T cells , analyzed by Western blotting with the indicated antibodies . ( B ) Streptavidin pull down assays with the BioGFP-tagged talin1 or the indicated talin1 mutants , co-expressed with HA-KANK1 in HEK293T cells , analyzed by Western blotting with the indicated antibodies . ( C ) TIRFM images of live HeLa cells depleted of KANK1 and KANK2 and co-expressing the indicated siRNA-resistant GFP-KANK1 fusions and TagRFP-paxillin . ( D ) Fluorescence profile of GFP-tagged mutants and TagRFP-paxillin based on line scan measurement across the FA area in TIRFM images as in panel ( C ) . ( E ) Widefield fluorescence images of HeLa cells depleted of endogenous talin1 and talin2 , rescued by the expression of the wild type GFP-tagged mouse talin1 or its G1404L mutant and labeled for endogenous KANK1 by immunofluorescence staining . ( F ) Quantification of peripheral clustering of KANK1 in cells treated and analyzed as in ( E ) ( n=12 , 6 cells per condition ) . Error bar , SEM; ***p<0 . 001 , Mann-Whitney U test . DOI: http://dx . doi . org/10 . 7554/eLife . 18124 . 01210 . 7554/eLife . 18124 . 013Figure 3—source data 1 . An Excel sheet with numerical data on the quantification of peripheral clustering of KANK1 represented as a plot in Figure 3F . DOI: http://dx . doi . org/10 . 7554/eLife . 18124 . 01310 . 7554/eLife . 18124 . 014Figure 3—figure supplement 1 . Validation of KANK1/2 and talin1/2 knockdown and effect of disrupted KANK/talin 1 binding in cell spreading and FA formation in HeLa cells . ( A–B ) Western blot analysis of KANK1/KANK2 ( A ) , and talin1/talin2 ( B ) co-depletion by independent siRNAs ( KANK1#1 , K1 #1 , KANK1#2 , K1 #2 and KANK2#1 , K2 #1; talin1 #2 , t1 #2 and talin2#1 , t2 #1 ) compared to control siRNA ( CT ) . Ku80 , loading control . ( C ) HeLa cells co-depleted of talin1 and talin2 were transfected with the indicated talin1 fusions and stained for the endogenous KANK1 . ( D ) Fluorescent F-actin staining in cells treated as in ( C ) . ( E–G ) Cell area ( E ) , FA count ( F ) and FA area ( G ) in cells treated as in ( C ) ( n=12 cells , 577–664 FAs analyzed ) . In all plots: error bar , SEM; ns , non-significant , *p<0 . 05 , Mann- Whitney U test . DOI: http://dx . doi . org/10 . 7554/eLife . 18124 . 01410 . 7554/eLife . 18124 . 015Figure 3—figure supplement 1—source data 1 . An Excel sheet with numerical data on the quantification of cell area , FA number and FA area represented as plots in Figure 3—figure supplement 1E–G . DOI: http://dx . doi . org/10 . 7554/eLife . 18124 . 015 To investigate the localization of KANK1-4A , we used HeLa cells depleted of endogenous KANK1 and KANK2 , the two KANK isoforms present in these cells based on our proteomics studies ( van der Vaart et al . , 2013 ) ( Figure 3—figure supplement 1A ) , in order to exclude the potential oligomerization of the mutant KANK1 with the endogenous proteins . Rescue experiments were performed using GFP-KANK1 , resistant for the used siRNAs due to several silent mutations ( van der Vaart et al . , 2013 ) , or its 4A mutant . We also included in this analysis a KANK1 mutant lacking the liprin-β1-binding coiled coil domain ( the △CC deletion mutant , Figure 1A ) , and the 4A-version of the KANK1-△CC deletion mutant . Total Internal Reflection Fluorescence Microscopy ( TIRFM ) -based live imaging showed that , consistent with our previous results , the GFP-tagged wild type KANK1 strongly accumulated in cortical patches that were tightly clustered around FAs ( Figure 3C , D ) . The KANK1-△CC mutant , which lacked the liprin-β1-binding site but contained an intact KN motif , showed highly specific ring-like accumulations at the rims of FAs ( Figure 3C , D ) . In contrast , KANK1-4A was not clustered anymore around FAs but was dispersed over the cell cortex ( Figure 3C , D ) . The KANK1-△CC-4A mutant , lacking both the liprin-β1 and the talin-binding sites , and the KN-4A mutant were completely diffuse ( Figure 3C , D ) . To test the impact of the talin1-G1404L mutant , we depleted both talin1 and talin2 , which are co-expressed in HeLa cells ( Figure 3—figure supplement 1B ) , and rescued them by introducing mouse GFP-talin1 , which was resistant to used siRNAs . The depletion of the two talin proteins resulted in a dramatic loss of FAs and cell detachment from coverslips ( data not shown ) , in agreement with the well-established essential role of talin1 in FA formation ( Calderwood et al . , 2013; del Rio et al . , 2009; Yan et al . , 2015; Yao et al . , 2014 ) . Therefore , in this experiment only cells expressing GFP-talin1 displayed normal attachment and spreading ( Figure 3—figure supplement 1C ) . The GFP-talin1-G1404L mutant could fully support cell attachment and spreading , although the cell area was slightly increased compared to cells rescued with the wild type GFP-talin1 ( Figure 3—figure supplement 1C–E ) . The number and size of focal adhesions were not significantly different between the cells rescued with the wild type talin1 or its G1404L mutant ( Figure 3—figure supplement 1F , G ) , indicating that the mutant is functional in supporting FA formation . Strikingly , while in cells expressing the wild-type talin1 , KANK1 was prominently clustered around FAs , it was dispersed over the plasma membrane in cells expressing talin1-G1404L ( Figure 3E , F , Figure 1—figure supplement 1 ) . We conclude that perturbing the KANK1-talin interaction , including the use of a single point mutation in the ~2500 amino acid long talin1 protein , which does not interfere with the talin function in FA formation , abrogates KANK1 association with FAs . We next tested whether mislocalization of KANK1 due to the perturbation of KANK1-talin1 binding affected other CMSC components . The localization of GFP-KANK1 and its mutants relative to FAs labeled with endogenous markers was very similar to that described above based on live imaging experiments ( Figure 4A ) . Co-depletion of KANK1 and KANK2 abolished clustering of CMSC components , such as LL5β and KIF21A at the cell edge ( Figure 4B , C ) . Wild type GFP-KANK1 could rescue cortical clustering of these proteins in KANK1 and KANK2-depleted cells ( Figure 4B , C ) . However , this was not the case for the KANK1-4A mutant , the KANK1-△CC mutant or the KANK1 version bearing both mutations ( Figure 4B , C ) . Importantly , the dispersed puncta of the KANK1-4A mutant still colocalized with LL5β , as the binding to liprin-β1 was intact in this mutant ( Figure 3A , Figure 4B , C ) , while the FA-associated rings of KANK1-△CC , the mutant deficient in liprin-β1 binding , showed a mutually exclusive localization with LL5β ( Figure 4B ) . In contrast , KIF21A , which binds to the ankyrin repeat domain of KANK1 , could still colocalize with KANK1-△CC at FA rims ( Figure 4B ) . The overall accumulation of KIF21A at the cell periphery was , however , reduced , in line with the strongly reduced KANK1 peripheral clusters observed with the KANK1-△CC mutant . The diffuse localization of the KANK1-4A-△CC mutant led to the strongly dispersed distribution of the CMSC markers ( Figure 4B , C ) . Furthermore , only the full-length wild type KANK1 , but neither the 4A nor △CC mutant could support efficient accumulation of CLASP2 at the peripheral cell cortex in KANK1 and KANK2-depleted cells ( Figure 4D , E ) , in line with the fact that cortical recruitment of CLASPs depends on LL5β ( Lansbergen et al . , 2006 ) . 10 . 7554/eLife . 18124 . 016Figure 4 . KANK1-talin interaction is required for recruiting CMSCs to FAs . ( A ) Widefield fluorescence images of HeLa cells depleted of KANK1 and KANK2 and expressing the indicated siRNA-resistant GFP-KANK1 fusions ( rescue ) , stained for the FA marker phospho-tyrosine ( pY ) . ( B ) Widefield fluorescence images of HeLa cells transfected with the control siRNA or siRNAs against KANK1 and KANK2 , expressing GFP alone or the indicated siRNA-resistant GFP-KANK1 fusions and stained for LL5β or KIF21A . ( C ) Quantification of peripheral clustering of LL5β and KIF21A in cells treated as in panel ( B ) ( n=12 , 5–6 cells per condition ) . ( D ) TIRFM images of live HeLa cells depleted of KANK1 and KANK2 and co-expressing the indicated siRNA-resistant GFP-KANK1 fusions and mCherry-CLASP2 . ( E ) Quantification of peripheral clustering of mCherry-CLASP2 in cells treated as in panel ( D ) ( n=20 , 8 cells per condition ) . ( F ) Widefield fluorescence images of HeLa cells transfected with the indicated GFP-KANK1 fusions and stained for endogenous LL5β . ( G ) Quantification of peripheral clustering of LL5β in cells treated as in panel ( F ) ( n=12 , 6 cells per condition ) . ( H ) Widefield fluorescence images of HeLa cells transfected with GFP-tagged KANK1 or its CC1 mutant and stained for LL5β . ( I ) Quantification of peripheral clustering of LL5β cells treated as in panel ( H ) ( n=12 , 6 cells per condition ) . Error bars , SEM; ns , non-significant; **p<0 . 005; ***p<0 . 001 , Mann-Whitney U test . DOI: http://dx . doi . org/10 . 7554/eLife . 18124 . 01610 . 7554/eLife . 18124 . 017Figure 4—source data 1 . An Excel sheet with numerical data on the quantification of peripheral clustering of different markers represented as plots in Figure 4C , E , G , I . DOI: http://dx . doi . org/10 . 7554/eLife . 18124 . 017 Next , we investigated whether disrupting the KANK1-talin1 interaction from the talin1 side would affect also CMSC localization and found that this was indeed the case: both LL5β and KIF21A were clustered around FAs in talin1 and talin2-depleted cells rescued with the wild type GFP-talin1 , but not in the cells expressing the GFP-talin1-G1404L mutant , deficient in KANK1 binding ( Figure 4F , G ) . Our data showed that KANK1-△CC could not support proper clustering of CMSC components at the cell edge in spite of its tight accumulation at the FA rims . These data indicate that in addition to binding to talin1 , the localization of CMSC clusters depends on the KANK1-liprin-β1 connection . This notion is supported by the observation that the overexpressed coiled coil region of KANK1 ( CC1 ) , which can compete for liprin-β1 binding but cannot interact with talin1 , acted as a very potent dominant negative , which suppressed accumulation of LL5β at the cell periphery ( Figure 4H , I ) . We conclude that the core CMSC protein LL5β as well as the microtubule-binding CMSC components KIF21A and CLASP2 depend on the KANK1 interaction with both talin1 and liprin-β1 for their efficient clustering in the vicinity of focal adhesions at the cell periphery . We next investigated the impact of the disruption of KANK1-talin1 interaction on microtubule organization . Due to their stereotypic round shape , HeLa cells represent a particularly convenient model for studying the impact of CMSCs on the distribution and dynamics of microtubule plus ends ( Lansbergen et al . , 2006; Mimori-Kiyosue et al . , 2005; van der Vaart et al . , 2013 ) . In this cell line , microtubules grow rapidly in the central parts of the cell , while at the cell margin , where CMSCs cluster in the vicinity of peripheral FAs , microtubule plus ends are tethered to the cortex and display persistent but slow growth due to the combined action of several types of microtubule regulators , including CLASPs , spectraplakins and KIF21A ( Drabek et al . , 2006; Mimori-Kiyosue et al . , 2005; van der Vaart et al . , 2013 ) . This type of regulation prevents microtubule overgrowth at the cell edge and results in an orderly arrangement of microtubule plus ends perpendicular to the cell margin ( van der Vaart et al . , 2013 ) ( Figure 5A ) . In cells with perturbed CMSCs , microtubule plus ends at the cell periphery become disorganized: the velocity of their growth at the cell margin increases , and their orientation becomes parallel instead of being perpendicular to the cell edge ( van der Vaart et al . , 2013 ) ( Figure 5A ) . 10 . 7554/eLife . 18124 . 018Figure 5 . The role of talin-KANK1 interaction in regulating microtubule plus end dynamics around FAs . ( A ) Schematic representation of the pattern of microtubule growth in control HeLa cells and in cells with perturbed CMSCs , based on ( van der Vaart et al . , 2013 ) . ( B ) TIRFM images of live HeLa cells depleted of KANK1 and KANK2 and co-expressing the indicated siRNA-resistant GFP-KANK1 fusions and EB3-mRFP . Images are maximum intensity projection of 241 images from time lapse recording of both fluorescence channels . ( C ) Distributions of microtubule growth rates at the 3 µm broad cell area adjacent to the cell edge , and in the central area of the ventral cortex for the cells treated as described in ( B ) ( n=87–153 , 7–8 cells ) . ( D ) Ratio of microtubule growth rate in the cell center and at the cell edge for the cells treated as described in B ( n=7–8 cells ) . ( E ) Angles of microtubule growth relative to the cell margin for the cells treated as described in B . Box plots indicate the 25th percentile ( bottom boundary ) , median ( middle line ) , mean ( red dots ) , 75th percentile ( top boundary ) , nearest observations within 1 . 5 times the interquartile range ( whiskers ) and outliers ( black dots ) ( n=93–114 , 7–8 cells ) . ( F ) TIRFM images of live HeLa cells depleted of talin1 and talin 2 and co-expressing the indicated GFP-talin1 fusions and EB3-mRFP . Images are maximum intensity projection of 241 images from time lapse recordings of both fluorescence channels . ( G ) Distributions of microtubule growth rates at the 3 µm broad cell area adjacent to the cell edge , and in the central area of the ventral cortex for the cells treated as described in F ( n=88–154 , 7 cells ) . ( H ) The ratio of microtubule growth rate in the cell center and at the cell edge for the cells treated as described in panel ( F ) ( n=7 cells ) . ( I ) Angles of microtubule growth relative to the cell margin for the cells treated as described in F . Box plots as in ( E ) ( n=155–166 , 10 cells ) . In all plots: error bars , SEM; ns , non-significant; **p<0 . 01; **p<0 . 005; ***p<0 . 001 , Mann-Whitney U test . DOI: http://dx . doi . org/10 . 7554/eLife . 18124 . 01810 . 7554/eLife . 18124 . 019Figure 5—source data 1 . An Excel sheet with numerical data on the quantification of different aspects of microtubule organization and dynamics represented as plots in Figure 5C–E , G–I . DOI: http://dx . doi . org/10 . 7554/eLife . 18124 . 019 Using live cell imaging of the microtubule plus end marker EB3-mRFP in KANK1/2 depleted cells rescued with the wild-type GFP-KANK1 , we could indeed show that microtubule plus end growth velocity was almost 2 . 5 times slower at the cell margin compared to the central part of the cell , and the majority of microtubules at the cell margin grew at a 60–80° angle to the cell edge ( Figure 5B–E ) . In the KANK1/2 depleted cells expressing KANK1 mutants , the velocity of microtubule growth in central cell regions was not affected , but the growth rate at the cell periphery increased , and microtubules were growing at oblique angles to the cell margin ( Figure 5B–E ) . The increase of the microtubule growth rate observed with the GFP-KANK1-ΔCC mutant was less strong than with the two 4A mutants ( Figure 5B–E ) . This can be explained by the fact that GFP-KANK1-ΔCC was strongly clustered at FA rims ( Figure 3C , Figure 5B ) , and , through its ankyrin repeat domain , could still recruit some KIF21A , a potent microtubule polymerization inhibitor ( van der Vaart et al . , 2013 ) . The results with rescue of talin1 and talin2 co-depletion with GFP-talin1 or its G1404L mutant fully supported the conclusions obtained with the KANK1-4A mutant: while in GFP-talin1-expressing cells microtubule growth at the cell edge was three fold slower than in the cell center , only a 1 . 5 fold difference was observed in GFP-talin1-G1404L expressing cells , and the proportion of microtubules growing parallel rather than perpendicular to the cell edge greatly increased ( Figure 5F–I ) . We conclude that a single point mutation in talin1 , which does not interfere with FA formation , is sufficient to perturb CMSC clustering and , as a consequence , induce microtubule disorganization in the vicinity of peripheral FAs .
In this study , we have shown that the conserved KN motif of KANK1 represents an LD-type ligand of talin , which allows this adaptor protein to accumulate in the vicinity of integrin-based adhesions . This function is likely to be conserved in the animal kingdom , as the KANK orthologue in C . elegans , Vab-19 , in conjunction with integrins , plays important roles in a dynamic cell-extracellular matrix adhesion developmental process ( Ihara et al . , 2011 ) . The exact impact of KANK1-talin binding likely depends on the specific system , as the loss of KANK proteins was shown to reduce motility of HeLa cells and podocytes ( Gee et al . , 2015; Li et al . , 2011 ) , but promote insulin-dependent cell migration in HEK293 cells ( Kakinuma et al . , 2008 ) . An important question is how KANK-talin1 binding mediates the localization of KANK1 to the rim but not the core of FAs . One possibility , suggested by our deletion analysis of KANK1 , is that while the KN peptide alone can penetrate into FAs , larger KN-containing protein fragments are sterically excluded from the dense actin-containing core of the FA . However , our experiment with the KN-LacZ fusion did not support this simple idea , indicating that the underlying mechanism is likely to be more complex and might involve specific disordered or ordered domains and additional partners of KANK1 , or other regulatory mechanisms . Interestingly , we found that reducing contractility with a ROCK1 inhibitor caused an increase in overlap of KANK1 with FA markers , suggesting that the interaction between KANK1 and talin might be mechanosensitive . An exciting possibility is that full length KANK1 can efficiently interact only with talin molecules at the periphery of focal adhesions because they are not fully incorporated into the focal adhesion structure and are thus less stretched . The KANK1 binding site on talin R7 overlaps with the high affinity actin binding site in talin ( which spans R4-R8 ) ( Atherton et al . , 2015 ) and it is possible that different conformational populations of talin exist within adhesions and link to different cytoskeletal components . Another important question is how KANK1 binding to the rim of focal adhesions can promote CMSC accumulation around these structures , a spatial arrangement in which most of the CMSC molecules cannot be in a direct contact with FAs . Previous work on CMSC complexes showed that they are formed through an intricate network of interactions . The 'core' components of these complexes , which can be recruited to the plasma membrane independently of each other , are LL5β ( and in some cells , its homologue LL5α ) , liprins and KANKs ( of which KANK1 seems to predominate in HeLa cells ) ( Astro and de Curtis , 2015; Hotta et al . , 2010; Lansbergen et al . , 2006; van der Vaart et al . , 2013 ) ( Figure 6A ) . The clustering of CMSC components is mutually dependent and relies on homo- and heterodimerization of liprins α1 and β1 , the association between KANK1 and liprin-β1 , the scaffolding protein ELKS , which binds to both LL5β and liprin-α1 , and possibly additional interactions ( Astro and de Curtis , 2015; Lansbergen et al . , 2006; van der Vaart et al . , 2013 ) , while the microtubule-binding proteins , such as CLASPs and KIF21A , seem to associate as a second 'layer' with the membrane-bound CMSC-assemblies ( Figure 6A ) . The CMSC 'patches' can remain relatively stable for tens of minutes , while their individual components are dynamic and exchange with different , characteristic turnover rates ( van der Vaart et al . , 2013 ) . 10 . 7554/eLife . 18124 . 020Figure 6 . Model of talin-directed assembly of cortical microtubule stabilizing complex . ( A ) Three-step CMSC clustering around focal adhesion: 1 ) KANK1 binds talin rod domain R7 via the KN motif , 2 ) KANK1 initiates a cortical platform assembly by binding liprin-β1 via its CC1 domain , 3 ) completion of CMSC assembly by further clustering of liprins , ELKS , LL5β , CLASP and KIF21A around FA . ( B ) KANK1 binding to nascent talin clusters acts as a 'seed' for macromolecular complex assembly and organization around a FA . DOI: http://dx . doi . org/10 . 7554/eLife . 18124 . 020 The dynamic assemblies of CMSC components , which are spatially separate from other plasma membrane domains and which rely on multivalent protein-protein interactions , are reminiscent of cytoplasmic and nucleoplasmic membrane-unbounded organelles such as P granules and stress granules , the assembly of which has been proposed to be driven by phase transitions ( Astro and de Curtis , 2015; Brangwynne , 2013; Hyman and Simons , 2012 ) . The formation of such structures , which can be compared to liquid droplets , can be triggered by local concentration of CMSC components . It is tempting to speculate that by concentrating KANK1 at the FA rims , talin1 helps to 'nucleate' CMSC assembly , which can then propagate to form large structures surrounding FAs ( Figure 6B ) . Additional membrane-bound cues , such as the presence of PIP3 , to which LL5β can bind ( Paranavitane et al . , 2003 ) , can further promote CMSC coalescence by increasing concentration of CMSC players in specific areas of the plasma membrane . This model helps to explain why the CMSC accumulation at the cell periphery is reduced but not abolished when PI3 kinase is inhibited ( Lansbergen et al . , 2006 ) , and why the clustering of all CMSC components is mutually dependent . Most importantly , this model accounts for the mysterious ability of the two large and spatially distinct macromolecular assemblies , FAs and CMSCs , to form in close proximity of each other . To conclude , our study revealed that a mechanosensitive integrin-associated adaptor talin not only participates in organizing the actin cytoskeleton but also directly triggers formation of a cortical microtubule-stabilizing macromolecular assembly , which surrounds adhesion sites and controls their formation and dynamics by regulating microtubule-dependent signaling and trafficking .
HeLa Kyoto cell line was described previously ( Lansbergen et al . , 2006; Mimori-Kiyosue et al . , 2005 ) . HEK293T cells were purchased from ATCC; culture and transfection of DNA and siRNA into these cell lines was performed as previously described ( van der Vaart et al . , 2013 ) . HaCaT cells were purchased at Cell Line Service ( Eppelheim , Germany ) and cultured according to manufacturer’s instructions . The cell lines were routinely checked for mycoplasma contamination using LT07-518 Mycoalert assay ( Lonza , Switzerland ) . The identity of the cell lines was monitored by immunofluorescence-staining-based analysis with multiple markers . Blebbistatin was purchased from Enzo Life Sciences and used at 50 μM . Serum starvation in HeLa cells was done for 48 hr and focal adhesion assembly was stimulated by incubation with fetal calf serum-containing medium with or without blebbistatin for 2 hr . ROCK1 inhibitor Y-27632 was purchased at Sigma-Aldrich and used at 1 or 10 μM . Double stable HeLa cell line expressing GFP-KANK1 and TagRFP-paxillin was made by viral infection . We used a pLVIN-TagRFP-paxillin-based lentivirus and a pQC-GFP-KANK1-based retrovirus packaged in HEK293T cells using respectively Lenti-X HTX packaging and pCL-Ampho vectors . Antibiotic selection was applied to cells 48 hr after infection using 500 μg/ml G418 ( Geneticin , Life Technologies ) and 1 μg/ml puromycin ( InvivoGen ) . BioGFP-tagged KANK1 mutants were constructed using PCR and pBioGFP-C1 vector as previously described ( van der Vaart et al . , 2013 ) . Rescue constructs for BioGFP-tagged KANK1 were either based on the version previously described ( van der Vaart et al . , 2013 ) or a version obtained by PCR-based mutagenesis of the sequence AGTCAGCGTCTGCGAA to GGTGAGTGTGTGTGAG . mCherry-tagged paxillin construct was made by replacing GFP from pQC-GPXN ( Bouchet et al . , 2011 ) by mCherry ( pmCherry-C1 , Clontech ) . TagRFP-tagged paxillin construct was made by PCR-based amplification and cloning in pTagRFP-T-C1 ( kind gift from Y . Mimori-Kiyosue , Riken Institute , Japan ) . HA-tagged KANK1 construct was generated by cloning KANK1 coding sequence into pMT2-SM-HA ( gift of C . Hoogenraad , Utrecht University , The Netherlands ) . pLVX-IRES-Neo ( pLVIN ) vectors was constructed by cloning the IRES-neomycin resistance cassette from the pQCXIN plasmid ( Clontech ) into the pLVX-IRES-Puro plasmid ( Clontech ) . The lentiviral Lenti-X HTX Packaging vector mix was purchased from Clontech . The retroviral packaging vector pCL-Ampho was kindly provided by E . Bindels , Erasmus MC , The Netherlands . The retroviral pQC-GFP-KANK1 vector was constructed by cloning GFP-KANK1 in pQCXIN and the lentiviral pLVIN-TagRFP-paxillin vector was constructed by cloning TagRFP-paxillin in pLVIN . BirA coding vector was described before ( van der Vaart et al . , 2013 ) . GFP-tagged mouse talin 1 construct was a kind gift from Dr . A Huttenlocher ( Addgene plasmid # 26724 ) ( Franco et al . , 2004 ) . GFP-tagged KN-LacZ fusion was made using PCR-based amplification of KN and LacZ ( kind gift , C . Hoogenraad , Utrecht University , The Netherlands ) , pBioGFP-C1 vector and Gibson Assembly mix ( New England Biolabs ) . Site directed mutagenesis of KANK1 and talin1 constructs was realized by overlapping PCR-based strategy and validated by sequencing . mCherry-tagged CLASP2 construct was a gift from A . Aher ( Utrecht University , The Netherlands ) . Single siRNAs were ordered from Sigma-Aldrich or Ambion , used at 5–15 nM , validated by Western blot analysis and/or immunofluorescence , and target sequences were the following: human KANK1 #1 , CAGAGAAGGACATGCGGAT; human KANK1#2 , GAAGTCAGCGTCTGCGAAA , human KANK2#1 , ATGTCAACGTGCAAGATGA; human KANK2 #2 , TCGAGAATCTCAGCACATA; human talin 1 #1 , TCTGTACTGAGTAATAGCCAT; human talin 1 #2 , TGAATGTCCTGTCAACTGCTG; human talin 2 #1 , TTTCGTTTTCATCTACTCCTT; human talin 2 #2 , TTCGTGTTTGGATTCGTCGAC . The combination of siRNAs talin1 #2 and talin2#1 was the most efficient and was used for the experiments shown in the paper . Streptavidin-based pull down assays of biotinylated proteins expressed using pBioGFP-C1 constructs transfected in HEK293T cells was performed and analyzed as previously described ( van der Vaart et al . , 2013 ) . For mass spectrometry sample preparation , streptavidin beads resulting from pull-downs assays were ran on a 12% Bis-Tris 1D SDS-PAGE gel ( Biorad ) for 1 cm and stained with colloidal coomassie dye G-250 ( Gel Code Blue Stain Reagent , Thermo Scientific ) . The lane was cut and treated with 6 . 5 mM dithiothreitol ( DTT ) for 1 hr at 60°C for reduction and 54 mM iodoacetamide for 30 min for alkylation . The proteins were digested overnight with trypsin ( Promega ) at 37°C . The peptides were extracted with 100% acetonitrile ( ACN ) and dried in a vacuum concentrator . For RP-nanoLC-MS/MS , samples were resuspended in 10% formic acid ( FA ) / 5% DMSO and was analyzed using a Proxeon Easy-nLC100 ( Thermo Scientific ) connected to an Orbitrap Q-Exactive mass spectrometer . Samples were first trapped ( Dr Maisch Reprosil C18 , 3 μm , 2 cm × 100 μm ) before being separated on an analytical column ( Agilent Zorbax 1 . 8 μm SB-C18 , 40 cm × 50 μm ) , using a gradient of 180 min at a column flow of 150 nl min-1 . Trapping was performed at 8 μL/min for 10 min in solvent A ( 0 . 1 M acetic acid in water ) and the gradient was as follows 15- 40% solvent B ( 0 . 1 M acetic acid in acetonitrile ) in 151 min , 40–100% in 3 min , 100% solvent B for 2 min , and 100% solvent A for 13 min . Nanospray was performed at 1 . 7 kV using a fused silica capillary that was pulled in-house and coated with gold ( o . d . 360 μm; i . d . 20 μm; tip i . d . 10 μm ) . The mass spectrometers were used in a data-dependent mode , which automatically switched between MS and MS/MS . Full scan MS spectra from m/z 350 – 1500 were acquired at a resolution of 35 . 000 at m/z 400 after the accumulation to a target value of 3E6 . Up to 20 most intense precursor ions were selected for fragmentation . HCD fragmentation was performed at normalized collision energy of 25% after the accumulation to a target value of 5E4 . MS2 was acquired at a resolution of 17 , 500 and dynamic exclusion was enabled . For data analysis , raw files were processed using Proteome Discoverer 1 . 4 ( version 1 . 4 . 1 . 14 , Thermo Scientific , Bremen , Germany ) . Database search was performed using the swiss-prot human database ( version 29th of January 2015 ) and Mascot ( version 2 . 5 . 1 , Matrix Science , UK ) as the search engine . Carbamidomethylation of cysteines was set as a fixed modification and oxidation of methionine was set as a variable modification . Trypsin was specified as enzyme and up to two miss cleavages were allowed . Data filtering was performed using a percolator ( Käll et al . , 2007 ) , resulting in 1% false discovery rate ( FDR ) . Additional filter was ion score >20 . Antibodies against HA and GFP tags , and liprin β1 used for Western blot analysis were previously described ( van der Vaart et al . , 2013 ) . Rabbit antibodies against KANK1 ( HPA005539 ) and KANK2 ( HPA015643 ) were purchased at Sigma-Aldrich . Western blot analysis of KANK1 was performed using rabbit polyclonal KANK1 antibody ( A301-882A ) purchased at Bethyl Laboratories . Talin immunofluorescence staining was performed using mouse monoclonal 8d4 antibody ( Sigma-Aldrich ) . Western blot analysis of talin 1 and 2 expression was performed using respectively the isotype specific mouse monoclonal 97H6 ( Sigma-Aldrich ) and 68E7 ( Abcam ) antibodies . Ku80 ( 7/Ku80 ) antibody was purchased from BD Biosciences . Immunofluorescence staining of KANK1 , LL5β , liprin β1 , KIF21A and CLASP2 in HeLa and HaCaT cells was performed using the antibodies and procedures previously described ( Lansbergen et al . , 2006; van der Vaart et al . , 2013 ) . F-actin was stained using Alexa Fluor 594-conjugated phalloidin ( Life Technologies ) . Phospho-tyrosine mouse antibody ( PT-66 ) was purchased from Sigma-Aldrich and rabbit FAK phospho-tyrosine 397 was purchased from Biosource . Fixed samples and corresponding immunofluorescence images were acquired using widefield fluorescence illumination on a Nikon Eclipse 80i or Ni upright microscope equipped with a CoolSNAP HQ2 CCD camera ( Photometrics ) or a DS-Qi2 camera ( Nikon ) an Intensilight C-HGFI precentered fiber illuminator ( Nikon ) , ET-DAPI , ET-EGFP and ET-mCherry filters ( Chroma ) , Nikon NIS Br software , Plan Apo VC 100x NA 1 . 4 oil , Plan Apo Lambda 100X oil NA 1 . 45 and Plan Apo VC 60x NA 1 . 4 oil ( Nikon ) objectives . TIRFM-based live cell imaging was performed using the setup described before ( van der Vaart et al . , 2013 ) or a similar Nikon Ti microscope-based Ilas2 system ( Roper Scientific , Evry , FRANCE ) equipped with dual laser illuminator for azimuthal spinning TIRF ( or Hilo ) illumination , 150 mW 488 nm laser and 100 mW 561 nm laser , 49 , 002 and 49 , 008 Chroma filter sets , EMCCD Evolve mono FW DELTA 512x512 camera ( Roper Scientific ) with the intermediate lens 2 . 5X ( Nikon C mount adapter 2 . 5X ) , CCD camera CoolSNAP MYO M-USB-14-AC ( Roper Scientific ) and controlled with MetaMorph 7 . 8 . 8 software ( Molecular Devices ) . Simultaneous imaging of green TIRFM imaging was performed as described before ( van der Vaart et al . , 2013 ) or using the Optosplit III image splitter device ( Andor ) on the Ilas2 system . For presentation , images were adjusted for brightness and processed by Gaussian blur and Unsharp mask filter using ImageJ 1 . 47v ( NIH ) . Fluorescence profiles are values measured by line scan analysis in ImageJ , normalized by background average fluorescence , expressed as a factor of the baseline value obtained for individual channel and plotted as a function of maximum length factor of the selection line ( distance ratio ) . Protein clustering at the cell edge is the ratio of the total fluorescence in the first 5 μm from the cell edge to the next 5 μm measured by line scan analysis in ImageJ after thresholding for cell outline marking and out-of-cell region value assigned to zero . The results were plotted as percentage of control condition average value . FA counting and area measurement was performed using Analyze Particles under ImageJ on focal adhesion binary mask obtained after Gaussian blur/threshold-based cell outline marking and background subtraction ( rolling ball radius , 10 pixels ) . KANK1/talin colocalization was analyzed using Pearson R value provided by Colocalization Analysis plugin under FiJi-ImageJ and a 3 μm diameter circular ROI centered on talin clusters detected by immunofluorescent staining . nEB3-mRFP dynamics was recorded by 0 . 5 s interval time lapse TIRF imaging . Microtubule growth was measured using kymographs obtained from EB3-mRFP time lapse image series , plotted and presented as previously described ( van der Vaart et al . , 2013 ) . Ratio of microtubule growth in cell center to periphery was obtained as values for individual cells . Microtubule growth trajectory angle to the cell edge was manually measured in ImageJ using tracks obtained by maximum intensity projection of EB3-mRFP image series . The cDNAs encoding murine talin1 residues 1357–1653 ( R7-R8 ) , 1357–1653 Δ1454–1586 ( R7 ) and 1461–1580 ( R8 ) were synthesized by PCR using a mouse talin1 cDNA as template and cloned into the expression vector pet151-TOPO ( Invitrogen ) ( Gingras et al . , 2010 ) . Talin mutants were synthesized by GeneArt . Talin polypeptides were expressed in E . coli BL21 ( DE3 ) cultured either in LB for unlabeled protein , or in M9 minimal medium for the preparation of isotopically labeled samples for NMR . Recombinant His-tagged talin polypeptides were purified by nickel-affinity chromatography following standard procedures . The His-tag was removed by cleavage with AcTEV protease ( Invitrogen ) , and the proteins were further purified by anion-exchange . Protein concentrations were determined using their respective extinction coefficient at 280 nm . KANK peptides with a C-terminal cysteine residue were synthesized by Biomatik ( USA ) : KANK1 ( 30–60 ) C – PYFVETPYGFQLDLDFVKYVDDIQKGNTIKKC KANK1 ( 30–68 ) C - PYFVETPYGFQLDLDFVKYVDDIQKGNTIKKLNIQKRRKC KANK1-4A - PYFVETPYGFQAAAAFVKYVDDIQKGNTIKKLNIQKRRKC KANK2 ( 31–61 ) C - PYSVETPYGYRLDLDFLKYVDDIEKGHTLRRC Fluorescence Polarization was carried out on KANK peptides with a carboxy terminal cysteine . Peptide stock solutions were made in PBS ( 137 mM NaCl , 27 mM KCl , 100 mM Na2HPO4 , 18 mM KH2PO4 ) , 100 mg/ml TCEP and 0 . 05% Triton X-100 , and coupled via the carboxy terminal cysteine to the Thiol reactive BIODIPY TMR dye ( Invitrogen ) . Uncoupled dye was removed by gel filtration using a PD-10 column ( GE Healthcare ) . The labeled peptide was concentrated to a final concentration of 1 mM using a centricon with 3K molecular weight cut off ( Millipore ) . The Fluorescence Polarization assay was carried out on a black 96well plate ( Nunc ) . Titrations were performed in triplicate using a fixed 0 . 5 μM concentration of peptide and an increasing concentration of Talin R7-R8 protein within a final volume of 100 μl of assay buffer ( PBS ) . Fluorescence Polarization measurements were recorded on a BMGLabTech CLARIOstar plate reader at room temperature and analyzed using GraphPad Prism ( version 6 . 07 ) . Kd values were calculated with a nonlinear curve fitting using a one site total and non-specific binding model . NMR experiments for the resonance assignment of talin1 R7 , residues 1357–1653 Δ1454–1586 were carried out with 1 mM protein in 20 mM sodium phosphate , pH 6 . 5 , 50 mM NaCl , 2 mM dithiothreitol , 10% ( v/v ) 2H2O . NMR spectra were obtained at 298 K using a Bruker AVANCE III spectrometer equipped with CryoProbe . Proton chemical shifts were referenced to external 2 , 2-dimethyl-2-silapentane- 5-sulfonic acid , and 15N and 13C chemical shifts were referenced indirectly using recommended gyromagnetic ratios ( Wishart et al . , 1995 ) . The spectra were processed using Topspin and analyzed using CCPN Analysis ( Skinner et al . , 2015 ) . Three-dimensional HNCO , HN ( CA ) CO , HNCA , HN ( CO ) CA , HNCACB , and CBCA ( CO ) NH experiments were used for the sequential assignment of the backbone NH , N , CO , CA , and CB resonances . The backbone resonance assignments of mouse talin1 R7 ( 1357–1653 Δ1454–1586 ) have been deposited in the BioMagResBank with the accession number 19139 . | Animal cells are organized into tissues and organs . A scaffold-like framework outside of the cells called the extracellular matrix provides support to the cells and helps to hold them in place . Cells attach to the extracellular matrix via structures called focal adhesions on the cell surface; these structures contain a protein called talin . For a cell to be able to move , the existing focal adhesions must be broken down and new adhesions allowed to form . This process is regulated by the delivery and removal of different materials along fibers called microtubules . Microtubules can usually grow and shrink rapidly , but near focal adhesions they are captured at the surface of the cell and become more stable . However , it is not clear how focal adhesions promote microtubule capture and stability . Bouchet et al . found that a protein called KANK1 binds to the focal adhesion protein talin in human cells grown in a culture dish . This allows KANK1 to recruit microtubules to the cell surface around the focal adhesions by binding to particular proteins that are associated with microtubules . Disrupting the interaction between KANK1 and talin by making small alterations in these two proteins blocked the ability of focal adhesions to capture surrounding microtubules . The next step following on from this work will be to find out whether this process also takes place in the cells within an animal’s body , such as a fly or a mouse . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2016 | Talin-KANK1 interaction controls the recruitment of cortical microtubule stabilizing complexes to focal adhesions |
Vestibular hair cells in the inner ear encode head movements and mediate the sense of balance . These cells undergo cell death and replacement ( turnover ) throughout life in non-mammalian vertebrates . However , there is no definitive evidence that this process occurs in mammals . We used fate-mapping and other methods to demonstrate that utricular type II vestibular hair cells undergo turnover in adult mice under normal conditions . We found that supporting cells phagocytose both type I and II hair cells . Plp1-CreERT2-expressing supporting cells replace type II hair cells . Type I hair cells are not restored by Plp1-CreERT2-expressing supporting cells or by Atoh1-CreERTM-expressing type II hair cells . Destruction of hair cells causes supporting cells to generate 6 times as many type II hair cells compared to normal conditions . These findings expand our understanding of sensorineural plasticity in adult vestibular organs and further elucidate the roles that supporting cells serve during homeostasis and after injury .
In the adult mammalian nervous system , production of sensory and neural cells is limited to a few areas , including the hippocampus , the olfactory epithelium , and the olfactory bulb ( reviewed in Beites et al . , 2005; Ming and Song , 2005 ) . When neurons die naturally in these areas , their debris is phagocytosed by microglia ( reviewed in Fu et al . , 2014 ) . Different glial populations then divide and generate replacement neurons ( Morshead et al . , 1994; Doetsch et al . , 1999; Johansson et al . , 1999 ) . In non-mammalian vertebrates , the mechanosensory receptors for balance , called vestibular hair cells ( HCs ) , undergo turnover ( cell death and replacement ) throughout life . In the avian utricle , a vestibular organ sensing linear head movements , HCs die and are replaced at a slow rate in adulthood , maintaining cell numbers ( Jørgensen and Mathiesen , 1988; Roberson et al . , 1992; Kil et al . , 1997; Goodyear et al . , 1999 ) . In mice , vestibular HC production is reported to occur only during gestation and the first two postnatal weeks ( Ruben , 1967; Rüsch et al . , 1998; Kirkegaard and Nyengaard , 2005; Burns et al . , 2012 ) . However , dying and immature-appearing HCs have been detected in utricles under normal conditions in adult guinea pigs ( Forge et al . , 1993; Li et al . , 1995; Rubel et al . , 1995; Lambert et al . , 1997; Forge et al . , 1998; Forge and Li , 2000 ) and bats ( Kirkegaard and Jørgensen , 2000 , 2001 ) . The ability of adult rodents to regenerate small numbers of utricular HCs after ototoxin-induced damage is another indicator of plasticity in mammalian vestibular epithelia ( Forge et al . , 1993; Warchol et al . , 1993; Forge et al . , 1998; Kawamoto et al . , 2009; Golub et al . , 2012 ) . Many tissues capable of regeneration after injury , such as integumentary , olfactory , and intestinal epithelia , also undergo cellular turnover under normal conditions ( Taylor et al . , 2000; Ito et al . , 2005; Barker et al . , 2007; Leung et al . , 2007 ) . Collectively , these studies suggest vestibular HC turnover may occur in adult mammals , but definitive evidence has not been presented . Supporting cells ( SCs ) , which surround HCs , are epithelial cells that have properties of glia . Among their many functions ( reviewed in Monzack and Cunningham , 2013; Wan et al . , 2013 ) , SCs serve as progenitors to new vestibular HCs during homeostasis in mature birds ( Jørgensen and Mathiesen , 1988; Roberson et al . , 1992; Kil et al . , 1997 ) and after ototoxin-induced death of HCs in mature birds and mammals ( Tsue et al . , 1994; Lin et al . , 2011 ) . SCs also act as phagocytes after ototoxic damage , clearing cellular debris in birds ( Bird et al . , 2010 ) and mammals ( Monzack et al . , 2015 ) . The capacity of SCs in adult mammals to phagocytose and renew vestibular HCs under normal conditions has not been explored . In this study , we demonstrate that a small , but significant proportion of HCs in utricles of adult mice are cleared and replaced by SCs in the absence of a damaging stimulus . There are two types of vestibular HCs , type I and type II , both of which are phagocytosed and removed from the sensory epithelium . However , fate-mapping indicates that SCs replace only type II HCs , and type II HCs do not convert into type I HCs , at least over the 8-month period of adulthood we examined . When HCs are killed by a toxin , SCs transdifferentiate into 6 times as many type II HCs compared to normal HC addition within the same 4-week period . This study demonstrates that the utricle is an additional neuroepithelium in mammals capable of generating sensory receptor cells throughout adulthood and defines the lineage of new HC production in this mature tissue under normal conditions and after HC damage in vivo .
Utricles are otolithic vestibular organs located in the inner ear ( Figure 1A ) . The utricular sensory epithelium ( macula ) is composed of an alternating array of SCs and two types of HCs ( Figure 1B–D ) and contains the processes of vestibular afferent and efferent nerves ( reviewed in Eatock and Songer , 2011 ) . Type I HCs have a flask-shaped body and a nucleus located at mid-epithelial depth , and they synapse onto large calyceal afferents . Type II HCs have complex shapes , with thick necks and basal cytoplasmic processes projecting from the cell body ( Figure 1C , D; Pujol et al . , 2014 ) . Each type II HC nucleus is located near the lumenal surface . Type II HCs synapse onto small ( bouton ) afferents . SCs span the basal-to-apical extent of the epithelium , and their nuclei reside below HC nuclei , near the basal lamina . The utricle is divided into two zones: a C-shaped central region called the striola , in which specialized type I HCs are enriched , and the extrastriolar region , which surrounds the striola and occupies the remainder of the utricle ( Figure 1B ) . 10 . 7554/eLife . 18128 . 003Figure 1 . Phagosomes target type I and type II HCs for clearance in adult mouse utricular maculae under normal conditions . ( A ) Schematic of the inner ear with the utricle highlighted in magenta . ( B ) Schematic of a surface view of a utricle ( xy view ) with HCs in magenta . Blue outlined region denotes the striola ( S ) . The region surrounding the striola is the extrastriola ( ES ) . Boxed area shown at higher magnification in C . ( C ) Higher magnification of schematic in B with slices through the level of type II HC nuclei ( left panel ) and the level of the SC nuclei ( right panel ) . Nuclei in blue , HC perinuclear cytoplasm ( left panel ) and basolateral processes ( right panel ) in magenta , and SC cytoplasm in grey ( xy view ) . ( D ) Schematic of a cross-section ( xz view ) of the adult mouse utricle . II , type II HCs; I , type I HCs . Double lines indicate positions of the xy views shown in C , with the upper set of lines referring the left panel in C , and the lower set of lines referring to the right panel in C . The bracket indicates the level of the SC nuclei , the focal plane of xy confocal optical images in E–E’’’ , H–J , K’–K’’ . In panels E–K” , F-actin was labeled with phalloidin ( green ) and HCs were labeled with anti-myosin VIIa antibodies ( Myo , magenta ) , except there is no myosin label in panel J . In E–E’” , blue label is DAPI . In H–K” , each blue label is a different cell-selective marker . ( E–E’’’ ) Confocal xy optical sections of the SC nuclear layer in an adult Swiss Webster utricle . ( E ) Two ectopic HCs ( arrowheads ) are located next to F-actin-rich phagosomes ( arrows ) . Inset , a ring-shaped phagosome not associated with a HC . See Video 1 for a 3D reconstruction of a phagosome targeting a HC , and see Video 2 for all xy images in the z-series of E . Myosin-labeled HC cytoplasm that is not surrounding a nucleus corresponds to type II HC basolateral processes ( see 1C , right panel ) . ( E’–E”’ ) Higher magnification of the two HCs indicated by arrowheads in E . Arrowheads in E” point to the nucleus of each ectopic HC . ( F ) Transverse section of an adult Swiss Webster mouse utricle showing an ectopic HC ( black arrow ) located in the SC nuclear layer that has condensed chromatin . The ectopic cell is surrounded by a calyx ( black arrowhead ) , typical of a type I HC . The calyx appears as an electron-lucent ring around the cell , which has minimal cytoplasm and contains numerous electron-dense mitochondria ( small gray dots ) . A calyx surrounding a normally localized type I HC ( I ) is indicated by the white arrow . Several normally positioned type I and II HCs are indicated ( I , II over nucleus ) . Inset , higher magnification of the ectopic type I HC . ( G ) Several HCs and F-actin-rich phagosomes are shown in this projection image of a Swiss Webster macula . Two F-actin spikes are indicated by arrowheads . White circles indicate the area of 3 type II HCs for reference . ( H–K’’ ) Phagosomes co-localized with markers of type I or II HCs in xy confocal slices at the level of SC nuclei in Swiss Webster utricles . ( H ) Two ectopic HCs ( arrowheads ) have POU4F3-positive nuclei ( blue ) and are connected to a large ring-shaped phagosome . ( I ) Two ectopic HCs associated with a basket-like phagosome . One HC has a SOX2-negative nucleus ( arrow , lacking blue label ) , and the other HC has a SOX2-positive nucleus ( arrowhead , blue ) . ( J ) Three ectopic HCs ( arrowheads ) associated with phagosomes are calretinin-positive ( blue ) . Arrow indicates an example of a F-actin spike . ( K ) Tenascin ( Ten , blue ) immunolabeling in normally localized HCs ( arrowhead ) . ( K’ , K” ) Tenascin labeling ( blue ) is evident in two ectopic HCs ( arrowheads ) co-localized with phagosomes ( green ) . Scale bar shown in E is 10 μm for E , 7 µm for E’–F , 3 . 5 µm for F inset , and 14 µm for G . Scale bar shown in H is 5 µm for H–K’’ . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 003 To address if vestibular HCs undergo turnover in normal adult mice , we sought evidence for programmed cell death in whole-mounted utricles utilizing terminal deoxynucleotidyl transferase dUTP nick end labeling ( TUNEL ) and immunolabeling for activated caspase-3 ( aCasp3 ) . Although we detected cells with TUNEL or aCasp3 immunolabeling in connective tissue below the macula and in intestinal epithelia labeled simultaneously with utricles , we detected no labeled cells in normal maculae using either method ( data not shown ) . We reasoned that HC removal may occur too infrequently to capture cell death , so we looked for other evidence . When SCs clear HCs after ototoxin-mediated injury , they first produce an actin cable to constrict the HC and extrude its most apical portion , including the stereocilia bundle ( Meiteles and Raphael , 1994; Li et al . , 1995; Bird et al . , 2010; Monzack et al . , 2015 ) . Then , one or more SCs form actin-rich phagosomes that engulf the remaining HC body . Actin filaments comprising the phagosome are interlaced around each HC as it is consumed , and the phagosome structure appears basket- or ring-like ( Bird et al . , 2010; Monzack et al . , 2015 ) . To determine if utricular SCs consume HCs under normal conditions in adult mice , we labeled utricles from 5- to 10-week-old Swiss Webster mice with phalloidin , which binds filamentous actin ( F-actin ) . In whole-mounted normal utricles , we detected numerous F-actin-rich structures that were basket-like or ring-shaped ( Figure 1E–E”’ , G–K” , Video 1 ) , resembling phagosomes described before ( Bird et al . , 2010; Monzack et al . , 2015 ) . The vast majority of these structures was restricted to the basal compartment of the epithelium , amongst SC nuclei . We counted 48 . 9 ± 14 . 8 ( mean ± standard deviation ) phagosomes per utricle ( n = 8; Figure 3A; Figure 3—source data 1 ) . Immunolabeling of myosin VIIa , a HC-specific marker ( Hasson et al . , 1995 ) , revealed that 21 . 7% ± 16 . 1% of phagosomes per utricle were clearly associated with HCs whose cell bodies had been translocated basally to the SC nuclear layer and were therefore considered ectopic ( 95% confidence interval: 8 . 8–34 . 6%; n = 6; Figure 1E–E”’ , H–K”; Video 2 ) . There were two primary types of associations between phagosomes and ectopic HCs . In 55 . 2% ( ±9 . 7%; 95% confidence interval: 47 . 5–63 . 0%; n = 6 ) of associations , phagosomes were composed of a ring-like structure that fully encircled a HC body ( Figure 1E–E’” , arrowheads ) and were connected to a basket-like structure devoid of HC material ( Figure 1E–E’” , arrows ) . Most other phagosomes associated with ectopic HCs consisted of a basket-like structure with one or more F-actin-rich processes that extended laterally and either contacted nearby HCs ( Figure 1G ) or pierced and entered their cytoplasm ( Figure 1H , J , K” , Video 1 ) . In Swiss Webster mice , we counted 12 ± 9 . 9 ectopic HCs per utricle that were associated with phagosomes ( 95% confidence interval: 4 . 3–20 . 1; n = 6 ) . However , the majority of phagosomes ( 71 . 7% ± 14 . 3%; 95% confidence interval: 60 . 3–83 . 1%; n = 6 ) lacked HC staining within or around them ( Figure 1E inset ) , suggesting they may be clearing other cell types or being retracted after completion of HC digestion . 10 . 7554/eLife . 18128 . 004Video 1 . 3D reconstruction of a phagosome in an adult Swiss Webster utricle under normal conditions . This movie is a y-axis rotation of a 3D reconstruction of an ectopic HC ( myosin VIIa , magenta ) associated with a F-actin phagosome ( phalloidin , green ) . The phagosome consists of a large basket-like structure with a spike , or process , that extends through the cytoplasm of the HC along its nucleus . Note the basket-like structure is not solid F-actin but has gaps that lack labeling suggestive of a lattice . All nuclei are labeled with DAPI ( blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 00410 . 7554/eLife . 18128 . 005Video 2 . Phagosomes in the SC nuclear layer of an adult Swiss Webster utricle under normal conditions . This movie was constructed from optical sections collected from the area shown in Figure 1E . It begins at the level of type II HC nuclei and progresses through type I HC and SC nuclei , ending at the basal lamina . All nuclei are labeled with DAPI ( blue ) , HCs are labeled with myosin VIIa antibodies ( magenta ) , and the F-actin in phagosome basket- and ring-like structures and spikes ( or processes ) is labeled with phalloidin ( green ) . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 005 Examination of serial transverse sections of a utricle from an adult Swiss Webster mouse confirmed that some HCs were located in a basal , ectopic position ( Figure 1F ) . Of 6 such cells we observed in the sectioned utricle , one HC appeared to have condensed chromatin characteristic of apoptosis ( Figure 1F ) . Although we examined hundreds of ectopic HCs in utricles , no other HCs had obvious abnormalities or signs of apoptosis other than their location . Indeed , 4' , 6-diamidino-2-phenylindole ( DAPI ) labeling revealed no variation in nuclear size , integrity , or density of chromatin in ectopically located HCs or HCs associated with phagosomes ( Figure 1E” ) . To examine the type of HCs targeted by phagosomes , we labeled utricles with phalloidin and antibodies to other HC markers ( Figure 1H–K” ) . The nuclei of basally translocated cells being engulfed by phagosomes were immunoreactive for antibodies against POU4F3 ( Figure 1H ) , a HC-specific transcription factor ( Erkman et al . , 1996; Xiang et al . , 1997 ) , confirming their identity . We found evidence that both type I and type II HCs were being targeted by phagosomes . SOX2 is a transcription factor that is abundant in SCs and type II HCs , but is not found in type I HCs ( Oesterle et al . , 2008 ) . Some phagosomes were associated with a myosin VIIa-positive cell whose nucleus was SOX2-positive ( Figure 1I ) . In other utricles , we found phagosome-associated HCs that were immunoreactive for the calcium-binding protein , calretinin ( Figure 1J ) , which is selectively elevated in type II HCs ( Desai et al . , 2005 ) . Some phagosomes engulfed cells that were immunoreactive for tenascin ( Figure 1K–K” ) , an extracellular matrix protein that lines the space between the type I HC plasma membrane and the afferent calyx ( Swartz and Santi , 1999 ) . Further , one ectopic HC in transverse plastic sections was wrapped by a calyceal afferent ( Figure 1F ) , identifiable by its electron-lucent appearance except for scattered mitochondria , which is typical of a type I HC . To address the mechanism of HC clearance , we examined phalloidin labeling in HCs located in their proper positions . We estimate that 2–10 normal-appearing HCs per utricle contained a long ‘spike’ of F-actin that coursed through the cytoplasm . Typically , the spike extended from the cell’s apex to its base , then exited the cell and descended toward the basal lamina ( Figure 2A–A” , Video 3 ) . It is notable that we detected a similar spike-like structure in the cytoplasm of 5 . 3 ( ±4 . 3; 95% confidence interval: 2 . 0–8 . 7; n = 6 ) basally translocated HCs per utricle that were associated with phagosomes ( Figure 2B–B” , and Figure 1H , J , K” ) . Some actin spikes extended laterally outside the HCs for several cell widths ( Figure 1G ) , and some spikes were not clearly associated with any HCs . 10 . 7554/eLife . 18128 . 006Figure 2 . F-actin spikes are present in HCs , and the HC bundle is likely ejected prior to HC body translocation under normal conditions . ( A–A” ) A F-actin spike ( green , arrowhead ) in a normally localized HC [myosin VIIa ( Myo ) magenta] from a C57Bl/6J mouse is shown in 3 views . ( A ) is a xy optical section at the level of type II HC nuclei ( DAPI , blue ) . ( A’ , A” ) are yz and xz optical ( cross ) sections of the HC indicated in A by the arrowhead . Note the F-actin spike is beneath the apical surface , which is highlighted by the row of brightly labeled stereocilia ( green ) at the top of A’ and A” . See Video 3 for all xy images in the z-series of A . HC , HC nuclear layer; SC , SC nuclear layer . ( B–B” ) A F-actin spike ( green , arrowhead ) in an ectopic HC ( Myo , magenta ) from a C57Bl/6J mouse is shown in 3 views . ( B ) is a xy optical section at the level of SC nuclei ( DAPI , blue ) . ( B’ , B” ) are yz and xz optical sections of the HC shown in B . Arrow points to a basket-like phagosome . ( C–D’’ ) Confocal optical sections from two normal Swiss Webster mouse utricles labeled with F-actin , antibodies to myosin VIIa , and with either anti-espin antibodies ( C–C” ) or anti-PMCA2 antibodies ( D–D” ) . The 3 panels for each utricle ( C–C” or D-D” ) show different label combinations for the same field , as indicated . All panels are focused on the SC layer , except the boxed insets , which are focused on stereocilia . Arrowheads in C , D and C” , D” point to ectopic HCs ( Myo , magenta in C and D ) , while arrows in all panels point to F-actin spikes ( green in C , D and white in C’ , D’ ) . Scale bar in A is 3 µm and applies to A–B” . Scale bar in C is 10 µm and applies to C–D” , including insets . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 00610 . 7554/eLife . 18128 . 007Video 3 . F-actin spike through a normally located HC in an adult mouse utricle under normal conditions . This movie was constructed from optical sections collected from the area shown in Figure 2A–A’’ . The movie begins above the stereocilia bundles ( F-actin , green ) , continues through the type II and type I HC nuclear layers ( DAPI , blue ) , and ends in the SC nuclear layer . A myosin VIIa labeled HC ( magenta ) , which has a normal-appearing stereocilia bundle ( below the arrow ) , has a F-actin spike ( green , arrow ) within its cytoplasm , extending the length of the HC and into the SC nuclear layer . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 007 During phagocytosis after HC damage by aminoglycosides , the apical portion of the HC is cleaved and ejected apically ( Meiteles and Raphael , 1994; Li et al . , 1995; Bird et al . , 2010; Monzack et al . , 2015 ) . To determine if this also occurs under normal conditions , we labeled utricles with antibodies to espin and plasma membrane calcium ATPase 2 ( PMCA2 ) , which are reported to label stereocilia , but not other structures in the sensory epithelium ( Dumont et al . , 2001; Li et al . , 2004 ) . As anticipated , stereocilia were brightly labeled with both antibodies ( Figure 2C , D insets ) . Occasional ectopic HCs were diffusely labeled by either antibody ( Figure 2C , C” , espin labeling ) , but none of them contained brightly labeled foci resembling stereocilia ( Figure 2C–D” ) . Thus , our observations suggest that the apical part of the HC is ejected apically prior to translocation or that stereocilia degenerate during translocation . Interestingly , espin antibodies labeled most actin-rich spikes ( Figure 2C–C” ) . Initially , we thought this could indicate that spikes are derived from stereocilia . However , spikes were not labeled for antibodies to PMCA2 ( Figure 2D–D” ) , suggesting this is not the case . Perhaps espin serves as an actin-bundling protein in phagosome spikes , as it does in stereocilia ( Zheng et al . , 2000 ) . In 5- to 10-week-old Swiss Webster mice , there were 48 . 9 ( ±14 . 8 ) phalloidin-labeled phagosomes per utricle ( Figure 3A , Figure 3—source data 1 ) , displaying a range of morphologies . To assess if phagosomes are unique to Swiss Webster mice at 5–10 weeks of age , we analyzed Swiss Webster utricles at 3 weeks and 43–46 weeks of age , as well as utricles from two strains of inbred mice ( C57Bl/6J and CBA/CaJ ) at all 3 ages ( Figure 3A , Figure 3—source 1 ) . The distribution of phagosomes in the utricular macula was similar across utricles and ages: phagosomes appeared to be concentrated in the peristriolar region ( Figure 3B , B’ ) . The morphologies of phagosomes were also similar in all 3 strains of mice ( not shown ) . However , C57Bl/6J and CBA/CaJ utricles consistently had fewer phagosomes than Swiss Webster utricles ( Figure 3A , Figure 3—source data 1 ) . Within each strain , the number of phagosomes was similar at all ages examined ( Figure 3A , Figure 3—source data 1 ) . These findings demonstrate that HC clearance occurs under normal conditions in different strains of mice at a range of ages . 10 . 7554/eLife . 18128 . 008Figure 3 . Phagosomes are present in several mouse strains across ages . ( A ) Number of phagosomes per utricle in 3 mouse strains ( CBA/CaJ , C57Bl/6J , and Swiss Webster ) at 3 weeks , 5–10 weeks , and 43–46 weeks of age . Data are presented as mean ± 1 standard deviation for n = 3–8 mice per group ( see Figure 3—source data 1 ) . Within each strain , the number of phagosomes did not increase significantly over time . CBA/CaJ and C57Bl/6J mice had similar phagosome numbers ( 10–15 per utricle ) , but Swiss Webster mice had significantly more phagosomes per utricle ( 45–65 ) than the other strains at every age , as determined by two-way ANOVA ( p=0 . 0189 for age and p<0 . 0001 for strain followed by Bonferroni’s multiple comparisons test; ****p<0 . 0001 ) . ( B , B’ ) Confocal projection image of the utricular macula from an adult Swiss Webster mouse showing that F-actin-rich phagosomes ( green ) are concentrated in the peristriolar region of the extrastriola ( ES ) but are largely absent from the calbindin-labeled striola ( S , blue ) and the peripheral-most portion of the ES . The projection image was constructed from just beneath the stereocilia through the SC layer to avoid obstruction of the phagosomes by the F-actin-rich stereocilia . Scale bar is 100 μm . Arrowheads in B , B’ point to a portion of the utricular macula that sustained damage during dissection; this region is rich in F-actin but is not a phagosome . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 00810 . 7554/eLife . 18128 . 009Figure 3—source data 1 . Quantification of phagosomes in the normal utricle of3 mouse strains at 3 ages . Mean ( one standard deviation , SD ) and 95% confidence interval ( CI ) of number of F-actin ( phalloidin ) -labeled phagosomes per utricle . n , number of mice . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 009 We examined whether phagosomes derive from SCs or from cells in the macrophage/monocyte lineage . First , we fluorescently labeled the cytoplasm of SCs to assess if they generate actin-rich phagosomes using Plp1-CreERT2 mice , which have been used previously to label SCs in mouse utricles ( Gómez-Casati et al . , 2010; Burns et al . , 2012; Wang et al . , 2015 ) . In 6-week-old Plp1-CreERT2:ROSA26CAG-loxP-stop-loxP-tdTomato mice ( hereafter referred to as Plp1-CreERT2:ROSA26tdTomato mice ) , the majority of SCs were tdTomato-positive at one week after injection of tamoxifen ( Figure 4B ) . A small number of cells in the transitional epithelium , which borders the sensory epithelium ( Figure 4B ) , and numerous cells in the stroma ( presumed Schwann cells , not shown ) were also tdTomato-positive . We sampled 8 regions of the macula and determined that 91 . 7% ( ±6 . 1%; n = 3 ) and 68 . 4% ( ±1 . 8%; n = 3 ) of SCs in the extrastiola and the striola , respectively , were tdTomato-positive ( Figure 4—source data 1 ) . In age-matched Plp1-CreERT2:ROSA26tdTomato mice that did not receive tamoxifen , <5% of SCs per utricle ( 126 . 8 ± 46 . 8; 95% confidence interval: 80 . 9–172 . 6; n = 4 ) were tdTomato-positive ( Figure 4A ) , revealing some tamoxifen-independent Cre activity . We labeled Plp1-CreERT2:ROSA26tdTomato utricles collected at one week post tamoxifen with phalloidin to visualize phagosomes and antibodies against myosin VIIa to visualize HCs . We detected an average of 27 . 8 ( ±4 . 3; 95% confidence interval: 23 . 0–32 . 6; n = 3 ) phagosomes per utricle , which were fewer than Swiss Webster mice , but more than CBA/CaJ and C57Bl/6J mice ( Figure 3A , Figure 3—source data 1 ) and 4 . 5 ( ±2 . 3; 95% confidence interval: 1 . 9–7 . 1; n = 3 ) phagosomes were associated with a HC . In some utricles , we detected overlap of tdTomato and phalloidin signals , indicating that some phagosomes were derived from SCs ( Figure 4C–E’ ) . It was unclear if phagosomes were generated by a single SC or by two or more adjacent SCs , but phagosomes were consistently derived from the most basal region of a SC ( Figure 4D–E’ ) . 10 . 7554/eLife . 18128 . 010Figure 4 . SCs , not macrophages , produce phagosomes in adult mouse utricles under normal conditions . ( A , B ) Confocal projection images of the utricular macula from 7-week-old Plp1-CreERT2:ROSA26tdTomatomice showing the numbers and distribution of tdTomato-positive SCs ( magenta ) in the utricular sensory epithelium of mice that received no tamoxifen ( A ) or mice that received tamoxifen ( B ) . S , striola; SE , sensory epithelium; TE , transitional epithelium . See Figure 4—source data 1 for quantification of tdTomato-positive SCs in extrastriolar and striolar regions . ( C–E’ ) Higher magnification optical sections of a Plp1-CreERT2:ROSA26tdTomatomouse utricular macula at one week post tamoxifen showing overlap between tdTomato-labeled SC cytoplasm ( magenta ) and a F-actin-rich phagosome ( green , arrows ) . ( C ) Xy view with double lines indicating where cross-sectional images were created in D-E’ . ( D , D’ ) Yz view of same area in C . ( E , E’ ) Xz view of same area in C . Very bright green labeling in D and E is F-actin in the stereocilia bundles of HCs . ( F , G ) Confocal optical sections of an adult Swiss Webster utricle . ( F ) Two F-actin-rich phagosomes ( green , arrows ) in the SC nuclear layer ( DAPI , blue ) did not co-label for antibodies to IBA1 , a macrophage/monocyte lineage marker ( magenta ) . ( F’ ) An IBA1-positive cell ( magenta , arrowhead ) resided in the connective tissue under the phagosomes shown in F . ( G ) Xz view of the field shown in F , F’ . The white dotted line indicates the border between sensory epithelium ( SE ) and connective tissue ( CT ) . Arrows and arrowhead indicate the same cells shown in F , F’ . ( H ) Confocal projection image of the utricular macula from an adult Lfng-eGFP mouse showing eGFP-positive SCs ( green ) in the sensory epithelium ( SE ) . S , Striola . ( I–I”’ ) Confocal optical sections of the same field from the extrastriolar region of a Lfng-eGFP utricle , with eGFP in green and F-actin in blue . ( I ) Xz view through the SE and CT of the utricle . ( I’ , I” ) Xy views of the SC layer in the SE . ( I”’ ) Xy view of the CT . The dotted line in I is the approximate location of the basal lamina , between the SE and the CT . Two phagosomes ( arrowheads , F-actin , blue ) are flanked by eGFP-positive SCs ( green ) and are also co-labeled with eGFP ( green ) ; co-labeled phagosomes appear cyan . One phagosome ( arrow in I’ , I” ) is eGFP-negative and not flanked by eGFP-positive SCs . The double lines in I’–I”’ indicate the position of the xz view shown in I . Scale bar in A is 100 µm and applies to A , B , H . Scale bar in C is 5 µm and applies to C–G . Scale bar in I is 5 µm and applies to I–I”’ . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 01010 . 7554/eLife . 18128 . 011Figure 4—source data 1 . Quantification of the percentage of SCs labeled with tdTomato in Plp1-CreERT2:ROSA26tdTomato utricles . Mean percentage { ( one standard deviation , SD ) ; [95% confidence interval , CI]} of tdTomato-labeled SCs in Plp1-CreERT2:ROSA26tdTomato mice given tamoxifen at 6 weeks ( wks ) of age and sacrificed 1 or 15 weeks post tamoxifen ( 7 or 21 weeks of age , respectively ) . SCs were sampled in two striolar regions and 6 extrastriolar regions in each utricle ( see Materials and methods ) . n = 3 mice at both timepoints . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 011 SCs , cells in the transitional epithelium , and cells in the connective tissue underlying the sensory epithelium ( presumed Schwann cells ) were labeled by tdTomato in Plp1-CreERT2:ROSA26tdTomato mice . To further delineate which cell types contribute to the production of phagosomes , we examined utricles from Lfng-eGFP mice , which express eGFP under control of Lfng regulatory elements ( Gong et al . , 2003 ) . In Lfng-eGFP utricles , eGFP was expressed in the majority of SCs , but not in HCs , cells in the transitional epithelium , or the underlying connective tissue ( Figure 4H–I”’; Burns et al . , 2015 ) . F-actin-labeled phagosomes were also detected in the basal compartment of the sensory epithelium of Lfng-eGFP utricles , and some were co-labeled with eGFP ( Figure 4I–I” ) , providing further evidence that phagosomes derive from SCs . To assess if cells from the macrophage/monocyte lineage also phagocytose HCs under normal conditions , we labeled adult Swiss Webster utricles with myosin VIIa to detect HCs and ionized calcium-binding adaptor molecule 1 ( IBA1 ) or CD68 to detect resting or activated macrophages , respectively ( Holness et al . , 1993; Imai et al . , 1996; Ito et al . , 1998 ) . We also applied phalloidin to label phagosomes and DAPI to label nuclei . F-actin-labeled phagosomes in the sensory epithelium never co-localized with IBA1 labeling ( Figure 4F , G ) or CD68 labeling ( not shown ) , although we consistently detected cells that were IBA1-positive ( Figure 4F’ , G ) or CD68-positive ( not shown ) in the connective tissue below the macula . These results are consistent with a previous study that did not detect macrophages in the utricular macula of normal adult mice ( Kaur et al . , 2015 ) . Our observations demonstrate that SCs , not macrophages , clear type I and II vestibular HCs from adult mouse utricles under normal conditions by creating F-actin-rich phagosomes . Numbers of total utricular HCs and SCs seem to be stable in adult mice at the ages we examined ( Kirkegaard and Nyengaard , 2005; Burns et al . , 2012 ) . Therefore we reasoned that if HCs are cleared from the macula , they must be replaced . To address this , we looked for evidence of ongoing HC addition by assessing if HCs in normal adult utricles express two markers specific for immature HCs , protocadherin15-CD2 ( PCDH15-CD2 ) and ATOH1 . PCDH15 is a protein localized to the tip links of HC stereocilia . Two alternatively spliced variants of PCDH15 ( CD1 and CD3 ) are present in mature stereocilia ( Ahmed et al . , 2006 ) . In contrast , PCDH15-CD2 is detected along the length of stereocilia during early development , but becomes confined to the kinocilium once HCs mature ( Ahmed et al . , 2006; Webb et al . , 2011 ) . Thus , PCDH15-CD2 is a selective marker of immature stereocilia . As positive controls , we labeled utricles from neonatal ( <8 day-old ) Swiss Webster mice with antibodies to PCDH15-CD2 and with phalloidin to mark stereocilia . In the periphery of the utricle , where newly formed HCs are differentiating in neonates ( Burns et al . , 2012 ) , we detected numerous HCs with PCDH15-CD2 labeling throughout the stereocilia ( Figure 5A ) . Next , we examined utricles from 6- to 9-week-old CBA/CaJ and Swiss Webster mice . We observed small numbers of HCs with PCDH15-CD2 labeling throughout stereocilia in adult utricles ( Figure 5B ) , primarily in the most peripheral portion of the extrastriolar region ( Figure 5C ) . CBA/CaJ and Swiss Webster mice had 17 . 2 ( ±4 . 0 ) and 24 . 0 ( ±8 . 8 ) HCs with the immature labeling pattern for PCDH15-CD2 , respectively ( Figure 5D , Figure 5—source data 1 ) . 10 . 7554/eLife . 18128 . 012Figure 5 . Immature HCs are found in adult mouse utricles under normal conditions . ( A , B ) Confocal xy optical sections of utricles from neonatal ( <postnatal day 8 ) ( A ) and adult ( >6 weeks ) ( B ) Swiss Webster mice . Some HC stereocilia bundles at each age ( arrows ) were co-labeled with F-actin ( green ) and antibodies to PCDH15-CD2 ( magenta ) . Some PCDH15-CD2 labeling occurred in other places on the epithelial surface , but our analysis was restricted to stereocilia only . White lines indicate the approximate location of edge of the utricular macula . ( C ) Schematized map of a utricle from a representative adult Swiss Webster mouse . HCs with PCDH15-CD2-positive stereocilia are depicted by black dots . S , striola . ( D ) Number of PCDH15-CD2-expressing HC stereocilia bundles per utricle in normal adult CBA/CaJ and Swiss Webster mice , where there was no statistically significant difference ( determined by an unpaired two-tailed Student’s t-test; p=0 . 1635 , see Figure 5—source data 1 ) . Data are expressed as mean +1 standard deviation . ( E ) Confocal projection image through the utricular macula of a 6-week-old Atoh1GFP/GFP mouse expressing ATOH1-GFP fusion protein ( green ) . S , striola . ( F , G ) Schematics of cross-sectional views through a 6-week-old Atoh1GFP/GFP mouse utricle shown in F’–F’” and G’–G’” . ( F’–F”’ ) Confocal optical section of a HC [arrowhead; myosin VIIa ( Myo ) , magenta; DAPI , blue] that expressed ATOH1-GFP ( green ) . ( G’–G”’ ) Confocal optical section of a SC ( arrow; DAPI , blue ) that expressed ATOH1-GFP ( green ) . HC , HC nuclear layer; SC , SC nuclear layer; BL , basal lamina . Scale bar in B is 12 µm and applies to A , B . Scale bar in E is 100 µm . Scale bar in G’’’ is 10 μm and applies to F’–G’’’ . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 01210 . 7554/eLife . 18128 . 013Figure 5—source data 1 . Quantification of immature HC markers in the normal adult mouse utricle . ( A ) Mean ( one standard deviation , SD ) and 95% confidence interval ( CI ) of number of PCDH15-CD2-labeled stereocilia bundles per utricle in two strains of mice . ( B ) Mean ( 1 SD ) and 95% CI of number of ATOH1-GFP-positive cells per utricle . HCs were identified as myosin VIIa-positive cells with nuclei in the apical two-thirds of the epithelium . SCs were identified as myosin VIIa-negative cells whose bodies extend across the entire macular depth , whose nuclei are smaller than HC nuclei , and are positioned near the basal lamina . Unknown cells did not meet criteria for HCs or SCs . n , number of mice . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 013 ATOH1 is a transcription factor that is abundant in early differentiating HCs and weakly expressed in mature HCs ( Lanford et al . , 2000; Chen et al . , 2002; Woods et al . , 2004; Cai et al . , 2013 ) . We examined utricles collected from 6-week-old Atoh1GFP/GFP mice , which express an ATOH1-GFP fusion protein under control of the endogenous Atoh1 promoter ( Rose et al . , 2009 ) . In Atoh1GFP/GFP mice , ATOH1-GFP is abundant in developing cochlear HCs but is lost once HCs mature ( Cai et al . , 2013 ) . Using antibodies to amplify the GFP signal , we detected numerous GFP-positive cells that were concentrated in the extrastriola ( Figure 5E ) . We counted 82 . 3 ( ±25 . 7 ) brightly labeled HCs and 8 . 3 ( ±2 . 9 ) brightly labeled SCs per utricle ( Figure 5F–G”’ , Figure 5—source data 1 ) . HCs were identified as myosin VIIa-positive cells with nuclei in the apical two-thirds of the epithelium ( Figure 5F–F”’ ) . SCs were identified as myosin VIIa-negative cells whose bodies extend across the entire macular depth , whose nuclei are smaller than HC nuclei , and are positioned near the basal lamina ( Figure 5G–G”’ ) . We also detected 8 . 0 ( ±2 . 6 ) brightly labeled cells per utricle that did not meet criteria for HCs or SCs . Often , these cells were myosin VIIa-negative with a nucleus located between the SC and HC nuclear layers and had thick apical necks . We postulate that these cells were SCs in the process of transitioning to a HC fate . Collectively , these observations demonstrate that immature HCs exist in adult mouse utricles under normal conditions . The presence of immature HCs in utricles from normal adult mice is consistent with the hypothesis that new HCs are added in adulthood . Based on previous work ( Lin et al . , 2011; Burns et al . , 2012 ) , the likely source of new HCs is neighboring SCs . To determine if utricular HCs derive from SCs in normal adult mice , we used Plp1-CreERT2:ROSA26tdTomato mice , in which tamoxifen induces tdTomato expression in SCs ( Figure 4B , Figure 4—source data 1 ) . Prior to fate-mapping , we determined that 91 . 7% ( ±6 . 1%; n = 3 ) of extrastriolar SCs were tdTomato-positive one week after tamoxifen injection ( Figure 4—source data 1 ) . We hypothesized that most new HCs would be added in this region , since we detected most HCs with immature markers ( PCDH15-CD2 and ATOH1-GFP ) there ( Figure 5C , E ) . At 15 weeks post tamoxifen , the latest time analyzed in this fate-mapping study , 85 . 9% ( ±3 . 2%; n = 3 ) of extrastriolar SCs were labeled ( Figure 4—source data 1 ) . The concentration of Plp1-CreERT2 activity in the extrastriola throughout the time-course of our experiment indicated that this mouse line could effectively record the transition of SCs into HCs . To fate-map the transition of SCs into HCs during adulthood , we gave tamoxifen to 6-week-old Plp1-CreERT2:ROSA26tdTomato mice and examined utricles at 1 , 4 , 10 , or 15 weeks later ( Figure 6A ) . We assessed whether the number of tdTomato-positive HCs increased over time after induction of tdTomato labeling , as predicted if SCs convert into HCs . Utricles were labeled with antibodies to myosin VIIa and DAPI , and we collected optical slices throughout the entire macula using confocal microscopy . tdTomato-positive HCs ( Figure 6B–C’’’ ) were identified in all utricles and were scattered throughout the macula , but they appeared to be most common in the extrastriola ( Figure 6F ) , which is consistent with findings for immature HC markers ( Figure 5C , E ) . 10 . 7554/eLife . 18128 . 014Figure 6 . SCs produce new type II HCs in adult mouse utricles under normal conditions . ( A ) Experimental timeline . Plp1-CreERT2:ROSA26tdTomato mice were injected with tamoxifen ( Tam ) at 6 weeks of age to label SCs with tdTomato and were sacrificed ( Sac ) at 7 , 10 , 16 , and 21 weeks of age ( corresponding to 1 , 4 , 10 , and 15 weeks post tamoxifen ) . Control animals ( age-matched Plp1-CreERT2:ROSA26tdTomato mice that did not receive tamoxifen ) were sacrificed at the same ages . ( B–C”’ ) Examples of a tdTomato-labeled ( magenta ) type I HC ( B–B”’ ) and type II HC ( C–C”’ ) from a Plp1-CreERT2:ROSA26tdTomato mouse utricle at 10 weeks post tamoxifen , which were classified according to criteria defined in Materials and methods , Figure 1D , and Video 4 . ( B , B’ and C , C’ ) xy slices taken at the levels indicated in the schematic to the left . Myosin VIIa ( Myo ) is in green and DAPI is in blue . Thin arrows in B , B’ point to two type I HCs , at the level of the neck ( B ) and the nucleus ( B’ ) . Only the type I HC on the right is tdTomato-positive , which is most evident in its nucleus ( B’ ) . Fat arrows in C , C’ point to two type II HCs , at the level of the neck ( C ) and the nucleus ( C’ ) . Only the type II HC on the left is tdTomato-positive . ( B” , C” ) Xz view of the same cells shown in B–B’ and C–C’ , providing perspective on their morphology and lamination . Labels indicate the approximate positions of the nuclei for each cell type ( HCII , type II HC; HCI , type I HC; and SC , SC ) . ( B”’ , C”’ ) : Same images as B” , C” , but with Myo labeling only . Scale bar in C is 10 µm and applies to B–C’ . Scale bars in B’’ , C’’ are 10 µm and apply respectively to B’’’ and C’’’ . It is important to note that , in thin optical slices such as these , myosin VIIa labeling intensity varied across cells , independent of tdTomato labeling intensity . For example , in panels C” and C”’ , the type II HC on the left is brighter than the type II HC on the right . Further , in panels B’ , B” , the myosin VIIa labeling for type I HC perinuclear cytoplasm was relatively weak , and labeling at the neck ( B ) was more robust . In cases such as this one , other morphological criteria—nuclear position , relative neck thickness , and presence/absence of a basolateral process—were essential for cell-typing . ( D ) Total number of tdTomato-expressing HCs per utricle , categorized by HC type [type I ( black ) , type II ( blue ) , and ‘unknown’ ( orange ) ] . Patterned bars = control Plp1-CreERT2:ROSA26tdTomato mice that did not receive tamoxifen . Solid bars = Plp1-CreERT2:ROSA26tdTomato mice that received tamoxifen at 6 weeks of age . Data are expressed as mean ±1 standard deviation for n = 4–8 mice ( see Figure 6—source data 1 ) . *p<0 . 05; ****p<0 . 0001 as determined by a two-way ANOVA ( p<0 . 0001 for treatment/age; p<0 . 0001 for HC type ) followed by Tukey’s multiple comparisons post-test . ( E ) Linear regression analysis of tdTomato-expressing type II HCs per utricle demonstrated an increase in labeled cells with time ( data correspond to D , blue solid bars ) . The calculated slope was 1 . 97 type II HCs per week , and the R2 value was 0 . 691 . Data are expressed as mean ±1 standard deviation with dotted lines representing the 95% confidence interval . ( F ) Map of a representative Plp1-CreERT2:ROSA26tdTomato utricle injected with tamoxifen at 6 weeks of age and analyzed 15 weeks later . tdTomato-labeled HCs are depicted by magenta dots . S , striola . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 01410 . 7554/eLife . 18128 . 015Figure 6—source data 1 . Quantification of tdTomato-labeled HCs in Plp1-CreERT2:ROSA26tdTomato utricles over time . Mean ( one standard deviation , SD ) and 95% confidence interval ( CI ) of the number of tdTomato-labeled HCs per utricle categorized by type in Plp1-CreERT2:ROSA26tdTomato mice given tamoxifen at 6 weeks ( wks ) of age ( right ) or in age-matched littermate controls that did not receive tamoxifen ( left ) . Un . , unknown . n , number of mice . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 015 We counted tdTomato-positive HCs in each utricle that were type I or type II at 1 , 4 , 10 , or 15 weeks post tamoxifen using several criteria ( see Materials and methods , Video 4 , Figure 1D ) . Any cells that did not match criteria were scored as ‘unknown . ’ Examples of tdTomato-positive type I and II HCs are shown in Figure 6B–B”’ and Figure 6C–C”’ , respectively . At one week post tamoxifen , we detected 4 . 0 ( ±4 . 5 ) tdTomato-positive type I and 8 . 5 ( ±4 . 8 ) tdTomato-positive type II HCs per utricle ( Figure 6D , Figure 6—source data 1 ) , which were not significantly different from age-matched no-tamoxifen controls ( Figure 6D , Figure 6—source data 1 ) . In tamoxifen-treated mice , the number of tdTomato-positive type II HCs per utricle increased over time from 8 . 5 ( ±4 . 8 ) cells at one week post tamoxifen to 36 . 6 ( ±7 . 7 ) cells at 15 weeks post tamoxifen ( p<0 . 0001 for effect of treatment/time post tamoxifen; p<0 . 0001 for effect of HC type , determined by two-way ANOVA followed by Tukey’s multiple comparisons test ) . However , there was no significant change in numbers of tdTomato-positive HCs that were type I or ‘unknown’ during this time ( Figure 6D , Figure 6—source data 1 ) . Linear regression analysis showed that the average number of tdTomato-positive type II HCs increased by ~2 cells per week ( Figure 6E ) . In mice that did not receive tamoxifen , there was no significant change in tdTomato-positive HCs of any type over time ( Figure 6D ) . 10 . 7554/eLife . 18128 . 016Video 4 . Clearly stratified type I and type II HCs in the adult mouse utricle . This movie begins at the cuticular plate of the HCs , proceeds through the level of HC necks , type II HC nuclei ( HCII ) , type I HC nuclei ( HCI ) , and ends in the SC nuclear layer where type II HC basolateral processes are located . An example of a type II HC basolateral process is shown by the arrow . Myosin VIIa antibodies were used to label all HCs ( green ) . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 016 To test if new HCs derive from division of SCs , we administered the thymidine analog 5’-bromodeoxyurdine ( BrdU ) to normal adult mice . BrdU was delivered using two different routes of administration ( intraperitoneal injection and drinking water ) and different dosing regimens for up to two weeks of continuous administration . We examined utricles after BrdU treatment from 18 adult mice , including both C57Bl/6J and Swiss Webster strains ( Table 1 ) . We routinely detected strongly BrdU-labeled nuclei in within-animal positive control tissue ( intestinal epithelium and utricular connective tissue ) , but we found only one example of a BrdU-labeled pair of cells in the utricular macula ( not shown ) . Therefore , if SCs divide during HC turnover , they do so rarely . These observations are consistent with the slow rate of HC addition that we observed and with previous studies of cell division in the macula of normal mice ( Li and Forge , 1997; Kuntz and Oesterle , 1998 ) . 10 . 7554/eLife . 18128 . 017Table 1 . Detection of utricular SC proliferation under normal conditions . We used two delivery methods and 3 different time periods of exposure to administer BrdU to mice to investigate cell proliferation in the utricular sensory epithelium ( SE ) in two strains of mice ( C57Bl/6J and Swiss Webster ) that were older than 6 weeks . BrdU was either injected intraperitoneally ( IP ) or administered per os ( PO ) in the drinking water . SD , one standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 017Number of BrdU+ cells per SE Mean ( SD ) Delivery route Delivery rate DoseDays of exposureNumber of mice/utricles ( strain ) 0 . 25 ( 0 . 50 ) IP1x/day50 mg/kg 72/4 ( C57Bl/6J ) 0 ( 0 ) IP2x/day50 mg/kg 33/3 ( Swiss Webster ) 0 ( 0 ) POContinuous2 mg/ml 77/14 ( C57Bl/6J ) 0 ( 0 ) POContinuous2 mg/ml 73/6 ( Swiss Webster ) 0 ( 0 ) POContinuous2 mg/ml 143/6 ( Swiss Webster ) Our observations indicate that , although type I and type II HCs are cleared from the macula under normal conditions , only type II HCs are replaced by Plp1-CreERT2-expressing SCs . This finding raised the question of how type I HCs are maintained in adulthood . We hypothesized that type II HCs might convert into type I HCs . To test this hypothesis , we fate-mapped type II HCs under normal conditions in adult utricles using Atoh1-CreERTM:ROSA26tdTomato mice in which CreER expression is driven by an Atoh1 enhancer ( Chow et al . , 2006 ) . After tamoxifen administration to 6-week-old Atoh1-CreERTM:ROSA26tdTomato mice , we used myosin VIIa immunolabeling and cell classification criteria ( Figure 1D , Materials and methods , Video 5 ) to determine that no SCs were tdTomato-positive at one week post tamoxifen . Using the same criteria , we determined that 92 . 7% ( ±2 . 8% ) of tdTomato-labeled cells were type II HCs and 5 . 9% ( ±3 . 0% ) were type I HCs ( Figure 7C–F” , Figure 7—source data 1 ) . Assuming a total utricular HC population of 3800 ( Golub et al . , 2012 ) and a type I:type II ratio of 1 . 17:1 ( Pujol et al . , 2014 ) , we estimate that ~2 . 5% of type I HCs and ~39 . 3% of type II HCs were labeled in Atoh1-CreERTM:ROSA26tdTomato mouse utricles . Age-matched control mice that did not receive tamoxifen had fewer than 15 labeled HCs per utricle , indicating a low level of Cre activity in the absence of tamoxifen ( Figure 7B , Figure 7—source data 1 ) . 10 . 7554/eLife . 18128 . 018Video 5 . Type I and type II HCs labeled with Atoh1-CreERTM:ROSA26tdTomato . This movie was constructed from optical sections collected from the area shown in Figure 7E–F’’ . The movie begins at the cuticular plate , passes through HC necks , type II HC nuclei , type I HC nuclei , SC nuclei , and ends at the basal lamina . About 10 type II HCs and one type I HC ( arrow ) are labeled with tdTomato ( red ) . All nuclei are labeled with DAPI ( blue ) . At the level of the type II HC nuclei , the type I HC neck ( arrow ) can be seen . At the level of the SC nuclei , tdTomato-labeled type II basolateral processes are observed . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 01810 . 7554/eLife . 18128 . 019Figure 7 . Atoh1-CreERTM-labeled type II HCs do not convert into type I HCs over 8 months in adult mouse utricles under normal conditions . ( A ) Experimental timeline . Atoh1-CreERTM: ROSA26tdTomato mice were injected with tamoxifen ( Tam ) at 6 weeks of age and were sacrificed ( Sac ) at 7 , 16 , 21 , and 38 weeks of age ( corresponding to 1 , 10 , 15 , and 32 weeks post tamoxifen ) . ( B , C ) Confocal projection images of whole utricles from 7-week-old Atoh1-CreERTM:ROSA26tdTomato mice showing the numbers and distribution of tdTomato-positive SCs ( magenta ) in the utricular sensory epithelium of mice that received no tamoxifen ( B ) or mice that received tamoxifen ( C ) S , striola . ( D–F” ) Examples of tdTomato-labeled type I and type II HCs . ( D ) A schematic of a cross-section through the utricular macula , with brackets indicating the optical sections generated for F–F” . I , type I HC; II , type II HC . ( E ) Xz view ( similar to D ) of images used to generate F–F” , taken from the extrastriolar region of an Atoh1-CreERTM:ROSA26tdTomato mouse at one week post tamoxifen . Brackets indicate the location of optical sections shown in F–F” . Arrow points to a tdTomato-positive type II HC ( II ) with a basolateral process ( p ) . Arrowhead points to a tdTomato-positive type I HC ( I ) with a thin neck and more basally located nucleus than the type II HC . tdTomato is shown in magenta; DAPI is shown in blue . ( F–F” ) Progressively deeper optical xy sections through the utricular macula . Arrows and arrowheads point to same HCs as shown in E . Note that the type II HC has an apically located nucleus ( arrow , [E , F] ) and a basolateral process ( arrows , E , F’ , F”; p , E , F’’ ) . Note that the type I HC ( arrowhead , [E] ) has a thin neck ( arrowhead , [F] ) , a basally located nucleus ( arrowhead , [F’] ) and no basolateral process ( arrowhead , [F”] ) . Scale bar in C is 100 μm and applies to B , C . Scale bar in E is 6 µm and applies to E–F” . ( G ) Number of tdTomato-positive type I HCs at 1 , 10 , 15 , and 32 weeks post tamoxifen normalized to the total number tdTomato-positive cells at each timepoint ( see Figure 7–source data 1 for raw data ) . No significant differences in tdTomato-positive type I HCs were observed across time ( determined by ANCOVA , p=0 . 103; n = 4 ) . Data are expressed as mean ±1 standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 01910 . 7554/eLife . 18128 . 020Figure 7—source data 1 . Quantification of tdTomato-labeled HCs in Atoh1-CreERTM:ROSA26tdTomato utricles . Mean ( one standard deviation , SD ) and 95% confidence interval ( CI ) of the number and percentage of tdTomato-labeled HCs per utricle , categorized by type [type I , type II , or unknown ( Un . ) ] . Atoh1-CreERTM:ROSA26tdTomato mice were given tamoxifen at 6 weeks ( wks ) of age ( right ) or were age-matched littermate controls that did not receive tamoxifen ( left ) . For the graph in Figure 7G , we present the number of tdTomato-positive type I HCs at 1 , 10 , 15 , and 32 weeks post tamoxifen ( shown here ) , normalized to the total number of tdTomato-positive cells at each timepoint . n , number of mice . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 020 Next , we tested if the number of tdTomato-labeled type I HCs increased over time , as one would predict if significant numbers of labeled type II HCs were to convert into type I HCs . We administered tamoxifen to Atoh1-CreERTM:ROSA26tdTomato mice at 6 weeks of age and euthanized them at 10 , 15 , or 32 weeks post tamoxifen ( Figure 7A ) . Control mice that did not receive tamoxifen were euthanized at 16 , 21 , and 38 weeks of age ( Figure 7A ) . Using analysis of covariance ( ANCOVA ) to control for variability in the total number of tdTomato-labeled cells across timepoints , we found no significant change in tdTomato-labeled type I HCs between 1 and 32 weeks post tamoxifen ( p=0 . 103; Figure 7G , Figure 7—source data 1 ) . These observations demonstrate that few , if any , type II HCs labeled in Atoh1-CreERTM:ROSA26tdTomato mice transdifferentiated into type I HCs during the 8-month period we examined . Our results demonstrate that SCs transdifferentiate into type II HCs at a low rate under normal conditions . To determine if SC transdifferentiation is increased after destruction of HCs , we bred Plp1-CreERT2:ROSA26tdTomato mice with Pou4f3DTR mice , in which the human diphtheria toxin receptor ( DTR ) is knocked into the endogenous coding region for Pou4f3 , a HC-specific transcription factor ( Golub et al . , 2012 ) . Injection of diphtheria toxin ( DT ) to mice with a single Pou4f3DTR allele kills all but 6% of HCs in the utricle , and over the next two months , HC numbers are restored to 17% of normal levels ( Golub et al . , 2012 ) . This study found that only type II HCs are regenerated . Since type II HCs constitute half of the normal HC population , we assume that ~34% of the type II HC population is regenerated . We injected Plp1-CreERT2:ROSA26tdTomato:Pou4f3DTR mice with tamoxifen at 6 weeks of age , then with DT one week later ( Figure 8A ) . Utricles were collected and labeled with antibodies against myosin VIIa at 3 weeks post DT ( equivalent to 4 weeks post tamoxifen ) . We expected , based on our prior study ( Golub et al . , 2012 ) , that HC regeneration would be well underway at this point and we would detect tdTomato-labeled HCs if SCs generated new HCs . Age-matched Plp1-CreERT2:ROSA26tdTomato mice ( lacking the Pou4f3DTR allele ) were treated identically and served as negative controls , since DT causes no vestibular HC loss when administered to mice lacking the Pou4f3DTR allele ( Golub et al . , 2012 ) . 10 . 7554/eLife . 18128 . 021Figure 8 . DT-mediated HC damage increases SC-to-HC transition in adult mouse utricles . ( A ) Experimental timeline . Plp1-CreERT2:ROSA26tdTomato:Pou4f3DTR ( damaged ) mice and Plp1-CreERT2:ROSA26tdTomato ( control ) mice were injected with tamoxifen ( Tam ) at 9 weeks of age to label SCs with tdTomato , injected with diphtheria toxin ( DT ) at 10 weeks of age to kill HCs , and sacrificed ( Sac ) at 13 weeks of age ( corresponding to 4 weeks post tamoxifen and 3 weeks post DT ) . ( B ) Utricles from control mice ( Plp1-CreERT2:ROSA26tdTomato ) that received DT injection but lacked the Pou4f3DTR allele exhibited normal-appearing HC densities [myosin VIIa ( Myo ) , green] . S , striola . ( C ) Utricles from damaged Plp1-CreERT2:ROSA26tdTomato:Pou4f3DTR mice sacrificed 3 weeks post DT had fewer HCs ( Myo , green ) . SCs were labeled with tdTomato ( magenta ) in both control and damaged adult mouse utricles ( B , C ) . S , striola . ( D , E ) Confocal optical images of the extrastriolar region from similar utricles as shown in B and C , acquired at higher magnification . ( D’ , E’ ) Higher magnifications of the boxed areas in D , E . tdTomato-expressing HCs ( arrows ) were detected in control ( D , D’ ) and damaged ( E , E’ ) utricles . Scale bar in C is 100 µm and applies to B , C . Scale bar in E is 20 µm and applies to D , E . Scale bar in E’ is 10 µm and applies to D’ , E’ . ( F ) Total number of tdTomato-expressing HCs at 13 weeks of age ( equivalent to 3 weeks post DT and 4 weeks post tamoxifen ) in control ( white bar , n = 3 ) and damaged ( black bar , n = 4 ) utricles ( see Figure 8—source data 1 ) . Damaged utricles had significantly more tdTomato-labeled HCs compared to control utricles determined by an unpaired , two-tailed Student’s t-test ( p=0 . 0346 ) . Data are expressed as mean ±1 standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 02110 . 7554/eLife . 18128 . 022Figure 8—source data 1 . Quantification of tdTomato-labeled HCs after HC damage in Plp1-CreERT2:ROSA26tdTomato:Pou4f3DTR and control utricles . Mean ( one standard deviation , SD ) and 95% confidence interval ( CI ) of the number and percentage of tdTomato-labeled HCs per utricle . Plp1-CreERT2:ROSA26tdTomato:Pou4f3DTR mice ( damaged ) were given tamoxifen at 9 weeks of age , DT at 10 weeks of age and analyzed at 13 weeks of age . Controls were littermates that did not contain the Pou4f3DTR allele but received both the tamoxifen and DT injections . n , number of mice . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 022 Utricles from control mice showed normal-appearing HC numbers ( Figure 8B , D–D’ ) . By contrast , utricles from Plp1-CreERT2:ROSA26tdTomato:Pou4f3DTR mice had a large reduction in HC numbers after DT treatment ( Figure 8C , E–E’ ) . We counted 424 . 8 ( ±119 . 0; 95% confidence interval: 308 . 1–541 . 4; n = 4 ) HCs per utricle ( ~12% of controls ) , the vast majority of which were type II , as expected . When we scored HCs for tdTomato labeling , we found that significantly more tdTomato-labeled type II HCs were present in damaged versus control utricles ( Figure 8F; p=0 . 0346; unpaired , two-tailed Student’s t-test ) . During the 4-week period after tamoxifen injection , SCs generated 6 times more tdTomato-labeled type II HCs in damaged utricles ( 101 . 3 ± 49 . 7 ) compared to controls ( 16 . 3 ± 4 . 5 ) ( Figure 8F , Figure 8—source data 1 ) . Assuming a HC population of 3800 ( Golub et al . , 2012 ) , 0 . 4% ( ±0 . 1% ) of HCs were tdTomato-labeled in control utricles . In contrast , 23 . 9% ( ±8 . 2% ) of HCs were tdTomato-labeled in damaged utricles ( Figure 8—source data 1 ) . These results demonstrate that SCs in adult mouse utricles mount a truly regenerative response to damage , significantly increasing transdifferentiation into type II HCs relative to baseline levels . Our findings raised the question: do mechanisms of HC clearance following DT-induced HC damage resemble those during normal conditions ? To address this , we labeled utricles from Pou4f3DTR mice at 4 , 7 , 14 , 40 , 90 , or 120 days post DT , as well as from littermates that lacked the Pou4f3DTR allele ( labeled as 0 day post DT ) with antibodies to myosin VIIa , TUNEL , and DAPI ( Figure 9A–D ) . Degenerating HCs were abundant at 4 and 7 days post DT ( Figure 9B–C ) , as described previously ( Golub et al . , 2012 ) . Degenerating HCs , seen in their normal positions , had myosin VIIa in their nuclei , which was not observed in utricles under normal conditions ( Figure 9A–B” ) . Further , degenerating HCs had nuclei with apoptotic features: some nuclei had abnormal shapes and/or condensed chromatin ( Figure 9B–B” ) , and some nuclei were TUNEL-positive ( Figure 9C ) . By contrast , HCs that were being cleared under normal conditions were ectopic ( located near the basal lamina ) , and most had healthy-appearing nuclei ( Figure 1E–E’” ) . 10 . 7554/eLife . 18128 . 023Figure 9 . DT-mediated HC damage induces apoptosis and a small increase in phagosome numbers in adult mouse utricles . ( A–A” ) The same field of a control ( Pou4f3DTR-negative ) utricle , focused on the type II HC layer . ( B–B” ) The same field of a Pou4f3DTR utricle at 4 days post DT , focused on the HC layer . Green arrowheads point to myosin VIIa-labeled HCs ( Myo , magenta in A , B; white in A’ , B’ ) , while green arrows point to a nucleus ( DAPI , blue in B; white in B’’ ) with condensed chromatin . ( C ) Projection image of a Pou4f3DTR utricle at 7 days post DT . Arrowheads point to F-actin-rich ( green ) phagosomes , while arrows point to TUNEL-labeled ( magenta ) DNA . Scale bar in A is 6 µm and applies to A–B” . Scale bar in C is 5 µm . ( D ) Total number of phagosomes per Pou4f3DTR utricle at different times post DT . See Figure 9–source data 1 for raw data . There were significantly more phagosomes at 4 and 7 days post DT compared to control littermates lacking the Pou4f3DTR allele ( 0 day post DT ) determined by a one-way ANOVA ( p<0 . 0001 ) followed by a Dunnett’s multiple comparisons test ( *p<0 . 05: n = 3–7 ) . Data are expressed as mean +1 standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 02310 . 7554/eLife . 18128 . 024Figure 9—source data 1 . Quantification of phagosomes in Pou4f3DTR mice after HC damage . Mean ( one standard deviation , SD ) and 95% confidence interval ( CI ) of number of F-actin ( phalloidin ) -labeled phagosomes per utricle . Littermates lacking the Pou4f3DTR allele were used as control and labeled as 0 day post DT . n , number of mice . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 024 Somewhat surprisingly , few TUNEL-positive nuclei seen after DT treatment were associated with F-actin-rich phagosomes ( Figure 9C ) . The number of phagosomes per utricle increased significantly over time ( p=0 . 0001; one-way ANOVA; n = 3–7 ) , with 4 and 7 days post DT showing a ~60% increase relative to controls ( p<0 . 05; Dunnett’s multiple comparisons test ) . Phagosome numbers returned to control levels at later times post DT ( Figure 9D , Figure 9—source data 1 ) . This increase was much lower than expected if F-actin-rich phagosomes play a major role in HC clearance after DT-induced HC death , since we previously showed a decrease of >2000 HCs during the first 7 days post DT in Pou4f3DTR mice ( Golub et al . , 2012 ) .
SCs , which are epithelial cells that resemble glia in several respects , serve many functions in HC epithelia , including maintaining structural integrity , clearing ions and neurotransmitters from extracellular space , and releasing neurotrophins to maintain synapses between HCs and neurons ( reviewed in Monzack and Cunningham , 2013; Wan et al . , 2013 ) . SCs also phagocytose HCs that are injured by ototoxins or intense noise ( Abrashkin et al . , 2006; Bird et al . , 2010; Anttonen et al . , 2014 ) . Our results demonstrate an additional , previously unknown role of SCs as phagocytes of type I and II vestibular HCs under normal conditions . While data from both Plp1-CreERT2:ROSA26tdTomato and Lfng-eGFP mice support the hypothesis that SCs act as phagocytes to remove HCs , we cannot rule out that other cell types in the organ also play this role . For instance , Schwann cells were also labeled by tdTomato in utricles of Plp1-CreERT2:ROSA26tdTomato mice , and they have phagocytic capabilities in the peripheral nervous system ( Band et al . , 1986; Reichert et al . , 1994 ) . We attempted to assess Schwann cell contribution to phagosomes , but we were unable to identify antibodies or CreER mouse lines that labeled Schwann cells without also labeling SCs . Plp1-CreERT2:ROSA26tdTomato mice also labeled cells in the transitional epithelium . However , these cells are unlikely sources of phagosomes , since the majority of phagosomes we observed were located in the central region of the macula , away from the transitional epithelium . Our study did not determine the reason why HC clearance is concentrated centrally . This finding is somewhat counterintuitive , since HC addition appears to be most abundant in the periphery . However , it is possible that HC clearance and HC addition are not spatially coupled . During development , addition of utricular HCs in rodents occurs in a central-to-peripheral gradient ( Sans and Chat , 1982; Burns et al . , 2012 ) . Since we found HC clearance to be focused centrally , we hypothesize that the oldest HCs are being removed . During adulthood , peripherally added HCs might migrate toward the center over time , where they are eventually cleared . These hypotheses remain to be tested . Since our study involved analysis of static images , we were unable to determine the temporal sequence of events , how quickly HCs are removed , how many HCs are cleared by SCs per week , the dynamics of phagosome formation , or the lifespan of phagosomes . However , based on our observations , we generated a model ( Figure 10A ) that should be tested using live-cell imaging and serial immuno-electron microscopy . We propose that the first sign a HC has been ‘selected’ for clearance is the appearance of a large actin spike in its cytoplasm . Actin-rich cables or ‘cytocauds’ are also observed in HC cytoplasm of adult utricles in mice ( Sobin et al . , 1982 ) , guinea pigs ( Flock et al . , 1979; Kanzaki et al . , 2002 ) , and humans ( Taylor et al . , 2015 ) . We suspect this actin spike is either derived from the HC itself or from a SC that has been signaled to ‘attack’ the HC . We hypothesize that this spike extends and becomes anchored basally , at which point the HC body is translocated to the basal compartment . Once translocated , adjacent SCs create basket-like phagosomes and consume the HC . However , it is also possible that the actin spike does not play a role in HC clearance . Because we did not detect stereocilia in translocated HCs , we presume that the apical portion of each HC is cleaved by SCs and ejected apically before translocation , similar to what occurs after aminoglycoside toxicity ( Meiteles and Raphael , 1994; Li et al . , 1995; Bird et al . , 2010; Monzack et al . , 2015 ) . However , it is also possible that stereocilia proteins are degraded in HCs prior to , or during , translocation . 10 . 7554/eLife . 18128 . 025Figure 10 . Model of HC turnover in adult mouse utricles under normal conditions . ( A ) Model of HC clearance . An actin spike ( green ) forms to connect SCs ( grey ) and a HC targeted for clearance ( magenta ) . The apical portion of SCs converges and cleaves off the top of the HC , including the stereocilia . The actin spike aids the translocation of the HC to the SC nuclear layer , where SCs form an actin-rich phagosome and engulf the HC . Once the HC is removed from the sensory epithelium , an empty ring- or basket-like actin structure may remain until it is resorbed by SCs . ( B , C ) Model of type II HC addition . ( B ) Plp1-CreERT2-expressing SCs ( grey ) give rise to type II HCs ( magenta ) by first translocating their nuclei into the HC layer and becoming an intermediate cell type ( dark magenta ) . Type II HCs that express Atoh1-CreERTM do not transdifferentiate into type I HCs . ( C ) Plp1-CreERT2-expressing SCs ( grey ) do not directly generate type I HCs ( magenta ) in adulthood . DOI: http://dx . doi . org/10 . 7554/eLife . 18128 . 025 SC-derived phagosomes that clear HCs during turnover share some features with those that clear HCs after drug damage , including their high actin composition and shape . However in normal utricles , we found that HCs are pierced by a single actin spike and translocated to a basal position , which has not been reported after damage . Although the number of phagosomes increased after DT-mediated HC death , the increase was not proportionate to HC loss , and many apoptotic HCs were not localized near phagosomes . These observations suggest that F-actin-rich phagosomes derived from SCs play a minor role in removing HCs in mouse utricles after DT treatment , which is consistent with previous observations ( Kaur et al . , 2015 ) . By contrast , F-actin-rich phagosomes derived from SCs actively clear utricular HCs following aminoglycoside treatments in vitro ( Bird et al . , 2010; Monzack et al . , 2015 ) . Further , we found that naturally occurring HC death does not trigger infiltration of immune-derived macrophages , which is robust after DT treatment ( Kaur et al . , 2015 ) . Additional studies are needed to understand why different forms of HC degeneration trigger distinct forms of cell clearance . It is not clear why certain HCs are removed from the epithelium under normal conditions . Most HCs associated with phagosomes appear normal aside from their unusual basal position . They have a typical nucleus and protein expression and show no nuclear condensation that occurs during apoptosis , nor cell swelling characteristic of necrosis ( reviewed in Nikoletopoulou et al . , 2013 ) . Nonetheless , HCs may be phagocytosed because they become damaged due to normal ‘wear and tear’ under physiological conditions . Thus , it is important to define the triggers for HC clearance . These could include degeneration of HCs or changes in innervation , cell membrane integrity , or other HC features . Further , SCs might act as primary phagocytes ( Brown and Neher , 2012; Monzack and Cunningham , 2013 ) . We demonstrate that SCs play a second critical role in maintaining vestibular epithelia under normal conditions: they form new HCs , presumably replacing those lost during clearance . Consistent with this , small numbers of immature HCs were detected in normal utricles of adult mice using two markers: PCDH15-CD2 and ATOH1-GFP . Recently , evidence for immature stereocilia bundles was found in utricles of aged humans ( >60 years old ) ( Taylor et al . , 2015 ) , suggesting HC turnover may also occur in adult primates . Two prior studies did not find evidence for ATOH1-immunolabeled cells ( Wang et al . , 2010 ) or cells with Atoh1 enhancer activity ( Lin et al . , 2011 ) in normal utricles from adult mice . These contrasting results may be due to different sensitivities of the methods used to assess Atoh1 expression . Fate-mapping of SCs in Plp1-CreERT2:ROSA26tdTomato mice provided additional evidence for new HC addition . We assume SCs are the sole source of new HCs in our experiments based on a large body of literature showing that SCs are progenitors to vestibular HCs in non-mammalian vertebrates during both regeneration and turnover ( reviewed in Corwin and Oberholtzer , 1997; Warchol , 2011 ) . However , transitional epithelial cells and Schwann cells were also labeled in utricles of Plp1-CreERT2:ROSA26tdTomato mice . Therefore , additional CreER lines are needed to fate-map these cell populations and determine if they are unexpected sources of HCs during turnover . Our data showed that SCs generate new type II HCs , but not type I HCs , under normal conditions ( Figure 10B , C ) . These findings were surprising , since we detected clearance of both HC types . We hypothesized that type I HCs might be replaced by type II HCs since type II-to-I HC conversion has been suggested to occur after ototoxin-induced HC loss in avian utricles ( Weisleder and Rubel , 1993; Zakir and Dickman , 2006 ) . Additional support for this hypothesis was provided by Kirkegaard and Jørgensen ( 2001 ) , who detected individual HCs that appeared to be a hybrid with properties of both type I and type II HCs in vestibular organs of adult bats . However , when we fate-mapped type II HCs in mice under normal conditions , there was no evidence of type II-to-I conversion over an 8-month period of early adulthood ( Figure 10B ) . However , the rate of conversion could take longer than 8 months . Further , type I HCs may derive from type II HCs that were not fate-mapped in Atoh1-CreERTM:ROSA26tdTomato mice or from SCs that were not fate-mapped in Plp1-CreERT2:ROSA26tdTomato mice . It should be noted , however , that only type II HCs are replaced after drug-induced damage in adult guinea pigs ( Forge et al . , 1998 ) and mice ( Kawamoto et al . , 2009; Golub et al . , 2012 ) , supporting the interpretation that type I HC replacement is attenuated or perhaps impossible in adult mice . Regenerated HCs share many features with neighboring mature type II HCs , including myosin VIIa immunoreactivity , properly positioned nuclei , proper relative neck thickness , well defined stereocilia bundles , and basolateral processes . However , since we could not birth-date HCs in this study , we were not able to determine the degree to which new type II HCs are mature . Further , we did not perform any analyses that would inform on the functional status of the new HCs . Therefore , additional work is required to assess if HCs replaced in adult animals under normal conditions or after damage are functionally mature . We were unable to distinguish whether HCs added during normal conditions derive from the progeny of SC division or via direct transdifferentiation , during which SCs phenotypically convert into HCs without dividing ( reviewed in Stone and Cotanche , 2007 ) . Using multiple methods for BrdU administration , we only detected BrdU labeling in rare cells within the utricular macula , and their identity was not established . This finding is consistent with other studies that assessed SC division in normal utricles in vivo ( Li and Forge , 1997; Kuntz and Oesterle , 1998 ) . In contrast , dividing SCs are readily detected during HC turnover in mature birds ( Roberson et al . , 1992; Kil et al . , 1997; Stone et al . , 1999 ) . We cannot rule out , however , that SCs divide at a very low rate and our methods were insufficient to record them . Indeed , mammalian utricles that were cultured and treated with aminoglycoside antibiotics generated new HC-like cells and SCs by mitotic division ( Warchol et al . , 1993 ) . One possible interpretation for the discrepancy is that culture conditions promote utricular SC division . Prior fate-mapping studies showed that SCs form new vestibular HCs in juvenile and adult mice after HC damage ( Lin et al . , 2011; Slowik and Bermingham-McDonogh , 2013; Wang et al . , 2015 ) . In the present study , we found that , upon HC destruction , SC transdifferentiation into type II HCs increases by 6-fold over type II HC addition under normal conditions . This demonstrates that adult mammals are able to upregulate vestibular HC addition , relative to normal levels , upon HC injury . We can use several approaches to estimate the rate of HC turnover . SC fate-mapping in mice with a mixed genetic background indicated that ~2 HCs are added to each utricle per week . This rate is likely a slight underestimate , since we only fate-mapped ~90% of extrastriolar and ~70% of striolar SCs in the utricle . Further , there could also be other sources of added HCs that were not fate-mapped , such as cells in the transitional epithelium or even in the stroma . We assessed the rate of HC addition by examining two markers for immature HCs ( PCDH15-CD2 and ATOH1-GFP ) in mice with a range of backgrounds . These markers labeled ~20 and ~82 HCs per utricle , respectively , which suggests a higher rate of HC addition than fate-mapping . However , we do not know how long each marker is expressed in differentiating cells , so we cannot use these analyses to define the rate of HC addition . We can also assess the rate of turnover by estimating how many HCs die in a given period . Swiss Webster mice had ~50 phagosomes per utricle , of which ~12 were associated with HCs , while Plp1-CreERT2:ROSA26tdTomato mice on a mixed background had ~28 phagosomes per utricle , of which ~5 were targeting HCs . Only 10–15 phagosomes were detected in CBA/CaJ or C57Bl/6J mice . It is unclear why Swiss Webster mice have significantly more phagosomes than the other tested mouse strains . They may have a higher rate of HC turnover , or their phagosomes may remain in the utricle for a longer period of time than in other strains . Although we cannot deduce the lifespan of a phagosome or the duration of HC clearance in the current study , Monzack et al . ( 2015 ) showed in aminoglycoside-treated mouse utricles that a phagosome engulfed a HC , cleared it , and began retracting within 8 hr . If this time frame is similar in normal utricles , and we use the conservative estimate from Plp1-CreERT2:ROSA26tdTomato ( mixed background ) mice that 5 HCs at any given time are being targeted by phagosomes generated within an 8 hr period , then we can estimate that one HC is targeted for removal every 1 . 6 hr , which is equivalent to 15 HCs per day or 105 HCs per week . Since numbers of HCs are maintained in adult mice , this could suggest that as many as 105 HCs are added per week , which is considerably more than indicated by SC fate-mapping with Plp1-CreERT2:ROSA26tdTomato mice . Accordingly , live-cell imaging must be performed to define the lifespan of a phagosome in normal utricles , which may differ from the lifespan of phagosomes after HC damage . If we make a conservative estimate , based on SC fate-mapping alone , that 2 HCs are turned over per week , and we consider the mouse’s lifespan to be 18 months , with an average of 4 . 3 weeks/month , then we can estimate that 155 HCs , or 4% of the HC population , are turned over during a mouse’s lifetime . It seems unlikely that such a small degree of HC turnover would have a significant impact on vestibular homeostasis . However , we found evidence that type I HCs are cleared , but no evidence that type I HCs are replaced . A progressive loss of type I HCs over time or disruption in the balance of type II HC clearance and addition , for example during aging , could result in vestibular dysfunction . Park et al . ( 1987 ) measured a ~14% reduction in utricular HCs in old C57Bl/6NNia mice , and this reduction was similar for both type I and II HCs . By contrast , Kirkegaard and Nyengaard ( 2005 ) found no evidence of HC loss in aged outbred mice . Declines in vestibular function accompanying aging have been reported in some strains of mice ( Shiga et al . , 2005; Mock et al . , 2011 ) and in humans ( e . g . , Agrawal et al . , 2009 ) , so it is important to further investigate if HC turnover is reduced with aging .
Swiss Webster ( stock #689 ) , CBA/CaJ ( stock #654 ) , C57Bl/6J ( stock #664 ) , Plp1-CreERT2 ( stock #5975; RRID:MGI:3696409; C57Bl/6J background; Doerflinger et al . , 2003 ) , ROSA26CAG-loxP-stop-loxP-tdTomato ( ROSA26tdTomato ) ( also called Ai14 , stock #7908; RRID:IMSR_JAX:007908; C57Bl/6J background; Madisen et al . , 2010 ) , and Atoh1GFP ( stock #13593; RRID:IMSR_JAX:013593; C57Bl/6J background; Rose et al . , 2009 ) mice were purchased from The Jackson Laboratory ( Bar Harbor , ME ) . Lfng-eGFP mice ( RRID:MMRRC_015881-UCD , CD1 background ) were generated by the GENSTAT program ( Gong et al . , 2003 ) , were provided by Dr . Andrew Groves ( Baylor College of Medicine , Waco , TX ) for this study , and are available from Mutant Mouse Resource and Research Centers ( strain name: B6;FVB-Tg[Lfng-EGFP]HM340Gsat/Mmucd ) . Pou4f3DTR mice ( C57Bl/6J or CBA/CaJ background; Golub et al . , 2012 ) were provided by Dr . Edwin Rubel ( University of Washington , Seattle , WA ) and Atoh1-CreERTM mice ( RRID:MMRRC_029581-UNC; FVB/NJ background; Chow et al . , 2006 ) were provided by Dr . Suzanne Baker ( St . Jude Children’s Research Hospital , Memphis , TN ) . Genotyping for Plp1-CreERT2 , ROSA26tdTomato , Atoh1GFP , and Atoh1-CreERTM mice was performed by Transnetyx , Inc . ( Cordova , TN ) . Genotyping for Pou4f3DTR and Lfng-eGFP mice was described previously ( Golub et al . , 2012; Burns et al . , 2015 ) . Both genders were used in all studies . Atoh1GFP/GFP mice carried two alleles ( homozygous ) , and Plp1-CreERT2 , ROSA26tdTomato , Pou4f3DTR , Atoh1-CreERTM , and Lfng-eGFP mice were heterozygotes . Crosses led to progeny with mixed strain background . All procedures were conducted in accordance with approved animal protocols from the Institutional Animal Care and Use Committees at the University of Washington ( Seattle , WA ) and Southern Illinois University School of Medicine ( Springfield , IL ) . For fate-mapping and phagosome analyses in normal mice , tamoxifen [9 mg/40 g body weight , intraperitoneal injection ( IP ) ; Sigma-Aldrich ( St . Louis , MO ) ] was injected once on two consecutive days in 6-week-old Plp1-CreERT2:ROSA26tdTomato and Atoh1-CreERTM:ROSA26tdTomato mice . For fate-mapping studies in damaged utricles , tamoxifen ( 6 mg/40 g body weight , IP ) was injected once on two consecutive days in 9-week-old Plp1-CreERT2:ROSA26tdTomato:Pou4f3DTR mice , followed one week later by two intramuscular ( IM ) injections of DT . Each injection was 25 ng/g body weight , and there was a 48 hr interval between injections . DT was purchased from Sigma-Aldrich ( St . Louis , MO ) or List Biological Laboratories , Inc . ( Campbell , CA ) . The different tamoxifen doses produced similar labeling of utricular SCs . Controls for normal utricles consisted of Plp1-CreERT2:ROSA26tdTomato and Atoh1-CreERTM: ROSA26tdTomato mice that did not receive tamoxifen injection and were housed separately from tamoxifen-treated mice . Controls for damaged utricles were littermates lacking the Pou4f3DTR allele ( Plp1-CreERT2:ROSA26tdTomato mice ) that received both tamoxifen and DT injections . For phagosome analyses in damaged utricles , DT was injected as described above , and mice were euthanized 4 , 7 , 14 , 40 , 90 , or 120 days later . Controls were littermates lacking the Pou4f3DTR allele ( Plp1-CreERT2:ROSA26tdTomato mice ) that received both tamoxifen and DT injections . Methods for administering BrdU [Sigma-Aldrich ( St . Louis , MO ) ] to adult mice are described in Table 1 . Temporal bones were dissected , and a small hole in the temporal bone was generated adjacent to the oval window . Fixative ( 2% paraformaldehyde/3% glutaraldehyde in 0 . 1 M sodium phosphate buffer , pH 7 . 4 ) was injected into the hole , and temporal bones were immersion-fixed overnight at 4°C . Utricles were removed , and the otoconial membrane and otoconia were dissected away . Utricles were post-fixed in 1% osmium tetroxide and embedded in Eponate 12 Kit with DMP-30 ( Ted Pella Inc . , Redding , CA ) . Semi-thin sections ( 2 µm ) were mounted onto gel-coated slides and stained with toluidine blue dye [Sigma-Aldrich ( St . Louis , MO ) ] . Sections were imaged using a Zeiss Axioplan microscope ( Zeiss , Jena , Germany ) . Temporal bones were removed and either submerged in electron microscopy grade 4% paraformaldehyde ( Polysciences , Inc . , Warrington , PA ) overnight at room temperature or partially dissected to remove otoconia from the utricle before fixation in 4% paraformaldehyde for two hours at room temperature . After fixation , temporal bones were stored in PBS until whole utricles were dissected out of temporal bones and placed in 96-well plates for free-floating immunofluorescent labeling . Utricles were washed twice in 1X PBS , then permeabilized with 2 mg/ml bovine serum albumin [BSA , Sigma-Aldrich ( St . Louis , MO ) ] and 1 . 0% Triton X-100 ( Sigma-Aldrich , St . Loius , MO ) in 1X PBS for 1 hr at room temperature . Utricles were then incubated in blocking buffer [10% normal horse serum ( Vector Laboratories , Burlingame , CA ) and 0 . 4% Triton X-100 in 1X PBS] at room temperature for 3 hr . Primary antibodies were diluted in blocking buffer and incubated overnight at 4°C , except for the anti-BrdU , anti-GFP , and mouse anti-myosin VIIa antibodies that were incubated for 48–72 hr . The following primary antibodies were used: mouse anti-BrdU [1:300 , RRID:AB_400326 , BD Biosciences ( San Jose , CA ) , #347580]; rabbit anti-calbindin [1:200 , RRID:AB_2068336 , EMD Millipore ( Billerica , MA ) , #AB1778]; rabbit anti-calretinin [1:100 , RRID:AB_2068506 , EMD Millipore ( Billerica , MA ) , #AB5054]; rabbit anti-aCasp3 [1:200 , RRID:AB_397274 , BD Biosciences ( San Jose , CA ) , #559565]; rat anti-CD68 [1:100 , RRID:AB_566872 , AbD Serotec ( Raleigh , NC ) , #MCA A341GA]; rabbit anti-espin [1:500 , RRID:AB_2630385 , gift from Dr . Stefan Heller , Stanford University]; rabbit anti-GFP [1:250 , RRID:AB_221570 , Invitrogen ( Carlsbad , CA ) , #A6455]; rabbit anti-IBA1 [1:1000 , RRID:AB_839505 , Wako Pure Chemical Industries , LLC ( Richmond , VA ) , #019–19741]; mouse anti-myosin VIIa [1:100 , RRID:AB_2282417 , Developmental Studies Hybridoma Bank ( Iowa City , IA ) , #138–1]; rabbit anti-myosin VIIa [1:100 , RRID:AB_2314839 , Proteus Biosciences , Inc . ( Ramona , CA ) , #25–6790]; rabbit anti-PMCA2 [1:1000 , RRID:AB_2630386 , gift from Dr . Peter Barr-Gillespie , Oregon Health Sciences University]; mouse anti-POU4F3 [1:500 , RRID:AB_2167543 , Santa Cruz Biotechnology , Inc . ( Dallas , TX ) , #sc-81980]; rabbit anti-PCDH15-CD2 [1:200 , RRID:AB_2630387 , gift from Drs . Tom Friedman ( National Institutes of Health ) and Zubair Ahmed ( University of Maryland ) ]; goat anti-SOX2 [1:200 , RRID:AB_2286684 . Santa Cruz Biotechnology , Inc . ( Dallas , TX ) , #sc-17320]; and rabbit anti-tenascin [1:100 , RRID:AB_2256033 , EMD Millipore ( Billerica , MA ) , #AB19013] . After rinsing thrice with 1X PBS , utricles were incubated with Alexa Flour-conjugated secondary antibodies [1:400 in blocking buffer , Invitrogen ( Carlsbad , CA ) ] for 2 hr at room temperature . Utricles were then incubated with Alexa Flour-conjugated phalloidin [1:100 , RRID:AB_2620155 or RRID:AB_2315147 , Invitrogen ( Carlsbad , CA ) , #A12379 or #A22287] in 1X PBS for 30 min followed by a second 30 min incubation in DAPI ( Sigma-Aldrich , St . Louis , MO ) at 1 µg/ml in 1X PBS . After rinsing thrice in 1X PBS , whole utricles were mounted in Fluoromount-G ( Southern Biotech , Birmingham , AL ) between two cover slips anchored to a glass slide with putty . To detect apoptosis in whole utricles and gut tissue ( positive control ) , we utilized the ApopTag Fluorescein In Situ Apoptosis Detection Kit [#S7110 , Chemicon ( now EMD Millipore , Billerica , MA ) ] according to the manufacturer’s instructions followed by immunostaining as described above . To detect proliferating cells in whole utricles and gut tissue ( positive control ) from mice treated with BrdU , fixed tissue was first labeled with antibodies for cell specific markers as described above . After secondary antibody incubation , tissue was fixed again in 4% paraformaldehyde for 20 min at room temperature . After rinsing in 1X PBS and a second permeabilization step , the tissue was incubated in 2 N HCl in 0 . 05% Triton X-100 for 1 hr at room temperature , then rinsed in 1X PBS , and placed in blocking buffer for 30 min . Then the tissue was immunostained with anti-BrdU primary antibodies as described above . Fluorescent images were obtained with an Olympus FV-1000 microscope ( Olympus , Center Valley , PA ) . In most cases , z-series images were collected with a 60x oil objective through the entire macula , from above stereocilia bundles to the stroma below the basal lamina , at 0 . 5 or 0 . 25 µm increments . For qualitative analyses , we generated z-series in 2–3 extrastriolar and striolar regions . For quantitative analysis for Figures 3 , 5 , 6 , 7 , 8 and 9 , we generated high-resolution images or montages of the entire utricular macula . Image analysis was performed using Fiji ( http://fiji . sc/ ) . Fiji’s Cell Counter plugin was used for quantitative analyses . To determine the percentage of SCs labeled in Plp1-CreERT2:ROSA26tdTomato mice one week after tamoxifen injection , eight 50 × 50 µm regions of each utricle ( 3 in the medial extrastriola , 3 in the lateral extrastriola , and two in the striolar area ) were sampled . All SCs within the sampled regions were counted and scored as tdTomato-positive or tdTomato-negative . The average percentage of extrastriolar SCs that were tdTomato-labeled was calculated by combining medial and lateral extrastriolar regions . To assess whether phalloidin-labeled phagosomes co-localized with IBA1 immunolabeling , 301 phalloidin-labeled phagosomes from 6 Swiss Webster mice at 9 weeks of age were closely examined . To quantify PCDH15-CD2-labeled stereocilia bundles in adult CBA/CaJ and Swiss Webster mice , the 100x oil objective on the Olympus FV-1000 confocal microscope was used . The observer looked over the entire utricle and counted phalloidin-labeled stereocilia bundles co-localized with multiple puncta of PCDH15-CD2 along the length and width of the bundle . HCs were distinguished from SCs by the presence of a HC marker and the location of their nuclei , which reside above the clear monolayer of SC nuclei along the basal lamina ( Figure 1D ) . We used several morphological criteria to classify HCs as type I or type II in whole-mounted utricles . These criteria for mice are defined in many publications , including Rüsch et al . ( 1998 ) and Pujol et al . ( 2014 ) , are described in Figure 1D , and shown in Videos 4 and 5 . Almost all morphological analyses were performed in whole-mount utricles labeled with antibodies to myosin VIIa , which label the cytoplasm of both type I and type II HCs . We acquired z-series images through the entire sensory epithelium ( from the lumen through the basal lamina ) in striolar and extrastriolar regions . Images were analyzed thoroughly off-line by scanning up and down through the z-series , examining the entire body of each HC . This approach allowed us to assess all criteria in each image and compare features amongst several cells . Four morphological criteria were used to classify HCs across all regions of the utricle: thickness of the apical cytoplasm , thickness of the perinuclear cytoplasm , size and shape of the nucleus , and presence/absence of basolateral processes . The apical cytoplasm of type I HCs constricts significantly below the cuticular plate , while the apical cytoplasm of type II HCs does not . The perinuclear cytoplasm is very thin in type I HCs and considerably thicker in type II HCs . The nuclei in type I HCs are smaller and round , while the nuclei of type II HCs are larger and oblong . The clearest distinguishing features between type I and II HCs throughout the entire utricle ( striola and extrastriola ) are basolateral processes ( Pujol et al . , 2014 ) , which are cytoplasmic portions of type II HCs that extend below the nucleus and are fully labeled by anti-myosin VIIa antibodies . In the extrastriola ( the largest area in the utricle ) , there is a fifth powerful criterion for HC typing: the nuclei of type I and II HCs are located in distinct near-monolayers . Type II HC nuclei are consistently located apical to type I HC nuclei ( Videos 4 and 5 ) . In the striola , where type II and type I HC nuclei are less clearly stratified , we used the 4 criteria listed above—thickness of the apical and perinuclear cytoplasm , nuclear size and shape , and presence or absence of a basolateral process . In studies with Plp1-CreERT2:ROSA26tdTomato and Plp1-CreERT2:ROSA26tdTomato:Pou4f3DTR mice , the cytoplasm of most SCs was labeled with tdTomato . As a result , we had another criterion to help us with typing: type II HCs were completely encompassed by the tdTomato-positive SC processes , while type I HCs had an unlabeled gap around them , which was the calyx . Any cells that did not clearly match criteria described above for any analysis were scored as an ‘unknown . ’ The Cell Counter plugin in Fiji was used to mark tdTomato-labeled HCs in one utricle from an adult Plp1-CreERT2:ROSA26tdTomato mouse ( Figure 6F ) . Results were imported into Microsoft Powerpoint 2011 ( Microsoft , Redmond , WA ) to assemble the images and reconstruct an entire utricle . The file was imported into Adobe Photoshop CS 4 ( Adobe , San Jose , CA ) , where an outline was drawn around the whole utricle and dots were drawn directly over the Cell Counter labels to create the maps of tdTomato-labeled HCs . To generate the map of PCDH15-CD2-labeled bundles ( Figure 5C ) , an adult Swiss Webster utricle was scanned at high magnification using confocal microscopy , and the position of each PCDH15-CD2-labeled bundle was noted for each field on a grid . The utricle was reconstructed using Adobe Photoshop CS 4 . Data were analyzed using Graphpad Prism 5 . 0 ( Graphpad Software , La Jolla , CA ) or SAS 9 . 4 ( Cary , NC ) . All data are reported as mean ±1 standard deviation . 95% confidence intervals were calculated with Microsoft Excel 2011 ( Microsoft , Redmond , WA ) . For all studies , one utricle per animal was assessed and the ‘n’ value represents the number of mice included in the study , unless otherwise noted . n’s for each data set are included in the Results section , source data tables , and/or the Figure Legends . | Cells in the inner ear called hair cells sense sound waves and head movements , allowing us to hear and maintain balance . In non-mammals such as birds and fish , the hair cells responsible for balance die and are replaced ( in a process known as turnover ) throughout life . However , it is largely assumed that no new balance hair cells are made in adult mammals such as humans and mice . This would mean that injured hair cells are never replaced , which could cause balance problems such as dizziness over time . There have been hints in past studies that perhaps some balance hair cells die or are newly made in adult mammals . Using a variety of new cell labeling and tracking methods in different types of mutant mice , Bucks et al . now show that the turnover of balance hair cells happens in adult mice under normal conditions . Both types of balance hair cells – known as type I and type II – are removed by supporting cells that surround the hair cells . In addition , the supporting cells can convert into new type II hair cells , but not type I hair cells , and type II hair cells do not convert into new type I hair cells . To compare these results with what happens after hair cell damage , Bucks et al . injected a toxin into mutant mice to kill most hair cells . This revealed that supporting cells make 6 times as many hair cells after severe damage than under normal conditions , but still only make type II hair cells . One important issue to study next is whether type I hair cells are ever created in adulthood . Many elderly people develop balance problems that lead to catastrophic falls . Perhaps one reason this occurs is because type I hair cells cannot be replaced in humans . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"neuroscience"
] | 2017 | Supporting cells remove and replace sensory receptor hair cells in a balance organ of adult mice |
Chromatin assembly involves the combined action of ATP-dependent motor proteins and histone chaperones . Because motor proteins in chromatin assembly also function as chromatin remodeling factors , we investigated the relationship between ATP-driven chromatin assembly and chromatin remodeling in the generation of periodic nucleosome arrays . We found that chromatin remodeling-defective Chd1 motor proteins are able to catalyze ATP-dependent chromatin assembly . The resulting nucleosomes are not , however , spaced in periodic arrays . Wild-type Chd1 , but not chromatin remodeling-defective Chd1 , can catalyze the conversion of randomly-distributed nucleosomes into periodic arrays . These results reveal a functional distinction between ATP-dependent nucleosome assembly and chromatin remodeling , and suggest a model for chromatin assembly in which randomly-distributed nucleosomes are formed by the nucleosome assembly function of Chd1 , and then regularly-spaced nucleosome arrays are generated by the chromatin remodeling activity of Chd1 . These findings uncover an unforeseen level of specificity in the role of motor proteins in chromatin assembly .
The assembly of nucleosomes is necessary for the regeneration of chromatin following DNA replication , transcription , and DNA repair , and is an active ATP-driven process , as originally discovered by Worcel et al . ( Glikin et al . , 1984; Ruberti and Worcel , 1986 ) . Nucleosome assembly is facilitated by the combined activities of ATP-dependent motor proteins ( reviewed in Haushalter and Kadonaga , 2003; Lusser and Kadonaga , 2004 ) and histone chaperones ( reviewed in Corpet and Almouzni , 2009; Campos and Reinberg , 2010; Das et al . , 2010; Ransom et al . , 2010; Avvakumov et al . , 2011; Elsässer and D’Arcy , 2012; Burgess and Zhang , 2013 ) . Histone chaperones initially deposit core histones onto DNA to form non-nucleosomal histone–DNA intermediates ( prenucleosomes ) , which can then be converted into periodic arrays of canonical nucleosomes by ATP-driven motor proteins ( Torigoe et al . , 2011 ) . ATP-dependent factors that participate in chromatin assembly include Chd1 ( chromo-ATPase/helicase-DNA-binding protein 1 ) , ATRX ( alpha thalassemia/mental retardation syndrome X-linked ) , and several ISWI ( imitation switch ) -containing complexes , such as ACF ( ATP-utilizing chromatin assembly and remodeling factor ) , CHRAC ( chromatin accessibility complex ) , RSF ( remodeling and spacing factor ) , and ToRC ( Toutatis-containing chromatin remodeling complex ) ( Ito et al . , 1997; Varga-Weisz et al . , 1997; Loyola et al . , 2001; Lusser et al . , 2005; Lewis et al . , 2010; Emelyanov et al . , 2012 ) . These motor proteins exhibit both chromatin assembly and remodeling activities , and are members of the SNF2 ( sucrose non-fermenting 2 ) protein family , which comprises the ATPases that are known to be involved in chromatin remodeling ( reviewed in Clapier and Cairns , 2009; Flaus and Owen-Hughes , 2011; Hargreaves and Crabtree , 2011; Ryan and Owen-Hughes , 2011 ) . With the Chd1 , ACF , and ToRC motor proteins , it has been shown that efficient chromatin assembly requires both a histone chaperone ( such as NAP1 ) and the motor protein ( Ito et al . , 1997 , 1999; Lusser et al . , 2005; Emelyanov et al . , 2012 ) . Hence , the Chd1 , ACF , and ToRC motor proteins are not able to catalyze chromatin assembly in the absence of a histone chaperone . To determine whether nucleosome assembly requires the ability to reposition nucleosomes , we examined the properties of mutant Chd1 proteins that are defective for chromatin remodeling activity yet retain a substantial amount of their ATPase activity ( Patel et al . , 2011 ) . These studies have enabled us to identify functionally distinct roles of chromatin assembly and remodeling in the formation of periodic nucleosome arrays .
To investigate the relation between ATP-dependent chromatin assembly and ATP-dependent chromatin remodeling , we analyzed two mutant versions of Saccharomyces cerevisiae Chd1 ( yChd1 ) that exhibit substantial ( ∼40% of wild-type ) nucleosome-stimulated ATPase activity but are nearly completely deficient ( <0 . 1% of wild-type ) in chromatin remodeling activity , as assessed by the nucleosome sliding assay ( Patel et al . , 2011 ) . These chromatin remodeling-defective yChd1 proteins contain either a deletion of residues 932–940 ( Δ932–940 ) or a Trp932 to Ala substitution ( W932A; Figure 1A ) . The Δ932–940 and W932A mutations of yChd1 block its ability to couple ATPase activity to nucleosome remodeling/sliding , but still allow for robust ATPase activity . In contrast , deletions of more C-terminal regions of yChd1 ( ranging from residues 939–1010 ) cause the near complete loss of the ATPase activity ( Patel et al . , 2011; Ryan et al . , 2011 ) . Hence , for our analysis , we employed the Δ932–940 and W932A mutant yChd1 proteins . 10 . 7554/eLife . 00863 . 003Figure 1 . ATP-dependent nucleosome assembly is functionally distinct from chromatin remodeling . ( A ) Diagram of Chd1 and the conserved coupling region . The numbers below the schematic diagram and above the amino acid sequences indicate positions in S . cerevisiae Chd1 ( yChd1 ) . ( B ) Chromatin remodeling-defective mutant yChd1 proteins ( yChd1Δ932–940; yChd1W932A ) can assemble nucleosomes in an ATP-dependent manner . Chromatin assembly reactions were performed with wild-type or mutant yChd1 proteins ( 120 nM ) in the presence of either adenylyl-imidodiphosphate ( AMP-PNP ) or ATP . The efficiency of nucleosome assembly was monitored by the DNA supercoiling assay . The positions of supercoiled ( SC ) , relaxed ( Rel ) , and nicked open circular ( N ) DNAs are indicated . ( C ) Quantitative analysis of the efficiency of nucleosome assembly by mutant vs wild-type yChd1 proteins . Chromatin assembly reactions with yChd1 proteins were analyzed by the DNA supercoiling assay . The change in % supercoiling ( [Δ supercoiled DNA/total DNA ) × 100% ) vs concentration of yChd1 ( nM ) is shown . The results are presented as the mean ± standard deviation ( N ≥ 3 ) . ( D ) Agarose gel electrophoresis of DNA fragments derived from chromatin assembled with wild-type or mutant yChd1 proteins . Chromatin assembly reactions were carried out as in ( B ) , except that the concentration of Chd1 proteins was 60 nM . The reaction products were digested extensively with micrococcal nuclease ( MNase ) and subsequently deproteinized . The resulting DNA fragments were resolved on a 3% agarose gel and visualized by staining with ethidium bromide . The arrows indicate the position of DNA fragments derived from core particles . ( E ) Native nucleoprotein gel analysis of nucleosomes assembled with wild-type or mutant yChd1 proteins . Chromatin assembly reactions were carried out as in ( D ) . The reaction products were digested extensively with MNase; the resulting nucleoprotein complexes were subjected to electrophoresis on a nondenaturing 5% polyacrylamide gel; and the DNA was stained with Sybr Green I ( Invitrogen ) . The positions of core particles ( CP ) and dinucleosomes ( D ) are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 00863 . 003 We first examined whether the chromatin remodeling-defective yChd1 proteins are able to function in the ATP-dependent assembly of nucleosomes . To this end , we used the purified , defined ATP-dependent chromatin assembly system that consists of NAP1 , core histones , Chd1 , ATP , and relaxed DNA ( Lusser et al . , 2005 ) . Chromatin assembly requires both the Chd1 motor protein as well as the NAP1 histone chaperone ( Lusser et al . , 2005 ) . The extent of nucleosome assembly was monitored by the DNA supercoiling assay , in which the change in the linking number of DNA that occurs upon formation of nucleosomes in the presence of topoisomerase I is observed ( Germond et al . , 1975; Simpson et al . , 1985 ) . In these experiments , we observed that the mutant yChd1 proteins are able to assemble nucleosomes in an ATP-dependent manner ( Figure 1B ) . In the presence of adenylyl imidodiphosphate ( AMP-PNP ) , a β-γ-non-hydrolyzable analog of ATP , chromatin assembly was not observed . The extent of nucleosome assembly by the mutant proteins was about 65% of that of wild-type yChd1 ( Figure 1C ) . We further tested whether the mutant yChd1 proteins catalyze the formation of nucleosomes by micrococcal nuclease ( MNase ) digestion analysis . First , we extensively digested the reaction products with MNase , deproteinized the samples , and analyzed the resulting DNA fragments by agarose gel electrophoresis ( Figure 1D ) . With the wild-type as well as the mutant yChd1 proteins , we observed the ATP-dependent formation of DNA fragments corresponding to the length of a core particle . However , in the absence of ATP , particularly with the mutant yChd1 proteins , there were DNA fragments that are longer than those obtained in the presence of ATP . We therefore examined whether or not the species formed in the absence of ATP contained canonical nucleosomes by nondenaturing gel electrophoresis of the MNase digestion products ( Figure 1E ) . In this assay , we observed ATP-dependent stimulation of the formation of core particles by wild-type as well as mutant yChd1 proteins . Thus , the mutant yChd1 proteins catalyze the ATP-dependent formation of nucleosomes , as assessed by the generation of chromatin that yields core particles upon extensive MNase digestion . To determine the relative rates of chromatin assembly by the wild-type and mutant yChd1 proteins , we performed kinetic analyses and found that the initial rates of nucleosome assembly by the Δ932–940 and W932A proteins were approximately 8 . 5% and 12% of the rate of wild-type yChd1 ( Figure 2 ) . Hence , the chromatin-remodeling defective yChd1 proteins exhibit substantial ATP-driven nucleosome assembly activity ( ∼10% of the rate of wild-type yChd1 ) that is at least 100-fold higher than their ATP-driven chromatin remodeling/sliding activity ( <0 . 1% of wild-type yChd1 ) . The properties of wild-type and mutant yChd1 proteins indicate that the ATP-dependent catalysis of nucleosome assembly appears to be a functionally distinct process from the ATP-dependent remodeling of chromatin . 10 . 7554/eLife . 00863 . 004Figure 2 . Analysis of the initial rates of nucleosome assembly by the wild-type , W932A , and Δ932–940 yChd1 proteins . The initial rates were measured as change in % supercoiling ( [Δ supercoiled DNA/total DNA] × 100% ) / ( nM protein ) vs time ( min ) . The table summarizes the nucleosome assembly rates as mean ± standard deviation ( N = 3 ) . The relative rates are given with respect to that of the wild-type protein . DOI: http://dx . doi . org/10 . 7554/eLife . 00863 . 004 Because ATP-dependent chromatin assembly factors such as Chd1 are able to catalyze the formation of regularly-spaced nucleosome arrays ( e . g . , Ito et al . , 1997; Varga-Weisz et al . , 1997; Loyola et al . , 2001; Lusser et al . , 2005; Lewis et al . , 2010; Emelyanov et al . , 2012 ) , we examined the periodicity of the nucleosomes assembled by the mutant yChd1 proteins by using the partial MNase digestion assay . These experiments revealed that the chromatin remodeling-defective yChd1 proteins are unable to generate periodic arrays of nucleosomes ( Figure 3 ) . However , because the mutant yChd1 proteins are not fully active for nucleosome assembly ( Figures 1 and 2 ) , it was difficult to attribute the absence of periodic nucleosome arrays in the MNase assay ( Figure 3 ) to decreased efficiency of assembly or to a defect in the formation of periodic nucleosome arrays . 10 . 7554/eLife . 00863 . 005Figure 3 . Chromatin remodeling-defective yChd1 proteins are unable to generate arrays of evenly-spaced nucleosomes during chromatin assembly . ( A ) Chromatin assembly reactions were performed with wild-type or mutant yChd1 proteins ( 30 nM ) in the presence of AMP-PNP or ATP . The reaction products were subjected to partial MNase digestion analysis . ( B ) Chromatin assembly reactions were performed with the indicated concentrations of yChd1 proteins and then subjected to partial MNase digestion analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 00863 . 005 To determine the ability of the mutant yChd1 proteins to yield periodic nucleosome arrays , we employed a nucleosome spacing assay , which is depicted in Figure 4A . In this assay , the extent of ATP-dependent conversion of randomly-distributed nucleosomes ( formed by the salt-dialysis method ) to evenly-spaced nucleosome arrays was monitored by the partial MNase digestion assay . As seen in Figure 4B , the mutant yChd1 proteins are defective in the ATP-dependent formation of periodic nucleosome arrays . To quantitate the ATP-dependent catalysis of nucleosome spacing , we devised a nucleosome spacing index ( Figure 4C ) that measures the formation of distinct di- and tri-nucleosome MNase digestion bands . The spacing index increases with the periodicity of the chromatin , and can be used for the comparison of series of samples that are under identical electrophoretic conditions . For example , quantitation of the nucleosome spacing data in Figure 4B reveals that wild-type yChd1 has strong ATP-dependent spacing activity , whereas the mutant yChd1 proteins exhibit little or no spacing activity ( Figure 4D ) . The essentially complete absence of spacing activity in the mutant yChd1 proteins was further observed in experiments that were carried out with concentrations of yChd1 proteins ranging from 2–120 nM ( Figure 4E ) . ( Examination of the spacing index also reveals that the periodicity of the partial MNase digestion array slightly decreases upon addition of Chd1 proteins in the presence of AMP-PNP . This effect may be due to the blockage of MNase digestion by static nonproductive binding of the Chd1 to DNA . ) Hence , the mutant yChd1 proteins are able to catalyze the ATP-dependent assembly of nucleosomes , but are not able to mediate the ATP-dependent formation of regularly-spaced nucleosomes . 10 . 7554/eLife . 00863 . 006Figure 4 . The chromatin remodeling-defective yChd1 proteins are not able to catalyze the formation of regularly-spaced nucleosomes . ( A ) Schematic representation of the nucleosome spacing reaction . Randomly-distributed nucleosomes were generated by salt dialysis reconstitution of chromatin . Wild-type or mutant Chd1 protein was added along with ATP , and the reaction products were characterized by partial MNase digestion analysis . ( B ) Mutant yChd1 proteins do not reposition nucleosomes into periodic arrays . Spacing reactions were performed by incubating salt-dialysis chromatin with wild-type or mutant yChd1 proteins ( 30 nM ) in the presence of either AMP-PNP or ATP . The reaction products were characterized by partial MNase digestion analysis . ( C ) Determination of the spacing index . Agarose gels from nucleosome spacing reactions were stained by ethidium bromide and then subjected to imaging and analysis on ImageQuantTL ( GE ) to obtain densiometry scans . The spacing index is the average height of the di- and tri-nucleosome peaks ( [P2 + P3 ) /2 ) minus the height of the valley ( V3 ) between the peaks , as indicated by the formula shown in the figure . ( D ) Quantitative analysis of nucleosome spacing by wild-type and mutant yChd1 proteins . The spacing indices were determined for the products of reactions such as those shown in ( B ) . The results are presented as the mean ± standard deviation ( N = 6 ) . ( E ) The chromatin remodeling-defective yChd1 proteins exhibit little to no activity in the nucleosome spacing assay at different concentrations . Nucleosome spacing reactions were performed with yChd1 proteins and subjected to partial MNase digestion analysis . The graph depicts the average spacing index ± standard deviation ( N = 4 ) vs concentration ( nM ) for each of the indicated proteins . DOI: http://dx . doi . org/10 . 7554/eLife . 00863 . 006 To determine whether the functionally distinct chromatin assembly and remodeling functions of Chd1 are conserved from yeast to metazoans , we generated and purified mutant versions of Drosophila melanogaster Chd1 ( dChd1 ) that correspond to the mutant yChd1 proteins ( Figures 1A and 5A ) . Specifically , dChd1 Δ1099–1107 is analogous to yChd1 Δ932–940 , and dChd1 W1099A is analogous to yChd1 W932A . Similar to that seen with the yChd1 proteins , the nucleosome-stimulated ATPase activities of the mutant dChd1 proteins are 37–38% of the wild-type activity , whereas the ATP-dependent chromatin remodeling activity of each of the mutant dChd1 proteins is about 2% of that of the wild-type protein ( Table 1 ) . The chromatin remodeling-defective dChd1 proteins exhibit substantial ( ∼65% of wild-type ) ATP-dependent nucleosome assembly activity , as measured by the DNA supercoiling assay ( Figure 5B ) , but are not able to convert naked DNA into periodic nucleosome arrays in chromatin assembly assays ( Figure 5C ) or to catalyze the ATP-dependent spacing of nucleosome ( Figure 5D ) . The magnitudes of the differences between the assembly and remodeling activities of the wild-type vs mutant dChd1 proteins is not as large as those seen with yChd1 . Nevertheless , it is evident that the functional distinction between nucleosome assembly and remodeling/sliding/spacing in Chd1 is conserved from yeast to Drosophila . 10 . 7554/eLife . 00863 . 007Figure 5 . ATP-dependent nucleosome assembly is observed with chromatin remodeling-defective Drosophila Chd1 ( dChd1 ) proteins . ( A ) Purification of recombinant wild-type and mutant dChd1 . The purified proteins were analyzed by polyacrylamide-SDS gel electrophoresis and visualized by staining with Coomassie Blue . ( B ) Chromatin remodeling-defective dChd1 proteins can assemble nucleosomes in an ATP-dependent manner . Chromatin assembly reactions were performed with wild-type or mutant dChd1 proteins ( 60 nM ) in the presence of either AMP-PNP or ATP . The reaction products were analyzed by the DNA supercoiling assay . The positions of supercoiled ( SC ) , relaxed ( Rel ) , and nicked open circular ( N ) DNAs are indicated . ( C ) Chromatin assembly with remodeling-defective dChd1 proteins does not yield regularly-spaced nucleosomes . Chromatin assembly reactions were performed with wild-type or mutant dChd1 proteins ( 60 nM ) in the presence of either AMP-PNP or ATP . Reaction products were subjected to partial MNase digestion analysis . ( D ) The chromatin remodeling-defective dChd1 proteins exhibit reduced nucleosome spacing activity . Spacing assays were performed with dChd1 proteins and subjected to partial MNase digestion analysis . This graph depicts the mean spacing index ± standard deviation ( N = 3 ) vs concentration ( nM ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00863 . 00710 . 7554/eLife . 00863 . 008Table 1 . Nucleosome sliding and ATP hydrolysis activities of wild-type and mutant dChd1 proteinsDOI: http://dx . doi . org/10 . 7554/eLife . 00863 . 008dChd1 proteinNucleosome sliding activityATP hydrolysis activity ( stimulated by DNA ) ATP hydrolysis activity ( stimulated by nucleosomes ) kcat ( min−1 ) Relative rate ( % ) kcat ( min−1 ) Relative rate ( % ) kcat ( min−1 ) Relative rate ( % ) Wild-type2 . 2±0 . 210061±2100115±3100W1099A0 . 045±0 . 0092 . 0±0 . 434±256±344±338±3Δ1099–11070 . 039±0 . 0111 . 8±0 . 530±249±342±437±3 The analysis of the mutant Chd1 proteins indicated that ATP-dependent chromatin assembly does not require chromatin remodeling activity . It remained formally possible , however , that chromatin remodeling is sufficient for chromatin assembly . To test this notion , we carried out parallel chromatin remodeling and assembly reactions with purified yChd1 ( as a positive control/reference ) and purified human Brg1 ( hBrg1; also known as SMARCA4 ) , which is a SNF2-like family ATPase found in human SWI/SNF chromatin remodeling complexes ( reviewed in Clapier and Cairns , 2009; Flaus and Owen-Hughes , 2011; Hargreaves and Crabtree , 2011; Ryan and Owen-Hughes , 2011 ) . The purified hBrg1 protein is active for chromatin remodeling ( e . g . , Phelan et al . , 1999 ) . By using the restriction enzyme accessibility assay for chromatin remodeling ( e . g . , as in Varga-Weisz et al . , 1997; Boyer et al . , 2000; Shen et al . , 2000; Alexiadis and Kadonaga , 2002 ) , we found that remodeling activity of yChd1 was comparable to that of hBrg1 over a range of concentrations ( Figure 6A ) . We then performed chromatin assembly reactions with the same concentrations of factors , and found that hBrg1 did not assemble chromatin under conditions in which efficient chromatin assembly was observed with the yChd1 control ( Figure 6B ) . Moreover , the hBrg1 caused a decrease in the supercoiling of DNA , which may be related to the nucleosome disruption activity of the SWI/SNF complex ( e . g . , Kwon et al . , 1994 ) . These results thus show that ATP-dependent chromatin remodeling does not necessarily result in the formation of nucleosomes from histones , histone chaperone ( NAP1 ) , DNA , and ATP . 10 . 7554/eLife . 00863 . 009Figure 6 . The Brg1 chromatin remodeling factor does not catalyze chromatin assembly . ( A ) Purified human Brg1 ( hBrg1 ) has a specific activity for chromatin remodeling that is similar to that of wild-type yChd1 . Restriction accessibility assays were performed with salt dialysis-reconstituted chromatin with Hae III restriction enzyme and the indicated concentrations of either yChd1 or hBrg1 . Naked DNA was used as a reference . After digestion with Hae III , the nucleic acids were deproteinized , subjected to agarose gel electrophoresis , and then visualized by staining with ethidium bromide . ( B ) Brg1 does not assemble chromatin . Chromatin assembly reactions were performed with the indicated concentrations of yChd1 or hBrg1 , and the extent of chromatin assembly was monitored by the DNA supercoiling assay . DOI: http://dx . doi . org/10 . 7554/eLife . 00863 . 009 In the context of chromatin assembly , our findings suggest that two ATP-dependent processes are involved in the formation of periodic arrays of nucleosomes ( Figure 7 ) . First , an ATP-driven chromatin assembly activity generates randomly-distributed nucleosomes from histones , histone chaperone ( s ) , and DNA , probably via a prenucleosome intermediate ( Torigoe et al . , 2011 ) . Then , the randomly-distributed nucleosomes are converted into periodic nucleosome arrays via an ATP-driven nucleosome spacing ( remodeling ) activity . Although we depict the two processes separately , they may occur concurrently with wild-type Chd1 . The basis for the formation of periodic nucleosomes as the final product of chromatin assembly is not known , but it is possible that attractive forces between nucleosomes are maximized when the nucleosomes are arranged and/or compacted in a periodic array . In addition , the internucleosomal spacing may be influenced by the interaction of the factors with DNA . 10 . 7554/eLife . 00863 . 010Figure 7 . A model for the functions of ATP-driven nucleosome assembly and remodeling activities in the assembly of chromatin . In this model , nucleosomes can be assembled by an ATP-dependent motor protein in the essentially complete absence of chromatin remodeling activity , as seen with the mutant Chd1 proteins . The resulting nucleosomes are , however , randomly distributed throughout the DNA template . These randomly-distributed nucleosomes can be converted into periodic nucleosome arrays by an ATP-dependent nucleosome remodeling ( spacing ) activity that can be distinguished from the chromatin assembly activity . These processes may occur concurrently with the wild-type Chd1 . DOI: http://dx . doi . org/10 . 7554/eLife . 00863 . 010 In this process , it might be expected that the individual rates of nucleosome assembly and nucleosome spacing would be at least as fast as the overall rate of assembly of periodic nucleosome arrays . We therefore compared the relative rates of the following: ( i ) nucleosome assembly ( from naked DNA to randomly-distributed nucleosomes; first step in Figure 7 ) ; ( ii ) nucleosome spacing ( from randomly-distributed nucleosomes to periodic nucleosomes; second step in Figure 7 ) ; and ( iii ) the overall assembly of naked DNA into periodic nucleosome arrays ( both steps in Figure 7 ) . We performed each of these processes under identical conditions ( Figure 8 ) , and found that the individual rates of nucleosome assembly and nucleosome spacing are indeed faster than the overall rate of assembly of spaced nucleosomes . Hence , the chromatin assembly process depicted in Figure 7 is compatible with the kinetic data . 10 . 7554/eLife . 00863 . 011Figure 8 . The individual rates of nucleosome formation and spacing are faster than the overall rate of assembly of periodic arrays of nucleosomes . ( A ) Determination of the rates of nucleosome formation , nucleosome spacing , and assembly of regularly-spaced nucleosomes . Nucleosome formation in chromatin assembly reactions was monitored by using the DNA supercoiling assay . The positions of supercoiled ( SC ) , relaxed ( Rel ) , and nicked open circular ( N ) DNAs are indicated . Nucleosome spacing reactions were analyzed by partial MNase digestion analysis . The overall assembly of periodic arrays of nucleosomes was determined by performing chromatin assembly reactions and analyzing the reaction products by partial MNase digestion analysis . All reactions were performed with wild-type yChd1 at 30 nM . ( B ) Quantitation of the nucleosome formation and spacing assays , such as those shown in ( A ) . The figure displays the change in % supercoiling ( [Δ supercoiled DNA/total DNA] × 100% ) and spacing indices vs reaction time ( min ) . The reactions followed first-order kinetics , and the first-order rate constants ( min−1 ) and half-times for reaction ( t1/2 , min ) are given in the table . The data points are presented as mean ± standard deviation ( N = 3 ) , and are depicted with the plots of the first order curves ( r2 > 0 . 95 for all graphs ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00863 . 011 In this study , we observed a conserved functional distinction between ATP-dependent nucleosome assembly and ATP-dependent chromatin remodeling by Chd1 . Specifically , the ATP-dependent assembly of nucleosomes by Chd1 occurs in the near complete absence of ATP-dependent chromatin remodeling . Hence , these findings revealed a level of specificity in the role of ATP-dependent motor proteins in chromatin assembly that had not been previously anticipated . In addition , the ATP-dependent chromatin remodeling protein , human Brg1 , does not assemble nucleosomes . We further examined the ability of the chromatin remodeling-defective Chd1 proteins to function in the assembly of periodic nucleosome arrays . The results suggest a process ( Figure 7 ) in which randomly-distributed canonical nucleosomes are generated by the nucleosome assembly function of Chd1 and are then converted into regularly-spaced nucleosome arrays by the chromatin remodeling/spacing/sliding activity of Chd1 . This model for chromatin assembly should provide a useful framework for the analysis of the regeneration of nucleosomes during the many processes in the eukaryotic nucleus that involve the disassembly and reassembly of chromatin .
D . melanogaster NAP1 and topoisomerase I ( ND432 N-terminally truncated form containing the catalytic domain ) were purified as described ( Fyodorov and Kadonaga , 2003 ) . Native D . melanogaster core histones were purified from embryos by the method of Fyodorov and Levenstein ( 2002 ) . Wild-type and mutant ( internal deletion Δ932–940 and point mutant W932A ) forms of truncated S . cerevisiae Chd1 ( amino acids 118–1274 ) , which possesses the core chromodomains , ATPase motor , and DNA-binding domain , were synthesized in bacteria and purified , as described ( Patel et al . , 2011 ) . Human Brg1 was purified as described previously ( Phelan et al . , 1999 ) . The coding sequence for full-length D . melanogaster Chd1 was subcloned into pDEST17 vectors ( Invitrogen , Carlsbad , CA ) . The internal deletion ( Δ1099–1107 ) and single amino acid substitution ( W1099A ) constructs were generated by PCR . The D . melanogaster Chd1 proteins were synthesized in Escherichia coli BL21-star ( DE3 ) cells ( Invitrogen ) containing two additional plasmids—a trigger factor chaperone overexpression plasmid ( Li Ma and Guy Montelione , Rutgers University ) and the RIL plasmid for rare tRNAs ( Stratagene , Santa Clara , CA ) . Two liters of bacterial culture were grown at 37°C to A600 ∼0 . 5–0 . 6 and induced with 0 . 5 mM IPTG . The cells were incubated at 17°C for 18 hr , harvested , and sonicated in Buffer A ( 50 mM Tris-HCl , pH 7 . 5 , 2 mM MgCl2 , 500 mM NaCl , 10% [vol/vol] glycerol , 10 mM β-mercaptoethanol , 1 mM benzamidine , and 0 . 2 mM PMSF ) containing 10 mM imidazole . The lysate was cleared by centrifugation ( 20 min; 45 , 000×g; 4°C ) and mixed with 3 ml Ni-NTA agarose ( Qiagen , Germantown , MD ) . After incubation on a rotating wheel for 3 hr at 4°C , the resin was loaded into a disposable 20-ml polypropylene column and washed with 30 column volumes of Buffer A containing 10 mM imidazole . His-tagged Chd1 was then eluted with 10 ml of Buffer A containing 300 mM imidazole . Following concentration with an Amicon Ultra-15 ( 30 kDa nominal molecular weight limit ) Centrifugal Filter Unit ( Millipore , Billerica , MA ) , Chd1 was applied to a Superdex 200 prep grade ( GE Healthcare , Piscataway , NJ ) size exclusion column ( [column volume , 120 ml]; column dimensions [diameter × length] , 1 . 6 cm × 60 cm; flow rate , 1 . 0 ml/min ) and eluted with 1 . 5 column volumes of Buffer B ( 50 mM Tris-HCl , pH 7 . 5 , 2 mM MgCl2 , 10% [vol/vol] glycerol , 10 mM β-mercaptoethanol , 1 mM benzamidine , and 0 . 2 mM PMSF ) containing 300 mM NaCl . The peak fractions were analyzed by polyacrylamide-SDS gel electrophoresis , pooled , and dialyzed against Buffer B containing 100 mM NaCl . The resulting sample was applied to a Source 15S ( GE Healthcare ) cation exchange column ( [column volume , 1 . 0 ml]; column dimensions [diameter × length] , 0 . 5 cm × 5 cm; flow rate , 1 . 0 ml/min ) . The column was washed with 10 column volumes of 100 mM NaCl in Buffer B , and the protein was eluted with a linear gradient ( 12 column volumes ) from 100 mM to 1 M NaCl in Buffer B . Peak fractions were pooled and dialyzed against Buffer B containing 50 mM NaCl . Chromatin assembly reactions were performed as described previously ( Fyodorov and Kadonaga , 2003; Lusser et al . , 2005 ) . All reactions contained core histones ( 0 . 353 µg ) , NAP1 ( 1 . 4 µg ) , relaxed circular DNA plasmid ( 0 . 294 µg ) , ATP ( 3 mM ) , topoisomerase I ( 1 nM ) , and an ATP regeneration system ( 3 mM phosphoenolpyruvate , 20 U/µl pyruvate kinase ) in a final volume of 70 µl . The buffer composition of the final reaction mixture was as follows: 15 mM Hepes ( K+ ) , pH 7 . 6 , 3 mM Tris , 100 mM KCl , 5 mM NaCl , 5 . 5 mM MgCl2 , 0 . 1 mM EDTA , 6 . 6% ( vol/vol ) glycerol , 1% ( wt/vol ) polyvinyl alcohol ( average MW 10 , 000 ) , 1% ( wt/vol ) polyethylene glycol 8000 , and 20 µg/ml bovine serum albumin . The reaction products were analyzed by DNA supercoiling and partial MNase digestion assays ( Fyodorov and Kadonaga , 2003 ) as well as by extensive MNase digestion of the reaction products followed by agarose gel electrophoresis of DNA fragments ( e . g . , Torigoe et al . , 2011 ) or native nucleoprotein gel electrophoresis of chromatin particles ( Varshavsky et al . , 1976 ) . The percent supercoiling ( [amount of supercoiled DNA/amount of total DNA species] × 100% ) in the DNA supercoiling assays was quantified with ImageQuantTL ( GE Healthcare ) . Because it is not possible to ascertain the fraction of nicked DNA that is packaged into chromatin , we included the nicked DNA in the ‘amount of total DNA species’ but not in the ‘amount of supercoiled DNA’ . Therefore , ‘percent supercoiling’ reflects the amount of closed circular plasmid DNA that is packaged into chromatin and does not include the nicked DNA that is packaged into chromatin . Nucleosomes were reconstituted onto plasmid DNA by the salt-dialysis method and purified by sucrose gradient sedimentation . The resulting chromatin ( 0 . 147 µg of DNA ) was incubated with topoisomerase I ( 1 nM ) , ATP ( 3 mM ) , an ATP regeneration system ( 3 mM phosphoenolpyruvate , 20 U/µl pyruvate kinase ) , and Chd1 proteins in a final volume of 35 µl . The reaction medium was identical to that used for chromatin assembly . The reaction products were analyzed by partial MNase digestion assays ( Fyodorov and Kadonaga , 2003 ) . Chromatin assembly and nucleosome spacing reactions were subjected to partial MNase digestion analysis . Gels were imaged and analyzed with ImageQuantTL ( GE Healthcare ) to obtain densitometry scans of the ethidium-stained bands . Following subtraction of background staining , three heights in the signal were determined: the maximum of the peak corresponding to dinucleosomes ( P2 ) , the maximum of the peak corresponding to trinucleosomes ( P3 ) , and the minimum of the valley between these two peaks ( V2 ) . The spacing index was calculated by using the following equation: ( 0 . 5 ) ( P2 + P3 ) − V2 . Nucleosomes were reconstituted by the gradient salt dialysis method by using S . cerevisiae core histones and a FAM-labeled 208 bp fragment with a 601 nucleosome positioning sequence at one end ( Patel et al . , 2011 ) . Nucleosomes were purified over a mini prep-cell ( Bio-Rad , Hercules , CA ) . Sliding reactions , which monitor the Chd1-catalyzed movement of nucleosomes on a 208 bp DNA fragment , were performed as previously described ( Patel et al . , 2011 ) , with 0-N-63 nucleosomes ( 100 nM ) incubated with dChd1 proteins ( 100 nM ) in buffer containing 20 mM Hepes-K+ , pH 7 . 6 , 50 mM KCl , 1 mM DTT , 0 . 1 mM EDTA , 5% sucrose , 0 . 1 mg/ml BSA , 2 . 5 mM ATP , and 5 mM MgCl2 . Reactions were carried out at 23°C and quenched at the indicated times with a stop solution containing 25 mM EDTA and 2 mg/ml DNA . Changes in nucleosome positions over time were resolved by native polyacrylamide gels , and quantified with ImageJ . Data are averages of three or more separate experiments , and sliding rates were calculated from single exponential fits to data . ATPase rates were determined using an NADH-coupled assay as previously described ( Patel et al . , 2011 ) . Briefly , dChd1 proteins ( 50 nM ) were incubated in the absence or presence of DNA or nucleosome substrates up to 500 nM concentration . Substrates were the same 208 bp DNA fragment either alone or reconstituted into nucleosomes with yeast histones . Data were fit to the Michalis-Menton equation in Kaleidagraph , kobs = ( kcat ) [S]/ ( Km + [S] ) , where [S] is the initial concentration of substrate . The DNA- and nucleosome-stimulated rate constants were subtracted from the basal rate constants in the absence of substrates , which were on the order of 10–30 min−1 . Restriction enzyme accessibility assays were performed as described previously ( Alexiadis and Kadonaga , 2002; Rattner et al . , 2009 ) . | In many cells , genomic DNA is wrapped around proteins known as histones to produce particles called nucleosomes . These particles then join together—like beads on a string—to form a highly periodic structure called chromatin . In the nucleus , chromatin is further folded and condensed into chromosomes . However , many important processes , including the replication of DNA and the transcription of genes , require access to the DNA . The cell must therefore be able to disassemble chromatin and remove the histones , and then , once these processes are complete , to reassemble the chromatin . Enzymes known as chromatin assembly factors are responsible for the disassembly and reassembly of chromatin . There are two main types of chromatin assembly factors in eukaryotic cells ( i . e . , cells with nuclei ) —histone chaperones and motor proteins . The histone chaperones escort histones from the cytoplasm , where they are made , to the nucleus . The motor proteins—using energy supplied by ATP molecules—then catalyze the formation of nucleosomes . This involves two activities: the motor proteins assemble nucleosomes by helping the DNA to wrap around the histones , and they also remodel chromatin by altering the positions of nucleosomes along the DNA to ensure that they are periodic—that is , regularly spaced . A conserved motor protein called Chd1 performs chromatin assembly and remodeling in eukaryotic cells . Chd1 works in conjunction with histone chaperones—both are needed for chromatin assembly , and so are DNA , histones and ATP . However , whether or not chromatin assembly and chromatin remodeling by Chd1 are identical or distinct processes is not well understood . Torigoe et al . have now discovered a mutant Chd1 protein that has nucleosome assembly activity ( i . e . , it can make nucleosomes ) but cannot remodel chromatin ( i . e . , it is unable to move nucleosomes ) , and thus have demonstrated that these two processes are functionally distinct . Torigoe et al . additionally have found that the mutant Chd1 proteins produce randomly distributed nucleosomes rather than the periodic arrays normally found in chromatin . Further analysis then revealed that the wild-type Chd1 protein , which can remodel chromatin , is able to convert randomly distributed nucleosomes into periodic arrays . These findings have led to a new model for chromatin assembly in which Chd1 initially generates randomly distributed nucleosomes ( via its assembly function ) , and then converts them into periodic arrays of nucleosomes ( via its remodeling function ) . Together , these studies shed light on the mechanisms by which chromatin is created and manipulated in cells . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"biochemistry",
"and",
"chemical",
"biology"
] | 2013 | ATP-dependent chromatin assembly is functionally distinct from chromatin remodeling |
Antimicrobial peptides ( AMPs ) are host-encoded antibiotics that combat invading microorganisms . These short , cationic peptides have been implicated in many biological processes , primarily involving innate immunity . In vitro studies have shown AMPs kill bacteria and fungi at physiological concentrations , but little validation has been done in vivo . We utilized CRISPR gene editing to delete most known immune-inducible AMPs of Drosophila , namely: 4 Attacins , 2 Diptericins , Drosocin , Drosomycin , Metchnikowin and Defensin . Using individual and multiple knockouts , including flies lacking these ten AMP genes , we characterize the in vivo function of individual and groups of AMPs against diverse bacterial and fungal pathogens . We found that Drosophila AMPs act primarily against Gram-negative bacteria and fungi , contributing either additively or synergistically . We also describe remarkable specificity wherein certain AMPs contribute the bulk of microbicidal activity against specific pathogens , providing functional demonstrations of highly specific AMP-pathogen interactions in an in vivo setting .
While innate immune mechanisms were neglected during the decades where adaptive immunity captured most of the attention , they have become central to our understanding of immunology . Recent emphasis on innate immunity has , however , mostly focused on the first two phases of the immune response: microbial recognition and associated downstream signaling pathways . In contrast , how innate immune effectors individually or collectively contribute to host resistance has not been investigated to the same extent . The existence of multiple effectors that redundantly contribute to host resistance has hampered their functional characterization by genetic approaches ( Lemaitre and Hoffmann , 2007 ) . The single mutation methodology that still prevails today has obvious limits in the study of immune effectors , which often belong to large gene families . As such , our current understanding of the logic underlying the roles of immune effectors is only poorly defined . As a consequence , the key parameters that influence host survival associated with a successful immune response are not well characterized . In this paper , we harnessed the power of the CRISPR gene editing approach to study the function of Drosophila antimicrobial peptides in host defence both individually and collectively . Antimicrobial peptides ( AMPs ) are small , cationic , usually amphipathic peptides that contribute to innate immune defence in plants and animals ( Imler and Bulet , 2005; Guaní-Guerra et al . , 2010; Rolff and Schmid-Hempel , 2016 ) . They display potent antimicrobial activity in vitro by disrupting negatively charged microbial membranes , but AMPs can also target specific microbial processes ( Park et al . , 1998; Kragol et al . , 2001; Rahnamaeian et al . , 2015 ) . Their expression is induced to very high levels upon challenge to provide microbicidal concentrations in the μM range . Numerous studies have revealed unique roles that AMPs may play in host physiology including anti-tumor activity ( Suttmann et al . , 2008; Kuroda et al . , 2015; Araki et al . , 2018; Parvy et al . , 2019 ) , inflammation in aging ( Cao et al . , 2013; Kounatidis et al . , 2017; E et al . , 2018 ) , involvement in memory ( Bozler et al . , 2017; Barajas-Azpeleta et al . , 2018 ) , mammalian immune signaling ( van Wetering et al . , 2002; Tjabringa et al . , 2003 ) , wound-healing ( Tokumaru et al . , 2005; Chung et al . , 2017 ) , regulation of the host microbiota ( Login et al . , 2011; Mergaert et al . , 2017 ) , tolerance to oxidative stress ( Zhao et al . , 2011; Zheng et al . , 2007 ) , and of course microbicidal activity ( Imler and Bulet , 2005; Wimley , 2010 ) . The fact that AMP genes are immune inducible and expressed at high levels has led to the common assumption they play a vital role in the innate immune response . However , little is known in most cases about how AMPs individually or collectively contribute to animal host defence . In vivo functional analysis of AMPs has been hampered by the sheer number and small size of these genes , making them difficult to mutate with traditional genetic tools ( but e . g . see Hoeckendorf et al . , 2012 ) . Since the first animal AMPs were discovered in silk moths ( Steiner et al . , 1981 ) , insects and particularly Drosophila melanogaster have emerged as a powerful model for characterizing their function . There are currently seven well-characterized families of inducible AMPs in D . melanogaster , but we note that many genes encoding small peptides are strongly upregulated upon infection and are awaiting description ( De Gregorio et al . , 2002 ) . The activities of the seven known AMP families of Drosophila have been determined either in vitro by using peptides directly purified from flies or produced in heterologous systems , or deduced by comparison with homologous peptides isolated in other insect species: Drosomycin and Metchnikowin show antifungal activity ( Fehlbaum et al . , 1994; Levashina et al . , 1995 ) ; Cecropins ( four inducible genes ) and Defensin have both antibacterial and some antifungal activities ( Hultmark et al . , 1980; Ekengren and Hultmark , 1999; Cociancich et al . , 1993; Tzou et al . , 2002 ) ; and Drosocin , Attacins ( four genes ) and Diptericins ( two genes ) primarily exhibit antibacterial activity ( Kragol et al . , 2001; Asling et al . , 1995; Cudic et al . , 1999; Hedengren et al . , 2000; Bulet et al . , 1996 ) . In Drosophila , these AMPs are produced either locally at various surface epithelia in contact with environmental microbes ( Tzou et al . , 2000; Basset et al . , 2000; Gendrin et al . , 2009 ) , or secreted systemically into the hemolymph , the insect blood . During systemic infection , these 14 antimicrobial peptides are strongly induced in the fat body , an organ analogous to the mammalian liver . The systemic production of AMPs is regulated at the transcriptional level by two NF-κB pathways , the Toll and Imd pathways , which are activated by different classes of microbes . The Toll pathway is predominantly responsive to Gram-positive bacteria and fungi , and accordingly plays a major role in defence against these microbes . In contrast , the Imd pathway is activated by Gram-negative bacteria and a subset of Gram-positive bacteria with DAP-type peptidoglycan , and mutations affecting this pathway cause profound susceptibility to Gram-negative bacteria ( De Gregorio et al . , 2002; Lemaitre et al . , 1997 ) . However , the expression pattern of AMP genes is complex as each gene is expressed with different kinetics and can often receive transcriptional input from both pathways ( De Gregorio et al . , 2002; Leulier et al . , 2000 ) . This ranges from Diptericin , which is tightly regulated by the Imd pathway , to Drosomycin , whose expression is mostly regulated by the Toll pathway ( Lemaitre et al . , 1997 ) , except at surface epithelia where Drosomycin is under the control of Imd signaling ( Ferrandon et al . , 1998; Tzou et al . , 2000 ) . While a critical role of AMPs in Drosophila host defence is supported by transgenic flies overexpressing a single AMP ( Tzou et al . , 2002 ) , the specific contributions of each of these AMPs has not been tested . Indeed loss-of-function mutants for most AMP genes were not previously available due to their small size , making them difficult to mutate before the advent of CRISPR/Cas9 technology . Despite this , the great susceptibility to infection of mutants with defective Toll and Imd pathways is commonly attributed to the loss of the AMPs they regulate , though these pathways control hundreds of genes awaiting characterization ( De Gregorio et al . , 2002 ) . Strikingly , Clemmons et al . ( 2015 ) recently reported that flies lacking a set of uncharacterized Toll-responsive peptides ( named Bomanins ) succumb to infection by Gram-positive bacteria and fungi at rates similar to Toll-deficient mutants ( Clemmons et al . , 2015 ) . This provocatively suggests that Bomanins , and not AMPs , might be the predominant effectors downstream of the Toll pathway; yet synthesized Bomanins do not display antimicrobial activity in vitro ( Lindsay et al . , 2018 ) . Thus , while today the fly represents one of the best-characterized animal immune systems , the contribution of AMPs as immune effectors is poorly defined as we still do not understand why Toll and Imd pathway mutants succumb to infection . In this paper , we took advantage of recent gene editing technologies to delete most of the known immune inducible AMP genes of Drosophila . Using single and multiple knockouts , as well as a variety of bacterial and fungal pathogens , we have characterized the in vivo function of individual and groups of antimicrobial peptides . We reveal that AMPs can play highly specific roles in defence , being vital for surviving certain infections yet dispensable against others . We highlight key interactions amongst immune effectors and pathogens and reveal to what extent these defence peptides act in concert or alone .
We generated null mutants for 10 of the 14 known Drosophila antimicrobial peptide genes that are induced upon systemic infection . These include five single gene mutations affecting Defensin ( DefSK3 ) , Attacin C ( AttCMi ) , Metchnikowin ( MtkR1 ) , Attacin D ( AttDSK1 ) and Drosomycin ( DrsR1 ) , respectively , and two small deletions removing both Diptericins DptA and DptB ( DptSK1 ) , or the gene cluster containing Drosocin , and Attacins AttA and AttB ( Dro-AttABSK2 ) . The function of Cecropins were not assessed in this manuscript . All mutations/deletions were made using the CRISPR editing approach with the exception of Attacin C , which was disrupted by insertion of a Minos transposable element ( Bellen et al . , 2011 ) , and the Drosomycin and Metchnikowin deletions generated by homologous recombination ( Figure 1A and Figure 1—figure supplement 1 ) . To disentangle the role of Drosocin and AttA/AttB in the Dro-AttABSK2 deletion , we also generated an individual Drosocin mutant ( DroSK4 ) ; for complete information , see Figure 1—figure supplement 1 . We then isogenized these mutations for at least seven generations into the w1118 DrosDel isogenic genetic background ( Ryder et al . , 2004 ) ( iso w1118 ) . Then , we recombined these seven independent mutations into a background lacking these 10 inducible AMPs referred to as ‘ΔAMPs . ’ ΔAMPs flies were viable and showed no morphological defects . To confirm the absence of AMPs in our ΔAMPs background , we performed a MALDI-TOF analysis of hemolymph from both unchallenged and immune-challenged flies infected by a mixture of Escherichia coli and Micrococcus luteus . This analysis revealed the presence of peaks induced upon challenge corresponding to AMPs in wild-type but not ΔAMPs flies . Importantly , it also confirmed that induction of most other immune-induced molecules ( IMs ) ( Uttenweiler-Joseph et al . , 1998 ) , was unaffected in ΔAMPs flies ( Figure 1B ) . Of note , we failed to observe two IMs , IM7 and IM21 , in our ΔAMPs flies , suggesting that these unknown peptides are secondary products of AMP genes . We further confirmed that Toll and Imd NF-κB signaling pathways were intact in ΔAMPs flies by measuring the expression of target genes of these pathways ( Figure 1C–D ) . This demonstrates that Drosophila AMPs are not signaling molecules required for Toll or Imd pathway activity . We also assessed the role of AMPs in the melanization response , wound clotting , and hemocyte populations . After clean injury , ΔAMPs flies survive as wild-type ( Figure 1—figure supplement 2A ) . We found no defect in melanization ( χ2 , p=0 . 34 , Figure 1—figure supplement 2B ) as both adults and larvae strongly melanize the cuticle following clean injury ( Figure 1—figure supplement 2C ) . Furthermore , we visualized the rapid formation of clot fibers ex vivo using the hanging drop assay and PNA staining ( Scherfer et al . , 2004 ) in hemolymph of both wild-type and ΔAMPs larvae ( Figure 1—figure supplement 2D ) . Hemocyte counting ( i . e . crystal cells , FACS ) did not reveal any deficiency in hemocyte populations of ΔAMPs larvae ( Figure 1—figure supplement 2E , F , and not shown ) . Altogether , our study suggests that Drosophila AMPs are primarily immune effectors , and not regulators of innate immunity . We used these ΔAMPs flies to explore the role that AMPs play in defence against pathogens during systemic infection . We first focused our attention on Gram-negative bacterial infections , which are combatted by Imd pathway-mediated defence in Drosophila ( Lemaitre and Hoffmann , 2007 ) . We challenged wild-type and ΔAMPs flies with six different Gram-negative bacterial species , using inoculation doses ( given as OD600 ) selected such that at least some wild-type flies were killed . In our survival experiments , we also include Oregon R ( OR-R ) as an alternate wild-type for comparison , and Relish mutants ( RelE20 ) that lack a functional Imd response and are known to be very susceptible to this class of bacteria ( Hedengren et al . , 1999 ) ( Figure 2 ) . Globally , ΔAMPs flies were extremely susceptible to all Gram-negative pathogens tested ( Figure 2 , light blue plots ) . The susceptibility of AMP-deficient flies to Gram-negative bacteria largely mirrored that of RelE20 flies . For all Gram-negative infections tested , ΔAMPs flies show a higher bacterial count at 18 hr post-infection ( hpi ) indicating that AMPs actively inhibit bacterial growth , as expected of ‘antimicrobial peptides’ ( Figure 2—figure supplement 1A ) . Use of GFP-expressing bacteria show that bacterial growth in ΔAMPs flies radiates from the wound site until spreading systemically ( Figure 2—figure supplement 1B , C ) . Collectively , the use of AMP-deficient flies reveals that AMPs are major players in resistance to Gram-negative bacteria , and likely constitute an essential component of the Imd pathway’s contribution for survival against these germs . Previous studies have shown that resistance to Gram-positive bacteria and fungi in Drosophila is mostly mediated by the Toll pathway , although the Imd pathway also contributes to some extent ( Lemaitre et al . , 1997; Leulier et al . , 2000; Rutschmann et al . , 2002; Tanji et al . , 2007 ) . Moreover , a deletion removing ten uncharacterized Bomanins ( BomΔ55C ) induces a strong susceptibility to both Gram-positive bacteria and fungi ( Clemmons et al . , 2015 ) , suggesting that Bomanins are major players downstream of Toll in the defence against these germs . This prompted us to explore the role of antimicrobial peptides in defence against Gram-positive bacteria and fungi . In these experiments , we additionally included spätzle mutant flies ( spzrm7 ) lacking Toll signaling as susceptible controls . We first challenged wild-type and ΔAMPs flies with two lysine-type ( E . faecalis , S . aureus ) and two DAP-type ( B . subtilis , L . innocua ) peptidoglycan-containing Gram-positive bacterial species . We observed that ΔAMPs flies display only weak or no increased susceptibility to infection with these Gram-positive bacterial species , as ΔAMPs survival rates were closer to the wild-type than to spzrm7 mutants lacking a functional Toll pathway ( Figure 2 , orange plots ) , with the exception of S . aureus . Meanwhile , BomΔ55C mutants consistently phenocopied spzrm7 flies , confirming the important contribution of these peptides in defence against Gram-positive bacteria ( Clemmons et al . , 2015 ) . Next , we monitored the survival of ΔAMPs to the yeast Candida albicans , the opportunistic fungus Aspergillus fumigatus and two entomopathogenic fungi , Beauveria bassiana , and Metarhizium anisopliae . For the latter two , we used a natural mode of infection by spreading spores on the cuticle ( Lemaitre et al . , 1997 ) . ΔAMPs flies were more susceptible to fungal infections with B . bassiana , A . fumigatus , and C . albicans , but not M . anisopliae ( Figure 2 , yellow plots ) . In all instances , BomΔ55C mutants were as or more susceptible to fungal infection than ΔAMPs flies , approaching Toll-deficient mutant levels . Collectively , our data demonstrate that AMPs are major immune effectors in defence against Gram-negative bacteria and have a less essential role in defence against bacteria and fungi . The impact of the ΔAMPs deletion on survival could be due to the action of certain AMPs having a specific effect , or more likely due to the combinatory action of co-expressed AMPs . Indeed , cooperation of AMPs to potentiate their microbicidal activity has been suggested by numerous in vitro approaches ( Rahnamaeian et al . , 2015; Yu et al . , 2016; Mohan et al . , 2014 ) , but rarely in an in vivo context ( Zanchi et al . , 2017 ) . Having shown that AMPs as a whole significantly contribute to fly defence , we next explored the contribution of individual peptides to this effect . To tackle this question in a systematic manner , we performed survival analyses using fly lines lacking one or several AMPs , focusing on pathogens with a range of virulence that we previously showed to be sensitive to the action of AMPs . This includes the yeast C . albicans and the Gram-negative bacterial species P . burhodogranariea , P . rettgeri , Ecc15 , and E . cloacae . Given seven independent AMP mutations , over 100 combinations of mutants are possible , making a systematic analysis of AMP interactions a logistical nightmare . Therefore , we designed an approach that would allow us to characterize their contributions to defence by deleting groups of AMPs . To this end , we generated three groups of combined mutants: A ) flies lacking Defensin ( Group A ) ; Defensin is regulated by Imd signalling but is primarily active against Gram-positive bacteria in vitro ( Imler and Bulet , 2005 ) . B ) Flies lacking three antibacterial and structurally related AMP families: the Proline-rich Drosocin and the Proline- and Glycine-rich Diptericins and Attacins ( Group B , regulated by the Imd pathway ) . C ) Flies lacking the two antifungal peptide genes Metchnikowin and Drosomycin ( Group C , mostly regulated by the Toll pathway ) . We then combined these three groups to generate flies lacking AMPs from groups A and B ( AB ) , A and C ( AC ) , or B and C ( BC ) . Finally , flies lacking all three groups are our ΔAMPs flies , which are highly susceptible to a number of infections . By screening these seven genotypes as well as individual mutants , we were able to assess potential interactions between AMPs of different groups , as well as decipher the function of individual AMPs . We first applied this AMP-groups approach to infections with the relatively avirulent yeast C . albicans . Previous studies have shown that Toll , but not Imd , contributes to defence against this fungus ( Gottar et al . , 2006; Glittenberg et al . , 2011 ) . Thus , we suspected that the two antifungal peptides , Drosomycin and Metchnikowin , could play a significant role in the susceptibility of ΔAMPs flies to this yeast . Consistent with this , Group C flies lacking Metchnikowin and Drosomycin were more susceptible to infection ( p<0 . 001 relative to iso w1118 ) with a survival rate similar to ΔAMPs flies ( Figure 3A ) . Curiously , AC-deficient flies that also lack Defensin survived better than Group C-deficient flies ( Log-Rank p=0 . 014 ) . We have no explanation for this interaction , but this could be due to i ) a better canalization of the immune response by preventing the induction of ineffective AMPs , ii ) complex biochemical interactions amongst the AMPs involved affecting either the host or pathogen or iii ) differences in genetic background generated by additional recombination . We then investigated the individual contributions of Metchnikowin and Drosomycin to survival to C . albicans . We found that both MtkR1 and DrsR1 individual mutants were somewhat susceptible to infection , but notably only Mtk; Drs compound mutants reached ΔAMPs levels of susceptibility ( Figure 3B ) . This co-occurring loss of resistance appears to be primarily additive ( Mutant , Cox Hazard Ratio ( HR ) , p-value: MtkR1 , HR =+1 . 17 , p=0 . 008; DrsR1 , HR =+1 . 85 , p<0 . 001; Mtk*Drs , HR = −0 . 80 , p=0 . 116 ) . We observed that Group C deficient flies eventually succumb to uncontrolled C . albicans growth by monitoring yeast titre , indicating that these AMPs indeed act by suppressing yeast growth ( Figure 3C ) . In conclusion , our study provides an in vivo validation of the potent antifungal activities of Metchnikowin and Drosomycin ( Fehlbaum et al . , 1994; Levashina et al . , 1995 ) , and highlights a clear example of additive cooperation of AMPs . We next analyzed the contribution of AMPs in resistance to infection with the moderately virulent Gram-negative bacterium P . burhodogranariea . We found that Group B mutants lacking Drosocin , the two Diptericins , and the four Attacins , were as susceptible to infection as ΔAMPs flies ( Figure 4A ) , while flies lacking the antifungal peptides Drosomycin and Metchnikowin ( Toll-regulated , Group C ) resisted the infection as wild-type . Flies lacking Defensin ( Group A ) showed an intermediate susceptibility , but behave as wild-type in the additional absence of Toll Group C peptides ( Group AC ) . Thus , we again observed a better survival rate with the co-occurring loss of Group A and C peptides ( see possible explanation above ) . In this case , Group A flies were susceptible while AC flies were not . Following the observation that Group B flies were as susceptible as ΔAMPs flies , we sought to better decipher the contribution of each Group B AMP to resistance to P . burhodogranariea . We observed that mutants for Drosocin alone ( DroSK4 ) , or the DiptericinA/B deficiency were not susceptible to this bacterium ( Figure 4B ) . We additionally saw no marked susceptibility of Drosocin-Attacin A/B-deficient flies , nor Attacin C or Attacin D mutants ( not shown ) . Interestingly , we found that compound mutants lacking Drosocin and Attacins A , B , C , and D ( Figure 4B: ‘ΔDro , ΔAtt’ ) , or Drosocin and Diptericins DptA and DptB ( ‘ΔDro , ΔDpt’ ) displayed an intermediate susceptibility . Only the Group B mutants lacking Drosocin , all Attacins , and both Diptericins ( ΔDro , ΔAtt , ΔDpt ) phenocopied ΔAMPs flies ( Figure 4B ) , with synergistic statistical interactions observed upon co-occurring loss of Attacins and Diptericins ( ΔAtt*ΔDpt: HR =+1 . 45 , p<0 . 001 ) ; we emphasize here that this synergistic interaction solely reflects that the effect on survival of combining these mutations is greater than the sum effect of the individual mutations ( discussed later ) . By 6hpi , bacterial titres of individual flies already showed significant differences in the most susceptible genotypes ( Figure 4C ) , although these differences were reduced by 18 hpi likely owing to the high chronic load P . burhodogranariea establishes in surviving flies ( Duneau et al . , 2017; also see Figure 2—figure supplement 1A ) . Collectively , the use of various compound mutants reveals that several Imd-responsive AMPs , notably Drosocin , Attacins , and Diptericins , jointly contribute to defence against P . burhodogranariea infection . A strong susceptibility of Group B flies was also observed upon infection with Ecc15 , another Gram-negative bacterium commonly used to infect flies ( Neyen et al . , 2014 ) ( Figure 4—figure supplement 1B ) . We continued our exploration of AMP interactions using our AMP groups approach with the fairly virulent P . rettgeri ( strain Dmel ) , a strain isolated from wild-caught Drosophila hemolymph ( Juneja and Lazzaro , 2009 ) . We were especially interested by this bacterium as previous studies ( Unckless et al . , 2015; Unckless and Lazzaro , 2016 ) have shown a correlation between susceptibility to P . rettgeri and a polymorphism in the Diptericin A gene pointing to a specific AMP-pathogen interaction . Use of compound mutants revealed only loss of Group B AMPs was needed to reach the susceptibility of ΔAMPs and RelE20 flies ( Figure 5A ) . Use of individual mutant lines , however , revealed a pattern overtly different from P . burhodogranariea , as the sole Diptericin A/B deficiency caused susceptibility similar to Group B , ΔAMPs , and RelE20 flies ( Figure 5B , C ) . We further confirmed this susceptibility using a DptA RNAi construct ( Figure 5—figure supplement 1A , B ) . Moreover , flies carrying the DptSK1 mutation over a deficiency ( Df ( 2R ) Exel6067 ) were also highly susceptible to P . rettgeri ( Figure 5D ) . Interestingly , flies that were heterozygotes for DptSK1 or the Df ( 2R ) Exel6067 that still have one copy of the two Diptericins were markedly susceptible to infection with P . rettgeri ( Figure 5D ) . This indicates that a full transcriptional output of Diptericin is required over the course of the infection to resist P . rettgeri ( Figure 5E ) . Altogether , our results suggest that only the Diptericin gene family , amongst the many AMPs regulated by the Imd pathway , provides the full AMP-based contribution to defence against this bacterium . To test this hypothesis , we generated a fly line lacking all the AMPs except DptA and DptB ( ΔAMPs+Dpt ) . Strikingly , ΔAMPs+Dpt flies have the same survival rate as wild-type flies , further emphasizing the specificity of this interaction ( Figure 5B ) . Bacterial counts confirm that the susceptibility of these Diptericin mutants arises from an inability of the host to suppress bacterial growth ( Figure 5C ) . Collectively , our study shows that Diptericins are critical to resist P . rettgeri , while they play an important but less essential role in defence against P . burhodogranariea infection . We were curious whether Diptericin’s major contribution to defence observed with P . rettgeri could be generalized to other members of the genus Providencia . An exclusive role for Diptericins was also found for the more virulent P . stuartii ( Figure 5—figure supplement 1C ) , but not for other Providencia species tested ( P . burhodogranariea , P . alcalifaciens , P . sneebia , P . vermicola ) ( data not shown ) . In the course of our exploration of AMP-pathogen interactions , we identified another highly specific interaction between E . cloacae and Drosocin . Use of compound mutants revealed that alone , Group B flies were already susceptible to E . cloacae . Meanwhile , Group AB flies additionally lacking Defensin reached ΔAMPs levels of susceptibility , while Group A and Group C flies resisted as wild-typeMeanwhile , Group AB flies reached ΔAMPs levels of susceptibility , while Group A and Group C flies resisted as wild-type ( Figure 6A ) . The high susceptibility of Group AB flies results from a synergistic statistical interaction amongst Group A ( Defensin ) and Group B peptides in defence against E . cloacae ( A*B , HR =+2 . 55 , p=0 . 003 ) . We chose to further explore the AMPs deleted in Group B flies , as alone this genotype already displayed a strong susceptibility . Use of individual mutant lines revealed that mutants for Drosocin alone ( DroSK4 ) or the Drosocin-Attacin A/B deficiency ( Dro-AttABSK2 ) , but not AttC , AttD , nor DptSK1 ( not shown ) , recapitulate the susceptibility observed in Group B flies ( Figure 6B ) . At 18 hpi , both DroSK4 and ΔAMPs flies had significantly higher bacterial loads compared to wild-type flies , while RelE20 mutants were already moribund with much higher bacterial loads ( Figure 6C ) . Indeed , the deletion of Drosocin alone alters the fly’s ability to control the otherwise avirulent E . cloacae upon inoculations using OD = 200 ( ~39 , 000 bacteria , Figure 6A–C ) or even OD = 10 ( ~7000 bacteria , Figure 6—figure supplement 1A ) . We confirmed the high susceptibility of Drosocin mutant flies to E . cloacae in various contexts: transheterozygote flies carrying DroSK4 over a Drosocin deficiency ( Df ( 2R ) BSC858 ) that also lacks flanking genes including AttA and AttB ( ( Figure 6D ) , the Dro SK4 mutations in an alternate genetic background ( yw , Figure 6E ) , and , Drosocin RNAi ( Figure 6—figure supplement 1B , C ) . Thus , we recovered two highly specific AMP-pathogen interactions: Diptericins are essential to combat P . rettgeri infection , while Drosocin is paramount to surviving E . cloacae infection .
Despite the recent emphasis on innate immunity , little is known on how immune effectors contribute individually or collectively to host defence , exemplified by the lack of in depth in vivo functional characterization of Drosophila AMPs . Taking advantage of new gene editing approaches , we developed a systematic mutation approach to study the function of Drosophila AMPs . With seven distinct mutations , we were able to generate a fly line lacking 10 AMPs that are known to be strongly induced during the systemic immune response . A striking first finding is that ΔAMPs flies were perfectly healthy and have an otherwise wild-type immune response . This indicates that in contrast to mammals ( van Wetering et al . , 2002 ) , these Drosophila AMPs are not likely to function as signaling molecules . Using a systemic mode of infection that induces AMP expression in the fat body and hemocytes , we found that most flies lacking a single AMP family exhibited a higher susceptibility to certain pathogens consistent with their in vitro activity . We found activity of Diptericins against P . rettgeri , Drosocin against E . cloacae , Drosomycin and Metchnikowin against C . albicans , and Defensin against P . burhodogranariea . In most cases , the susceptibility of single mutants was slight , and the contribution of individual AMPs could be revealed only when combined to other AMP mutations as illustrated by the susceptibility of Drosocin , Attacin , and Diptericin combined mutants to P . burhodogranariea . Thus , the use of compound rather than single mutations provides a better strategy to decipher the contribution of AMPs to host defence . Our findings are consistent with a previous study using flies that constitutively expressed individual peptides ( Tzou et al . , 2002 ) , which showed an activity of Drosomycin against A . fumigatus and Attacin against Ecc15 . Beyond the systemic immune response , AMPs are also expressed in many tissues such as the gut and trachea ( Ferrandon et al . , 1998; Tzou et al . , 2000 ) . Future studies should investigate the role of AMPs in these local epithelial immune responses . The Toll and Imd pathways provide a paradigm of innate immunity , illustrating how two distinct pathways link pathogen recognition to distinct but overlapping sets of downstream immune effectors ( Lemaitre and Hoffmann , 2007; Buchon et al . , 2014 ) . However , a method of deciphering the contributions of the different downstream effectors to the specificity of these pathways remained out of reach , as mutations in these immune effectors were lacking . Our study shows that AMPs contribute greatly to resistance to Gram-negative bacteria . Consistent with this , ΔAMPs flies are almost as susceptible as Imd-deficient mutants to most Gram-negative bacteria . In contrast , flies lacking AMPs were only slightly more susceptible to Gram-positive bacteria and fungal infections compared to wild-type flies , and this susceptibility rarely approached the susceptibility of Bomanin mutants . It is possible that additional loss of Cecropins would further increase the sensitivity of ΔAMPs flies to bacteria or fungi . This may be due to the cell walls of Gram-negative bacteria being thinner and more fluid than the rigid cell walls of Gram-positive bacteria ( Fayaz et al . , 2010 ) , consequently making Gram-negative bacteria more prone to the action of pore-forming cationic peptides . It would be interesting to know if the specificity of AMPs to primarily combatting Gram-negative bacteria is also true in other species . Based on our study and Clemmons et al . ( 2015 ) , we can now explain the susceptibility of Toll and Imd mutants at the level of the effectors , as we show that mutations affecting Imd-pathway responsive antibacterial peptide genes are highly susceptible to Gram-negative bacteria while the Toll-responsive targets Drosomycin , Metchnikowin , and especially the Bomanins , confer resistance to fungi and Gram-positive bacteria . Thus , the susceptibility of these two pathways to different sets of microbes not only reflects specificity at the level of recognition , but can now also be translated to the activities of downstream effectors . It remains to be seen how Bomanins contribute to the microbicidal activity of immune-induced hemolymph , as attempts to synthesize Bomanins have not revealed direct antimicrobial activity ( Lindsay et al . , 2018 ) . It should also be noted that many putative effectors downstream of Toll and Imd remain uncharacterized , and so could also contribute to host defence beyond these AMPs and Bomanins . In the last few years , numerous in vitro studies have focused on the potential for synergistic interactions of AMPs in microbial killing ( Rahnamaeian et al . , 2015; Yu et al . , 2016; Zanchi et al . , 2017; Yan and Hancock , 2001; Nuding et al . , 2014; Zerweck et al . , 2017; Chen et al . , 2005; Stewart et al . , 2014; Zdybicka-Barabas et al . , 2012 ) . Our collection of AMP mutant fly lines placed us in an ideal position to investigate AMP interactions in an in vivo setting . While Toll-responsive AMPs ( Group C: Metchnikowin , Drosomycin ) additively contributed to defence against the yeast C . albicans , we found that certain combinations of AMPs have synergistic contributions to defence against P . burhodogranariea . Synergistic loss of resistance may arise in two general fashions: first , co-operation of AMPs using similar mechanisms of action may breach a threshold microbicidal activity whereupon pathogens are no longer able to resist . This may be the case for the synergistic effect of Diptericins and Attacins against P . burhodogranariea , as only co-occurring loss of both these related glycine-rich peptide families ( Hedengren et al . , 2000 ) led to complete loss of resistance . Alternatively , synergy may arise due to complementary mechanisms of action , whereupon one AMP potentiates the other AMP’s ability to act . For instance , the action of the bumblebee AMP Abaecin , which binds to the molecular chaperone DnaK to inhibit bacterial DNA replication , is potentiated by the presence of the pore-forming peptide Hymenoptaecin ( Rahnamaeian et al . , 2016 ) . Drosophila Drosocin is highly similar to Abaecin and the related peptide Apidecin , including O-glycosylation of a critical threonine residue ( Imler and Bulet , 2005; Hanson et al . , 2016 ) , and thus likely acts in a similar fashion . Furthermore , Drosophila Attacin C is maturated into both a glycine-rich peptide and a Drosocin-like peptide called MPAC ( Rabel et al . , 2004 ) . As such , co-occuring loss of Drosocin , MPAC , and other possible MPAC-like peptides encoded by the Attacin/Diptericin superfamily may be responsible for the synergistic loss of resistance in Drosocin , Attacin , Diptericin combined mutants . It is commonly thought that the innate immune response lacks the specificity of the adaptive immune system , which mounts directed defences against specific pathogens . Accordingly for innate immunity , the diversity of immune-inducible AMPs can be justified by the need for generalist and/or co-operative mechanisms of microbial killing . However , an alternate explanation may be that innate immunity expresses diverse AMPs in an attempt to hit the pathogen with a ‘silver bullet:’ an AMP specifically attuned to defend against that pathogen . Here , we provide a demonstration in an in vivo setting that such a strategy may actually be employed by the innate immune system . Remarkably , we recovered not just one , but two examples of exquisite specificity in our laborious but relatively limited assays . Diptericin has previously been highlighted for its important role in defence against P . rettgeri ( Unckless et al . , 2016 ) , but it was previously unknown whether other AMPs may confer defence in this infection model . Astoundingly , flies mutant for the other inducible AMPs resisted P . rettgeri infection as wild-type , while only Diptericin mutants succumbed to infection . This means that Diptericins do not co-operate with these other AMPs in defence against P . rettgeri and are solely responsible for defence in this specific host-pathogen interaction . Moreover , +/DptSK1 heterozygote flies were nonetheless extremely susceptible to infection , demonstrating that a full transcriptional output over the course of infection is required to effectively prevent pathogen growth . A previous study has shown that ~7 hpi appears to be the critical time point at which P . rettgeri either grows unimpeded or the infection is controlled ( Duneau et al . , 2017 ) . This time point correlates with the time at which the Diptericin transcriptional output is in full-force ( Lemaitre et al . , 1997 ) . Thus , a lag in the transcriptional response in DptSK1/+ flies likely prevents the host from reaching a competent Diptericin concentration , indicating that Diptericin expression level is a key factor in successful host defence . We also show that Drosocin is specifically required for defence against E . cloacae . This striking finding validates previous biochemical analyses showing Drosocin in vitro activity against several Enterobacteriaceae , including E . cloacae ( Bulet et al . , 1996 ) . As ΔAMPs flies are more susceptible than Drosocin single mutants , other AMPs also contribute to Drosocin-mediated control of E . cloacae . As highlighted above , Drosocin is similar to other Proline-rich AMPs ( e . g . Abaecin , Pyrrhocoricin ) that have been shown to target bacterial DnaK ( Kragol et al . , 2001; Rahnamaeian et al . , 2015 ) . Alone , these peptides still penetrate bacteria cell walls through their uptake by bacterial permeases ( Rahnamaeian et al . , 2016; Narayanan et al . , 2014 ) . Thus , while Drosocin would benefit from the presence of pore-forming toxins to enter bacterial cells ( Rahnamaeian et al . , 2016 ) , the veritable ‘stake to the heart’ is likely the plunging of Drosocin itself into vital bacterial machinery . It has often been questioned why flies should need so many AMPs ( Lemaitre and Hoffmann , 2007; Rolff and Schmid-Hempel , 2016; Unckless and Lazzaro , 2016 ) . A common idea , supported by in vitro experiments ( Rahnamaeian et al . , 2015; Yan and Hancock , 2001; Zdybicka-Barabas et al . , 2012 ) is that AMPs work as cocktails , wherein multiple effectors are needed to kill invading pathogens . However , we find support for an alternative hypothesis that suggests AMP diversity may be due to highly specific interactions between AMPs and subsets of pathogens that they target . Burgeoning support for this idea also comes from recent evolutionary studies that show Drosophila and vertebrate AMPs experience positive selection ( Unckless et al . , 2015; Unckless and Lazzaro , 2016; Hanson et al . , 2016; Chapman et al . , 2018; Hellgren and Sheldon , 2011; Tennessen and Blouin , 2008; Sackton , 2019 ) , a hallmark of host-pathogen evolutionary conflict . Our functional demonstrations of AMP-pathogen specificity , using naturally relevant pathogens ( Juneja and Lazzaro , 2009; Cox and Gilmore , 2007 ) , suggest that such specificity is fairly common , and that certain AMPs can act as the arbiters of life or death upon infection by certain pathogens . This stands in contrast to the classical view that the AMP response contains such redundancy that single peptides should have little effect on organism-level immunity ( Rolff and Schmid-Hempel , 2016; Unckless et al . , 2015; Tzou et al . , 2000; Unckless and Lazzaro , 2016 ) . Nevertheless , it seems these immune effectors play non-redundant roles in defence . By providing a long-awaited in vivo functional validation for the role of AMPs in host defence , we also pave the way for a better understanding of the functions of immune effectors . Our approach of using multiple compound mutants , now possible with the development of new genome editing approaches , was especially effective to decipher the logic of immune effectors . Understanding the role of AMPs in innate immunity holds great promise for the development of novel antibiotics ( Chung et al . , 2017; Mylonakis et al . , 2016; Mahlapuu et al . , 2016 ) , insight into autoimmune diseases ( Schluesener et al . , 1993; Gilliet and Lande , 2008; Sun et al . , 2015; Kumar et al . , 2016 ) , and given their potential for remarkably specific interactions , perhaps in predicting key parameters that predispose individuals or populations to certain kinds of infections ( Unckless et al . , 2015; Unckless and Lazzaro , 2016; Chapman et al . , 2018 ) . Finally , our set of isogenized AMP mutant lines provides long-awaited tools to decipher the role of AMPs not only in systemic immunity , but also in local immune responses , and the various roles that AMPs may play in aging , neurodegeneration , anti-tumor activity , regulation of the microbiota and more , where disparate evidence has pointed to their involvement .
The DrosDel ( Ryder et al . , 2004 ) isogenic w1118 ( iso w1118 ) wild type was used as a genetic background for mutant isogenization . Alternate wild-types used throughout include Oregon R ( OR-R ) , w1118 from the Vienna Drosophila Resource Centre , and the Canton-S isogenic line Exelexis w1118 , which was kindly provided by Brian McCabe . BomΔ55C mutants were generously provided by Steven Wasserman , and BomΔ55C was isogenized into the iso w1118 background . RelE20 and spzrm7 iso w1118 flies were provided by Luis Teixeira ( Hedengren et al . , 1999; Ferreira et al . , 2014 ) . Prophenoloxidase mutants ( ΔPPO ) are described in Dudzic et al . ( 2015 ) . P-element mediated homologous recombination according to Baena-Lopez et al . ( 2013 ) was used to generate mutants for Mtk ( MtkR1 ) and Drs ( DrsR1 ) . Plasmids were provided by Mickael Poidevin . Attacin C mutants ( AttCMi , #25598 ) , the Diptericin deficiency ( Df ( 2R ) Exel6067 , #7549 ) , the Drosocin deficiency ( Df ( 2R ) BSC858 , #27928 ) , UAS-Diptericin RNAi ( DptRNAi , #53923 ) , UAS-Drosocin RNAi ( DroRNAi , #67223 ) , and Actin5C-Gal4 ( ActGal4 , #4414 ) were ordered from the Bloomington stock centre ( stock #s included ) . CRISPR mutations were performed by Shu Kondo according to Kondo and Ueda ( Kondo and Ueda , 2013 ) , and full descriptions are given in Figure 1—figure supplement 1 . In brief , flies deficient for Drosocin , Attacin A , and Attacin B ( Dro-AttABSK2 ) , and Diptericin A and Diptericin B ( DptSK1 ) were produced by gene region deletion specific to those AMPs without affecting other genes . Single mutants for Defensin ( DefSK3 ) , Drosocin ( DroSK4 ) , and Attacin D ( AttDSK1 ) are small indels resulting in the production of short ( 80–107 residues ) nonsense peptides . Mutations were isogenized for a minimum of seven generations into the iso w1118 background prior to subsequent recombination . It should be noted that Group A flies were initially thought to be a double mutant for both Defensin and the Cecropin cluster , resulting from a combination of DefSK3 and a CRISPR-induced Cecropin deletion ( called CecSK6 ) . It was subsequently shown that CecSK6 is a complex aberration at the Cecropin locus that retains a wild-type copy of the Cecropin cluster . This re-arranged Cecropin locus does not contribute significantly to the susceptibility of Group A flies , as Group A was not different from DefSK3 alone ( Log-Rank p=0 . 818; Figure 4—figure supplement 1A ) . Thus , group A flies were considered as single DefSK3 mutants . Bacteria were grown overnight on a shaking plate at 200 rpm in their respective growth media and temperature conditions , and then pelleted by centrifugation at 4°C . These bacterial pellets were diluted to the desired optical density at 600 nm ( OD ) as indicated . The following bacteria were grown at 37°C in LB media: Escherichia coli strain 1106 , Salmonella typhimurium , Enterobacter cloacae β12 , Providencia rettgeri strain Dmel , Providencia burhodogranariea strain B , Providencia stuartii strain DSM 4539 , Providencia sneebia strain Dmel , Providencia alcalifaciens strain Dmel , Providencia vermicola strain DSM 17385 , Bacillus subtilis , and Staphylococcus aureus . Erwinia carotovora carotovora ( Ecc15 ) and Micrococcus luteus were grown overnight in LB at 29°C . Enterococcus faecalis and Listeria innocua were cultured in BHI medium at 37°C . Candida albicans was cultured in YPG medium at 37°C . Aspergillus fumigatus was grown at room temperature on Malt Agar , and spores were collected in sterile PBS rinses , pelleted by centrifugation , and then resuspended to the desired OD in PBS . The entomopathogenic fungi Beauveria bassiana and Metarhizium anisopliae were grown on Malt Agar at room temperature until sporulation . Systemic infections were performed by pricking 3- to 5-day-old adult males in the thorax with a 100-μm-thick insect pin dipped into a concentrated pellet of bacteria or fungal spores . Infected flies were subsequently maintained at 25°C for experiments . For infections with B . bassiana and M . anisopliae , flies were anesthetized and then shaken on a sporulating plate of fungi for 30 s . At least two replicate survival experiments were performed for each infection , with 20–35 flies per vial on standard fly medium without yeast . Survivals were scored twice daily , with additional scoring at sensitive time points . Comparisons of iso w1118 wild-type to ΔAMPs mutants were made using a Cox-proportional hazard ( CoxPH ) model , where independent experiments were included as covariates , and covariates were removed if not significant ( p>0 . 05 ) . Direct comparisons were performed using Log-Rank tests in Prism seven software . The effect size and direction is included as the CoxPH hazard ratio ( HR ) where relevant , with a positive effect indicating increased susceptibility . CoxPH models were used to test for synergistic contributions of AMPs to survival in R 3 . 4 . 4 . Total sample size ( N ) is given for each experiment as indicated . The native Drosophila microbiota does not readily grow overnight on LB , allowing for a simple assay to estimate bacterial load . Flies were infected with bacteria at the indicated OD as described , and allowed to recover . At the indicated time post-infection , flies were anesthetized using CO2 and surface sterilized by washing them in 70% ethanol . Ethanol was removed , and then flies were homogenized using a Precellys bead beater at 6500 rpm for 30 s in LB broth , with 300 μl for individual samples , or 500 μl for pools of 5–7 flies . These homogenates were serially diluted and 150 μl was plated on LB agar . Bacterial plates were incubated overnight , and colony-forming units ( CFUs ) were counted manually . Statistical analyses were performed using One-way ANOVA with Sidak’s correction . p-Values are reported as <0 . 05 = * , <0 . 01 = ** , and <0 . 001 = *** . For C . albicans , BiGGY agar was used instead to select for Candida colonies from fly homogenates . Flies were infected by pricking flies with a needle dipped in a pellet of either E . coli or M . luteus ( OD600 = 200 ) , and frozen at −20°C 6 hr and 24 hr post-infection , respectively . Total RNA was then extracted from pooled samples of five flies each using TRIzol reagent , and re-suspended in MilliQ dH2O . Reverse transcription was performed using 0 . 5 mg total RNA in 10 μl reactions using PrimeScript RT ( TAKARA ) with random hexamer and oligo dT primers . Quantitative PCR was performed on a LightCycler 480 ( Roche ) in 96-well plates using Applied Biosystems SYBR Select Master Mix . Values represent the mean from three replicate experiments . Error bars represent one standard deviation from the mean . Primers used in this study can be found in Supplementary file 1 . Statistical analyses were performed using one-way ANOVA with Tukey post-hoc comparisons . p-Values are reported as not significant = ns , <0 . 05 = * , <0 . 01 = ** , and <0 . 001 = *** . qPCR primers and sources ( Kounatidis et al . , 2017; Hanson et al . , 2016; Iatsenko et al . , 2016 ) are included in Supplementary file 1 . Two methods were used to collect hemolymph from adult flies: in the first method , pools of five adult females were pricked twice in the thorax and once in the abdomen . Wounded flies were then spun down with 15 μl of 0 . 1% trifluoroacetic acid ( TFA ) at 21000 RCF at 4°C in a mini-column fitted with a 10 μm pore to prevent contamination by circulating hemocytes . These samples were frozen at −20°C until analysis , and three biological replicates were performed with four technical replicates . In the second method , approximately 20 nl of fresh hemolymph was extracted from individual adult males using a Nanoject , and immediately added to 1 μl of 1% TFA , and the matrix was added after drying . Peptide expression was visualized as described in Uttenweiler-Joseph et al . ( 1998 ) . Both methods produced similar results , and representative expression profiles are given . Melanization assays ( Dudzic et al . , 2018 ) and peanut agglutinin ( PNA ) clot staining ( Scherfer et al . , 2004 ) was performed as previously described . In brief , flies or L3 larvae were pricked , and the level of melanization was assessed at the wound site . We used FACS sorting to count circulating hemocytes . For sessile crystal cell visualization , L3 larvae were cooked in dH2O at 70°C for 20 min , and crystal cells were visualized on a Leica DFC300FX camera using Leica Application Suite and counted manually . | All animals – from humans to mice , jellyfish to fruit flies – are armed with an immune system to defend against infections . The immune system’s first line of defence often involves a group of short proteins called antimicrobial peptides . These proteins are found anywhere that germs and microbes come into contact with the body , including the skin , eyes and lungs . In many cases , it is unclear how individual antimicrobial peptides work . For example , which germs are they most effective against ? Do they work alone , or in a mixture of other antimicrobial peptides ? To learn more about a protein , scientists can often delete the gene that encodes it and observe what happens . Antimicrobial peptides , however , are small proteins encoded by a large number of very short genes , which makes them difficult to target with most genetic tools . Fortunately , gene editing via the CRISPR/Cas9 system can overcome many of the limitations of more traditional methods; this allowed Hanson et al . to systematically remove the antimicrobial peptide genes from fruit flies to explore how these proteins work . In the experiments , ten antimicrobial peptide genes known from fruit flies were removed , and the flies were then infected with a variety of bacteria and fungi . Hanson et al . found that the antimicrobial peptides were effective against many bacteria , but unexpectedly they were far more important for controlling one general kind of bacterial infection , but not another kind . Further experiments showed that some of these proteins work alone , targeting only a particular species of microbe . This finding suggested that animals might fight infections by very specific bacteria with a very specific antimicrobial peptide rather than with a mixture . By understanding how antimicrobial peptides work in more detail , scientists can learn what types of microbes they are most effective against . In the future , this information may eventually lead to the development of new types of antibiotics and better management of diseases that affect important insects , like bumblebees . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"immunology",
"and",
"inflammation"
] | 2019 | Synergy and remarkable specificity of antimicrobial peptides in vivo using a systematic knockout approach |
Signal recognition particles ( SRPs ) are universal ribonucleoprotein complexes found in all three domains of life that direct the cellular traffic and secretion of proteins . These complexes consist of SRP proteins and a single , highly structured SRP RNA . Canonical SRP RNA genes have not been identified for some Thermoproteus species even though they contain SRP19 and SRP54 proteins . Here , we show that genome rearrangement events in Thermoproteus tenax created a permuted SRP RNA gene . The 5'- and 3'-termini of this SRP RNA are located close to a functionally important loop present in all known SRP RNAs . RNA-Seq analyses revealed that these termini are ligated together to generate circular SRP RNA molecules that can bind to SRP19 and SRP54 . The circularization site is processed by the tRNA splicing endonuclease . This moonlighting activity of the tRNA splicing machinery permits the permutation of the SRP RNA and creates highly stable and functional circular RNA molecules .
Cells of T . tenax Kra1 ( DSM 2078 ) were a kind gift of R . Hensel ( Essen ) and grown heterotrophically in Thermoproteus medium ( Brock et al . , 1972 ) . For the preparation of T . tenax total ( >200 nt ) and small RNAs ( <200 nt ) , 0 . 1 g pelleted cells were lysed by homogenization and the RNA was subsequently isolated using the mirVanaTM miRNA Isolation Kit ( Ambion , Germany ) according to the manufacturer’s instructions . To generate the SRP S-domain RNA constructs ( WT , Open , GNAR , h8b ) , two forward and reverse complementary DNA oligonucleotides with 15 bases overhangs were synthesized ( Eurofins MWG Operon , Germany Supplementary file 1B ) . The oligonucleotides were phosphorylated , hybridized and cloned upstream of a T7 RNA polymerase promoter into pUC19 . The full-length SRP RNA gene was synthesized with a T7 RNA polymerase promoter and cloned into pMA-RQ ( Life Technologies , Germany ) . The SRP and S-domain RNA constructs were prepared by run-off transcription of 1 µg linearized plasmid as template DNA in a buffer containing 40 mM HEPES/KOH pH 8 . 0 , 22 mM MgCl2 , 5 mM dithiothreitol ( DTT ) , 1 mM spermidine , 4 mM of ATP , CTP , GTP and UTP , 20 U RNase inhibitor and 1 µg T7 RNA polymerase at 37°C for 4 hr . The BHB substrate for RNA cleavage assays was synthesized ( Eurofins MWG Operon , Supplementary file 1B ) . All RNA molecules were purified by phenol/chloroform extraction ( pH 5 . 2 ) , EtOH precipitated , separated by denaturing-PAGE ( 8 M Urea , 1× TBE , 10% polyacrylamide ) next to a RNA marker ( low range ssRNA ladder , NEB , Germany ) and visualized by toluidine blue staining . The gel bands were cut out and eluted overnight on ice in 500 µl elution buffer ( 20 mM Tris/HCl pH 7 . 5 , 250 mM sodium acetate , 1 mM EDTA pH 8 . 0 , 0 . 25% SDS ) and EtOH precipitated . Preparations of T . tenax small RNAs were treated with T4 polynucleotide kinase ( T4 PNK , Ambion ) to ensure proper adapter ligation of RNA termini . The cDNA libraries were prepared according to the TruSeq Small RNA Library Protocol ( Illumina ) and sequenced using HiSeq2000 technology ( Illumina ) at the Max-Planck Genome Centre ( Cologne ) . Sequencing data processing and analyses were performed using CLC Genomics Workbench 8 ( Qiagen , Germany ) . The sequences were trimmed by quality score ( limit: 0 . 05; max . ambiguities: 2 ) , adapter trimming and filtered by length ( 15 nt cutoff ) . The trimmed sequences were mapped to the T . tenax reference genome ( FN869859 ) using default settings ( mismatch cost set to 4 ) . The T . tenax RNA-Seq data is available at Gene Expression Omnibus ( GSE72127 , [http://www . ncbi . nlm . nih . gov/geo/query/acc . cgi ? acc=GSE72127] ) . Fractionation of 1–4 µg T . tenax total RNA was performed by electrophoretic separation on 8% denaturing-polyacrylamide gels next to an RNA and DNA marker ( low molecular weight DNA ladder , NEB ) . Additionally , to verify the circularization of SRP RNA , 15 ng of full-length SRP RNA in vitro transcript and circularized SRP RNA served as a running control . To obtain a circular SRP RNA control , full length SRP RNA transcript was 5'-dephosphorylated for 6 hr at 37°C using alkaline phosphatase ( NEB ) and the reaction was stopped by heating for 10 min at 65°C . Then , the RNA was 5'-phosphorylated using T4 PNK ( NEB ) and ATP for 1 hr at 37°C and ligated using T4 RNA ligase 1 ( NEB ) for 1 hr at 37°C . The reaction was stopped by incubation at 95°C for 5 min in 2× formamide loading buffer ( 95% formamide , 5 mM EDTA pH 8 . 0 , 2 . 5 mg bromophenol blue , 2 . 5 mg xylene cyanol ) . The gel was SYBR Gold stained ( Life Technologies ) and bands were visualized via UV irradiation . The RNA was blotted by capillary transfer onto nylon membranes ( Roti-Nylon plus , Roth ) and immobilized by UV crosslinking . Hybridization was performed at 42°C for 18 hr in ULTRAhyb-Oligo buffer ( Ambion ) with two 5'-labeled probes ( the label was added using T4 PNK [NEB] and γ-[32P]-ATP [5000 ci/mmol , Hartmann Analytic] ) that were complementary to different regions in the SRP S-domain ( 42 or 45-mer probe , Supplementary file 1B ) . The blot was washed once ( 2× SSC and 0 . 1% SDS ) at 42°C for 30 min followed by a second washing step ( 1× SSC and 0 . 1% SDS ) at equal conditions . Radioactive signals were visualized by phosphorimaging . The T . tenax genes srp19 ( TTX_2083 ) and srp54 ( TTX_0615 ) were cloned into pET20b creating a fusion with a C-terminal 6× His-tag . The plasmids were transformed into E . coli strain DH5α ( Invitrogen ) and protein production carried out in E . coli Rosetta2 ( DE3 ) pLysS cells ( Stratagene ) . Cultures were grown in LB medium at 37°C , shaking at 200 rpm . For protein production , 1 mM IPTG was added to a growing culture ( OD600: 0 . 6 ) which was incubated for 3 hr . SRP19-His expression cells were homogenized in buffer 1 ( 50 mM Tris/HCl pH 8 , 300 mM NaCl ) , lysed by sonication and cleared by centrifugation ( 45 , 000 × g , 1 hr , 4°C ) . The protein was heat-precipitated ( 30 min , 70°C ) and centrifuged ( 14 , 000 × g , 30 min , 4°C ) . The SRP19-His protein was further purified by Ni-NTA affinity chromatography ( HisTrap HP , GE Healthcare ) and eluted with a linear imidazole gradient ( 0–500 mM ) at 400 mM imidazole using a FPLC Äkta system ( GE Healthcare ) . SRP54-His was purified by cell lysis in buffer 2 ( 100 mM potassium phosphate pH 7 . 5 , 500 mM NaCl , 10% glycerol , 1mM ß-Me ) , lysed by sonication and cleared by centrifugation ( 45 , 000 × g , 1 hr , 4°C ) . The protein was loaded onto a Ni-NTA affinity chromatography column and eluted with a 500 mM imidazole step . Protein fractions were dialysed in buffer 3 ( 50 mM Tris/HCl pH 7 . 5 , 10% glycerol , 1 mM DTT ) and purified over a HiTrap Heparin Sepharose HP column ( GE Healthcare ) eluting with a linear salt gradient ( 0–1 M ) at 650 mM NaCl . The formation of SRP RNA-protein complexes was tested using a DEAE-Sepharose assay as described earlier ( Bhuiyan , 2000; Gowda , 1997 ) . 2 µM of the SRP S-domain RNA constructs ( WT , Open , GNAR , h8b ) were folded in 55 µl binding buffer ( 50 mM Tris/HCl pH 8 , 300 mM KOAc , 5 mM MgCl2 , 1 mM DTT ) , heated for 10 min at 65°C and cooled down to room temperature for 1 hr . 2 µM of SRP19 or SRP54 were added and the reaction volume was adjusted to 70 µl . The mixture was incubated for 45 min at 37°C and immediately loaded onto an 80 µl bed-volume DEAE-Sepharose column ( GE Healthcare ) equilibrated in binding buffer . The flow-through ( F ) was combined with a 70 µl binding buffer wash . Bound complex ( E ) was eluted twice with 70 µl of elution buffer ( binding buffer with 1 M KOAc ) . Fractions F and E were TCA-precipitated , separated on 15% SDS-polyacrylamide gels , stained with InstantBlue and analyzed with the ImageJ software ( imagej . nih . gov/ij/ ) . The binding efficiency was tested in three independent experiments . The T . tenax genes TTX_1594 ( catalytic subunit ) and TTX_1893 ( structural subunit ) were cloned together into pETDuet-1 and the protein was purified as described earlier ( Zwieb , 2005 ) . Briefly , cells were resuspended in buffer 4 ( 50 mM Tris/HCl pH 7 . 5 , 500 mM NaCl , 3 mM DTT ) , sonicated and cleared by centrifugation ( 45 , 000 × g , 1 hr , 4°C ) . The cell lysate was heat-precipitated ( 30 min , 80°C ) and centrifuged ( 14 , 000 × g , 30 min , 4°C ) . 10 pmol of the BHB substrate for RNA cleavage assays was 5'-labeled with T4 PNK ( Ambion ) and γ-[32P]-ATP ( 5000 ci/mmol , Hartmann Analytic ) and gel-purified as described above . 1–1000 nM purified splicing endonuclease was incubated with 2 nM of 5'-labeled RNA in nuclease buffer ( 40 mM Tris/HCl pH 8 , 2 mM MgCl2 , 1 mM EDTA ) at 60°C for 20 min . The cleavage reaction was stopped by adding formamide loading buffer and incubating at 95°C for 5 min . The reaction was loaded onto a 20% denaturing polyacrylamide gel running in 1× TBE , 6 W for 4 hr next to a low molecular weight marker ( 10–100 nt , Affymetrix ) and an alkaline hydrolysis ladder of the BHB substrate . The cleavage products were visualized by phosphorimaging . | Cells make many proteins that are eventually released outside the cell or inserted into the cell’s membrane . As these proteins are still being made , they are captured by a “signal recognition particle” ( or SRP ) ; this molecular machine then guides the newly forming protein to the cell’s membrane . SRPs are found in all living organisms on Earth and contain several different proteins and a short RNA molecule . However , a few species belonging to the archaeal domain of life did not seem to contain an identifiable gene for the RNA component of the SRP . Now Plagens et al . have sought to solve the mystery of the “missing” component of this essential protein-targeting machine . This involved searching through the RNAs that are produced by an archaeon called Thermoproteus tenax , a single-celled microbe which grows in the absence of oxygen and at temperatures of up to 95°C . Plagens et al . discovered that the “missing” SRP RNA gene had not yet been identified because rearrangements in this archaeon’s genome had swapped the left and right portions of the SRP RNA gene . Further experiments revealed that the correct sequence order is restored in mature SRP RNA molecules by the two ends of the molecule being linked to form a circle . These RNA circles are made by the cellular machinery that normally removes the unneeded sections from other RNA molecules ( called transfer RNAs ) . Circular RNA is much more stable at high temperatures and does not degrade easily , and Plagens et al . suggest that this particular arrangement is therefore especially advantageous for this species . Future work will now aim to work out which selective pressures favor the evolution of such fragmented RNAs . | [
"Abstract",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"short",
"report"
] | 2015 | Circularization restores signal recognition particle RNA functionality in Thermoproteus |
We report that Histone Deacetylase 7 ( HDAC7 ) controls the thymic effector programming of Natural Killer T ( NKT ) cells , and that interference with this function contributes to tissue-specific autoimmunity . Gain of HDAC7 function in thymocytes blocks both negative selection and NKT development , and diverts Vα14/Jα18 TCR transgenic thymocytes into a Tconv-like lineage . Conversely , HDAC7 deletion promotes thymocyte apoptosis and causes expansion of innate-effector cells . Investigating the mechanisms involved , we found that HDAC7 binds PLZF and modulates PLZF-dependent transcription . Moreover , HDAC7 and many of its transcriptional targets are human risk loci for IBD and PSC , autoimmune diseases that strikingly resemble the disease we observe in HDAC7 gain-of-function in mice . Importantly , reconstitution of iNKT cells in these mice mitigated their disease , suggesting that the combined defects in negative selection and iNKT cells due to altered HDAC7 function can cause tissue-restricted autoimmunity , a finding that may explain the association between HDAC7 and hepatobiliary autoimmunity .
To become mature T cells , thymocytes must navigate through a complex process of selection and instruction , centered around signals received through their newly created T cell antigen receptors ( TCRs ) . For thymocytes destined to become conventional naïve CD4 or CD8 T cells ( Tconv ) , this requires passing two key checkpoints: positive selection , in which cortical CD4/CD8 double-positive ( DP ) thymocytes must receive a minimum level of TCR stimulation from self peptide-MHC complexes in order to adopt the appropriate lineage and continue maturation , and negative selection , in which thymocytes with self-reactivity above a critical threshold are deleted from the repertoire by activation-induced apoptosis . While the elucidation of these mechanisms decades ago established a basic conceptual framework for the creation of a competent and self-tolerant T cell repertoire , the years since have brought to light an ever-increasing variety of alternate developmental programs that produce specialized populations of mature T cells functionally distinct from Tconv . These populations , critical for both effective host defense and self-tolerance , are elicited from the diverse pool of T cell precursors by specialized selection mechanisms , mostly involving strong recognition of noncanonical ligands , as in the case of NKT cells ( Kronenberg , 2014 ) , or recognition of peptide-MHC ligands at high TCR avidities near the threshold of negative selection , as in the case of nTreg or CD8αα IEL ( Klein et al . , 2014; Moran et al . , 2011 ) . The TCR signals involved in their development are generally stronger than those that mediate positive selection to the Tconv lineage , and the process is thus termed agonist selection ( Stritesky et al . , 2012 ) . One feature that distinguishes many of these specialized cell types from Tconv is the thymic acquisition of constitutive effector function , a phenotype shared with innate immune cells and thus giving rise to the term ‘innate-like’ or ‘innate effector’ T cells . Whereas Tconv exit the thymus with a naive phenotype , circulate broadly , and require a several days long , orchestrated process of priming and clonal expansion to become fully functional effector/memory cells , innate-like T cells are often constitutively tissue-resident and make mature effector responses to their cognate stimuli immediately ( Kang and Malhotra , 2015 ) . Innate-like T cells exit the thymus larger than Tconv , with an antigen-experienced phenotype and an expanded secretory apparatus , allowing them to rapidly elaborate robust cytokine responses after brief TCR stimulation ( Kang and Malhotra , 2015; Chandra and Kronenberg , 2015; Brennan et al . , 2013 ) . These differences arise due an alternative thymic maturation process that parallels the priming of naïve T-cells in the periphery . For NKT cells , this includes a ~ 100 fold intra-thymic proliferative expansion to generate pre-established clonal populations ( Benlagha et al . , 2002 ) . Maintenance of the innate effector phenotype in NKT cells can at least partially be attributed to stable expression of their signature transcription factor Promyelocytic Leukemia Zinc Finger Protein ( PLZF , ZBTB16 ) ( Kovalovsky et al . , 2008; Savage et al . , 2008 ) . PLZF expression is established during thymic development of NKT cells , via a cellular mechanism that involves strong recognition of glycolipid ligands on the non-canonical CD1D MHC molecule by a clonally restricted ( in mice , Vα14/Jα18 with one of several possible β chains ) TCR , together with homotypic co-stimulation through the SAP family of co-receptors ( Bendelac et al . , 2007 ) . However , what downstream factors link these surface signals to stable PLZF expression and what other pathways may be involved are still open questions . We have previously described how Tconv development is regulated by the class IIA histone deacetylase Histone Deacetylase 7 ( HDAC7 ) , a TCR signal-regulated corepressor abundantly expressed in thymocytes ( Dequiedt et al . , 2003; Kasler and Verdin , 2007 ) . The activity of HDAC7 is controlled by nuclear exclusion in response to phosphorylation of conserved serine residues in their N-terminal adapter domains ( Verdin et al . , 2003 ) . In thymocytes , TCR stimulation results in HDAC7 phosphorylation and nuclear exclusion via Protein Kinase D ( Parra et al . , 2005 ) . CD4/CD8 double-positive ( DP ) thymocytes lacking HDAC7 are much more likely than WT thymocytes to die before becoming positively selected , significantly impeding their development into mature Tconv ( Kasler et al . , 2011 ) . Conversely , if a transgene encoding a phosphorylation-deficient , constitutively nuclear version of human HDAC7 ( HDAC7-ΔP ) is transiently expressed in the thymus at sub-endogenous levels ( Kasler et al . , 2012 ) , deletion of autoreactive thymocytes by negative selection is strongly blocked and the hosts develop lethal autoimmunity . Consistent with broad blockade of negative selection , we observed autoantibodies to a comprehensive array of tissue antigens in these mice ( Kasler et al . , 2012 ) . However , for reasons that were not clear to us at the time actual tissue destruction occurred almost exclusively in a gastrointestinal/hepatobiliary compartment that is anatomically tied together by the contiguous epithelial surfaces of the GI lumen and the pancreatic and biliary ductal systems ( Kasler et al . , 2012 ) . The potential significance of this peculiar pattern of HDAC7-mediated autoimmunity for human disease has recently been brought into sharp focus by two separate studies identifying polymorphisms at the loci of HDAC7 as well as several of its upstream regulatory kinases as independent risk factors in human inflammatory bowel disease ( IBD ) , and also in primary sclerosing cholangitis ( PSC ) , a destructive autoimmune syndrome of the hepatobiliary system , which is additionally associated with increased IBD risk ( Liu et al . , 2013; Jostins et al . , 2012 ) . The striking parallel between these human syndromes and the autoimmunity observed in HDAC7-ΔP transgenic mice suggested to us a connection between HDAC7 and these types of autoimmunity that goes beyond simply blocking thymic negative selection . This led us to undertake a more thorough phenotypic characterization of mice with altered HDAC7 function during T cell development , revealing that HDAC7 has a key role in the regulation of the innate effector programming of iNKT cells , at least in part via direct modulation of the transcriptional activity of PLZF . Both gain and loss of HDAC7 function in thymocytes resulted in aberrant effector programming of T cells in both the Tconv and innate-like lineages , leading to multiple abnormalities in peripheral populations . These studies shed new light on the molecular pathways that regulate the effector programming of innate-like T cells , reveal a new key molecular target of HDAC7 in T cell development , and set forth a novel cellular model of tissue-specific autoimmunity , in which one genetic lesion mediates multiple defects in thymic selection , which then converge in the periphery to produce a unique , tissue-restricted pattern of disease . Given the established genetic association between HDAC7 variants and very similar human syndromes , our findings are likely to be of considerable significance in the understanding of these diseases .
We previously showed that if a constitutively nuclear mutant of human HDAC7 ( HDAC7-ΔP ) is transiently expressed at normal levels during thymic T cell development but not in mature T cells , autoreactive cells that would normally die by negative selection instead exit the thymus as naïve Tconv ( Kasler et al . , 2012 ) . However in our previous study we did not assess the fates of most cells destined to become innate effectors . Analyzing these populations , we noted a modest suppression of Treg ( Kasler et al . , 2012 ) and CD8αα IEL ( Figure 1—figure supplement 1A ) , but the most striking observation we made was the near total absence of invariant Natural Killer T cells ( iNKT ) , an oligoclonal population that is reactive to α-galactosyl ceramide ( αGalCer ) presented by the CD1D non-canonical MHC molecule ( CD1D/αGalCer ) ( Kronenberg , 2014 ) . Cells positive for staining with CD1D/αGalCer tetramers represent approximately 3% of TCRβ-positive cells in wild type C57BL/6 ( B6 ) thymus and 30% in liver , however they are nearly undetectable in either of these tissues or in the spleens or livers of HDAC7-ΔP mice ( Figure 1A , B; Figure 1—figure supplement 1B , for full gating see Figure 1—figure supplement 1A ) , suggesting a profound deficiency in iNKT development . There were however consistently more cells detected in the thymus of HDAC7-ΔP transgenic mice with αGalCer-loaded tetramer than with empty tetramer ( Figure 1A , Figure 1—figure supplement 1C–D ) , suggesting that iNKT cells are not entirely absent in this background . Analyzing these cells according to the conventional staging system for iNKT development ( Stritesky et al . , 2012 ) , we found that rather than being predominantly CD44hi/NK1 . 1+ ( Stage 3 ) , as in the case of WT iNKT cells , the few thymic tetramer-reactive cells from HDAC7-ΔP mice were evenly distributed between the CD44hi/NK1 . 1+ , CD44hi/NK1 . 1- ( Stage 2 ) , and CD44lo/NK1 . 1- ( Stage 0–1 ) populations ( Figure 1—figure supplement 1D ) . Further analysis of the Stage 0–1 population showed these cells to be predominantly CD24hi , indicating a profound reduction in numbers at all developmental stages that were detectable above background ( Figure 1—figure supplement 1C ) . Examining these stages after ~20 fold enrichment of iNKT cells using tetramer and magnetic beads , we noted the same pattern , with all populations other than CD24 hi /CD44lo/NK1 . 1- cells being highly underrepresented ( Figure 1C , D ) . These results are consistent with either a developmental block before Stage one or a severe defect in the survival or normal proliferation of iNKT cells from Stage one onwards . We also evaluated the prevalence of CD44/NK1 . 1-expressing T cells that were not tetramer-reactive , and noted a marked reduction in their numbers in liver and spleen as well ( Figure 1—figure supplement 2E–F ) , suggesting a broad defect in the development of the NKT lineage . To rule out cell-extrinsic mechanisms for this phenotype , we generated mixed hematopoietic chimeras reconstituted with a 1:1 mixture of wild-type ( WT ) and HDAC7-ΔP bone marrow . As we previously reported ( Kasler et al . , 2012 ) , the HDAC7-ΔP transgenic population contributed robustly to the pool of CD4 and CD8 SP thymocytes , although there was a transient reduction in prevalence at the immature single positive ( ISP ) stage ( Figure 1—figure supplement 2A ) . At early time points post-reconstitution ( 6-8wk ) , the distributions of naïve and memory T-cells in peripheral CD4 +and CD8+Tconv subsets were equivalent as well ( Figure 1—figure supplement 2B ) . However , while the wild type-derived population reconstituted hepatic iNKT cells efficiently , HDAC7-ΔP bone marrow made almost no contribution to this compartment in the liver , where iNKT cells are most abundant ( Figure 1E ) . This was also true in the thymus and spleen ( Figure 1—figure supplement 2C ) , demonstrating that the abnormalities observed in the intact transgenic mice were due to a cell-autonomous mechanism . We next examined the effects of loss of HDAC7 in the thymus on these phenotypes , using our previously characterized strain that deletes loxp-flanked Hdac7 under the control of the Lck proximal promoter ( Hdac7flox:/:-::lckcre , henceforth Hdac7-KO ) ( Kasler et al . , 2011 ) . We previously reported that loss of HDAC7 during T cell development increased apoptosis of DP thymocytes leading to inefficient positive selection . This shortened thymocyte lifespan resulted in a truncation of the TCR Jα repertoire , with distal rearrangements underrepresented ( Kasler et al . , 2011 ) . It was thus not surprising to find that Hdac7-KO mice with an endogenous TCR repertoire had fewer iNKT cells than WT controls; for example , Hdac7-KO thymus had a 2–5 fold lower abundance of iNKT cells than WT littermates ( Figure 2A ) . This reduction , consistent with the degree of underrepresentation of the relatively distal Jα18 TCR segment we previously noted ( Kasler et al . , 2011 ) , was similarly observed in the spleen and liver ( Figure 3—figure supplement 1A , B ) . Importantly , unlike the residual tetramer-reactive cells in HDAC7-ΔP mice , when staged after magnetic enrichment , iNKT calls in Hdac7-KO mice had normal expression of CD44 and NK1 . 1 , suggesting that their development was not functionally altered . ( Fig , 2A , at right ) . Although deletion of Hdac7 did not result in expansion of NK1 . 1-expressing T cells , we did observe significant abnormalities in the effector programming of non-tetramer-reactive thymocytes . We noted a substantial expansion of a CD44hi Eomes +population in the mature CD8 SP compartment in the thymus ( Figure 2B , Figure 2—figure supplement 1B ) . Examination of the peripheral CD8 T cells in these animals also showed a substantial increase in CD44 expression , suggesting an expansion of innate effector CD8 cells ( Figure 2C ) . These cells resemble Eomes +innate memory CD8 +cells that are typically generated in trans , in response to IL4 secretion by thymic-resident iNKT cells ( Lee et al . , 2013; Weinreich et al . , 2010 ) , however as previously noted iNKTcells are depleted rather than expanded in Hdac7-KO mice , and we did not observe a consistent increase in the proportion of PLZF- or Vγ6 . 3-positive γδ T cells ( Figure 2—figure supplement 1A , C ) , suggesting a different mechanism . To clarify this question , we examined the phenotypes resulting from loss of HDAC7 in WT: Hdac7-KO mixed hematopoietic chimeras . In 1:1 chimeras , Hdac7-KO thymocytes competed equally through the ISP stage , but thereafter competed poorly and became steadily less abundant . This substantial underrepresentation of Hdac7-KO CD4 SP and mature CD8 SP thymocytes in 1:1 chimeras ( Figure 2—figure supplement 2A–B ) made analysis at this stage difficult , however their representation in the periphery was sufficient . In the spleen , we saw a strong increase in CD44 expression in the Hdac7-KO-derived vs . to the WT-derived CD8 T cell population ( Figure 2D , E ) , duplicating what we saw in the intact mice and indicating that the phenotypes we observed are likely cell-autonomous . To further characterize the phenotype of these cells , we briefly stimulated splenocytes from these chimeras ex vivo , and found that Hdac7-KO-derived CD8 +T cells produced much more IFNγ than WT-derived CD8 +T cells in the same culture , assessed both as percent cytokine-positive ( Figure 3F , G ) and by median fluorescence intensity ( MFI ) of cytokine staining ( Figure 3H ) . CD8 +T cells from Hdac7-KO population also had increased expression of the Eomes-associated chemokine receptor CXCR3 and the trafficking receptor Ly6C ( Figure 2—figure supplement 1D ) . Loss of HDAC7 thus appears to result in the aberrant adoption of innate effector programming by CD8 SP thymocytes that would otherwise have exited the thymus as naive Tconv . We observed a much more modest degree of abnormality in the CD4 compartment , comprising a 20–30% increase in the frequency of memory and IL4-secreting cells ( Figure 2—figure supplement 2B–C ) , which we hypothesize is due to the greater similarity that CD8 thymic selection bears to NKT selection , in terms of both the similarity of CD1D to Class I MHC and the availability of selecting ligands on all thymocytes rather than just on specialized thymic APC . Loss of HDAC7 may thus allow some DP thymocytes to aberrantly adopt this lineage through some partial analogue of NKT selection . Collectively , our findings with both the Hdac7-KO and HDAC7-ΔP transgenic strains suggest that HDAC7 may function as a gatekeeper of innate effector programming , blocking the functional maturation of iNKT cells when constitutively expressed in the nucleus , and conversely allowing the aberrant acquisition of innate effector characteristics in Conventional T cells when it is conditionally deleted . To generate a larger population of iNKT precursors for more in-depth evaluation the role of HDAC7 , we employed the Vα14-Jα18 TCRα transgene ( henceforth ‘Vα14’ ) , encoding the invariant TCRα chain that when paired with the appropriate endogenous β chains allows iNKT cells to bind glycolipids with high affinity ( Griewank et al . , 2007 ) . Expressing this TCR transgene greatly increases the frequency of CD1D/αGalCer-reactive thymocytes , which arise naturally only at only around 1 in 104 cells . As expected , mice expressing only the Vα14 transgene had many more iNKT cells in thymus and spleen than WT mice ( Figure 3A , Figure 3—figure supplement 1A–B ) . Also consistent with our expectations , when we crossed the Vα14 TCRα transgene into the Hdac7-KO strain , we observed a complete rescue of iNKT cell abundance in the thymus and periphery ( Figure 3—figure supplement 1A–B ) , resulting in identical numbers between Vα14 and Vα14 Hdac7-KO mice . These cells were phenotypically similar to Vα14 iNKT cells in terms of CD44/NK1 . 1 expression ( Figure 3—figure supplement 1A , bottom ) , suggesting that shortened thymocyte lifespan was indeed the main cause of the reduced iNKT abundance in Hdac7-KO mice . In contrast to this finding , when the Vα14 transgene was co-expressed with HDAC7-ΔP , the rescue in the numbers of CD1D/αGalCer-reactive cells was incomplete ( Figure 3A , Figure 3—figure supplement 1C ) , and the cells were phenotypically abnormal ( Figure 3B–F ) . This result suggests that rather than blocking the maturation of CD1D/αGalCer-reactive cells categorically , HDAC7-ΔP blocked one or more steps normally associated with post-positive selection iNKT differentiation ( Benlagha et al . , 2002 ) , directing the cells instead to mature as if they were positively selected Tconv . Consistent with this idea , other characteristics of CD1D/αGalCer-reactive Vα14 x HDAC7-ΔP T cells were similar to those of naïve Tconv . Flow analysis revealed that like the residual tetramer-reactive cells present in the HDAC7-ΔP mice ( Figure 1C ) the rescued iNKT cells in Vα14 x HDAC7-ΔP mice failed to upregulate the memory marker CD44 or the NKT marker NK1 . 1 in the thymus like their Vα14-only counterparts ( Figure 3C , top row ) , although they did downregulate CD24 nearly as efficiently as Vα14 iNKT cells ( Figure 3—figure supplement 1D ) , suggesting that they were able to mature to stage 1 . This phenotype persisted in the spleen , after the HDAC7-ΔP transgene was turned off ( Figure 3B , bottom row ) , suggesting that the cells had failed to complete effector programming in the thymus . We next examined their cytokine responses to brief ex-vivo stimulation . When stimulated for 4 hr with PMA/ionomycin , CD1D/αGalCer-reactive WT and Vα14 transgenic iNKT cells exhibited a robust cytokine response , secreting both IFNγ and IL-4 . In contrast , Vα14 x HDAC7-ΔP iNKT were far less likely to make IFNγ or IL-4 ( Figure 3C , D ) , as would be expected for naïve Tconv . Additionally , iNKT cells typically express high levels of the integrin LFA-1 ( CD11a/CD18 ) , allowing them to remain localized in tissue-specific vascular beds such as hepatic sinusoids ( Thomas et al . , 2011 ) . In contrast , Vα14 x HDAC7-ΔP iNKT cells exhibited far lower expression levels ( Figure 3E , F ) , comparable to those seen in circulating non-CD1D/αGalCer-reactive CD4+ ( mainly naïve ) T-cells ( Figure 3F , right ) . Moreover , while Vα14 x HDAC7-ΔP iNKT cells were found at comparable frequency in spleen to WT iNKT cells , they failed to concentrate in peripheral tissues such as the liver ( Figure 3—figure supplement 1E , F ) , a behavior more characteristic of naïve Tconv rather than iNKT cells . These data are most consistent with a model in which HDAC7-ΔP prevents iNKT precursors from initiating innate effector development: Since Vα14 x HDAC7-ΔP iNKT cells have low CD44 expression , produce few cytokines after brief stimulation , and freely recirculate , they appear to become diverted into functionally naïve-like T-cells . When considering how both gain and loss of thymic HDAC7 function alter innate effector development , we were struck by how closely our results mirrored findings reported in similar studies of the transcription factor PLZF . Specifically , the severe depletion of iNKT cells ( Figure 1A ) and loss of effector memory phenotype in peripheral iNKT cells observed in gain-of-function HDAC7-ΔP ( Figure 3B ) strongly resembles the iNKT defect observed in PLZF knockouts ( Kovalovsky et al . , 2008; Savage et al . , 2008 ) . Conversely , the consequences of loss of HDAC7 function – notably expansion of IFNγ-secreting CD8 +and IL4-secreting CD4 +memory cells ( Figure 2D–H , Figure 1—figure supplement 2B–C ) mirror results reported in gain-of-function PLZF transgenic mice ( Kovalovsky et al . , 2010; Savage et al . , 2011 ) . Polyclonal ( non-invariant ) type II NKT cells are also thought to be PLZF-dependent ( Zhao et al . , 2014 ) , and we similarly noted a near absence of tissue-resident type II NKTs in HDAC7-ΔP mice , defined by a Tet-TCRβ+CD8-CD44hiNK1 . 1+ profile ( Figure 1—figure supplement 1E–F ) . HDAC7 and PLZF thus appear to play nearly inverse roles in iNKT development ( Figure 4E ) . One possible mechanism for this inverse relationship is that nuclear HDAC7 represses the expression of PLZF , preventing HDAC7-ΔP thymocytes from expressing PLZF ( Seiler et al . , 2012 ) . Indeed , we observed a pronounced reduction in PLZF expression in TCRβ+T cells from HDAC7-ΔP mice in all organs examined , including thymus , spleen and liver ( Figure 4A , B ) . However , PLZF was still detected in CD4 +SP thymocytes from Vα14 x HDAC7-ΔP mice; although expression was restricted compared to Vα14-only thymus ( Figure 4C ) , Interestingly , PLZF expression was maintained in roughly half of splenic Vα14 x HDAC7-ΔP iNKT cells ( Figure 4D , right panel ) . Thus , transcriptional repression of PLZF expression by HDAC7 is probably insufficient to fully explain the iNKT phenotype , as even PLZF +Vα14 x HDAC7-ΔP iNKT cells exhibit naïve-like characteristics ( Figure 3B ) . The remarkable inverse relationship between the phenotypes mediated by alterations of HDAC7 and PLZF function in iNKT cell development prompted us to take an unbiased , genome-wide approach to understanding how these two factors might coordinately regulate the transcriptional landscape of this process . To this end , we generated gene expression profiles by RNA-seq of PBS-57 tetramer-reactive Vα14 Tg and Vα14 X HDAC7-ΔP Tg thymocytes and splenocytes , as well as polyclonal naïve ( i . e . CD44- ) conventional CD4 SP thymocytes and splenocytes . Differential gene expression profiles were constructed for Vα14 Tg vs . naïve Tconv , Vα14 X HDAC7-ΔP Tg vs naïve Tconv , and Vα14 X HDAC7-ΔP vs . Vα14 Tg , by comparing the normalized scalar expression values for three biological replicates of each condition , based on roughly 40 million mapped reads per sample ( See Supplementary file 1 , Materials and methods ) . When we plotted significant expression changes for tetramer-reactive Vα14 Tg cell vs . Tconv ( Figure 5A , left and right panels , horizontal axes ) against the corresponding changes for Vα14 X HDAC7-ΔP vs Tconv ( vertical axes ) , it was evident in both thymus and spleen that HDAC7-ΔP makes the overall gene expression pattern of tetramer-reactive cells more similar to that of Tconv , as shown by the clockwise shift of the plot trend line from the diagonal in both tissues ( Figure 5A , solid plot diagonal vs . dotted trend line ) . Reflecting this effect , iNKT development-associated gene expression changes ( both up and down ) that were suppressed by HDAC7-ΔP ( Figure 5A , green plot points and numbers ) greatly outnumbered those enhanced by HDAC7-ΔP ( Figure 5A , red plot points and numbers ) in both spleen and thymus . Strongly induced genes involved in iNKT cell development that were suppressed by HDAC7-ΔP included Id2 , Zbtb16 ( PLZF ) , Klrb1c ( NK1 . 1 ) , Tbx21 ( T-bet ) , Gata3 , Il4 , Ifng , and Zfp683 , which encodes HOBIT , a zinc-finger transcription factor recently shown to be essential for the acquisition of tissue-resident effector function ( Mackay et al . , 2016 ) ( Figure 5A , labeled points ) . This pattern of suppression was established in the thymus ( Figure 5A , left ) , but persisted in the spleen ( Figure 5A , right ) , after expression of HDAC7-ΔP was turned off . Blocking HDAC7 nuclear export in the thymus thus apparently programs a more naïve-like state of differentiation into tetramer-reactive cells that persists even after HDAC7-mediated repression is removed . Although some of the changes in gene expression that we observe , especially in the spleen , may be due to the different population distributions with respect to conventional iNKT staging that were sampled between the Vα14 Tg and Vα14 X HDAC7-ΔP tetramer-reactive cells , for many of the genes we identified ( e . g . Hobit , T-bet , Figure 5A ) , the magnitude of the suppression , i . e . lower than in the WT cells , is still greater than could be accounted for by this explanation . These data were also helpful in identifying key candidate molecular targets of HDAC7 . Ingenuity Pathway Analysis ( IPA , Qiagen ) analysis of putative upstream regulators of the HDAC7-affected gene set identified multiple targets highly relevant to iNKT development and function , including Zbtb16 ( PLZF ) , Id2 , Il4 , Ifng , Tbx21 ( T-bet ) , and Gata3 ( Figure 5—figure supplement 1A , for a complete list of putative upstream regulators see Supplementary file 3 ) . The downstream targets of these were almost universally affected in a manner that suggests inhibition rather than activation of the putative upstream regulator ( Figure 5—figure supplement 1a , column 2 ) . The expression of most of these upstream regulators was itself suppressed by HDAC7 , suggesting an obvious mechanism of regulation ( Figure 5—figure supplement 1A , column 3 ) , however the Tec kinase Itk , the most highly correlated upstream regulator of HDAC7 targets in both thymus and spleen , was only modestly suppressed in spleen and not significantly suppressed in thymus , suggesting that HDAC7 might regulate its activation more than its expression . ITK has a well-characterized role in the maturation of conventional CD8 T cells , CD8 innate effectors , and iNKT cells ( Atherly et al . , 2006; Felices and Berg , 2008 ) . Similarly , PLZF ( Zbtb16 ) expression was relatively modestly repressed by HDAC7-ΔP ( e . g , 12-fold , vs . 30 fold induction in in thymus , Figure 5A ) , yet its downstream targets were very highly correlated with the HDAC7 target gene set , based on both IPA analysis and comparison of HDAC7-regulated genes with genes identified in a recent , comprehensive study of PLZF-regulated genes in iNKT cell development ( Mao et al . , 2016 ) ( Figure 5B–C , Figure 5—figure supplement 1C ) . Gene expression changes due to loss of PLZF function in iNKT cells show a clear positive correlation with changes caused by expression of HDAC7-ΔP ( Figure 5B , Figure 5—figure supplement 1C ) , while changes caused by expression of a PLZF transgene show a clear negative correlation ( Figure 5B , Figure 5—figure supplement 1C ) , demonstrating an inverse relationship between HDAC7 and PLZF function . Genes that were found to associate directly with PLZF by chIP-seq ( Mao et al . , 2016 ) , cluster strongly around the HDAC7-PLZF diagonals , and are highly concentrated among the most negatively correlated genes in terms of the effects of HDAC7 vs . PLZF function ( Figure 5D–E , labeled genes , Figure 5—figure supplement 1C , red asterisks ) . Out of the 31 genes reported by Mao , et al . to be both bound by PLZF and differentially expressed in iNKT cells due to alteration of PLZF function , 17 were found on the PLZF-HDAC7 inverse diagonals and only four on the positive diagonals ( Figure 5B–C , labeled genes , Figure 5—figure supplement 1C , red asterisks ) . An additional four genes were negatively correlated with HDAC7 function but not differentially expressed during iNKT development ( see Supplementary file 1 ) , while one was positively correlated . Additionally , Mao , et al . identified BACH2 as a crucial interaction partner of PLZF , and our own data show BACH2 as not differentially expressed but nonetheless as one of the strongest putative upstream regulators of the HDAC7-regulated gene set ( Figure 5—figure supplement 1A ) , suggesting that HDAC7 may modulate its targets via a ternary interaction with PLZF . This remarkable degree of overlap strongly supports the idea that HDAC7 is a negative regulator of iNKT cell development that functions at least in part by negatively regulating PLZF-dependent transcription . Ontologic analysis of HDAC7-regulated genes using IPA provided strong evidence for their association with both innate-like effector function and inflammatory disease . Canonical pathways associated with the HDAC7-regulated gene set included multiple pathways associated with innate immune signaling and T cell effector function ( Figure 5—figure supplement 1B , green-shaded pathways , see Supplementary file 2 for a complete list of pathways and associated genes ) , as well as with inflammation and inflammatory disease states ( Figure 5—figure supplement 1A , blue-shaded pathways ) , particularly hepatic inflammation . This connection was brought into even sharper relief by two recent GWAS studies of primary sclerosing cholangitis ( PSC ) and inflammatory bowel disease ( IBD ) , which both identified HDAC7 among the disease-associated loci , and also individually its immediate upstream kinases PKD and SIK2 , as well as two isoforms of PKC that are upstream of PKD ( Figure 6A ) ( Liu et al . , 2013; Jostins et al . , 2012 ) . Moreover , a remarkably high proportion of the other hits from these studies are downstream of HDAC7 , i . e . their expression in iNKT cells is altered by HDAC7-ΔP . Of the 176 GWAS risk loci mapping to genes that were expressed in our RNA-seq data , 81 ( 46% ) were regulated by HDAC7 in NKT cells , a much higher degree of overlap than would be expected by chance ( p=3 . 49×10−16 , binomial distribution ) ( Figure 6A ) . Of the 16 strongest risk loci identified by the Liu , et al . study of PSC , 10 were differentially expressed due to expression of HDAC7-ΔP , and four more comprised HDAC7 itself , as well as its upstream regulators PRKD2 and SIK2 , and also PLZF interaction partner BACH2 ( Parra et al . , 2005; Mao et al . , 2016; Liu et al . , 2013 ) ( Figure 6 ) . To gain a better understanding of the significance of this overlap , we evaluated the 81 risk loci that were regulated by HDAC7 with respect to regulation by PLZF , differential expression in iNKT cells vs . Tconv , functional role in iNKT cell development , and functional role in autoimmune disease ( Figure 6B ) . This analysis revealed that a large proportion of these genes were functionally important in iNKT development ( Figure 6B , sixth row ) , while relatively fewer were identified as PLZF targets ( Figure 6B , third row ) , suggesting that HDAC7 affects autoimmunity and iNKT development via both PLZF-dependent and independent mechanisms . We then further filtered the genes for significance in at least 4 of the eight criteria examined ( Figure 6B ) , and then manually mapped the resulting 56 genes to their associated signaling pathways ( Figure 6C ) . Remarkably , all but 13 of these genes could be mapped to one of five interconnected signaling networks , comprising Th1 and Th2 cytokine signaling , chemokine signaling , TCR signaling with its associated costimulatory pathways , and signaling through cell membrane-associated TNF superfamily members ( Figure 6C , dark-colored symbols with white label , gray-shaded areas ) . These signaling networks are also heavily populated with HDAC7 targets that were not identified in the GWAS studies ( Figure 6C , light-colored symbols ) , an observation that is confirmed by IPA analysis of canonical signaling pathways and upstream regulators among HDAC7 targets ( Figure 5—figure supplement 1A , B ) . In nearly all cases , HDAC7 regulates these targets in a manner opposite to their regulation during iNKT cell development ( Figure 6C , symbol border vs . fill colors ) . This regulation by HDAC7 by clearly suppresses downstream signaling in all cases except for TNF superfamily costimulatory signaling , which is mostly potentiated ( Figure 6C , color of arrows per legend ) . Consistent with our phenotypic findings , nearly half of these genes have positive roles in NK/NKT development/function ( Figure 6B ) , showing that HDAC7 broadly suppresses several key signaling pathways that are highly important in both NKT cells and in human autoimmune diseases that are similar to the pathology observed in HDAC7-ΔP transgenic mice . This remarkable concordance strongly supports the idea that the role of HDAC7 in these cells is involved in the pathogenesis of PSC and IBD , and identifies a few key signaling pathways as candidates for further interrogation . HDAC7 is a class IIA histone deacetylase that lacks intrinsic DNA binding capacity and requires binding to target transcription factors to modulate transcription at specific loci ( Yang and Seto , 2008 ) . Class IIA HDACs typically act as dominant corepressors , as in the case of MEF2 , which is converted from a transcriptional activator to a repressor upon class IIA HDAC binding ( McKinsey et al . , 2000 ) . PLZF belongs to the BTB-ZF family of transcription factors ( Beaulieu and Sant'Angelo , 2011 ) previously reported to interact with class IIA HDACs ( Verdin et al . , 2003; Chauchereau et al . , 2004 ) ; indeed , one group has even demonstrated in vitro and in vivo binding of HDAC7 to PLZF in a separate cell type ( Lemercier et al . , 2002 ) . This suggested that HDAC7 might modulate PLZF activity in thymocytes through direct physical binding . Determining if this is the case directly is somewhat challenging however , as the abundance of PLZF in wild-type thymocytes is very low , being restricted to a small population of iNKT precursors . To circumvent this difficulty , we made cell lysates from PLZF-transgenic thymocytes and immunoprecipitated them with antibodies to endogenous HDAC7 . These experiments showed a specific interaction between HDAC7 and PLZF in thymocytes ( Figure 7A ) . To further define this interaction , we co-transfected FLAG-tagged full-length or truncated HDAC7 with full length HA-tagged PLZF ( Figure 7B , D ) , or conversely different truncations of PLZF with the ( interacting ) HDAC7 N-terminal adapter domain ( residues 1–497 , Figure 7C , E ) . After Immunoprecipitation of transfected lysates with anti-FLAG agarose beads , we quantified the amount of PLZF protein pulled down vs . input levels over 3–6 separate experiments for each construct , using the LiCor Odyssey system ( Figure 7D–E ) . The results of this analysis identify residues 65–200 of HDAC7 , containing the MEF2-interacting domain through the first PKD phosphorylation site , as the interacting region ( Figure 7D ) . Analysis of the PLZF deletions identified a region from residues 320–450 , encompassing a proline-rich tract and the first two zinc finger domains , as critical for interaction ( Figure 7E ) . Although the precise mode of transcriptional regulation by PLZF remains unclear , with different domains exhibiting activating and repressive activity in varying contexts ( Sadler et al . , 2015; Puszyk et al . , 2013; Melnick et al . , 2002 ) , we next wanted to examine if HDAC7 physical binding to PLZF could modulate its transcriptional activity . We transfected 293 T cells with fusions of the GAL4 DNA binding domain ( residues 1–142 ) to full-length PLZF , an HDAC7-interacting mutant of PLZF ( 1-460 ) , or a non-interacting mutant ( 1-318 ) , together with a SV40 minimal promoter-Gal4 ( 5 ) -firefly luciferase reporter and an EF-1α promoter-driven Renilla luciferase reporter ( Figure 7F , see Figure 7—figure supplement 71 for a diagram ) . To these constructs were added empty vector ( Figure 6F , light blue bars ) , a vector encoding full-length HDAC7 ( Figure 7F , dark blue bars ) , or one encoding a fusion of the HDAC7 1–497 interacting domain with the VP16 transcriptional activation domain ( HDAC7-VP16 , Figure 7F , medium blue bars ) . Measurement of luciferase activity in lysates from these cells showed that co-transfection of FL HDAC7 with FL PLZF or the 1–460 truncation reduced transcription from the Gal4-luc construct , while it did not affect transcription when co-transfected with the non-interacting 1–318 mutant ( Figure 7F ) . Conversely , HDAC7-VP16 increased transcription from the interacting PLZF constructs but not the non-interacting one ( Figure 7F ) . These experiments , together with our characterization of the HDAC7-PLZF interaction and transcriptional targets above , provide strong evidence that in thymocytes HDAC7 regulates PLZF in the same manner as MEF2 and other transcription factors , functioning as a TCR signal-dependent co-repressor that helps to silence PLZF-associated promoters in the absence of appropriate signals . This mechanism is likely to account for at least part of the effect of HDAC7 on iNKT cells . We earlier reported that HDAC7-ΔP mice develop spontaneous tissue-specific autoimmunity , with about 80% developing obliterative exocrine pancreatitis and concomitant T-cell infiltration in stomach , liver and small intestine within eight months ( Kasler et al . , 2012 ) . Although this had been previously attributed solely to a defect in negative selection of conventional thymocytes , the striking absence of iNKT cells in HDAC7-ΔP mice spurred us to consider whether disrupted innate effector development might also contribute to this autoimmune syndrome . Indeed , the very tissues vulnerable to T-cell infiltration in HDAC7-ΔP mice , notably the small intestine , liver and hepatobiliary mucosa , are typically populated by PLZF-dependent innate effectors such as iNKT and mucosal-associated invariant T ( MAIT ) cells ( Fan and Rudensky , 2016 ) . We thus set out to determine if restoring iNKT cells could alter the course of HDAC7-ΔP–induced autoimmunity . In our earlier studies , we found that that HDAC7-ΔP-mediated autoimmunity is dominantly transferable in mixed BM chimeras if a 5-fold excess of HDAC7-ΔP-derived bone marrow is used . While engraftment at these ratios produced comparable populations of WT and HDAC7-ΔP Tconv in peripheral tissues , we did not assess the reconstitution of the iNKT compartment in those studies ( Kasler et al . , 2012 ) , leaving open the possibility that there was an uncharacterized recessive component to the autoimmunity . Attempts to adoptively transfer mature iNKT cells directly into HDAC7-ΔP mice failed to effectively restore tissue-resident iNKT populations ( Figure 8—figure supplement 81A–C ) . Instead , we generated two sets of hematopoietic chimeras to determine if restoring iNKT cells using Vα14 bone marrow could ameliorate disease compared to WT bone marrow ( Figure 8A ) . When irradiated recipients were reconstituted with a 1:5 mixture of Vα14: HDAC7-ΔP bone marrow , peripheral iNKT cells were effectively rescued to normal levels , while in recipients receiving a 1:5 WT: HDAC7-ΔP mixture they were still essentially absent ( Figure 8B ) . Comparing these cohorts over time , we noted Vα14: HDAC7-ΔP chimeras had significantly lower peak plasma levels of ALT and AST , commonly used as an indication of liver damage , than WT: HDAC7-ΔP chimeras ( Figure 8C ) . Both cohorts eventually perished from exocrine pancreatitis and had similar pancreatic lipase levels in plasma ( Figure 8—figure supplement 81D ) , yet Vα14: HDAC7-ΔP chimeras exhibited significantly improved body weight maintenance in the first two months post-engraftment ( Figure 8D , left ) and a reduced overall mortality rate ( Figure 8D , right ) compared to WT: HDAC7-ΔP chimeras . These results provide evidence that disruptions in innate effector development , particularly the loss of iNKT cells in the hepatobiliary tract , exacerbates tissue specific autoimmunity in the HDAC7-ΔP setting . Restoring this missing innate effector population resulted in enhanced survival and a significant reduction in the severity of disease .
The discovery and characterization of innate effector lymphocytes has transformed our understanding of T-cell receptor signaling , barrier protection at mucosal surfaces , and the evolutionary origins of the vertebrate immune system , yet the identification of key regulatory factors that control naïve versus innate effector development in thymocytes is far from complete . We demonstrate here that the epigenetic regulator HDAC7 serves as a gatekeeper of this developmental fate decision in the thymus . When HDAC7 is prevented from releasing its genomic targets in response to TCR stimulation , PLZF-dependent innate effector development appears to be blocked , and iNKT cells appear to become diverted to a naïve-like fate , characterized by lack of expression of memory or NK markers and a failure to produce effector cytokines . Conversely when HDAC7 function is lost , naïve development is reduced , more thymocytes develop as EOMES-expressing CD8 innate effectors , and the fraction of peripheral CD4 and CD8 T cells expressing memory markers and primed for cytokine production increases . Thus , appropriately regulated nuclear export of HDAC7 appears to be a licensing step that permits both negative selection and the acquisition of alternative cell fates , such as PLZF-dependent agonist selection to the iNKT lineage . In this study , we focused on iNKT cells due to their relatively high abundance and easy identification using CD1D tetramers , but we suspect that HDAC7-ΔP similarly abrogates development of other PLZF-dependent innate effector subtypes , including rare MR1-restricted MAIT cells and γδ NKT cells ( Chandra and Kronenberg , 2015; Fan and Rudensky , 2016 ) . In contrast , another well-described innate effector type , CD8αα + IELs localized in small intestine ( Mayans et al . , 2014 ) , are only slightly reduced in HDAC7-ΔP mice ( Figure 1—figure supplement 1A ) , consistent with their PLZF-independent derivation ( Cheroutre et al . , 2011 ) . The identification of a committed precursor to innate lymphoid cells that transiently expresses high amounts of PLZF ( Constantinides et al . , 2014 ) also raises the intriguing possibility that development of these cell types may be regulated by class IIA HDACs as well . Furthermore , the main mechanism of action we investigate here , HDAC7 antagonism of PLZF via direct interaction , may be generalizable to other members of the BTB-POZ-ZF family . For example , the signature transcription factor of Tfh cells , Bcl6 , is known to associate with HDAC4 ( Lemercier et al . , 2002 ) ( Crotty , 2014 ) . A class IIA HDAC/BTB-ZF axis may thus regulate T cell or ILC development at additional branch points . Additionally , in recent years a number of transcriptional regulators and epigenetic modifiers – including JARID2 , NKAP , HDAC3 , and EZH2 ( Pereira et al . , 2014; Thapa et al . , 2013; Dobenecker et al . , 2015 ) – have been identified that regulate iNKT ontogeny . At least one member , HDAC3 , physically associates with class IIA HDACs as part of a larger co-repressive complex ( Fischle et al . , 2002 ) . Devising systems to investigate these relationships as well as HDAC7 association with PLZF via ChiP-Seq and other genomic-scale approaches is a current priority in our laboratory . By restoring the missing iNKT population with the use of Vα14 donor bone marrow , we significantly attenuated the severity and time course of HDAC7-ΔP-mediated autoimmune liver disease , resulting in improved liver function , better body weight maintenance , and reduced overall mortality . Although specific rescue of iNKT cells did not provide protection in all tissues – almost all Vα14: HDAC7-ΔP chimeras eventually developed the same ultimately lethal exocrine pancreatitis as WT: HDAC7-ΔP chimeras – our studies nonetheless reveal an important new role for impaired iNKT development as an exacerbating factor in liver autoimmunity . Since both HDAC7 and PLZF influence the development of several non-iNKT innate effector subtypes that would not have been restored with Vα14 bone marrow , it is tempting to speculate that restoring these other subsets might ameliorate tissue destruction and T-cell infiltration due to HDAC7-ΔP in other organs . Innate effector T-cells are often considered frontline first-responders to infection that amplify and orchestrate the early immune response to invading pathogens . Thus , it was somewhat surprising to uncover a protective or anti-inflammatory role for iNKT cells in attenuating tissue destruction . Additional studies will be required to uncover the mechanisms through which iNKT cells provide protection , but for now we favor a model in which innate effectors occupy tissue niches at their sites of residence , limiting access of other immune cells into those sites . Alternatively , the loss of iNKT cells in these tissues may compromise normal mucosal barrier function in a manner that promotes inflammation and the subsequent recruitment of autoreactive Tconv . In HDAC7-ΔP mice , escape of autoreactive Tconv due to impaired negative selection may thus produce a potentially but not necessarily pathogenic population , which requires the additional loss of PLZF-dependent innate effectors from their target tissues to create an opening for infiltration . This ‘two-hit’ model may explain multiple types of tissue-specific autoimmunity , in which genetic lesions that generate excess self-reactive lymphocytes are coupled with separate or related defects in tissue-resident innate effector populations at specific sites , rendering these tissues particularly vulnerable to attack . Our findings likely hold considerable relevance to understanding the etiology and mechanisms contributing to some types of human autoimmunity . Indeed , common variant single nucleotide polymorphisms ( SNPs ) in the HDAC7 gene are significantly associated with human autoimmune and auto-inflammatory diseases , namely primary sclerosing cholangitis ( Liu et al . , 2013 ) and inflammatory bowel disease ( Jostins et al . , 2012 ) . Additional common variant SNPs in kinases known to export Class IIA HDACs via phosphorylation , including SIK2 and PRKD2 , are also associated with primary sclerosing cholangitis ( Liu et al . , 2013 ) , suggesting aberrant regulation of HDAC7 nuclear export as a causative mechanism . Moreover , the genes that we identified as regulated by HDAC7 in iNKT development show a striking overlap with other risk loci from these GWAS studies ( Figure 6A ) , suggesting that the broad HDAC7 regulatory network may be a crucial nexus that underlies susceptibility to several autoimmune diseases of considerable clinical importance . Indeed , mapping the overlapping GWAS loci to their associated signaling networks revealed a remarkable clustering around a few important signaling pathways in iNKT and effector development , including IL12 , IL21 , IL18 , IFNG , and IL4 , as well as Ig- and TNF-superfamily costimulatory pathways . Deciphering the complex relationship between HDAC7 , PLZF and other HDAC7 interaction partners , the observed modulation of these pathways , and the resulting cellular and pathologic phenotypes will be a major task for us going forward . We hope that this effort will illuminate the way forward in translating our finding that reestablishing missing iNKT cells can ameliorate HDAC7-mediated hepatic autoimmunity into potential therapeutic modalities for the analogous human diseases , based on the restoration of innate effector function .
The initial objective of this work was to investigate the molecular mechanisms behind the control of iNKT development by HDAC7 , which was an observation we made incidentally in our prior characterization of the general role of HDAC7 in thymic T cell development . The idea that HDAC7 might do this at least in part via interaction with PLZF arose from the review of older literature on these molecules showing they interact . The notion that the role of HDAC7 in iNKT cells has a bearing on the tissue distribution of autoimmunity due to altered HDAC7 function arose from the concordance between NKT-populated tissues and those showing disease in HDAC7-ΔP transgenic mice . This idea was highlighted in importance by the publication of GWAS studies after the initiation of our work that statistically associated HDAC7 and its regulatory network with human diseases affecting the same tissues . We investigated these questions using a combination of cell culture and transgenic mouse models in which the function of HDAC7 and/or PLZF was altered in thymocytes . Parameters measured include cellular abundance in different tissues , T cell effector function after ex-vivo stimulation , luciferase expression , protein-protein interactions , global transcript abundance , and various clinical measures associated with autoimmune disease , as detailed in the following sections . With the exception of our RNA-seq study , which was done in one experiment using three biological replicates for each condition , all results depicted in this work are based on at least two completely independent trials , comprising at least three biological replicates , that is data from three separate animals of each genotype or from three separate transfections of reporter/expression constructs . Larger sample sizes than this were used as feasible , based on the availability of experimental genotypes of interest , prospective estimates of the statistical power required to show significance for effects of the magnitudes initially observed , and the constraints of time and resources required for analysis . No data that were collected were excluded from the study unless there was clear evidence of a technical failure in data collection , or in the case of the animal studies , morbidity/mortality that was clearly unrelated to the pathologic conditions under study . Except where otherwise indicated in the figure legends , all control-experimental pairs were composed of sex-matched littermates , and all primary immune phenotypes were measured in animals between 4 and 8 weeks of age . All mice were housed in a specific pathogen-free barrier facility at the Gladstone Institutes , in compliance with NIH guidelines and a UCSF IACUC animal use protocol . All experimental strains were on a C57BL/6 ( B6 ) genetic background . B6 , BoyJ , Vα14/Jα18 transgenic ( Tg ( Cd4-TcraDN32D3 ) 1Aben ) and PLZF transgenic ( C57BL/6-Tg ( Cd4-Zbtb16 ) 1797Aben/J ) mice were obtained from The Jackson Laboratory , Bar Harbor , ME . Mice deleting Hdac7 ( Hdac7flox:-::lckcre ) or expressing the HDAC7-ΔP transgene under the control of the Lck proximal promoter were prepared as described elsewhere ( Kasler et al . , 2011 ) ( Kasler et al . , 2012 ) . Hematopoietic chimeras were prepared as follows: Recipients ( 8–10 wk-old BoyJ or BoyJ X B6 ) mice were irradiated with a split dose of 700 + 500 Rads , 3 hr . apart , from a 137Cs source ( J . L . Shepherd and Associates , San Fernando , CA ) . Mice were reconstituted with 5 × 106 bone marrow cells from WT ( Boyj or B6 X BoyJ heterozygote ) , HDAC7-ΔP TG ( CD45 . 2 ) , Hdac7-KO ( CD45 . 2 ) , or Vα14/Jα18 ( CD45 . 2 ) transgenic donors , injected retro-orbitally in 200 μl of PBS . Bone marrow cell suspensions were prepared by crushing tibias and femurs , dissociating marrow cells in PBS , and purifying mononuclear cells by Ficoll gradient centrifugation . Serum for AST/ALT/lipase analysis was collected by tail vein incision and analyzed by the UCSF Clinical Laboratory at SFGH . Cell suspensions were prepared from mouse thymus and spleen by crushing , dissociation of cells by pipetting , straining through 40 μm nylon mesh , and ACK lysis . Magnetic enrichment of iNKT cells from ~2×107 thymocytes was performed using APC-conjugated PBS-57 tetramers with the Easy-Sep ( StemCell Technologies , Cambridge , MA ) APC Positive Selection Kit , according to the package directions . Lymphocytes were prepared from liver by mincing of the tissue , straining through a 40 μm nylon mesh , and discontinuous Percoll gradient centrifugation . Intestinal intra-epithelial lymphocytes were prepared by extensive flushing of whole small intestines with cold PBS , excision of Peyer’s patches under magnification using a Trypan Blue-filled pipet as contrast medium , cutting into ~5 mm longitudinally opened segments , and incubation at 37◦C with rocking in PBS with 2 mM DTT and 5 mM EDTA for 30 min . IEL were then further purified from the dissociated epithelium by Percoll gradient centrifugation . For analysis of cytokine expression , cells were cultured for 4 hr post-isolation with 50 ng/ml PMA ( MilliporeSigma , St . Louis , MO ) plus 0 . 5 μM ionomycin ( MIlliporeSigma ) and for 1 hr with 0 . 5 μg/ml Brefeldin A ( MilliporeSigma ) prior to staining . Viability staining was performed for 15 min in the dark at room temperature using eFluor 520 or eFluor 780 fixable viability dyes ( Thermo Fisher Scientific , Waltham , MA ) at 1:1000 in PBS . Surface staining with CD1D tetramers and fluorochrome-conjugated antibodies was performed for 30 min on ice in PBS with 2% FCS , followed by either fixation in PBS/1% PFA or fixation/permeabilization with the eBioscience FOXP3 intracellular staining kit ( Thermo Fisher ) . Intracellular staining for cytokines or transcription factors was performed for 1 hr on ice in eBioscience FOXP3 Perm/wash buffer . Analytical flow cytometry was performed using a BD ( Becton Dickinson , Franklin Lakes , NJ ) LSRII Cytek ( Cytek Bioscinces , Fremont , CA ) FACS Calibur DxP8 instrument . Data processing for presentation was done using FlowJo 10 . 0 ( BD ) . Cell sorting was performed using a BD FACS-Aria II instrument . CD1D-αGalCer tetramers ( PBS-57 ) , conjugated with either phycoerythrin ( PE ) or allophycocyanin ( APC ) were obtained from the NIH tetramer core ( http://tetramer . yerkes . emory . edu/ ) . The following commercial antibodies were used for flow cytometry: CD11a-PE-Cy7 ( Thermo Fisher ) , clone M17/4; CD18-PE ( Thermo Fisher ) , clone M18/2; CD24-PE-Cy7 ( BD ) , clone M1/69; CD3-APC-EF780 ( Thermo Fisher ) , clone 2C11; CD4-BV650 ( BD ) , clone RM4-5; CD4-PE ( BD ) , clone GK1 . 5; CD4-APC ( BD ) , clone RM4-5; CD44-PE-Cy7 ( Thermo Fisher ) , clone IM7; CD44-APC-Cy7 ( BD ) , clone IM7; CD44-APC ( Thermo Fisher , clone IM7; CD45 . 1-Pacific Blue ( Thermo Fisher ) , clone A20; CD45 . 1-FITC ( Thermo Fisher ) , clone A20; CD45 . 2-V500 ( BD ) , clone 104; CD45 . 2-PE-Cy7 ( BioLegend , San Diego , CA ) , clone 104; CD5-APC ( BD ) , clone 53–7 . 3; CD62L-APC-Cy7 ( BD ) , clone MEL-14; CD69-PE ( Thermo Fisher ) , clone H1 2F3; CD8-Alexa 700 ( Tonbo Biosciences , San Diego , CA ) , clone 53–6 . 7; CD8-PerCP ( BioLegend ) , clone 53–6 . 7; CXCR3-PE ( Thermo Fisher ) , clone cxcr3-173; Eomes-PE ( Thermo Fisher ) , clone Dan11mag; Ly6C-APC ( Thermo Fisher ) , clone hk1 . 4; NK1 . 1-PE-Cy7 ( BD ) , clone pk136; NK1 . 1-APC-Cy7 ( BD ) , clone pk136; PLZF-PE ( Thermo Fisher ) , clone Mags . 21f7; T-bet-PE-Cy7 ( Thermo Fisher ) , clone ebio4b10; TCRβ-PerCP-5 . 5 ( BD ) , clone H57-597; TCRβ-APC-Cy7 ( BD ) , clone h57-597; TCRγδ-APC ( BD ) , clone GL3; Vg6 . 3/6 . 2-PE ( BD ) , clone 8f4h7b7 . Cell suspensions were prepared from thymus and spleen of 6–8 week old wild type B6 , Vα14/Jα18 transgenic , or Vα14/Jα18 X HDAC7-ΔP mice . iNKT cells were sorted by FACS using antibodies to TCRβ ( + ) and the PBS-57 CD1D-αGalCer tetramer ( + ) . Naïve Tconv were sorted using antibodies to CD4 ( + ) , CD8 ( - ) , TCRβ ( + ) , and CD44 ( - ) . Cells ( 250 , 000–2 , 000 , 000 ) were purified from three littermate triplets for each strain ( 18 samples total ) , and total RNA ( 200 ng to 4 μg ) was prepared using the Rneasy Plus Mini Kit ( Qiagen inc . , Venlo , The Netherlands ) . Double-stranded cDNA libraries were prepared by the Gladstone Institutes Genomics Core using the Nugen Ovation kit ( Nugen , San Carlos , CA ) . The Libraries were sequenced by the UCSF Center for Advanced Technology using the Illumina HiSeq 4000 instrument ( Illumina , San Diego , CA ) . Six barcoded samples were loaded per lane . FASTQ files ( approximately 5 . 5 × 107 reads each ) were mapped to the UCSC Mouse genome Build 37 ( Mm . 9 ) using Bowtie2 ( Johns Hopkins University ) . Approximately 4 × 107 ( ~75% ) of reads per sample were mapped uniquely to the mouse genome . Gene-level tabulation , quality control , and expression analysis was done on . SAM format files generated by BOWTIE2 using SeqMonk 0 . 33 ( http://www . bioinformatics . babraham . ac . uk/projects ) . Ontologic analysis and pathway mapping were performed using Ingenuity Pathway Analysis ( http://www . ingenuity . com/ ) . All primary data associated with these experiments have been deposited at GEO ( https://www . ncbi . nlm . nih . gov/geo , accession GSE105026 ) , and a summary of all gene expression data and statistics for differentially expressed genes is provided in Supplementary file 1 . The human PLZF coding sequence ( RCAS ( B ) -Flag-PLZF ) , deposited by Peter Vogt ( Shi and Vogt , 2009 ) was obtained from Addgene . N-terminally HA-tagged full-length PLZF was amplified from this coding sequence using the following Primers: N-terminal PLZF Bam HI , HA tag , Eco RV , Bsa BI , Hpa I: 5’ aaaaaaggatccacc atg tat ccc tac gat gtt cca gat tat gcg ata tca atc gtt aac atg gat ctg aca aaa atg gg; C-terminal Swa 1 , stop , Not 1: 5’ cct cta cct gtg cta tgt gtg att taa atgattagataagcggccgcaaaaaa 3’ . This amplification product was subcloned into pCDNA3 . 1 ( + ) ( Thermo Fisher ) using Bam H1 and Not one sites . Different PLZF truncations were amplified from this construct and sub-cloned into the introduced flanking sites ( further details on request ) . For the GAL4 DNA-binding domain-PLZF fusion constructs , the GAL4 DNA-binding sequence was amplified from a plasmid template and inserted into the Eco RV sites of full-length or truncated PLZF expression constructs described above . Construction of full-length human HDAC7 and HDAC7-VP16 fusion-encoding expression vectors is described elsewhere ( Dequiedt et al . , 2003 ) . Other truncated , FLAG-tagged HDAC7 constructs were amplified from these templates and re-ligated into pCDNA3 . 1 ( + ) . The GAL4 UAS ( 5 ) SV40-Firefly luciferase reporter construct was prepared by ligation of an oligonucleotide cassette containing 5 GAL4 recognition sites into the Sma 1 site of pGL2 Promoter ( Promega Corp . , Fitchburg , WI ) . For pulldown experiments , 10 cm dishes seeded the previous day with 3 . 2 × 106 HEK 293 T cells were transfected with 20 μg of total DNA , consisting of 10 μg each of PLZF and HDAC7 constructs or the corresponding empty vectors , using CaPO4/chloroquine . HEK-293T cells , were originally obtained from ATCC , and had been confirmed by PCR testing to be mycoplasma-free within 6 months of their use for these experiments . After 48 hr , cells were harvested for interaction analysis . For reporter assays , 6-well dishes seeded with 0 . 8 × 106 HEK 293 T cells/well were transfected using CaPO4/chloroquine with 6 . 1 μg of total DNA , consisting of 2 μg each of gal4 ( 5 ) luc , gal4-PLZF fusion construct , and empty vector or HDAC7 expression construct , plus 100 ng of EF1α Renilla luciferase . Cells were harvested for luciferase assay 48 hr after transfection , and luciferase activity was measured using the Promega Dual-Luciferase assay kit . For the co-immunoprecipitation of endogenous HDAC7 with transgenic PLZF in thymocytes , thymocyte lysates from wild-type and PLZF transgenic mice were prepared using p300 lysis buffer ( 250 mM NaCl , 0 . 1% NP-40 , 20 mM NaH2PO4 , pH 7 . 5 , 5 mM EDTA , 30 mM sodium pyrophosphate , 10 mM NaF , and HALT protease/phosphatase inhibitors ( Thermo Fisher ) . After clarification ( 5 min , 13 , 000Xg ) and pre-clearing ( 3 hr at 4°C with proteinA/G agarose beads ) , lysates were immunoprecipitated with either 1 µg/ml of α-HDAC7 antibody ( H-273 , Santa Cruz Biotechnology , Dallas , TX ) or 1 µg/ml of rabbit IgG isotype control antibody ( Cell Signaling Technologies , Danvers , MA ) at 4°C overnight . The lysates were then incubated with 50 µl of protein A/G agarose beads ( Santa Cruz Biotechnology ) at 4°C for 4 hr , followed by washing five times with p300 lysis buffer . Immunoprecipitated proteins from the beads were eluted with non-reducing Laemmli SDS PAGE sample buffer by boiling for 3 min . For pulldown analysis of HDAC7-PLZF truncation mutants , 10 cm dished were harvested and lysed in 0 . 8 mL of P300 buffer , clarified by spinning 5 min . at 13 , 000 g , then incubated for 4 hr at 4°C with 30 μL/sample of FLAG M2-agarose beads ( MilliporeSigma ) . After four washes with p300 buffer , bound proteins were eluted from the beads by addition of 100 μL of reducing Laemmli SDS-PAGE sample buffer , followed by a 5 min incubation at 95°C . After SDS PAGE and transfer to nitrocellulose , membranes were probed with antibodies against HDAC7 ( H-273 , Santa Cruz Biotechnology ) , PLZF ( D9 , Santa Cruz Biotechnology ) , and β-actin ( Abcam , Cambridge , MA ) , HA epitope ( Cell Signaling ) , or FLAG epitope ( MilliporeSigma ) , overnight at 4°C . After washing and incubation with HRP- or IRDye-conjugated antibodies ( Li-Cor Biotechnology , Lincoln , NE ) , signal was detected using chemiluminescence and film or a Li-Cor Odyssey scanner respectively . Bands for quantitative pulldown analysis were quantified from the scanner output using ImageJ ( Wayne Rasband , National Insititutes of Health ) | To protect us , our immune system must walk a narrow line: while it eliminates all external threats , it also has to refrain from attacking the healthy tissues of our body . When such misdirected attacks do take place , they can result in life-threatening autoimmune diseases . T cells are a highly diverse population of immune cells that can recognize and orchestrate the body’s response against infected or ‘abnormal’ cells . Early in the development of most types of T cells , the body normally weeds out the ones that target healthy tissues . A gene known as Histone Deacetylase 7 ( HDAC7 ) regulates this process . However , when HDAC7 carries a specific mutation called HDAC7-ΔP , dangerous T cells that can attack healthy tissues ‘escape’ this selection . The HDAC7-ΔP mutation allows T cells that react to many different tissues to survive . However , in mice with this genetic change , only the liver , the digestive system and the pancreas are actually damaged by the immune system and show signs of autoimmune diseases . Why are these organs affected , and not the others ? Here , Kasler , Lee et al . find that HDAC7 also helps another type of T cell to develop . Known as invariant natural killer T – or iNKT – cells , these cells specialize in defending the gut , liver and pancreas against bacteria . Mice with the HDAC7-ΔP mutation can no longer produce iNKT cells . Remarkably , restoring normal levels of these cells in the HDAC7-ΔP animals reduces the symptoms of their autoimmune diseases , even though the mice are still carrying the T cells that have escaped selection and can attack healthy tissues . Taken together , these results explain why a mutation in HDAC7 can create problems only for specific organs in the body . However , it is still not clear exactly why losing iNKT cells increases autoimmune attacks of the tissues they normally occupy . One possibility is that these cells limit access to the organs by other immune cells that could cause damage . Another option is that , when iNKT cells are absent , gut bacteria can attack and create an inflammation . This recruits T cells to the site , including the ones that can attack healthy organs . In humans , mutations in HDAC7 , as well as in other genes that regulate it , are also associated with autoimmune disorders of the digestive tract and liver . These include inflammatory bowel diseases such as ulcerative colitis or Crohn’s disease . Ultimately the findings presented by Kasler , Lee et al . could be a starting point for finding new treatments for these illnesses . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"immunology",
"and",
"inflammation"
] | 2018 | Histone Deacetylase 7 mediates tissue-specific autoimmunity via control of innate effector function in invariant Natural Killer T Cells |
Ribonucleotide reductase ( RNR ) converts ribonucleotides to deoxyribonucleotides , a reaction that is essential for DNA biosynthesis and repair . This enzyme is responsible for reducing all four ribonucleotide substrates , with specificity regulated by the binding of an effector to a distal allosteric site . In all characterized RNRs , the binding of effector dATP alters the active site to select for pyrimidines over purines , whereas effectors dGTP and TTP select for substrates ADP and GDP , respectively . Here , we have determined structures of Escherichia coli class Ia RNR with all four substrate/specificity effector-pairs bound ( CDP/dATP , UDP/dATP , ADP/dGTP , GDP/TTP ) that reveal the conformational rearrangements responsible for this remarkable allostery . These structures delineate how RNR ‘reads’ the base of each effector and communicates substrate preference to the active site by forming differential hydrogen bonds , thereby maintaining the proper balance of deoxynucleotides in the cell .
Deoxyribonucleotides , the building blocks for DNA biosynthesis , are produced in the cell from ribonucleotide precursors by ribonucleotide reductase ( RNR ) ( Figure 1 ) . Three classes of RNRs are known , categorized by the cofactor they use to generate a protein radical required for catalysis . The best characterized of all RNRs is the class Ia enzyme from Escherichia coli that employs a di-iron-tyrosyl-radical cofactor to initiate chemistry and requires two dimeric protein subunits for enzymatic activity . The α2 subunit contains two ( β/α ) 10 barrels , which house the active sites at the barrel centers ( Eriksson et al . , 1997; Uhlin and Eklund , 1994 ) , and the β2 subunit utilizes a largely helical secondary structure to house the radical cofactor ( Sjöberg and Reichard , 1977 ) ( Figure 1B–C ) . As a central controller of nucleotide metabolism , RNR uses multiple allosteric mechanisms to maintain the balanced deoxyribonucleoside triphosphate ( dNTP ) pools that are required for accurate DNA replication . First , allosteric activity regulation modulates the overall size of dNTP pools . ATP or dATP binding at an allosteric activity site , found at the N-terminus of α2 ( Figure 1D ) , leads to up-regulation or down-regulation of enzyme activity , respectively ( Brown and Reichard , 1969 ) . In E . coli class Ia RNR , this regulation is achieved by changes in the oligomeric arrangement of the α2 and β2 subunits ( Brown and Reichard , 1969; Rofougaran et al . , 2008; Ando et al . , 2011 ) . When ATP is bound at the activity site , an α2β2 complex is favored . Although no X-ray structure of the active complex has been determined , low resolution models have been generated using small-angle X-ray scattering ( Ando et al . , 2011 ) , electron microscopy ( Minnihan et al . , 2013 ) , and distance measurements made through spectroscopic analyses ( Seyedsayamdost et al . , 2007 ) ( Figure 1D ) . This active α2β2 complex is capable of a long-range proton coupled electron transfer from β2 to α2 , forming a transient thiyl radical on Cys439 to initiate catalysis ( Licht et al . , 1996 ) . Alternatively , when concentrations of dATP become too high in the cell , dATP binds at the allosteric activity site and formation of an α4β4 complex is promoted . The structure of this complex was recently solved ( Ando et al . , 2011 ) , revealing a ring of alternating α2 and β2 units that cannot form a productive electron transfer path , thus inhibiting the enzyme ( Figure 1D ) . 10 . 7554/eLife . 07141 . 003Figure 1 . Escherichia coli class Ia RNR regulation is achieved through allostery . ( A ) E . coli RNR catalyzes reduction of nucleoside diphosphates using a radical , formed on an active site cysteine , to initiate catalysis . ( B ) A ribbon representation of the catalytic subunit ( α2 , a 172-kDa homodimer ) is shown with one α chain colored blue and the other cyan ( this work ) . Nucleotides are shown as spheres with NDP substrate in yellow and dNTP specificity effector in purple . Loop 1 and loop 2 , which are involved in specificity effector binding and recognition , are colored in red . Cys439 , where the active site thiyl radical is formed , is shown in orange spheres . ( C ) A ribbon representation of the radical generating subunit ( β2 , an 87-kDa homodimer ) is shown with one β chain colored orange and the other tan ( this work ) . The di-iron cofactor that generates the initial tyrosyl radical required for RNR activity is shown in green spheres . ( D ) Allosteric activity regulation is achieved by interconversion between an active α2β2 complex in the presence of the allosteric activity effector ATP and an inactive α4β4 species when dATP binds to the allosteric activity site ( PDB ID: 3UUS ) . The model for the α2β2 complex was created using small-angle X-ray scattering data ( Ando et al . , 2011 ) to fit the previously solved structure of α2 ( PDB ID: 3R1R ) and β2 ( PDB ID: 1RIB ) together . The α2 subunit is shown in grey surface representation , at a 90° angle from the representation shown in ( B ) and the β2 subunit is shown in orange surface representation . Allosteric activity sites are shown with ATP modeled in cyan and dATP in red spheres . ( E ) Allosteric specificity regulation is governed by the binding of deoxynucleoside triphosphates to RNR , influencing the preference for one substrate over another ( see Table 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07141 . 00310 . 7554/eLife . 07141 . 004Table 1 . Previously determined binding affinities for substrates in the absence and presence of specificity effectors or analogs . DOI: http://dx . doi . org/10 . 7554/eLife . 07141 . 004SubstrateKd with nospecificity effectorKd with effector or effector analogCDP1 mM ( von Döbeln and Reichard , 1976 ) , 0 . 3 mM ( Crona et al . , 2010 ) dAMP-PNP: 88 μM ( von Döbeln and Reichard , 1976 ) UDPn . d . KM with ATP: 220 μM ( Larsson and Reichard , 1966 ) ADP420 μM ( von Döbeln and Reichard , 1976 ) dGTPγS: 70 μM ( von Döbeln and Reichard , 1976 ) GDP110 μM , 80 μM ( von Döbeln and Reichard , 1976; Crona et al . , 2010 ) TTP: 22 μM ( von Döbeln and Reichard , 1976; Crona et al . , 2010 ) The second form of allosteric regulation is specificity regulation , which maintains the proper relative ratios of dNTPs in the cell . Briefly , the binding of ( d ) NTP effectors to an allosteric specificity site in α2 influences the preference of RNR for its four nucleoside diphosphate ( NDP ) substrates . Whereas high levels of dATP inhibit class Ia RNR , at lower levels , dATP promotes CDP or UDP reduction . Likewise , TTP promotes GDP reduction , and dGTP promotes ADP reduction ( Figure 1E ) ( Brown and Reichard , 1969; Rofougaran et al . , 2008; von Döbeln and Reichard , 1976 ) . Importantly , the affinity of the α2 and β2 subunits for each other is weak ( ~0 . 4 μM ) in the absence of effectors , whereas the binding of a complementary substrate/specificity effector pair increases the affinity of the class Ia RNR subunits fivefold ( Crona et al . , 2010; Hassan et al . , 2008 ) . Previous structural work , which includes: X-ray structures of GDP and TTP bound to E . coli α2 ( Eriksson et al . , 1997 ) , structures of all four substrate/effector pairs bound to class Ia α2 from Saccharomyces cerevisiae ( Xu et al . , 2006 ) , and class II α2 from Thermotoga maritima ( Larsson et al . , 2004 ) , revealed the location of the allosteric specificity sites at the ends of a four helix bundle at the dimer interface ( Figure 1B ) . These data , and accompanying in vitro and in vivo studies on S . cerevisiae ( Ahmad et al . , 2012; Kumar et al . , 2010; Kumar et al . , 2011 ) , also implicated which residues ( Gln294 and Arg298 , E . coli numbering ) and which regions of the structure are involved in the communication between the specificity site and the active site . A flexible loop , termed loop 2 ( residues 292–301 in E . coli ) ( Figure 1B ) , bridges the two sites , which are approximately 15 Å apart , and becomes more ordered upon effector and substrate binding ( Eriksson et al . , 1997 ) . An additional loop , termed loop 1 ( residues 259–278 ) , is also near the effector-binding site and is stabilized upon effector binding ( Eriksson et al . , 1997 ) . Thus , previous work established the players involved in allosteric specificity regulation in RNR . This work supports those assignments , and goes on to provide a novel set of crystallographic snapshots that reveal how these residues in the prototypic RNR from E . coli are able to communicate and thereby regulate substrate preference .
Each substrate/effector bound α4β4 structure reveals a closing of the ( β/α ) 10 barrel around the bound substrate that is not observed in the absence of substrate . An initial superposition of one α chain from the previously reported free α2 soaked with AMP-PNP ( PDB ID: 3R1R ( Eriksson et al . , 1997 ) ) onto the α4β4 crystal form was performed using only residues 432–446 , that comprise the conserved active site loop ( ‘finger loop’ ) found in the center of the ( β/α ) 10 barrel . In this structural alignment , one half of the active site barrel overlays exactly , whereas the second half of the barrel and the N-terminal portion of the structure has undergone a clamping movement ( Figure 3 ) . To better characterize this movement , a difference distance matrix ( DDM ) was calculated . This plot reveals relative movements in a reference-independent manner . Briefly , the distance between every possible pair of atoms in one chain of α from the α4β4 structure is subtracted from the same distance measured in the free α2 structure , giving a representation of how much the distances between pairs of atoms change between the two structures . The pattern of movements seen in the plot indicates that the N-terminal 220 residues ( region 1 of Figure 3B , C ) shift substantially ( >3Å ) , bringing them closer to half of the active site barrel ( residues ~224–439 , region 2 of Figure 3B , C ) . We also observe that loop 2 ( region 3 of Figure 3B , C ) and a flexible β hairpin that sits adjacent to the N-terminus ( residues 646–651 , region 5 of Figure 3B , C ) undergo substantial motions . 10 . 7554/eLife . 07141 . 009Figure 3 . Cα difference distance matrix plot reveals movements that occur concurrent with substrate binding for E . coli class Ia RNR . ( A ) Superposition of α from our CDP/dATP structure ( N-terminus and one half of the active site barrel in purple and the other half barrel in blue ) and α from a substrate-free E . coli RNR structure ( PDB ID: 3R1R , ( Eriksson et al . , 1997 ) ) ( N-terminus and one half of active site barrel in pink and the other half barrel in red ) . The two chains were aligned by the active site finger loop , which is colored green . CDP is shown as sticks . ( B ) Distances in chain A of the CDP/dATP structure ( this work ) were subtracted from chain A of the substrate-free α2 structure ( PDB ID: 3R1R ) for residues 4–737 . Scale is shown on the top and is ±3 . 0 Å ( positive values in blue indicate a shorter distance in the CDP/dATP structure and negative values in red indicate a longer distance ) . Regions that move in a concerted fashion are indicated with colored lines and residue ranges are listed to the left of the plot . ( C ) One α chain is shown in ribbons with residue ranges from ( B ) colored . Region 1 ( blue ) , the N-terminal 225 residues , contracts towards region 2 ( yellow ) . Region 3 ( red ) includes loop 2 residues and moves towards the active site ( in region 4 ) . A flexible loop , region 5 ( green ) , undergoes a large motion towards regions 2 and 4 whereas region 4 undergoes little movement with respect to the rest of the structure . DOI: http://dx . doi . org/10 . 7554/eLife . 07141 . 009 Closer inspection of the active site in the superposition ( again , aligned by the active site finger loop ) reveals that the movements described above directly affect substrate binding ( Figure 4A ) . Residues Ser622 , Ser625 , and Thr209 form hydrogen bonds with the NDP substrate , whereas in the substrate-free state ( pink ribbon in Figure 4A ) , these residues are shifted away from the center of the barrel . Thr209 is part of the N-terminal region that is shown to move in the DDM analysis . The structure superposition and DDM analysis also indicate that loop 2 shifts , bringing Arg298 closer to the active site . The Arg298 side chain now forms a cation-π interaction with the substrate base and a charge-charge interaction with the substrate β-phosphate . In the substrate-free structure , this side chain is pointing in the opposite direction . Protein contacts like Arg298 to the substrate diphosphate are expected to be particularly important for charge neutralization in enzymes , such as E . coli RNR , that do not employ metal cations such as Mg2+ for this purpose . The positions of residues that contact the ribose O3' ( Glu441 and Asn437 ) are virtually unchanged between the NDP-free and NDP-bound structures ( Figure 4A ) . All four of our substrate-bound structures show almost identical contacts between the enzyme and the substrate ribose and phosphates as described here and shown in Figure 4A , C and Figure 5 for GDP and CDP , respectively . 10 . 7554/eLife . 07141 . 010Figure 4 . Local movements stabilize substrate binding to the active site of α2 . ( A ) GDP/TTP-bound structure ( this work , cyan ) is overlaid with substrate free α2 ( PDB ID: 3R1R , pink ) . Distances are given in Å . GDP and residues that form hydrogen bonds to GDP are shown as sticks with GDP carbons in yellow , protein carbons in cyan or pink . Ser622 , Ser625 and Thr209 move to form hydrogen bonds to the phosphates of GDP . Arg298 of loop 2 reaches over the guanine base to contact the phosphates . ( B ) Previously reported α2 structure with GDP bound at an occupancy of 0 . 5 ( PDB ID: 4R1R ) . GDP and residues that form hydrogen bonds to GDP are shown in sticks with GDP carbons in yellow and protein carbons in green . ( C ) Van der Waals packing around GDP in GDP/TTP structure ( this work ) showing a tightly packed active site . ( D ) Van der Waals packing around GDP from previously reported α2 structure ( PDB ID: 4R1R ) showing that the active site is still open . DOI: http://dx . doi . org/10 . 7554/eLife . 07141 . 01010 . 7554/eLife . 07141 . 011Figure 5 . Details of CDP binding to a clamped-down active site in E . coli class Ia RNR . ( A ) Wall-eyed stereo view of CDP ribose and phosphate interactions with protein . CDP is shown as sticks with carbons in yellow . Protein side chain carbons are colored light purple and loop 2 residue , Arg298 , carbons are colored tan . A putative water molecule is shown as a red sphere . Hydrogen-bonding interactions , shown with black dashed lines , include: O3' of ribose to Glu441 , the proposed general base ( van der Donk et al . , 1996; Persson et al . , 1997 ) , and O2' to backbone carbonyl of Ser224 . Cys225 is the proposed proton donor for the 2'-OH that is lost as H2O . Cys225 is 3 . 4–3 . 6 Å from the O2' . The distance between the sulfur atom of Cys439 , where the thiyl radical is formed , and C3' of the ribose , where a hydrogen atom is abstracted to initiate catalysis , is 3 . 5–3 . 7 Å . ( B ) Wall-eyed stereo view of omit electron density for CDP structure shown in A . Orientation is tilted and rotated slightly to show the water density . A water molecule is present in this position in some substrate-bound RNR structures and not others , the significance of which is not clear . Arg298 is not shown for simplicity . DOI: http://dx . doi . org/10 . 7554/eLife . 07141 . 011 In contrast to the compact α2 barrel observed in our substrate/effector-bound α4β4 structures , no clamping motion of the barrel is observed between substrate-free α2 and a previously reported structure of α2 with GDP substrate and TTP effector bound ( Eriksson et al . , 1997 ) . In this structure , Ser622 , Ser625 , and Thr209 are all 3 . 0 Å or greater from the substrate diphosphate , and Arg298 points away from the active site ( Figure 4B ) . These weaker ( or missing ) interactions between protein and substrate in this previously determined structure are consistent with the fact that substrate is observed in an altered orientation in the active site ( Figure 4C , D ) and is only present at half occupancy . We attribute the differences in substrate positioning in this structure to the inability of the active site barrel to clamp around substrate due to crystal lattice contacts . Symmetry-related molecules are closer to the active site in the α2-only crystalline state than they are in the α4β4 crystal form . Importantly , differences in flexibility cannot be attributed to the fact that α4β4 is an inactive state of E . coli RNR because α2 is also an inactive state; the former because β2 is held at arm’s length from α2 , and the latter because β2 is absent altogether . Additionally , other than contributing to an amenable crystal lattice , there is nothing to suggest that β2 must be present for α2 to clamp . When lattice contacts do not restrain α2 movement , α2 should be able to clamp in the presence of a cognate substrate/effector pair regardless of β2 . Thus , it seems that different crystal forms have given us two distinct snapshots of substrate-bound states for the E . coli class Ia enzyme: a high affinity state which is closed and ready for radical-based chemistry when β2 becomes available ( Figure 4C ) and a lower affinity state in which substrate is still exposed to solvent and not ready to undergo catalysis , regardless of the availability of β2 ( Figure 4D ) . Regardless of the identity of the dNTP base , the ribose and phosphates of the three specificity effectors ( dATP , dGTP , and TTP ) form almost identical contacts with RNR ( Figure 6 ) . Despite the presence of the specificity effector site at the dimer interface , these common interactions are made with only one chain of the α2 dimer . Asp232 and His275 are within hydrogen-bonding distance of O3' of the ribose . One Mg2+ ion appears to form an octahedral coordination complex with three phosphate oxygens of the specificity effector and three water molecules ( Figure 6 ) . There are also hydrogen bonds between the phosphates and the backbone amides of residues Arg269 and Leu234 and charge-charge interactions with the side chains of Arg269 and/or Arg262 . 10 . 7554/eLife . 07141 . 012Figure 6 . Interactions that anchor specificity effectors in E . coli class Ia RNR involve residues outside of loop 2 . Interactions are shown for ( A ) CDP/dATP , ( B ) UDP/dATP , ( C ) ADP/dGTP , and ( D ) GDP/TTP . The dNTP effector carbons are colored cyan and the protein carbons are colored grey . Magnesium ions are colored grey and waters in red . The side chain of Leu234 is omitted for clarity . Hydrogen-bonding interactions are shown with black dashed lines . Figures are displayed in wall-eyed stereo . DOI: http://dx . doi . org/10 . 7554/eLife . 07141 . 012 Loop 2 is ordered in all four structures as indicated by 2Fo–Fc composite omit electron density ( Figure 2 ) . As noted above , Arg298 of loop 2 forms a conserved interaction with the β-phosphate of all four substrates ( Figures 7 , 8 ) . On the effector side of loop 2 , Cys292 packs against the base of all three effectors ( Figures 7 , 8 ) , consistent with binding studies that show the affinity for all specificity effectors is decreased when Cys292 is mutated to alanine ( Ormö and Sjöberg , 1996 ) . The remaining interactions with loop 2 are distinct for each substrate/effector pair ( Figures 7 , 8 ) . dATP , which preferentially increases RNR activity for CDP and UDP substrates , makes two hydrogen bonds with loop 2 . The backbone amide and carbonyl of Ser293 are in position to hydrogen bond to N1 and N6 , respectively , of the adenine base of dATP . These hydrogen bonds hold Ser293 in place and position the adjacent Gln294 in to a conformation that points into the substrate-binding site ( Figure 8A , B ) . In this conformation , Gln294 can form a hydrogen bond with O2 of CDP or UDP . There is no difference in substrate positioning between CDP and UDP , as the unique positions of the pyrimidine ring ( N2 and N4/O4 ) are not involved in any contacts with the protein . Furthermore , the positioning of Gln294 by dATP selects for pyrimidine binding , as binding of a purine base would be disfavored due to steric occlusion by Gln294 . 10 . 7554/eLife . 07141 . 013Figure 7 . Conformations of E . coli class Ia RNR loop 2 in the presence and absence of substrate-effector pairs . Structures are shown in wall-eyed stereo as sticks for ( A ) CDP/dATP , ( B ) UDP/dATP , ( C ) ADP/dGTP , and ( D ) GDP/TTP . ( E ) α2 with no substrates or effectors bound ( PDB ID: 3R1R ) . Substrates are shown with carbons in yellow , effectors are shown with carbons in cyan , and loop 2 is shown with carbons in grey . Other atoms colored as in previous figures . Hydrogen-bonding interactions are shown in black dashed lines with distances given in Å . Only interactions between protein residues are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 07141 . 01310 . 7554/eLife . 07141 . 014Figure 8 . Molecular basis for communication between substrate and effector binding sites in E . coli class Ia RNR . Structures are shown as sticks on the left with 2D representation of hydrogen-bonding interactions on the right for ( A ) CDP/dATP , ( B ) UDP/dATP , ( C ) ADP/dGTP , and ( D ) GDP/TTP . Atoms are colored as in Figure 7 . Hydrogen-bonding interactions are shown in black dashed lines with distances given in Å . In the 2D representation , each residue is colored a different color , residues 295–297 are shown as a black line , and hydrogen bonds are indicated with black dashed lines . DOI: http://dx . doi . org/10 . 7554/eLife . 07141 . 014 dGTP , which preferentially increases RNR activity for substrate ADP , forms one hydrogen bond with loop 1 residue Thr276 ( Figure 6C ) in addition to its loop 2 contacts . It is the only effector for which a hydrogen bond is formed between the base and loop 1 . Because the position of hydrogen bond donors and acceptors at N1 and O6 in dGTP is opposite from that of N1 and N6 of dATP , the interaction with the backbone amide of Ser293 in loop 2 is unfavorable in this structure . Instead , Ser293 is shifted away from the base , and N1 and N2 – both hydrogen bond donors – form an interaction with the backbone carbonyl of Gln294 , which has shifted away from the substrate ( Figure 7A–C and 8A–C ) . Ser293 thus makes no hydrogen bond contacts to dGTP and instead provides hydrogen bonds to Gly300 on the other side of loop 2 through both its side chain and backbone . Stabilization of Gly300 in this position in turn allows for a hydrogen bond between the carbonyl of Gly299 with the N6 position of the preferred substrate , ADP . The movement of Gln294 away from the substrate-binding site allows room for the larger ADP to bind . Thus , Ser293 and Gln294 are responsible both for differentiating between the purine bases adenine and guanine at the effector-binding site and for differentiating pyrimidines from purines at the substrate site , either directly ( in the case of Gln294 hydrogen bonding to CDP or UDP ) or indirectly through stabilization of Gly299 and Gly300 . With dATP bound , Ser293 plays a role in effector recognition and Gln294 in substrate recognition; these roles are reversed for ADP/dGTP . TTP , which preferentially increases RNR activity for substrate GDP , binds to the specificity site such that only the carbonyl of Cys292 from loop 2 is positioned to form a hydrogen-bonding interaction with the thymine base ( Figure 8D ) . As in the ADP/dGTP structure , Gln294 is swung away from the active site to make room for the binding of a purine substrate ( Figure 8D ) . On the substrate side of loop 2 , N1 and N2 of GDP are both within hydrogen-bonding distance of the backbone carbonyl of Gly299 . N2 of GDP is also within hydrogen-bonding distance of the backbone carbonyl of Ala252 ( not shown ) , giving GDP one more hydrogen bond than the other substrates For the purine substrates ADP and GDP , preference appears to rely on a shift of the carbonyl of Gly299 in loop 2 forward or backward in the active site , enabled by the binding of dGTP or TTP , respectively , at the specificity site . With dGTP bound , Ser293 is positioned such that it pushes the carbonyl of Gly299 'forward' ( away from the effector-side of loop 2 , see Figure 8C ) , allowing a hydrogen bond to form between the carbonyl of Gly299 and N6 of adenine . With the smaller TTP in the effector site , Ser293 rearranges such that the carbonyl of Gly299 can relax ‘backward’ ( toward the effector-side of loop 2 , see Figure 8D ) , allowing for a hydrogen bond between the carbonyl of Gly299 and N1 and N2 of guanine . Specificity is imparted by the fact that the ‘forward’ position of the Gly299 carbonyl interacts with the 6-position of the purine , which is a hydrogen bond donor ( NH2 ) in adenine and a hydrogen bond acceptor ( O ) in guanine; and the ‘backward’ position of the Gly299 carbonyl interacts with the 1-position and 2-position of the purine , which are both hydrogen bond donors ( NH and NH2 ) in guanine but not in adenine ( Figure 8C , D ) . To briefly summarize purine specificity , purines are favored over pyrimidines when Gln294 is positioned away from the active site creating a larger substrate-binding pocket , and ADP is favored over GDP when dGTP binding stabilizes Ser293 close to the Gly299 backbone pushing its carbonyl forward , and GDP is favored over ADP when TTP allows Ser293 to fall back . Our sets of structures implicate Arg298 and Gln294 as key residues in the specificity regulation of E . coli RNR . To confirm this prediction , we prepared Gln294Ala and Arg298Ala mutant enzymes and tested the activity for each substrate/effector pair . Because dATP at high concentrations inhibits RNR activity by binding to the allosteric activity site , whereas at lower concentrations it is a specificity effector , we first carried out control experiments to determine activity levels for CDP reduction in the absence of dATP and at an inhibitory dATP concentration ( 175 μM ) ( Figure 9 ) . Results show that wild-type RNR is as active ( CDP/dATP ) or more active ( UDP/dATP ) at 1 μM dATP than it is in the positive control ( CDP/ATP ) , indicating that at 1 μM , dATP is acting as a specificity effector and not as an allosteric inhibitor of activity , consistent with previous studies ( Birgander et al . , 2004 ) . The negative control shows the substantial decrease in wild-type RNR activity on CDP when the concentration of dATP is high enough ( 175 μM ) for dATP to bind to the allosteric activity site and inhibit the enzyme . With these controls in place , the activity of mutant RNRs with all four substrate/effector pairs can be established . Our structures predict that mutation of Arg298 to alanine should decrease activity for all substrates , and that is exactly what we observe ( Figure 9 ) . The effect is dramatic for this mutant , with negligible activity observed in all cases . In fact , the small amount of activity detected could be due to low levels of contamination of wild-type E . coli RNR instead of the Arg298Ala mutant protein . Thus , Arg298Ala-RNR is either mostly or completely inactive . In contrast , for Gln294Ala-RNR , we would expect purine and pyrimidine substrates to be differentially affected , with minimal or no loss of activity on ADP/GDP and measurable loss on CDP/UDP , and that is again what we observe . For Gln294Ala , activity is the same as wild-type for ADP/dGTP and perhaps even higher than wild-type for GDP/TTP , whereas for both CDP and UDP with 1 μM dATP , activity is decreased ( Figure 9 ) . These results are consistent with the removal of the Gln294 side chain from the active site to create room for the larger substrates ( ADP/GDP ) to bind and also with the positioning of Gln294 into the active site to stabilize CDP/UDP binding through hydrogen-bonding to the respective bases . 10 . 7554/eLife . 07141 . 015Figure 9 . Specific activity for wild-type and mutant forms of E . coli RNR in the presence of different substrate/effector pairs . Wild-type is shown in black , Gln294Ala in grey , and Arg298Ala in white . Activity was measured by a coupled assay that follows nicotinamide adenine dinucleotide phosphate ( NAPDH ) consumption ( see Materials and methods ) for the following substrate and effector concentrations: 1 mM CDP and 3 mM ATP ( far left ) , 1 mM CDP and 175 μM dATP ( second to left ) , 1 mM ADP and 120 μM dGTP , 1 mM GDP and 250 μM TTP , and 1 mM CDP/UDP and 1 μM dATP ( far right ) . Since dATP at high concentrations ( 175 μM ) inhibits the enzyme , the sets of bars at the far left represent control experiments to show activity levels under active ( CDP/ATP ) and inactive ( CDP/dATP ) conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 07141 . 015
In this work , we have sought to decipher the molecular rules of substrate/effector recognition in RNR and determine the basis for increased substrate affinity in the presence of a cognate effector . We observe that in the absence of substrates or effectors ( for example , in the previously reported structure of α2 ( PDB ID: 3R1R ( Eriksson et al . , 1997 ) ) , the barrel is not clamped and loop 2 is not found in any of the conformations seen in our effector-bound structures ( Figure 7E ) . However , in the presence of substrate/effector pairs , we find a clamped down barrel and a stabilized loop 2 that adopts three different conformations depending on the allosteric effector that is bound ( Figure 7A–D ) ( Video 2 ) . Loop 2 residues Cys292 , Ser293 , and Gln294 appear to be involved in specific interactions that read out the identity of the base and communicate that identity to the active site . Briefly , dATP , which enhances reduction of CDP/UDP ( von Döbeln and Reichard , 1976 ) , makes specific contacts with Ser293 , orienting Gln294 toward the active site and stabilizing the binding of both CDP and UDP ( Figure 8A–B ) . The lack of discrimination between CDP and UDP substrates by RNR is not problematic for the cell as cytidine deaminase provides another level of control for dCTP/TTP ratios ( Wang and Weiss , 1992; O'Donovan et al . , 1971 ) . In contrast , dGTP binding stabilizes Gln294 away from the RNR active site , so that CDP/UDP binding is not stabilized ( although it is not prohibited ) and there is room for the larger purine substrates to bind and to hydrogen bond to the carbonyl of Gly299 ( Figure 8C–D ) . The creation of a more expansive active site is consistent with the ability of RNR to reduce both purine substrates in the presence of dGTPγS , albeit with a much greater fold activity increase for preferred substrate ADP ( von Döbeln and Reichard , 1976 ) . Specificity for ADP versus GDP appears to be modulated by whether effector-loop contacts stabilize the carbonyl of Gly299 in a forward or a backward position , respectively ( Figure 8C–D ) . Thus , our structural data suggest that movement of Gln294 in and out of the active site alternatively shrinks and expands the active site for CDP/UDP versus ADP/GDP , whereas a more modest shift of the Gly299 carbonyl is involved in ADP/GDP selectivity . Mutagenesis of Gln294 to Ala is consistent with this proposal , showing decreased activity on CDP/UDP and no change or increased activity on ADP/GDP ( Figure 9 ) . 10 . 7554/eLife . 07141 . 016Video 2 . Loop 2 movements responsible for allosteric specificity regulation in E . coli class Ia RNR . DOI: http://dx . doi . org/10 . 7554/eLife . 07141 . 016 When a cognate substrate/effector pair is bound and the barrel is clamped , Arg298 is able to reach across the active site , stack against the NDP base , and hydrogen bond to the NDP β-phosphate , thereby sequestering the substrate in the active site , and neutralizing the substrate negative charge . These interactions yield what appears to be a high affinity substrate-bound state that is sequestered by solvent and thus amenable to radical-based chemistry . To invoke an analogy , Arg298 is like a latch on a suitcase , locking the active site barrel in cases in which the active site is appropriately packed ( Figure 10A ) . When a suitcase is packed with too many clothes , the latch cannot reach the lock , and the clothes are not secured . Similarly , a mismatched substrate/effector pair such as a purine nucleotide with dATP would not be expected to allow the barrel to clamp and loop 2 to rearrange such that Arg298 can reach the substrate phosphate and thus form a ‘latched’ complex ( Figure 10B ) . In the latter case , one would expect the mismatched substrate to be released from the enzyme , so that a complementary one can bind . Chemical logic tells us that allosteric regulation of specificity would require both low and high affinity substrate/effector-enzyme states; low affinity states to sample substrate/effector pairs , and high affinity states to capture the correct pair , recruit β2 ( class I ) or adenosylcobalamin ( class II ) , and initiate catalysis ( Figure 10 ) . Although the molecular basis for the fivefold increase in binding affinity of β2 for α2 in E . coli class Ia RNR in the presence of bound substrate/effector pairs is not understood , the clamping of the barrel that we observe here may be part of the molecular explanation . The use of Arg298 as a molecular latch to secure the barrel in this closed state is beautiful in its simplicity . It is also consistent with both our structural and our biochemical data that show almost complete loss of activity in the Arg298Ala variant ( Figure 9 ) . 10 . 7554/eLife . 07141 . 017Figure 10 . Snapshots of higher and lower affinity substrate-bound states of RNR . ( A ) Cartoon of a high-affinity complex for CDP/UDP bound to RNR . ( B ) Packing of active site in E . coli class Ia RNR CDP/dATP structure ( this work ) . ( C ) Packing of active site for UDP/dATP structure ( this work ) . ( D ) Cartoon of a lower-affinity complex in which positioning of Gln into the active site holds loop 2 away such that Arg cannot reach the substrate diphosphate . ( E ) Packing of active site of ADP-bound S . cerevisiae RNR structure ( PDB ID: 2CVX ) . With Gln288 ( Gln294 in E . coli ) in the active site , Arg293 ( Arg298 in E . coli ) does not reach the diphosphate of substrate . ( F ) Same structure as in ( E ) , but Gln is not shown . Shape complementary suggests that a tighter complex could form than the one that is visualized in this crystal structure . ( G ) Cartoon of a high-affinity complex for ADP/GDP bound to RNR . ( H ) Packing of active site in GDP structure of class II RNR from T . maritima ( PDB ID: IXJE ) is similar to that of the E . coli class Ia RNR with ADP/dGTP bound ( shown in panel I ) and the E . coli GDP/TTP structure that is shown in Figure 4C . ( I ) Packing of active site in E . coli class Ia RNR with ADP/dGTP bound ( this work ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07141 . 017 Though it is generally accepted that the rules of specificity regulation are conserved among class I , II , and III RNRs , this hypothesis is based on a relatively small number of characterized RNRs . Loop 2 is clearly an important player in specificity regulation , and yet it is not highly conserved , even among characterized RNRs ( Figure 11 ) . This being said , our structures show that backbone atoms of loop 2 make the vast majority of contacts to the substrate base and the effector base , limiting the number of residues that need to be strictly conserved to allow for the same molecular mechanism . Also , despite the fairly low sequence identities between the E . coli and the T . maritima α subunits ( 20 . 5% ) and the E . coli and the S . cerevisiae α subunits ( 29 . 3% ) , a structural comparison shows that all three enzymes interact with the phosphate groups and ribose moieties of their substrates and effectors in similar ways ( Table 4 ) . 10 . 7554/eLife . 07141 . 018Figure 11 . Structure-based sequence alignment of loop 2 residues of characterized class Ia , class Ib , and class II RNRs , with asterisk denoting RNRs for which structures are available . Absolutely conserved residues are starred and highlighted . Arg298 ( E . coli numbering ) stabilizes substrate binding and Gln294 stabilizes pyrimidine binding when dATP is bound to the specificity allosteric site . The eukaryotic RNRs have one additional residue inserted into the loop . Characterized monomeric class II RNRs are not included in this alignment . Beyond these characterized RNRs , sequence alignments are more challenging and conservation is less clear . Although Arg298 may be strictly conserved , Gln294 is unlikely to be . DOI: http://dx . doi . org/10 . 7554/eLife . 07141 . 01810 . 7554/eLife . 07141 . 019Table 4 . Interactions that anchor the ribose and phosphate moieties of substrate and specificity effector molecules to RNRs ( bound waters are not included ) DOI: http://dx . doi . org/10 . 7554/eLife . 07141 . 019E . coli ( this work ) S . cerevisiae ( PDB ID: 2CVX ) T . maritima ( PDB ID: 1XJE ) Substrates NDP ribose 3'-OHH bonds to:Asn437Glu441Asn426Glu430Asn320Glu324NDP ribose 2'-OHH bonds to:Backbone carbonylBackbone carbonylBackbone carbonylNDP phosphatesH bonds to:Backbone amidesThr209Ser625Arg298Backbone amidesSer202Thr611Arg293*Backbone amidesSer91-–Arg207**Effectors dNTP ribose 3'-OHH bond to:Asp232His 275Asp226-–Asp141-–dNTP phosphatesH-bond to:Backbone amidesArg262Arg269Backbone amidesArg256Lys243 ( from a different helix ) Backbone amidesArg171Lys158 ( from a different helix ) *Arg293 residue does not hydrogen bond to the substrate phosphate , but this observed lack of interaction may be due to crystal packing ( see text ) . **Arg207 hydrogen bonds to the substrate phosphate in the one structure ( GDP/TTP ) that has loop 2 fully modeled . The only noteworthy difference is that Arg298 , critical for reduction of all four NDPs in E . coli , directly contacts the β-phosphate of all four substrates . In contrast , none of the four S . cerevisiae RNR structures show a direct contact to the β-phosphate by the equivalent arginine , although one structure shows a through-water contact , and only one of the substrate/effector bound structures of T . maritima RNR ( GDP/TTP ) shows direct contact . This GDP/TTP-bound structure is the only T . maritima RNR structure in which loop 2 is fully ordered and it is remarkably similar to our E . coli structures ( Figure 10H , I ) . All contacts made to GDP are the same: Arg side chain to β-phosphate , and loop 2 carbonyl to N1 and N2 of base ( Figure 10H , 4C ) . The TTP contacts are not identical , but in both cases , the resulting loop 2 conformations allow the carbonyl of 299 ( E . coli numbering ) and 208 ( T . maritima numbering ) to ‘fall back’ to contact N1 and N2 of GDP . Excitingly , the T . maritima class II RNR structures also reveal a movement of Gln294 in and out of the active site in response to effector binding . In particular , the GDP/TTP and CDP/dATP bound structures from T . maritima class II RNR , which have an ordered and semi-ordered loop 2 , respectively , show that Gln203 ( equivalent to Gln294 in E . coli ) interacts directly with CDP when dATP is bound , and moves away from the active site when GDP/TTP are bound ( Larsson et al . , 2004 ) . Unfortunately , many loop 2 residues are disordered and thus not modeled in the T . maritima structures with CDP/dATP , UDP/dATP , and ADP/dGTP , preventing further comparison to our E . coli structures . In stark contrast to the behavior of Gln294 in E . coli and Gln203 in T . maritima , Gln288 in S . cerevisiae class Ia RNR is observed to point into the active site in all structures , regardless of which substrate/effector pairs are bound . Structural comparisons show that the protein does not pack as tightly around purine substrates when Gln points into the active site as it does when Gln is flipped away ( Figure 10E , F compared with 10H , I ) . It is possible that this difference in Gln positioning in S . cerevisiae , and differences described above for Arg298 ( E . coli numbering ) , indicate that the mechanism of allosteric regulation is not conserved among RNRs . However , it is also possible that these observed structural differences are a result of crystal packing variations or other deviations in how crystals were prepared . Additional structures of high-affinity RNR-substrate complexes will help to clarify the degree to which the roles of Gln294 and Arg298 are conserved across RNR species . Regardless of their exact roles in S . cerevisiae RNR , there are a number of studies that support the importance of these residues ( Ahmad et al . , 2012; Kumar et al . , 2010; Kumar et al . , 2011 ) . In particular , when mutant RNR is the only RNR being expressed in S . cerevisiae , mutation of Arg293 ( the Arg298 equivalent ) to Ala is lethal , and mutation of Gln288 ( the Gln294 equivalent ) to Ala yields a severe S phase defect ( Ahmad et al . , 2012 ) . Gln288Ala mutation in S . cerevisiae also leads to substantially elevated dGTP/dATP levels compared with dCTP/TTP levels when one compares wild-type S . cerevisiae expressing two wild-type RNRs with mutant S . cerevisiae expressing one wild-type RNR and one Gln288Ala mutant RNR ( Kumar et al . , 2010 ) . Under the same experimental conditions , all four dNTPs are elevated by similar amounts ( within 1–3% ) for the Arg293Ala mutation in S . cerevisiae ( Kumar et al . , 2010 ) . Although more studies are clearly needed to confirm or refute that the molecular basis of allosteric specificity regulation among RNRs is conserved , it is interesting to note that our in vitro data on E . coli ( Figure 9 ) is consistent with the more severe phenotype in S . cerevisiae for Arg than Gln mutation and is consistent with the observation that mutation of Gln in S . cerevisiae elevates purines over pyrimidines whereas mutation of Arg does not show differential elevation . In conclusion , the work presented here provides a unifying mechanism for substrate specificity regulation in the most studied RNR , the E . coli class Ia enzyme . Our structures show how each specificity effector is read out at a distal allosteric site and how that information is communicated to the active site where residues rearrange such that specific hydrogen bonds can be formed with the cognate substrate base . When an effector/substrate match is discovered , the barrel is clamped and latched in preparation for catalysis . Just as DNA replication and transcription take advantage of the unique hydrogen-bonding properties of each nucleotide base , enzymatic ribonucleotide reduction also employs these unique hydrogen-bonding properties for specificity regulation . Through an elegant set of protein rearrangements , E . coli RNR screens and selects its substrate from the four potential NDPs , ensuring appropriate pools of deoxynucleotides are available for DNA biosynthesis and repair .
For the enzyme assays , sodium salts of CDP , ADP , UDP , GDP , dATP , dGTP , and TTP were purchased from Sigma-Aldrich ( St . Louis , MO ) and dissolved into assay buffer ( 50 mM HEPES pH 7 . 6 , 15 mM MgCl2 , 1 mM ethylenediaminetetraacetic acid ( EDTA ) ) . The pH of each solution was slowly adjusted to 7–8 with NaOH , and the nucleotide concentrations were determined spectroscopically , using ε271 of 9 . 1 mM-1 cm-1 for CDP , ε259 of 15 . 4 mM-1 cm-1 for ADP , ε262 of 10 . 0 mM-1 cm-1 for UDP , ε253 of 13 . 7 mM-1 cm-1 for GDP , ε259 of 15 . 4 mM-1 cm-1 for ATP , ε259 of 15 . 2 mM-1 cm-1 for dATP , ε253 of 13 . 7 mM-1 cm-1 for dGTP , and ε262 of 9 . 6 mM-1 cm-1 for TTP . Preparation of nucleotides for crystallography was the same as described above except that 100 mM solutions of sodium salts of dATP , dGTP , and TTP were purchased from USB Corporation ( Cleveland , OH ) or Invitrogen ( Carlsbad , CA ) . The α2 and β2 proteins were prepared as described ( Salowe et al . , 1987; Salowe and Stubbe , 1986 ) . The concentrations of α2 and β2 were determined using ε280 of 189 and 131 mM-1cm-1 , respectively; unless noted otherwise , all molar concentrations are dimer concentrations ( i . e . α2 or β2 ) . For all structures , hydroxyurea-inactivated β2 ( met-β2 ) was used in place of active β2 . Met-β2 was prepared from purified active β2 as described ( Ando et al . , 2011 ) . α2 had a specific activity of 3800 nmol min-1 mg-1 . Prior to hydroxyurea treatment , β2 had a specific activity of 7700 nmol min-1 mg-1 as determined by a coupled spectrophotometric assay ( Ge et al . , 2003 ) . Crystals were grown using the hanging drop vapor diffusion technique by mixing 1 μL of protein ( 25 μM α2 and 50 μM met-β2 in 50 mM HEPES , 15 mM MgCl2 , and 1 mM EDTA , pH 7 . 6 , supplemented with 10 mM dATP and 5 mM DTT , 1% ( w/v ) isopropyl-β-thiogalactopyranoside ) with 1 μL of precipitant solution and equilibrating over a reservoir of 500 μL of precipitant at room temperature ( ~25°C ) . For the CDP/dATP structure , the precipitant solution was 9 . 5% ( w/v ) PEG 3350 , 100 mM MOPS pH 7 . 5 , 250 mM Mg ( CH3COO ) 2 , 25 mM MgCl2 , and 5% ( v/v ) glycerol . For the UDP/dATP , ADP/dGTP , and GDP/TTP structures , the precipitant solution was 12% ( w/v ) PEG 3350 , 100 mM MOPS pH 7 . 5 , 300 mM Mg ( CH3COO ) 2 , 30 mM MgCl2 , and 5% ( v/v ) glycerol . Freshly prepared drops were streak seeded with microcrystals grown under the same conditions with the exception of the MgCl2 concentration in the precipitant being 100 mM MgCl2 . Streak seeding was used to obtain larger single crystals . After 2 days of growth , crystals were transferred to a drop of soaking solution . For the CDP/dATP structure , dATP was co-crystallized , and CDP was soaked into the structure using a solution containing 10 . 5% ( w/v ) PEG 3350 , 100 mM MOPS pH 7 . 5 , 25 mM MgCl2 , 250 mM Mg ( CH3COO ) 2 , 5% ( v/v ) glycerol , 5 mM DTT , and 10 mM CDP . For the GDP/TTP , UDP/dATP and ADP/dGTP structures , the soaking solutions were the same as for CDP except for having 13% ( w/v ) PEG 3350 , and each nucleotide at 10 mM . Crystals were left in the soaking solution for 2 min and then cryoprotected . CDP-soaked crystals were looped through a solution of 12% ( w/v ) PEG 3350 , 100 mM MOPS pH 7 . 5 , 250 mM Mg ( CH3COO ) 2 , 60 mM MgCl2 and 10% , 15% , and 20% ( v/v ) glycerol in succession and then plunged directly into liquid N2 . GDP/TTP , UDP/dATP , and ADP/dGTP soaked crystals were looped through a solution of 14% ( w/v ) PEG 3350 , 100 mM MOPS pH 7 . 5 , 300 mM Mg ( CH3COO ) 2 , 100 mM MgCl2 , and 10% , 15% , and 20% ( v/v ) glycerol in succession and then plunged directly into liquid N2 . Diffraction data were collected at the Advanced Photon Source at Argonne National Laboratory . The CDP/dATP data set was collected on beamline 24ID-C at 100 K on an ADSC Q315 detector . The UDP/dATP , ADP/dGTP , and GDP/TTP data sets were collected on beamline 24ID-C at 100 K on a Pilatus 6M detector . All data were processed using HKL2000 ( Otwinowski and Minor , 1997 ) ( Table 2 ) . All four α4β4 complex structures soaked with substrates and effectors were solved to the full extent of the data resolution using the previously published 3 . 95-Å resolution structure ( Zimanyi et al . , 2012 ) with all nucleotides removed . Rfree test sets were chosen to contain the same reflections across all data sets . For all structures , initial refinement was carried out in CNS 1 . 3 ( Brünger et al . , 1998 ) and later in Phenix ( Afonine et al . , 2012 ) with model building performed in COOT ( Emsley et al . , 2010 ) . Refinement consisted of rigid body , simulated annealing , positional and individual B-factor refinement with no sigma cutoff . Loose non-crystallographic symmetry restraints were used throughout refinement . Simulated annealing composite omit maps generated in CNS 1 . 3 and Phenix were used to verify the models . All figures were made using the PyMOL Molecular Graphics System , version 1 . 5 . 0 . 4 ( Schrödinger , LLC ) . Final refinement statistics are shown in Table 2 . In addition to the substrate/specificity effector pairs bound in each structure , all structures have dATP or its hydrolysis product dADP in the allosteric activity site . In some cases , the high concentration of nucleotide used ( 10 mM ) has resulted in more than one nucleotide molecule bound near this allosteric site . Difference distance matrix plots were produced using the DDMP program from the Center for Structural Biology at Yale University ( New Haven , CT ) . Residues 4 to 737 of chain A from each structure were used for the analysis . Gln294Ala-α2 and Arg298Ala-α2 were constructed from the previously published vector pET-nrdA ( Minnihan et al . , 2011 ) through QuikChange site-directed mutagenesis ( Agilent , Santa Clara , CA ) using the following primers from Integrated DNA Technologies ( Coralville , IA ) : Gln294Ala forward primer ( 5'’-AAATCCTGCTCTGCGGGCGGTGTGC-3’ ) , Gln294Ala reverse primer ( 5’-GCACACCGCCCGCAGAGCAGGATTT-3’ ) , Arg298Ala forward primer ( 5’-CGTTGCCGCACCGCCAGCCACACCGC-3’ ) , and Arg298Ala reverse primer ( 5’-GCGGTGTGGCTGGCGGTGCGGCAACG-3’ ) . All mutations were confirmed by DNA sequencing performed by Genewiz ( South Plainfield , NJ ) . Expression and purification of N-terminal hexahistidine tagged E . coli wild-type α2 and N-terminal hexahistidine tagged mutant α2 variant proteins ( Gln294Ala-α2 and Arg298Ala-α2 ) were carried out as previously described ( Minnihan et al . , 2011 ) . Briefly , cells were resuspended in 40 mL of Buffer A ( 50 mM Tris pH 7 . 6 , 300 mM NaCl , 1 mM TCEP ) , lysed by sonication , and clarified by centrifugation at 29 , 000 × g . Lysate was applied to a 5 mL HisTrap HP column ( GE Healthcare Life Sciences , Pittsburg , PA ) , washed with Buffer A supplemented with 30 mM imidazole , and eluted with Buffer A supplemented with 300 mM imidazole . Protein was further purified on a Superdex 200 16/60 size exclusion column ( GE Healthcare Life Sciences ) and transferred to a final storage buffer of 20 mM HEPES 7 . 6 , 100 mM NaCl , and 5% glycerol . A final yield of ~25–50 mg/L of culture for wild-type α2 is typical . The purification for the mutants was identical , with similar yields . All proteins were judged as purified to homogeneity by sodium dodecyl sulfate/polyacrylamide gel electrophoresis ( SDS/PAGE ) , and their concentrations were determined using ε280 of 189 mM-1cm-1 . Hexahistidine tags were not removed since previous studies showed that these tags on E . coli α2 do not significantly alter activity ( Minnihan et al . , 2011 ) . Untagged E . coli β2 was purified as previously described ( Salowe and Stubbe , 1986 ) and contained ~1 . 1 radicals per dimer as estimated by UV-visible spectroscopy of the Y122 radical ( ε411 of 1760 mM-1 cm-1 ) after drop-line subtraction of the diferric cluster absorbance ( Bollinger et al . , 1995 ) . Untagged E . coli β2 was exchanged into a storage buffer containing 50 mM HEPES 7 . 6 and 5% glycerol , and its concentration determined using ε280 of 131 mM-1 cm-1 . For the coupled assay described below , E . coli thioredoxin reductase ( TrxR ) and E . coli thioredoxin ( Trx ) were prepared . The gene for TrxR was amplified from E . coli genomic DNA using primers with ends suitable for a second round of PCR to generate the plasmid borne gene in the pRham SUMO fusion vector ( Lucigen , Middleton , WI ) . The primers , ordered from Integrated DNA Technologies , were: forward 5’-CGCGAACAGATTGGAGGTGGCACGACCAAACACAGTAAACTG-3’ , and reverse 5’-GTGGCGGCCGCTCTATTATTTTGCGTCAGCTAAACCATCGAG-3’ . DNA sequencing performed by Genewiz was used to confirm the sequence of the resulting construct . This construct contains a hexahistidine tag and SUMO protein fusion at the N-terminus under a rhamnose promoter . The resulting protein was expressed according to the plasmid manufacturer protocol ( Lucigen ) and purified as described above for hexahistidine-tagged α2 , using the same buffers . The amount of flavin cofactor was quantified by absorption at 440 nm ( Thelander , 1967; Gleason et al . , 1990 ) , and a specific activity of 38 , 200 nmol min-1 mg-1 was determined by a coupled assay with 30 µM TR and 150 µM Ellman’s reagent ( 5 , 5'-dithiobis[2-nitrobenzoic acid] ) ( Pierce , Rockford , IL ) ( Luthman and Holmgren , 1982 ) . The final concentration for assays ( 0 . 5 µM ) was determined by flavin concentration as the as-expressed protein was only 60% loaded with flavin as determined by the A280/A440 ratio . The resulting protein behaved identically in assays as untagged TrxR purified from an overexpressing E . coli strain ( Russel and Model , 1985 ) . E . coli Trx was purified as previously described ( Chivers et al . , 1997 ) . Activity assays for wild-type α2 and the Gln294Ala and Arg298Ala mutants were performed using a continuous , coupled , spectrophotometric assay monitoring the consumption of NADPH by the Trx/TrxR system ( Ge et al . , 2003 ) . All experiments were performed on a Cary Bio300 spectrometer ( Agilent ) with data analysis performed using the Cary WinUV Kinetics program ( Agilent ) and Microsoft Excel . The assay buffer consisted of 50 mM HEPES pH 7 . 6 , 15 mM MgCl2 , 1 mM EDTA , and the following substrate and effector concentrations were used: 1 mM CDP and 3 mM ATP ( control for presence of activity ) , 1 mM CDP and 175 μM dATP ( control for inactivation by dATP ) , 1 mM ADP and 120 μM dGTP , 1 mM GDP and 250 μM TTP , and 1 mM CDP/UDP and 1 μM dATP . Substrate , effector , E . coli Trx ( 30 μM ) , E . coli TrxR ( 0 . 5 μM ) , and NADPH from Sigma-Aldrich ( 200 μM ) were mixed in assay buffer , and the reaction was initiated by the addition of α2 ( 0 . 1 μM ) and wild-type β2 ( 1 μM ) to a final volume of 120 μL . The decrease in NADPH absorbance at 340 nm was monitored over 1 min . The basal level of NADPH oxidation was monitored over 30 s prior to the addition of enzyme . | DNA contains the instructions required to make proteins and other molecules in cells . DNA is made of four building blocks called deoxyribonucleotides , which are in turn made from molecules called ribonucleotides by enzymes known as ribonucleotide reductases ( RNRs for short ) . RNR enzymes are responsible for maintaining a good balance in the levels of the different deoxyribonucleotides in cells , which is essential for DNA to be made and repaired correctly . Previous work has shown that each RNR can act on all four ribonucleotides . However , these enzymes become more selective for certain ribonucleotides depending on which deoxyribonucleotide is most common within the cell . For example , when a deoxyribonucleotide called dGTP is plentiful , it binds to a so-called “specificity site” on the enzyme and alters the shape of the enzyme’s active site . This then means that a ribonucleotide called ADP will bind in preference to the other ribonucleotides . However , it was not clear how the binding of deoxyribonucleotides to the enzyme influences the shape of the active site . Zimanyi et al . used a technique called X-ray crystallography to determine the three-dimensional structures of a bacterial RNR enzyme when it is bound to all four different combinations of deoxyribonucleotides and ribonucleotides . In the absence of nucleotides , the active site adopts a shape that resembles an open barrel . However , when RNR is bound to a deoxyribonucleotide at the specificity site and a ribonucleotide at the active site , the barrel clamps down , bringing the specificity site and the active site closer together . Additionally , a loop of the protein interacts with each of the deoxyribonucleotides in a different way and communicates their identity directly to the active site , which rearranges itself to hold on to the corresponding preferred ribonucleotide . Zimanyi et al . ’s findings provide an explanation for how RNRs can select between ribonucleotides so that they produce a good balance of deoxyribonucleotides in cells . This will inform future efforts to develop molecules that inhibit RNRs , which may have the potential to be used to treat bacterial infections or to kill cancer cells . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2016 | Molecular basis for allosteric specificity regulation in class Ia ribonucleotide reductase from Escherichia coli |
The glycosyltransferase EOGT transfers O-GlcNAc to a consensus site in epidermal growth factor-like ( EGF ) repeats of a limited number of secreted and membrane proteins , including Notch receptors . In EOGT-deficient cells , the binding of DLL1 and DLL4 , but not JAG1 , canonical Notch ligands was reduced , and ligand-induced Notch signaling was impaired . Mutagenesis of O-GlcNAc sites on NOTCH1 also resulted in decreased binding of DLL4 . EOGT functions were investigated in retinal angiogenesis that depends on Notch signaling . Global or endothelial cell-specific deletion of Eogt resulted in defective retinal angiogenesis , with a mild phenotype similar to that caused by reduced Notch signaling in retina . Combined deficiency of different Notch1 mutant alleles exacerbated the abnormalities in Eogt−/− retina , and Notch target gene expression was decreased in Eogt−/−endothelial cells . Thus , O-GlcNAc on EGF repeats of Notch receptors mediates ligand-induced Notch signaling required in endothelial cells for optimal vascular development .
N-acetylglucosamine linked to Ser or Thr ( O-GlcNAc ) is a rare form of post-translational modification specifically modifying epidermal growth factor-like ( EGF ) domains in secreted or membrane proteins ( Alfaro et al . , 2012; Matsuura et al . , 2008; Stanley and Okajima , 2010 ) . In contrast to O-GlcNAc modification of nuclear and cytoplasmic proteins ( Ma and Hart , 2014 ) , EGF-specific O-GlcNAcylation occurs in the endoplasmic reticulum ( ER ) by the action of the EGF-domain-specific O-GlcNAc transferase , EOGT ( Sakaidani et al . , 2011 ) . To date , only a small number of secreted or membrane proteins have been shown to be O-GlcNAcylated , including Notch receptors and their ligands ( Alfaro et al . , 2012; Müller et al . , 2013; Tashima and Stanley , 2014 ) . Notch receptors are also modified by O-fucose and O-glucose glycans at independent consensus sites on EGF repeats ( Figure 1A ) , and both types of O-glycan regulate the strength of Notch signaling ( Haltom and Jafar-Nejad , 2015 ) . Loss of O-fucose , but not O-glucose glycans reduces binding of canonical Notch ligands to Notch receptors , and consequently Notch signaling . Removing one class of O-glycan will affect many EGF repeats but is not expected to affect modification of EGF repeats by unrelated O-glycans . In this paper , we report that Notch receptors lacking only O-GlcNAc glycans have reduced ligand-induced Notch signaling . 10 . 7554/eLife . 24419 . 003Figure 1 . EOGT promotes NOTCH1 binding to Delta ligands . ( A ) Diagram of predicted O-glycans on mouse NOTCH1: red triangle , O-fucose glycans; blue circle , O-glucose glycans; blue square , O-GlcNAc glycans . Individual sugar residues that may extend O-fucose , O-glucose or O-GlcNAc to varying degrees are: yellow circle , galactose; pink diamond , sialic acid; orange star , xylose . ( B ) Flow cytometry of Lec1 CHO cells expressing vector control or Eogt siRNA with NOTCH1 mAb , DLL1-Fc , DLL4-Fc or JAG1-Fc . ( C ) Relative mean fluorescence index ( MFI ) for binding of DLL1-Fc and DLL4-Fc to control and Eogt-siRNA Lec1 cells . Concentrations of ligand varied from 100 to 750 ng/ml . MFI values for binding to control cells taken as 1 . 0 were 455 ± 55 ( DLL1-Fc ) and 1215 ± 49 ( DLL4-Fc ) . Data from 9 to 10 independent experiments are average , normalized MFI ± SEM; significance determined by paired , two-tailed Student’s t-test , *p<0 . 05 , ****p<0 . 0001 . ( D ) MFI values obtained for binding of DLL1-Fc or DLL4-Fc ( 750 ng/ml ) to control and Eogt knockdown Lec1 CHO cells , before and after transfection of a human EOGT cDNA . Data are mean ± SEM from three independent experiments . Significance determined by unpaired , two-tailed Student’s t-test , *p<0 . 05 . Western blot analysis of transfectants . ( E ) DLL4 and JAG1 beads bound to wild-type , EOGT-null , or NOTCH1-null HEK293T cells were observed by microscopy and counted ( n = 50 ) . Data are mean ± S . D . from three independent experiments . Statistical analysis was by Welch's t-test . **p<0 . 01; ***p<0 . 001 . ( F ) Wild-type or EOGT-null HEK293T cells were transfected with Notch1 alone or together with Eogt followed by incubation with DLL4 beads . The number of DLL4 beads bound to cells was markedly increased by co-transfection of Eogt and Notch1 . ( G ) Wild-type or EOGT-null HEK293T cells or cells transiently transfected with Notch1 with or without Eogt , were incubated with DLL4 or JAG1 beads . The number of Dynabeads bound per transfected cell marked by GFP expression was determined ( n = 50 ) . Data are mean ± S . D . from three independent experiments . *p<0 . 05; **p<0 . 01; ***p<0 . 001; #p<0 . 05; ###p<0 . 001 compared with the left-most wild type ( * ) or EOGTnull ( # ) bar by Welch's t test . ( H ) Wild-type or EOGT-null HEK293T cells were transfected with Notch1 with or without Eogt and subjected to flow cytometry using 8G10 NOTCH1 Ab . Mock transfectants were analyzed with ( Mock ) or without primary antibody ( Cont ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 00310 . 7554/eLife . 24419 . 004Figure 1—source data 1 . Raw data for Figure 1C , D , E , G . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 00410 . 7554/eLife . 24419 . 005Figure 1—figure supplement 1 . Generation and characterization of EOGT- and NOTCH1-null HEK293T cells . ( A ) Schematic diagram of the CRISPR/Cas9 genome editing strategy to generate EOGT-null HEK293T cells . CRISPR/Cas9-mediated DNA cleavage caused frameshift mutations in all three EOGT alleles in HEK293T cells . ( B ) Wild type or EOGT-null HEK293T cells with Notch1-EGF-mycHis alone or together with EOGT . CTD110 . 6 immunoblotting revealed the lack of O-GlcNAc on NOTCH1 EGF repeats in the absence of EOGT . ( C ) Total ER fraction was obtained using Endoplasmic Reticulum Enrichment Extraction Kit ( Novus Biologicals [NBP2-29482] ) from parental HEK293T cells or EOGT-null cells , and subjected to immunoblotting with EOGT ( 1:2000 dilution ) or BiP ( 1:10000 dilution ) antibodies . ( D ) Schematic diagram of the CRISPR/Cas9 genome editing strategy to generate NOTCH1-null HEK293T cells . CRISPR/Cas9-mediated DNA cleavage caused frameshift mutations in the single NOTCH1 allele in HEK293T cells . ( E ) Screening for CRISPR/Cas9-mediated genomic deletion at the Notch1 locus . A clone 1g1 was selected and deletion of the target sequence was confirmed by direct sequencing analysis . ( F ) Total cell lysates from parental HEK293T cells or NOTCH1-null cells were subjected to immunoprecipitation ( IP ) using NOTCH1 ECD antibody ( H-131 ) . The immunoprecipitates were analyzed by immunoblotting with NOTCH1 ICD antibody ( D6F11 ) . Aliquots of total cell lysates were immunoblotted with β-tubulin antibody . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 005 The physiological importance of EOGT was revealed in patients with Adams-Oliver syndrome ( AOS ) , a rare congenital disorder characterized by aplasia cutis congenita and terminal transverse limb defects , often accompanied by cardiovascular malformations and brain anomalies ( Algaze et al . , 2013; Piazza et al . , 2004 ) . Although the pathogenesis of AOS is broad , it could arise from small-vessel vasculopathy ( Piazza et al . , 2004 ) . AOS is a heterogeneous disorder caused by mutation in one of at least six different genes . Among these , loss-of-function mutations of EOGT and DOCK6 were identified as the basis of an autosomal-recessive form of AOS ( Shaheen et al . , 2013 , 2011 ) . In addition , autosomal dominant mutations of NOTCH1 , RBPJ , DLL4 and ARHGAP31 give rise to AOS ( Aminkeng , 2015; Hassed et al . , 2012; Meester et al . , 2015; Southgate et al . , 2015; Stittrich et al . , 2014 ) . Gain-of-function mutation of ARHGAP31 and loss-of-function mutation of DOCK6 suggested that inactivation of Cdc42/Rac1 functions underlies the molecular basis for AOS . In contrast , loss-of-function mutations of DLL4 , RBPJ and NOTCH1 in AOS patients suggest that impaired Notch signaling is an alternative basis of the pathogenesis of AOS . Here , we investigate the hypothesis that loss of EOGT affects Notch signaling using cell-based Notch ligand binding and signaling assays and Eogt mutant mice . We show that EOGT-catalyzed NOTCH1 O-GlcNAcylation potentiates DLL1- and DLL4-NOTCH1 binding and Notch signaling , whereas JAG1-NOTCH1 binding remains unaffected . Using retinal angiogenesis as a sensitive assay of Notch signaling in vivo ( Roca and Adams , 2007 ) , we show that mice lacking EOGT have impaired retinal vascular development , with a phenotype characteristic of Notch pathway deficiencies in retina ( Benedito et al . , 2009 ) . Moreover , we show that endothelial functions of EOGT are responsible for the retinal vascular phenotype . Thus , O-GlcNAc on the EGF repeats of Notch receptors is required for optimal Notch signaling in developing retina , and likely in other Notch-dependent processes in mammals .
To address whether EOGT regulates physical interactions between Notch receptors and ligands , Notch ligand binding assays were performed on control and Eogt–siRNA Lec1 Chinese hamster ovary ( CHO ) cells . Eogt transcripts determined by quantitative RT-PCR were reduced by ~60% . Eogt-siRNA Lec1 cells exhibited reduced binding of soluble DLL1-Fc and DLL4-Fc ( Figure 1B and Figure 1C ) . However , binding of soluble JAG1-Fc was not altered by knockdown of Eogt ( Figure 1B ) . Overexpression of an EOGT cDNA rescued DLL1 and DLL4 binding ( Figure 1D ) . Moreover , cell surface expression of NOTCH1 was not reduced in Lec1 cells with reduced Eogt ( Figure 1B ) . A second ligand binding assay used soluble Notch ligands attached to Protein A Dynabeads via their Fc domain , and EOGT- and NOTCH1-null HEK293T cells generated by CRISPR/Cas9 gene editing ( Figure 1—figure supplement 1 ) . Deletion of EOGT was verified using anti-EOGT antibody and by the lack of O-GlcNAc on a NOTCH1 extracellular domain fragment ( Figure 1—figure supplement 1 ) . Both DLL4 and JAG1 beads/cell were decreased in NOTCH1-null cells ( Figure 1E ) , confirming that NOTCH1 mediates ligand biding in HEK293T cells . The residual binding capacity of NOTCH1-null cells implicated the contribution of other Notch receptors . Similar to Eogt-siRNA CHO Lec1 cells , EOGT-null HEK293T cells exhibited decreased binding to DLL4 beads , but not to JAG1 beads ( Figure 1E ) . To determine whether overexpression of EOGT potentiates DLL4-NOTCH1 binding , HEK293T cells or EOGT-null cells were transfected with Notch1 and Eogt cDNA individually or together , and the ligand binding assay was performed . Notch1 overexpression led to increased binding of both DLL4 and JAG1 beads to HEK293T cells ( Figure 1F and G ) . In addition , the effect of Notch1 overexpression on DLL4 bead binding was selectively impaired in EOGT-null cells ( Figure 1F and G ) . Furthermore , simultaneous expression of Notch1 and Eogt enhanced DLL4 but not JAG1 bead binding , in both HEK293T and EOGT-null cells ( Figure 1F and G ) . The synergistic effect of Notch1 and Eogt on DLL4 bead binding provide strong evidence that EOGT potentiates DLL4-NOTCH1 physical interactions . As observed in Lec1 CHO cells ( Figure 1B ) , neither Eogt overexpression nor EOGT loss affected cell surface NOTCH1 expression ( Figure 1H ) . Thus , EOGT is not required for NOTCH1 trafficking to the plasma membrane . To determine whether it is the O-GlcNAc transferred by EOGT to NOTCH1 that directly affects the binding of DLL4 , we generated NOTCH1 site-specific mutants by Ala substitution of Ser/Thr in predicted O-GlcNAcylation sites . Alignment of previously reported O-GlcNAc-modified proteins including Notch and Dumpy in Drosophila , and Hspg2 , Nell1 , Lama5 , and Pamr1 in mouse brain suggested that the potential consensus sequence is C5XXG ( Y/F/L ) ( T/S ) GX2-3C6 ( Alfaro et al . , 2012; Sakaidani et al . , 2011 ) . A relatively small number of the EGF domains in mouse NOTCH1 contain the consensus sequence C5XXG ( Y/F ) ( T/S ) GXXC6 , and these are conserved among mouse , human , rat , Chinese hamster and zebrafish in EGF10 , 14 , 15 , 17 , 20 , 23 , 26 , 27 and 29 ( Figure 2—figure supplement 1 ) . Mutation of two or four Ser/Thr to Ala in EGF2 , 10 , 17 and 20 ( Notch1Δ2xO-GlcNAc; Notch1Δ4xO-GlcNAc; Figure 2A ) , outside the canonical ligand-binding region ( EGF11 and 12 ) , was performed . Cell surface expression of the NOTCH1 mutants was indistinguishable from wild-type NOTCH1 ( Figure 2B ) , consistent with unchanged NOTCH1 cell surface expression in EOGT-null HEK293T cells ( Figure 1H ) . Decreased O-GlcNAcylation of Notch1Δ2xO-GlcNAc and Notch1Δ4xO-GlcNAc mutants was confirmed by immunoblotting with CTD110 . 6 O-GlcNAc antibody ( Figure 2C ) . Expression of Notch1Δ4xO-GlcNAc in Eogt-transfected cells resulted in an ~80% decrease in O-GlcNAc immunostain signal ( Figure 2D ) . By contrast , Ala substitution in only EGF2 and 10 ( Notch1Δ2xO-GlcNAc ) removed ~60% of the signal in Eogt transfectants ( Figure 2D ) . Moreover , the number of DLL4 beads bound to NOTCH1Δ4xO-GlcNAc transfectants was significantly decreased relative to wild-type NOTCH1 ( Figure 2E ) , similar to the decrease observed in EOGT-null cells ( Figure 1E ) . This suggests that O-GlcNAc in EGF2 , 10 , 17 and/or 20 are important contributors to DLL4/NOTCH1 interactions . When NOTCH1Δ4xO-GlcNAc was simultaneously expressed with Eogt , Notch1Δ4xO-GlcNAc/Eogt cotransfectants also exhibited impaired binding to DLL4 beads . These results demonstrate that DLL4/NOTCH1 interactions mediated by EOGT require O-GlcNAc on sites that are located outside the canonical ligand-binding region . In contrast , the number of JAG1 beads bound to Notch1 transfectants was not reduced in Notch1Δ4xO-GlcNAc transfectants , irrespective of Eogt overexpression ( Figure 2E ) . These results provide strong evidence that O-GlcNAc on NOTCH1 EGF2 , 10 , 17 and/or 20 selectively affect DLL4/NOTCH1 but not JAG1/NOTCH1 physical interactions . The substantial decrease in O-GlcNAc signal following removal of only four of the nine potential O-GlcNAcylation sites , suggests that a limited number of sites are O-GlcNAcylated in NOTCH1 . 10 . 7554/eLife . 24419 . 006Figure 2 . O-GlcNAc on NOTCH1 EGF repeats promotes DLL4-NOTCH1 interactions . ( A ) Ala substitution of Thr/Ser in the O-GlcNAc consensus site C5XXG ( Y/F ) ( T/S ) GXXC6 in EGF2 , 10 , 17 , and 20 in the NOTCH1Δ4xO-GlcNAc mutant . ( B ) Cell surface expression of NOTCH1Δ4xO-GlcNAc is comparable to that of wild-type NOTCH1 . EOGT-null transfectants expressing Notch1 or Notch1Δ4xO-GlcNAc with or without Eogt were subjected to flow cytometry using 8G10 NOTCH1 antibody . Control transfectants were with ( Mock ) or without ( Cont ) primary antibody . Similar results were obtained for wild-type HEK293T cells ( not shown ) . ( C ) Expression and O-GlcNAcylation of Notch1 , Notch1Δ2xO-GlcNAc and Notch1Δ4xO-GlcNAc were analyzed by immunoprecipitation ( IP ) using NOTCH1 ( 8G10 ) antibody , followed by immunoblotting with CTD110 . 6 O-GlcNAc or NOTCH1 antibodies . ( D ) HEK293T transfectants expressing NOTCH1 , NOTCH1Δ2xO-GlcNAc or NOTCH1Δ4xO-GlcNAc were immunostained for O-GlcNAc ( CTD110 . 6 mAb; red ) or NOTCH1 ( 8G10 mAb; green ) . Merged images include DAPI ( blue ) staining . Note that CTD110 . 6 mAb binding is markedly decreased in the NOTCH1Δ4xO-GlcNAc mutant . ( E ) Quantification of the number of DLL4- or JAG1-coated Dynabeads bound to Notch1 versus Notch1Δ4xO-GlcNAc transfected cells ( marked with GFP ) , in the presence and absence of Eogt . Data are mean ± S . D from three independent experiments . Each experiment analyzed 50 cells . *p<0 . 05; **p<0 . 01; ***p<0 . 001 ( Welch's t test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 00610 . 7554/eLife . 24419 . 007Figure 2—source data 1 . Raw data for Figure 2E . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 00710 . 7554/eLife . 24419 . 008Figure 2—figure supplement 1 . NOTCH1 O-GlcNAc site mutants . ( A ) Schematic of the EGF repeats of mammalian NOTCH1 identifying the potential O-GlcNAc consensus site , C5XXG ( Y/F ) ( T/S ) GXXC6 , on EGF2 , 10 , 11 , 14 , 15 , 17 , 20 , 23 , 26 , 27 and 29 . These consensus sites are partly conserved in zebrafish and Drosophila . Recent studies of Drosophila Notch identified EGF4 , 11 , 12 , 14 and 20 as major O-GlcNAcylated sites ( Harvey et al . , 2016 ) . ( B ) Schematic of EGF repeats of mouse Notch1Δ2xO-GlcNAc and Notch1Δ4xO-GlcNAc . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 008 To investigate effects of EOGT on Notch ligand-induced signaling , we analyzed NOTCH1 activation and signaling in HeLa and Lec1 CHO cells with reduced Eogt . HeLa cells stably expressing four different shRNA constructs targeted to the coding or 3’UTR region of human EOGT were co-cultured with L cells , or DLL1-expressing L cells ( D1/L ) , in the presence and absence of a gamma-secretase inhibitor ( GSI ) . Ligand-induced activation of NOTCH1 generates the release of NOTCH1 intracellular domain ( ICD ) , identified by Western analysis using mAb Val1744 specific for the cleaved N-terminus of NOTCH1 ICD ( Huppert et al . , 2000 ) . NOTCH1 cleavage was stimulated poorly , or not at all , by L cells ( Figure 3A ) . However , co-culture with D1/L cells caused robust NOTCH1 activation in cells expressing a GAPDH control shRNA ( Figure 3A ) . Importantly , NOTCH1 cleavage was inhibited in D1/L co-cultures containing the GSI ( Figure 3A ) . All four EOGT-targeted shRNA constructs specifically reduced NOTCH1 activation . These findings were replicated and also reproduced in an independent set of HeLa cell transductants . Because endogenous levels of EOGT were not readily detectable in HeLa cells by Western analysis , and reduced transcript levels do not reflect enzyme activity , knockdown efficiency was determined by Western blot detection of the O-GlcNAc product of EOGT . The TA197 and TA198 shRNAs that were most effective at reducing NOTCH1 activation ( Figure 3A ) , exhibited marked loss of O-GlcNAc on species comigrating with NOTCH1 ( Figure 3B; the lower band of the O-GlcNAc-positive doublet ) . HeLa cells express NOTCH2 and NOTCH3 , which may also be modified by O-GlcNAc . Alternatively , the O-GlcNAc mAb may detect a glycoform of NOTCH1 that is not detected by the NOTCH1 extracellular domain ( ECD ) mAb . Image J analysis of immunoprecipitated Notch1-MycHis using CTD110 . 6 , NOTCH1 and Myc mAbs confirmed the reduction in O-GlcNAc on NOTCH1 in EOGT-knockdown HeLa cells . To investigate specificity , HeLa cells expressing GAPDH or TA197 shRNA against the 3’UTR of EOGT , were transfected with empty vector or a human EOGT cDNA , and DLL1-induced activation of NOTCH1 was determined . Overexpression of EOGT increased the amount of activated NOTCH1 induced by D1/L cells in control HeLa cells , including in the presence of the GSI ( Figure 3C , D ) . Partial rescue of EOGT knockdown by TA197 ( and TA198 , not shown ) was achieved by a human EOGT cDNA ( Figure 3C , D; representative of three independent experiments ) . 10 . 7554/eLife . 24419 . 009Figure 3 . Notch signaling is reduced in EOGT-deficient cells . ( A ) Knockdown of EOGT inhibits NOTCH1 activation cleavage . HeLa cells stably expressing shRNAs targeting GAPDH or EOGT ( TA195 , TA196 , TA197 , TA198 ) were co-cultured with L cells or D1/L ( D1 ) cells in the presence and absence of 1 µM DAPT ( GSI ) . After 6 hr , lysates were subjected to Western blot analysis using Abs to detect activated NOTCH1 ( N1-act ) and NOTCH1 full length ( N1–FL ) on the relevant section of the PVDF membrane . ( B ) Western blot analysis of samples from ( A ) using Ab to detect O-GlcNAc , followed after stripping by Ab to detect N1-FL . ( C ) HeLa cells stably expressing shRNA against GAPDH or EOGT ( TA197 ) were transfected with vector control or a human EOGT cDNA . After 4 days , co-culture was performed with D1/L ( D1 ) cells in the presence and absence of the DAPT . After ~7 hr , lysates were subjected to Western analysis to detect N1-FL , N1-act and EOGT on the relevant section of the PVDF membrane . O-GlcNAc was detected after stripping the N1-FL membrane section . * non-specific band . ( D ) The same as ( C ) except co-culture was with L cells or D1/L cells ( D1 ) . The second lane ( # ) was left empty . * non-specific band . ( E ) Knockdown of Eogt reduces ligand-induced Notch signaling . Lec1 CHO cells stably expressing siRNAs targeted against Eogt were transfected with TP1-luciferase and TK-renilla luciferase , co-cultured for 30 hr with L cells or D1/L cells , with and without GSI IX ( 12 . 5 μM ) or a human EOGT cDNA , and dual firefly and renilla luciferase assays were performed . Normalized firefly luciferase activity in L versus D1/L cell co-cultures was plotted as fold-change . The number of independent experiments , each performed in duplicate , are shown in the histogram . Error bars are mean ± standard error and p values were determined by two-tailed paired Student’s t-test ( ****p<0 . 0001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 00910 . 7554/eLife . 24419 . 010Figure 3—source data 1 . Raw data for Figure 3E . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 010 Ligand-induced Notch signaling was also investigated in CHO Lec1 Eogt-siRNA cells ( Tashima and Stanley , 2014 ) using the NOTCH1 ICD-responsive TP1-luciferase reporter . Reduction of Eogt correlated with reduced Notch signaling ( Figure 3E ) and transfection of a human EOGT cDNA largely rescued signaling in the Eogt knockdown cells . Taken together , the combined data provide strong evidence that EOGT regulates DLL1-induced Notch signaling . To investigate biological functions of O-GlcNAc on EGF repeats in vivo , Eogt mutant mice were generated . A floxed Neo allele ( EogtflNeo ) with loxP sites flanking exon 10 and a Neo cassette flanked by loxP/FRT sites was constructed ( Figure 4A ) . Southern blotting showed successful homologous recombination in an EogtflNeo mouse ( Figure 4B ) . EogtflNeo mice were crossed to Flp-deleter mice to obtain an EogtF mouse ( Figure 4A ) . Alternatively , EogtflNeo mice were crossed to global Cre-deleter mice to delete exon 10 ( Figure 4C ) . The amino acid sequence encoded by exon 10 constitutes part of the putative catalytic domain , and Cre recombinase-mediated excision of the sequence between the loxP sites would cause a frame-shift mutation , generating a catalytically inactive protein ( Figure 4—figure supplement 1 ) . RT-PCR detected low levels of aberrant transcripts lacking exon 10 ( Figure 4D ) and quantitative RT ( qRT ) -PCR confirmed that Eogt transcripts were low , possibly due to nonsense-mediated mRNA decay ( Figure 4E ) . Therefore , the Eogt gene lacking exon 10 is effectively a null allele . Indeed , immunoblotting confirmed the absence of EOGT protein in the lungs of Eogt−/− mice ( Figure 4F ) . 10 . 7554/eLife . 24419 . 011Figure 4 . Generation of Eogt-targeted mice . ( A ) Schematic drawing of the wild-type mouse Eogt allele ( WT ) , the targeting vector , the floxed allele with the neomycin ( Neo ) -resistance gene ( EogtflNeo ) , the floxed allele without Neo ( EogtF ) , and the deleted allele ( Eogt− ) . Eogt exons ( closed boxes ) ; Neo and diphtheria toxin ( DT3 ) genes served as positive and negative selection markers , respectively ( open boxes ) ; loxP sites ( gray triangles ) ; FRT sites ( open triangles ) ; KpnI ( K ) , HindIII ( H ) , and SalI ( S ) restriction sites . Homologous recombination between the WT allele and the targeting vector generated the EogtflNeo allele . The EogtF and Eogt− alleles were obtained by Flp-mediated and Ayu1-Cre-mediated recombination , respectively . Red lines indicate positions of probes used for Southern blotting . Positions of primers used for genotyping are indicated by closed triangles . ( B ) HindIII- or KpnI/SalI-digested genomic DNA isolated from WT ( +/+ ) or heterozygous floxed neo ( +/flNeo ) ES cells was analyzed by Southern blotting using the short arm or long arm probe , respectively . In addition to signal from the WT allele ( arrow ) , the flNeo mouse shows an additional band corresponding to the recombinant allele ( arrowhead ) . ( C ) Genomic DNA isolated from WT , Eogt+/− , and Eogt−/− mice was subjected to genotyping using 3rdloxFw , 3rdloxRv , and 25307Rv primers . ( D ) Semi-quantitative RT-PCR analysis of total RNA from WT or Eogt−/− brain ECs using primers targeting exons 9 and 11 . Minor transcripts lacking exon 10 were detected in Eogt−/− mice ( arrow ) . Gapdh was amplified as an internal control . ( E ) Quantification of Eogt transcripts in Eogt−/− and WT mouse . qRT-PCR analysis of brain ECs showed a marked decrease in Eogt transcripts . Data are mean ± S . D . from three independent experiments performed in triplicate . Each experiment analyzed pooled total RNA obtained from 10 mice . ( F ) Lack of EOGT protein expression in Eogt−/− mouse . Lung lysates prepared from adult WT or Eogt−/− mice were analyzed in parallel with cell lysates from HEK293T cells overexpressing Eogt . Immunoblotting was performed using anti-EOGT or β-tubulin antibodies . ( G ) Whole-mount in situ hybridization for Eogt in the P5 or P15 retina . Vascular staining of Eogt is evident in Eogt+/− retina . Eogt expression in vascular sprouts is indicated by arrows . ( H ) P5 control Tek-Cre and Tek-Cre EogtF/F retinas were subjected to in situ hybridization . Counter staining with Dylight 594-conjugated IB4 is shown below . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 01110 . 7554/eLife . 24419 . 012Figure 4—source data 1 . Raw data for Figure 4E . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 01210 . 7554/eLife . 24419 . 013Figure 4—figure supplement 1 . Deletion of exon 10 in the Eogt gene causes exon nine to be spliced to exon 11 leading to a frame shift encoding six novel amino acids before a stop codon . Truncated EOGT protein was not detected in lung indicating nonsense-mediated decay of mutant transcripts . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 01310 . 7554/eLife . 24419 . 014Figure 4—figure supplement 2 . Whole-mount in situ hybridization for Eogt . ( A ) In situ hybridization was performed in Eogt+/− or Eogt−/− retinas using anti-sense or sense probes . ( B ) P5 control Tek-Cre and Tek-Cre EogtF/F retinas were subjected to in situ hybridization using anti-sense or sense probes . Scale bars for ( A ) and ( B ) , 250 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 014 Previous studies revealed that Eogt is widely expressed in various tissues in the mouse , although specific cell types expressing Eogt were not identified ( Sakaidani et al . , 2012 ) . We now show by in situ hybridization that Eogt is expressed in endothelial cells ( EC ) of retinal arteries , capillaries and veins at P5 , and mainly in arteries and veins at P15 ( Figure 4G ) . At the P5 vascular front , Eogt signal was apparent in tip cells ( Figure 4G ) . These signals were absent in Eogt−/− retinas and control retinas stained with sense probes ( Figure 4—figure supplement 2A ) . To prove EC expression of Eogt , in situ hybridization was performed in retinas from conditional mutant mice in which Eogt expression was specifically eliminated in ECs . Tek-Cre;EogtF/F retinas exhibited a marked decrease in vascular staining with the Eogt probe ( Figure 4H and Figure 4—figure supplement 2B ) . Consistently , double staining showed that Eogt signal partially overlaps with the signal from isolectin-B4 ( IB4 ) that stains ECs ( Figure 4H ) , confirming Eogt expression in ECs . Residual staining in tissue surrounding the retinal vasculature suggested weaker Eogt expression in neuronal tissues . Eogt−/− mice were obtained at the expected Mendelian ratio of ~25% ( total n = 244 ) , and did not exhibit obvious abnormalities . Since Eogt is highly expressed in ECs , and retinal angiogenesis requires Notch signaling ( Roca and Adams , 2007 ) , we focused our analysis on angiogenesis and vessel formation in the postnatal mouse retina . Retinal angiogenesis is induced just after birth in mice when a single layer of superficial retinal plexus grows from the center toward the periphery until P7 . Although a vascular plexus was formed in P5 Eogt−/− retina , vascular progression toward the periphery was delayed ( Figure 5A ) . Vessel maturation measured by the association of blood vessels with mural cells was compromised , as evident by the reduced length of αSMA-positive vessels in P5 Eogt−/− retinas ( Figure 5A ) . However , the distribution of anti-NG2-positive pericytes was unaffected at P5 in Eogt−/− retinas ( data not shown ) . By contrast , the density and number of branch points were increased in P5 Eogt−/− retinas ( Figure 5B ) . At the vascular front , the number of filopodia was also increased in P5 Eogt−/− retinas ( Figure 5C ) . An increase in vascular branching was also observed in P15 Eogt−/− retinas ( Figure 5D ) . While some variation in retinal angiogenesis was observed in heterozygote Eogt+/− retinas , the phenotype was not autosomal dominant . These data suggest that EOGT is required for optimal retinal vascular development . Moreover , the phenotype exhibited by Eogt−/− retinas is similar to , although weaker than , that observed in retinas of Notch pathway mutants ( Benedito et al . , 2009; Kofler et al . , 2015; Phng and Gerhardt , 2009 ) . 10 . 7554/eLife . 24419 . 015Figure 5 . EOGT regulates retinal angiogenesis . ( A ) Whole-mount images of Eogt+/+ or Eogt−/− P5 retinas stained with IB4 and anti-αSMA antibody . Bars represent 1000 μm . Scatter plots at right show vascular progression per quadrant length , per retina , and αSMA+ vessel length from the optic nerve per retina , normalized to Eogt+/+ retinas ( taken as 1 . 0 ) . Each symbol represents average vessel progression from 3 to 4 quadrants , or the average length from the optic nerve of 3–4 αSMA+ vessels , per retina . For Eogt+/+ , average vascular progression was 0 . 63 ± 0 . 03 , and the average αSMA+ vessel length was 962 ± 58 µm . ( B ) Images of P5 vascular front of Eogt+/+ or Eogt−/− retinas stained with IB4 . Scatter plot at right shows the normalized number of vascular branch points ( 500 × 500 µm field ( n = 3–8 fields per retina ) , N = 11 mice ) . Each symbol represents the average per retina , per mouse . The average Eogt+/+ branch points were 302 ± 27 per retina , taken as 1 . 0 for normalization . ( C ) Images of filopodia emanating from tip cells ( red dots ) at the vascular front . The scatter plot at right shows the number of filopodia in P5 retinas ( 250 × 250 µm field ( n = 4–12 fields per retina ) , N = 6 mice ) . The average filopodia per mm vascular front for Eogt+/+ was 33 ± 2 per mm , taken as 1 . 0 for normalization . ( D ) Images of P15 retinas stained with IB4 . The scatter plot at right shows number of branch points in P15 retinas ( 500 × 500 µm field ( n = 3–8 fields per retina ) , N = 8 mice ) . The average for Eogt+/+ was 74 ± 3 per field . Data were normalized from mean ± standard error; p values were calculated by unpaired two-tailed Student’s TTEST . *p≤0 . 05; **p≤0 . 01; ***p≤0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 01510 . 7554/eLife . 24419 . 016Figure 5—figure supplement 1 . Images of control Tek-Cre and Tek-Cre EogtF/F retinas stained with isolectin B4 ( IB4 ) and quantification of branch points in P5 ( N = 6 mice ) and P15 ( N = 3 mice ) retinas and numbers of filopodia per mm vascular front in P5 retinas ( N = 6 mice ) . Data were normalized from mean ± standard error; p values were calculated by unpaired two-tailed Student’s t-test . *p≤0 . 05; **p≤0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 016 Loss of Notch signaling specifically in ECs also results in an increase in vascular branching ( Hellström et al . , 2007; Leslie et al . , 2007; Lobov et al . , 2007; Ridgway et al . , 2006; Siekmann and Lawson , 2007; Suchting et al . , 2007 ) . To investigate Eogt functions in ECs , retinas with EC-specific deletion of Eogt were examined . In both P5 and P15 retinas , vessel branching was increased in Tek-Cre:EogtF/F retinas compared with control Tek-Cre retinas ( Figure 5—figure supplement 1 ) . Therefore , EOGT in ECs regulates vascular branching during retinal angiogenesis . Interestingly , however , the length of αSMA-positive vessels and vessel progression in P5 Tek-Cre:EogtF/F retinas were not significantly changed compared with Tek-Cre control retinas ( data not shown ) . Since Notch signaling in pericytes and macrophages also regulates retinal angiogenesis ( Kofler et al . , 2015; Outtz et al . , 2011 ) , this result suggests additional Eogt functions in other cell types that were not affected in the EC-specific knockout of Eogt . To investigate whether EOGT interacts with NOTCH1 to control retinal angiogenesis , we determined the effects of decreasing NOTCH1 or Notch signaling on the retinal vascular phenotype in an Eogt-null background . We chose mutant alleles of Notch1 and Rbpj for these analyses because their haploinsufficiency causes AOS , as does mutant EOGT homozygosity . As expected from previous reports ( Hellström et al . , 2007; Kofler et al . , 2015 ) , Notch1 heterozygous retinas showed increased vessel branching at P5 , and a higher number of filopodia ( Figure 6A , B and C ) , similar to that observed in retinas of Eogt-null mice ( Figure 5 ) . These phenotypes were exaggerated in an Eogt−/− background . Enhanced vascular defects in Eogt−/−Notch1+/− compound mutant retinas were also observed in P15 retinas ( Figure 6D and E ) . Similar to Notch1 heterozygotes , Rbpj+/− retinas exhibited elevated vessel branching at P5 , which was further enhanced in Eogt−/−Rbpj+/− compound mutant retinas ( Figure 6A and B ) . These results are consistent with EOGT and Notch signaling acting in the same or parallel pathways . However , heterozygotes expressing the hypomorphic alleles Notch112f and Notch1lbd ( Ge and Stanley , 2008 , 2010 ) did not have a retinal phenotype and yet , they augmented the Eogt−/− phenotype at P5 and P15 , respectively ( Figures 6B , C , D and E ) . The enhancement in the Eogt−/− retinal vascular phenotype in two different Notch1 hypomorphic backgrounds that do not themselves exhibit retinal defects , strongly suggests that the addition of O-GlcNAc by EOGT is important in regulating the Notch signaling pathway . 10 . 7554/eLife . 24419 . 017Figure 6 . Reduced Notch signaling augments the loss of Eogt in retinal angiogenesis . ( A ) Images of vessel density in P5 retinas from Eogt−/− compared to compound mutant mice . ( B ) Scatter plots represent branch point numbers in P5 retinas from Eogt−/− compared with compound mutant mice , normalized to Eogt+/+ mice . The average number of branch points for the Eogt+/+ used to compare mice with an N1− or Rbpj− allele was 237 ± 5 , and for the Eogt+/+ compared to mice with an N12f or N1lbd allele was 380 ± 36 ( 500 X 500 μm field ( n = 3–8 fields per retina ) , N = 2–6 mice ) . ( C ) Scatter plots showing the number of filopodia in P5 compound mutant mice as compared to Eogt−/− mice , normalized to Eogt+/+ mice . Each symbol represents the average number of filopodia per mouse ( 250 × 250 µm field ( n = 4–12 fields per mouse ) , N = 2–4 mice . The average for Eogt+/+ compared to mice with a N1- or Rbpj- allele was 30 ± 1 per mm and for mice with a N112f or N1lbd allele was 36 ± 4 per mm , taken as 1 . 0 for normalization . ( D ) Images of branch points in P15 retinas comparing Eogt−/− to compound mutant mice as indicated . ( E ) Scatter plots shows branch points in compound mutants compared to Eogt−/− P15 retinas , normalized to Eogt+/+ mice . The average of the Eogt+/+ used to compare mice with a with a N1− or Rbpj− allele was 73 ± 3 , and for mice with a N112f or N1lbd allele it was 69 ± 12 ( 500 x 500 µm field ( n = 3–8 fields per retina ) , N = 3–6 mice . Data represent mean ± standard error except for P5 Eogt+/+Notch1+/− filopodia and P5 Eogt+/+Notch1+/lbd branch points and filopodia that are represented as mean ± range; p values were calculated by unpaired two-tailed Student’s t-test . *p≤0 . 05; **p≤0 . 01; ***p≤0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 017 Notch is an essential contributor to vascular integrity ( Hofmann and Iruela-Arispe , 2007 ) . Increased vascular permeability is manifested by extravasation of plasma proteins such as plasminogen , fibrinogen and albumin . Immunostaining of P15 wild-type retina with anti-fibrinogen antibody detected sporadic signals in the vascular area , whereas no extravascular signal was observed . In contrast , Eogt−/− retina exhibited diffuse fibrinogen staining outside blood vessels , and prominent staining on vessel walls ( Figure 7A and B ) , showing the leakage of plasma proteins . Similar to the blood-brain barrier , retinal vessels limit nonspecific transport between circulating blood and neural tissues of the retina ( Campbell et al . , 2009 ) . Accordingly , perfused sulfo-NHS-LC-biotin remained within vessels in control P15 retinas , confirming an intact blood-retina barrier . In contrast , sulfo-NHS-LC-biotin was occasionally detected in extravascular spaces in perfused Eogt−/− retinas ( Figure 7C ) . Taken together , these data suggested a partly impaired vascular integrity in the absence of Eogt . 10 . 7554/eLife . 24419 . 018Figure 7 . Reduced vessel integrity in the Eogt−/− retina . ( A ) Immunostaining with fibrinogen ( green ) and α-SMA ( magenta ) antibodies in P15 wild-type , Eogt−/− , Tek-Cre , Tek-Cre:EogtF/F , Notch1+/− , and Rbpj+/− retinas . Arrows indicate fibrinogen staining outside vessels stained by IB4 ( white ) . Three-dimensional images were constructed from confocal images by maximum intensity projection . ( B ) Higher magnification three-dimensional images of Eogt−/− retina constructed from confocal images using the Alpha-blend method . Below , single channel images showing fibrinogen ( green ) and IB4 ( white ) staining . ( C ) Sulfo-NHS-LC-biotin was perfused into P15 wild-type and Eogt−/− mice and extravasation determined immediately after perfusion by staining with CF488A-conjugated streptavidin ( green ) and Dylight594-conjugated IB4 ( white ) . Three-dimensional reconstructions were created by maximum intensity projection . Enlarged images of boxed area are shown ( right ) . ( D ) Sulfo-NHS-LC-biotin was perfused into P15 wild-type , Eogt−/− , Notch1+/− , Eogt−/−Notch1+/− mice as in ( C ) . Quantification of the number of extravasation sites in 210 × 210 μm squares ( n = 6 per retina per mouse ) is shown . Note that sulfo-NHS-LC-biotin extravasation in Eogt−/− retina is augmented in compound mutant mice . Data represent mean ± standard error; p values determined by Welch's t test . ***p≤0 . 001 . ( E ) Whole-mount images of wild-type or Eogt−/− P15 retinas stained with IB4 ( cyan ) and anti-αSMA ( magenta ) antibody . ( F ) Whole-mount staining of wild-type and Eogt−/− P15 retinas using IB4 ( white ) together with anti-fibrinogen ( green ) and anti-NG2 ( magenta ) antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 01810 . 7554/eLife . 24419 . 019Figure 7—source data 1 . Raw data for Figure 7D . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 01910 . 7554/eLife . 24419 . 020Figure 7—figure supplement 1 . Sulfo-NHS-LC-biotin was perfused into P15 wild-type , Eogt−/− , Tek-Cre , Tek-Cre;EogtF/F , Notch1+/− , Rbpj+/− , Eogt−/−Notch1+/− , Eogt−/−Rbpj+/− mice and stained with CF488A-conjugated streptavidin ( green ) and Dylight594-conjugated IB4 ( white ) immediately after perfusion . The extravasation sites are shown by arrows . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 020 Similar to Eogt−/− retinas , extravascular fibrinogen staining and extravasation of perfused sulfo-NHS-LC-biotin were observed in Notch1+/− or Rbpj+/− retinas ( Figure 7A and Figure 7—figure supplement 1 ) . Importantly , the combined loss of Eogt and a single Notch1 allele in compound mutant Eogt−/−Notch1+/− retinas gave a synergistic increase in extravasation of sulfo-NHS-LC-biotin along vessels ( Figure 7D ) . The synergistic effect was also observed in Eogt−/−Rbpj+/− retinas ( Figure 7D ) . These results suggest that Eogt interacts with the Notch signaling pathway , thereby affecting retinal vascular integrity . It has been reported that Notch activity is required for coverage of mural cells and vascular integrity ( Henshall et al . , 2015; Liu et al . , 2010; Wang et al . , 2014 ) . In particular , lack of Notch3 expression in mural cells results in decreased coverage of arteries with vascular smooth muscle cells ( VSMC ) ( Henshall et al . , 2015; Liu et al . , 2010 ) and capillaries with pericytes ( Kofler et al . , 2015 ) in retinal vasculatures . Unlike Notch3−/− retinas , αSMA staining revealed intact coverage of retinal arteries with VSMC in P15 Eogt−/− mice ( Figure 7E ) . Moreover , apparently normal pericyte investment was observed in P15 Eogt−/− retinas , regardless of the presence or absence of elevated fibrinogen staining ( Figure 7F ) , suggesting that sufficient Notch activity is maintained in mural cells . In contrast , inactivation of Eogt in ECs in P15 Tek-Cre:EogtF/F retinas showed aberrant fibrinogen staining and sulfo-NHS-LC-biotin extravasation ( Figure 7A and Figure 7—figure supplement 1 ) . Taken together , these results show that Eogt functions in ECs contribute to the integrity of retinal vessels by acting on Notch signaling activity . To investigate whether reduced EOGT directly impacts Notch target genes in ECs , brain ECs were isolated using anti-CD31 beads and gene expression was analyzed by qRT-PCR . Eogt−/− ECs showed reduced expression of Notch pathway targets ( Shawber et al . , 2003 ) , including Hes1 and Hey1 ( Figure 8A ) . Given that Notch signaling positively regulates Dll4 expression ( Sacilotto et al . , 2013 ) ( Figure 8—figure supplement 1 ) , loss of EOGT would disrupt a positive feedback loop sustaining Dll4 expression and Notch signaling . In fact , Dll4 expression was markedly decreased in Eogt−/− ECs , whereas expression of Jag1 and Notch receptors was maintained ( Figure 8A ) . Previous reports showed that Notch signaling represses expression of vascular endothelial growth factor receptors ( VEGFR ) in ECs ( Ehling et al . , 2013; Jakobsson et al . , 2010; Suchting et al . , 2007; Tammela et al . , 2008 ) , which have crucial roles in the formation , function and maintenance of the vasculature ( Simons et al . , 2016 ) . In Eogt−/− ECs , expression of Vegfr2 and Vegfr3 , but not Vegfr1 , was slightly increased compared with wild-type ECs . In cerebral ECs , Notch and TGF-β signaling regulate the expression of N-cadherin , which is associated with maintaining vascular integrity ( Li et al . , 2011 ) . Similarly , N-cadherin expression was relatively decreased in Eogt−/− ECs ( Figure 8A ) . 10 . 7554/eLife . 24419 . 021Figure 8 . EOGT acts on Notch receptors to regulate ligand-induced Notch signaling in ECs . ( A ) Relative mRNA expression in purified brain EC cells . WT ( gray ) and Eogt−/− ( dark gray ) EC cells isolated from cerebrum using anti-CD31 antibody were analyzed for gene expression related to the Notch signaling pathway . Gene transcript levels were normalized to Gapdh . Data are mean ± S . D . from three independent experiments performed in triplicate . Each experiment analyzed pooled total RNA obtained from 10 mice . *p<0 . 05; **p<0 . 01 ( Welch's t test ) . ( B ) qRT-PCR analysis of Hes1 and Hey1 in wild type or Eogt−/− lung EC cells following stimulation with immobilized DLL4-Fc or JAG1-Fc . Gene transcripts were normalized to Gapdh . Data are mean ± S . D . from three independent experiments performed in triplicate . Each experiment analyzed total RNA obtained from a single mouse . Significance determined by Welch's t-test . *p<0 . 05; **p<0 . 01; #p<0 . 05; ##p<0 . 01 compared with the left-most Hes1 ( * ) or Hey1 ( # ) histogram . ( C ) Cells were transfected to express Dll4 or Jag1 and subjected to flow cytometry using DLL4 or JAG1 antibody , respectively . Comparison with mock-transfected cells indicated the amount of exogenously expressed DLL4 and JAG1 on the cell surface in wild-type or EOGT-null HEK293T cells . Mock-transfected HEK293T cells were labeled without primary antibody ( Cont . ) . ( D ) Dll4/Jag1-transfected cells or Mock transfectants ( signal-sending cells ) were co-cultured with wild type or Eogt−/− lung EC cells ( signal-receiving cells ) . qRT-PCR analysis of mouse Hey1 expression suggested that EOGT is required for the ability to receive DLL4-mediated Notch signaling , and that EOGT is dispensable for DLL4 and JAG1 as inducers of Notch signaling . Gene transcript levels were normalized to PECAM1 ( CD31 ) . Data are mean ± S . D . from three independent experiments performed in triplicate . Significance determined by Welch's t-test . ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 02110 . 7554/eLife . 24419 . 022Figure 8—source data 1 . Raw data for Figure 8A , B , D . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 02210 . 7554/eLife . 24419 . 023Figure 8—figure supplement 1 . Dll4 expression is suppressed by inhibiting Notch signaling in ECs . Cultured lung ECs from wild-type or Eogt−/− lungs were treated with DAPT ( 2 μM ) or DMSO for 16 hr . Dll4 expression was analyzed by qRT-PCR . Gene transcripts were normalized to Gapdh . Data are mean ± S . D . from three independent experiments performed in triplicate . Each experiment analyzed total RNA obtained from a single mouse . Significance was calculated using Welch's t test . **p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 023 To further investigate effects of EOGT on Notch target gene expression in ECs , DLL4-Fc or JAG1-Fc were coated on culture dishes and incubated with ECs derived from wild type or Eogt−/− lung . In unstimulated , control ECs , the expression level of Notch target genes including Hey1 and Hes1 was higher in wild type compared with Eogt-null cells ( Figure 8B ) . Upon DLL4 or JAG1 stimulation , a further increase in Hey1 and Hes1 expression was observed in wild-type ECs ( Figure 8B ) . In contrast , DLL4 failed to induce Hey1 and Hes1 expression in Eogt−/− ECs , whereas JAG1 induced a robust induction of these genes ( Figure 8B ) . Taken together , these data suggest that EOGT acts as a positive regulator for DLL4-induced Notch signaling in EC cells . It has been reported that Notch ligands are also modified with O-glycans , which could affect Notch signaling activity ( Serth et al . , 2015; Thakurdas et al . , 2016 ) . To address the possibility that EOGT regulates Notch ligand function , HEK293T and EOGT-null cells were transfected to express JAG1 or DLL4 for co-culture with wild-type or Eogt−/− ECs . Exogenously expressed DLL4 and JAG1 in EOGT-null cells were expressed at the cell surface similarly to control HEK293T cells ( Figure 8C ) . Thus , Notch ligands are presented on the cell surface in the absence of EOGT . To determine if Notch ligands lacking O-GlcNAc stimulate Notch receptors , control and EOGT-null cells expressing DLL4 or JAG1 ( Sending cells ) were cultured with WT or Eogt-null ECs ( Receiving cells ) , and the expression of Notch target gene Hey1 was determined ( Figure 8D ) . Both DLL4 and JAG1 lacking O-GlcNAc induced upregulation of Hey1 in WT ECs similarly to controls . Thus , neither ligand needs O-GlcNAc to induce Notch signaling . Hey1 expression was reduced in Eogt−/− ECs that also failed to respond to DLL4 sending cells . By contrast , co-culture of Eogt−/− ECs with JAG1 sending cells induced mouse Hey1 expression to a level comparable with wild-type ECs ( Figure 8D ) . Thus , Notch receptors in Eogt-null cells responded similarly to controls to JAG1 sending cells . These results suggest that EOGT is dispensable for DLL4 and JAG1 as inducers of Notch signaling , and that O-GlcNAc on Notch receptors is required for optimal DLL- but not JAG-induced Notch activation .
Here , we show that O-GlcNAc on Notch EGF repeats is a functional modification of Notch receptors that regulates Notch signaling and is required for optimal stimulation by DLL ligands . DLL4 constitutes the critical Notch ligand in regulating angiogenesis and vascular development ( Benedito et al . , 2009; Hellström et al . , 2007 ) . By contrast , JAG1 antagonizes Notch signaling during vasculogenesis ( Benedito et al . , 2009; Pedrosa et al . , 2015 ) . We show here that EOGT-catalyzed O-GlcNAc modification potentiates DLL1- and DLL4-mediated NOTCH1 signaling in co-culture assays and is required in vivo during retinal angiogenesis and the maintenance of retinal vascular integrity in mice . Site-directed mutagenesis of O-GlcNAc consensus sites in mouse NOTCH1 revealed that O-GlcNAc on EGF repeats located outside the canonical ligand-binding region affects DLL4-NOTCH1 binding . This O-GlcNAc in one or more EGF repeats may function directly in the binding of DLL4 . Alternatively , O-GlcNAcylation may induce a conformational change of NOTCH1 that promotes DLL4 binding . Future structural analysis of O-GlcNAcylated NOTCH1 in complex with DLL4 will reveal how O-GlcNAc potentiates DLL4-NOTCH1 interactions . Such a study revealed that DLL4 makes specific contacts with the O-fucose glycan in the ligand binding domain ( EGF11 and 12 ) of NOTCH1 ( Luca et al . , 2015 ) . O-fucose glycans on EGF12 affect both Delta and Jagged ligand-induced signaling as evidenced by the EGF12 O-fucosylation site mutant of NOTCH1 which has reduced signaling in response to DLL1 and JAG1 ligands ( Rampal et al . , 2005; Shi et al . , 2007 ) . An O-GlcNAc consensus site is present in EGF11 , but not EGF12 in mammals , but the EGF11 site is not conserved among vertebrates . The ability of EOGT to potentiate Delta ligands is similar to that of Fringe , a glycosyltransferase that transfers GlcNAc to O-fucose on Notch receptors ( Brückner et al . , 2000; Moloney et al . , 2000 ) . In mammals , three Fringe genes , Mfng , Lfng , and Rfng , are expressed in endothelial cells ( Benedito et al . , 2009 ) . Lfng−/− mice exhibit enhanced sprouting and increased density of the vascular area in retinal angiogenesis ( Benedito et al . , 2009 ) , a similarly mild phenotype to that in Eogt−/− retina . Thus , Fringe and Eogt might play complementary functions during retinal angiogenesis . However , Mfng inhibits JAG1-Notch signaling ( Benedito et al . , 2009 ) , whereas we found no effect on JAG1-induced Notch signaling in assays using Eogt-deficient or Eogt-overexpressing cells . Thus , the distinct effects of the different GlcNAc modifications of Notch receptors indicate that EOGT and Fringe glycosyltransferases possess different physiological functions . Indeed , we observed no obvious abnormality in the skeleton ( data not shown ) of Eogt-null mice , making them phenotypically distinguishable from Lfng−/− mice ( Evrard et al . , 1998; Zhang and Gridley , 1998 ) . The phenotypic similarity and genetic interaction between Eogt and Notch1 in retinas suggest that decreased Notch signaling caused by Eogt mutation is mediated by impaired O-GlcNAcylation of Notch1 . However , we cannot formally exclude the possibility of involvement of other Notch receptors as EOGT substrates . In fact , all Notch receptors are potentially modified with O-GlcNAc and NOTCH2 was reported to be O-GlcNAcylated in mouse cerebrocortical tissue ( Alfaro et al . , 2012 ) . Although roles for Notch2 in vascular morphogenesis are not reported , Notch1 and Notch4 have partially overlapping roles ( Krebs et al . , 2000 ) . However , a recent report revealed that DLL4 and JAG1 may fail to activate NOTCH4 ( James et al . , 2014 ) . A detailed analysis will be required to address the contribution of other Notch receptors to EOGT-dependent regulation of Notch signaling . The GlcNAc transferred by Fringe or EOGT can be extended with galactose ( Chen et al . , 2001; Sakaidani et al . , 2012 ) . Interestingly , inactivation of galactosyltransferase B4galt1 in Lec20 CHO cells abrogated the effects of LFNG on Notch signaling ( Hou et al . , 2012 ) . Consistent with roles for galactose in Notch regulation , embryos lacking B4galt1 exhibit subtle Notch signaling defects during somitogenesis ( Chen et al . , 2006 ) . It would be of interest to investigate roles for galactose in retinal angiogenesis and vascular integrity . Our prediction of O-GlcNAc sites on Notch receptors relied on the consensus sequence C5XXG ( Y/F/L ) ( T/S ) GX2-3C6 derived from previously reported O-GlcNAcylated sequences ( Alfaro et al . , 2012; Sakaidani et al . , 2011 ) . In the course of preparation of this manuscript , a study in Drosophila Notch revealed that a Y/F/L residue at the −1 site is not absolutely required for the modification ( Harvey et al . , 2016 ) . Therefore , C5XXGX ( T/S ) GX2-3C6 can be proposed as a broad consensus site . Nonetheless , removing only four modification sites ( EGF2 , 10 , 17 , 20 ) out of 11 C5XXG ( Y/F ) ( T/S ) GXXC6 sequences drastically decreased the O-GlcNAc level on NOTCH1 ( Figure 2 ) . These results suggest that preferred O-GlcNAcylation requires additional constraints . Consistent with this view , five domains ( EGF4 , 11 , 12 , 14 and 20 ) out of 15 Drosophila Notch EGF repeats containing the C5XXG ( Y/F ) ( T/S ) GXXC6 consensus were found to be major O-GlcNAcylation sites in embryos ( Harvey et al . , 2016 ) ( Figure 2—figure supplement 1 ) . It is also noteworthy that Drosophila Eogt mutants fail to display an obvious loss of Notch signaling phenotypes ( Müller et al . , 2013; Sakaidani et al . , 2011 ) . Nevertheless , Notch pathway mutations were found to suppress the wing blistering phenotype in Eogt knockdown flies ( Müller et al . , 2013 ) . Thus , Eogt may affect Notch signaling in a tissue-specific manner . However , O-GlcNAc consensus sites on EGF repeats are poorly conserved between Drosophila Notch and mammalian NOTCH1 . Moreover , the elongation of O-GlcNAc by galactose is not observed in Drosophila embryos ( Harvey et al . , 2016 ) . It would be of interest to determine whether the O-GlcNAc Notch-ligand interaction we report here is conserved in Drosophila . The combined data on roles for O-glycans in Notch signaling suggest that each of the responsible glycosyltransferases may have different effects on Notch receptor functions . Thus , POFUT1 ( Ofut1 in Drosophila ) and POGLUT1 ( Rumi in Drosophila ) , which catalyze O-fucosylation and O-glucosylation , respectively , regulate the strength of Notch activation . Genetic inactivation of Pofut1/ofut1 in Drosophila results in diminished ligand binding to Notch and Poglut1/rumi mutation results in impaired Notch processing ( Haltom and Jafar-Nejad , 2015; Stanley and Okajima , 2010 ) . Subsequent additions to O-fucose and O-glucose regulate subcellular localization or change the affinity toward Notch ligands . In Drosophila , xylose on O-glucose negatively regulates Notch signaling by reducing cell surface expression of Notch ( Lee et al . , 2013 ) . The distinct effects of O-glycans on Notch receptors , coupled with differential expression of the relevant glycosyltransferases explain why different human diseases arise from mutation of POFUT1 ( Dowling-Degos disease ) , LFNG ( Spondylocostal Dysostosis ) and EOGT ( AOS ) ( Basmanav et al . , 2014 ) . Unexpectedly , Eogt-null mice do not exhibit abnormalities predicted from the symptoms of AOS patients . This was also the case in Eogt-null mice homozygous for Eogt gene disruption by a neo cassette ( H . Yagi and K . Kato , personal communication ) . However , the vascular defects observed in Eogt−/− mice are consistent with the view that small-vessel vasculopathy may be a basis for pathologies in AOS .
Antibodies used in microscopy , flow cytometry and Western blot experiments: biotinylated isolectin B4 ( IB4; Vector [B-1105] ) , Cy3-conjugated anti–α-smooth muscle actin ( αSMA ) antibody ( clone 1A4; Sigma [C6198] or fluorescein isothiocyanate ( FITC ) -conjugated anti-αSMA antibody ( clone 1A4; Sigma [F3777] ) , rabbit anti-human EOGT antibody ( Sigma [HPA019460] ) , mouse anti-O-GlcNAc antibody ( CTD110 . 6; Thermo Scientific [24565] or Sigma [07764] ) ( Comer et al . , 2001 ) , hamster anti-mouse NOTCH1 ECD antibody ( 8G10 , Santa Cruz [sc-32756] ) , sheep anti-hamster NOTCH1 ECD antibody ( R and D Systems , AF5267 ) , rabbit anti-human NOTCH1 ECD antibody ( H-131 , Santa Cruz [sc-9170] ) , rabbit anti-human NOTCH1 ICD antibody ( D6F11 , Cell signaling ) , rabbit anti-mouse activated NOTCH1 ( Val1744 , Cell Signaling Technology [4147] ) ( Huppert et al . , 2000 ) , sheep anti-BiP antibody ( BD Biosciences [51–9001980] ) , rabbit anti-NG2 chondroitin sulfate antibody ( Millipore [AB5320] ) , rat anti-NG2 chondroitin sulfate antibody ( R and D Systems [MAB6689] ) , mouse anti-Myc antibody ( 4A6 , Upstate [05-724] ) , hamster anti-mouse JAG1 antibody ( HMJ1-29; Biolegend [130902] ) , goat anti-mouse DLL4 antibody ( R and D Systems [AF1389] ) , rabbit anti-fibrinogen antibody ( Dako [A0080] ) , FITC-conjugated anti-hamster IgG antibody ( Cappel [55400] ) , CF488A-conjugated anti-goat IgG antibody ( Sigma [SAB4600032] ) , Dylight488–conjugated anti-rabbit IgG ( Vector Labs [DI-1488] ) , CF488A-conjugated streptavidin ( Biotium [29034] ) , CF640R-conjugated anti-rat IgG ( Sigma [SAB4600156] ) , Rhodamine Red-X-conjugated donkey anti-sheep IgG ( Jackson ImmunoResearch ) , Dylight649-conjugated streptavidin ( Vector Labs [SA5649] ) , horseradish peroxidase ( HRP ) -conjugated goat anti-mouse IgM ( Thermo Scientific [31444] ) , HRP-conjugated goat anti-rabbit IgG ( Invitrogen [65–6120] ) , HRP-conjugated goat anti-Armenian hamster IgG ( Santa Cruz [sc-2443] ) , HRP-conjugated horse anti-mouse IgG antibody ( Cell signaling [7076S] ) , alkaline phosphatase ( AP ) -conjugated anti-FITC antibody ( Roche [11-426-338-910] ) , FITC-conjugated IB4 ( Vector Labs [FL-1201] ) , and Dylight594-conjugated IB4 ( Vector Labs [DL-1178] ) . A new antibody was raised in rabbit against mouse EOGT lacking the first 19 amino acids ( aa 20–527 ) , produced in E . coli . For the purification of IgG , the rabbit serum was subjected to affinity chromatography on Protein A–Sepharose CL-4B ( GE Healthcare Life Science ) . A mouse Notch1 expression plasmid ( pTracer-CMV/Notch1 ) encoding full-length native Notch1 ( nucleotide numbers from −82 to 9378 ) was obtained from S . Chiba . For generating in situ hybridization probes , full length mouse Eogt cDNA created by PCR was cloned into pBluscript II SK ( - ) vector . For transfection of Eogt , the cDNA was cloned into pSectag2-IRES-GFP and pEF1 vectors ( Sakaidani et al . , 2012 ) . A human EOGT cDNA in pCR3 . 1 was previously described ( Tashima and Stanley , 2014 ) . Notch1 with O-GlcNAc site mutations were generated by the Site-Directed Mutagenesis Kit ( Stratagene [200518] ) to introduce the following mutations: 280T→G ( S94A in EGF2 ) ; 1213A→G ( T405A in EGF10 ) ; 2011A→G ( T671A in EGF17 ) ; 2350AG→GC ( S784A in EGF20 ) . An expression vector for Dll4 ( pSport6-DLL4 ) and Jag1 ( pBOB-Jag1 ) were obtained from DNAFORM ( Clone ID: 4017786 ) or provided by Gerry Weinmaster , respectively . HEK293T cells were kindly provided by Professor Tsukasa Matsuda ( Nagoya University ) and maintained in Dulbecco’s modified Eagle’s medium ( DMEM ) supplemented with 7 . 5% fetal bovine serum ( FBS ) . Transfection into HEK293T cells was performed using polyethylenimine , MW 4000 ( Polysciences [24885] ) . In brief , 4 µg of plasmid DNA was diluted in 200 µl of Opti-MEM I medium ( Gibco [31985] ) , which was then mixed with 12 µg of polyethylenimine and incubated for 30 min . Prior to transfection , HEK293T cells cultured in 6-well dish were pretreated with 800 µl of Opti-MEM I medium and supplemented with the DNA/ polyethylenimine mixture . After 4 hr of incubation , the medium was replaced with 2 ml of 7 . 5% FBS in DMEM containing penicillin/streptomycin followed by incubation for 48 hr before analysis . HeLa and CHO Lec1 cells were maintained in α-MEM containing 10% FBS . Transfection into HeLa and CHO Lec1 cells was performed using X-treme Gene 9 ( Roche [06365779001] ) as described by the manufacturer . Absence of mycoplasma contamination was confirmed upon receipt of the cell lines , and all the experiments were completed within 6 months from the contamination test . HEK293T cells were fixed with 4% paraformaldehyde in PBS for 20 min and ice-cold methanol for 5 min . After washing with PBS , cells were blocked with 5% FBS in PBS for 30 min . Immunostaining was performed overnight at 4°C by incubation with anti-O-GlcNAc mAb CTD110 . 6 ( 1:1000 ) and anti-NOTCH1 ECD Ab 8G10 ( 1:200 ) diluted in 5% FBS/PBS . After washing with cold PBS , cells were incubated with Alexa Fluor 555-conjugated goat anti-Mouse IgM heavy chain antibody ( 2 µg/ml; Invitrogen [A21426] ) and DyLight 488-conjugated goat anti-rabbit IgG antibody ( 3 µg/ml; Vector Labs [DI-1488] ) diluted in 5% FBS/PBS for 2 hr at RT , washed with cold PBS . Cells were mounted on slides using DAPI-Fluoromount-G ( Southern Biotech , [0100–20] ) for observation with a FSX 100 fluorescent microscope ( Olympus ) . Cultured HEK293T cells were detached by pipetting 10 times . After washing with cold PBS , cells were incubated with hamster anti-Notch1 ( 8G10; 4 µg/ml ) antibody , goat anti-DLL4 ( 0 . 1 μg/ml ) , or hamster anti-JAG1 ( 0 . 5 μg/ml ) antibody diluted in 5% FBS in PBS on ice for 30 min . After washing with cold PBS , cells were incubated with FITC-labeled anti-hamster IgG ( 40 µg/ml ) or FITC-labeled anti-goat IgG ( 10 μg/ml ) antibody in 5% FBS in PBS on ice for 30 min , washed with cold PBS , and subjected to flow cytometry using a FACSCalibur cytometer ( Becton Dickinson ) . For immunoprecipitation of mouse NOTCH1 from transfected HEK293T cells , cells were lysed in Cell Lysis Buffer ( Cell signaling [9803S] ) . Cell lysates were incubated with 8G10 Notch1 antibody at the concentration of 2 μg/ml for 2 hr at 4°C , and incubated with Protein G Sepharose ( GE Healthcare [17-0618-01] ) for 2 hr at 4°C . After extensive wash with 50 mM Tris-HCl , pH 7 . 4 , 150 mM NaCl , and 0 . 25% Triton X-100 , NOTCH1 was eluted with SDS-PAGE sample buffer containing 2% SDS and 70 mM 2-mercaptoethanol . Immunoblotting for the detection of O-GlcNAc on NOTCH1 was performed using CTD110 . 6 O-GlcNAc ( 1:2000 ) or 8G10 NOTCH1 ( 1:500 ) antibodies , followed by HRP-conjugated anti-mouse IgM ( 1:4000 ) or anti-hamster IgG antibodies ( 1:1000 ) , respectively , and enhanced chemiluminescence as described previously ( Ogawa et al . , 2015 ) . For purification of NOTCH1 EGF repeats ( Notch1-EGF1-36-MycHis ) from cell lysates , cells were transfected with pSegtag2/Hygro/Notch1-EGF:mycHis ( Hou et al . , 2012 ) alone or together with pSecTag2/Hygro/Eogt . Cell lysates were added with 2 μg of anti-Myc antibody ( 4A6 ) . After incubation for 2 hr at 4°C , Protein G Sepharose was added to the lysates , which were then incubated for 2 hr at 4°C . Immunoprecipitates were washed and eluted with SDS-PAGE sample buffer as described above . Immunodetection was performed using anti-Myc antibody ( 1:1000 ) or CTD110 . 6 ( 1:2000 ) antibodies , followed by HRP-conjugated anti-mouse IgM ( 1:4000 ) or anti-mouse IgG antibodies ( 1:2000 ) , respectively , and enhanced chemiluminescence . BAC clone RP24-388B15 derived from C57BL/6J mice and containing a genomic region that encodes Eogt was obtained from the BACPAC Resource Center . Conditional knockout ( cKO ) targeting vector ( pDT-loxP-loxP-FRT-PGKneo-loxPFRT ) was obtained from Satoru Takahashi ( Tsukuba University ) . Mouse genomic DNA fragments including Eogt exon 10 , a 7 . 8 kb upstream region ( long arm ) , or 1 . 9 kb downstream region ( short arm ) were amplified by PCR with PrimeSTAR Max polymerase ( Takara [R045A] ) . Each fragment was cloned into the PmlI site , PmeI/KpnI sites , or the XhoI site of the cKO vector , respectively , using the In-Fusion Advantage PCR Cloning Kit ( Clontech [639619] ) . All insert sequences were confirmed by DNA sequencing . Eogt-targeting vector DNA was linearized with AscI , and introduced into C57BL/6J ES cells ( Tanimoto et al . , 2008 ) by electroporation . Genomic DNA isolated from G418-resistant clones was subjected to PCR screening to confirm homologous recombination using KOD FX polymerase ( Toyobo [KFX-101] ) , followed by a second screening to confirm the position of the third loxP sequence using LA Taq polymerase ( Takara [RR002A] ) . Selected ES clones were subjected to Southern blotting using long-arm ( 900 bp ) or short-arm ( 1 kb ) probes radiolabelled with [32P]dCTP using the Megaprime Kit ( GE Healthcare [RPN1606] ) . Genomic DNA was digested with KpnI/SalI and hybridized with the long-arm probe , or digested with HindIII and hybridized with the short-arm probe . Two clones were microinjected into blastocysts that were implanted into pseudopregnant female mice in the Tsukuba University transgenic core facility using standard methodologies . Primers used for PCR are shown in Table 1 . 10 . 7554/eLife . 24419 . 024Table 1 . Primers used for cloning DNA fragments , screening ES cells , or genotyping by PCR . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 024Target region Exon10 Fw 5’–ATACGAAGTTATCACCGAACCTAGCCCATATTT–3’ Exon10 Rv2 5’–ACGAAGTTATGTCGACGACTGAGCATTGCTGTT–3’Long arm region Long arm Fw1 5’–CGAATCAAGCTGTTTGGTCCATTCTCTGCTCCA–3’ Long arm Rv 5’–ACGAAGTTATGGTACGGTCAACTTGAAGAAGTA–3’Short arm region Short arm Fw 5’–TAGGAACTTCCTCGAAATTCAGTGCTTAGAAGT–3’ Short arm Rv1 5’–GCGCGCCTTTCTCGAACACTGTGTACAGTGACA–3’Long-arm probe larm16380Fw 5’–CTGCCTCAGCTTCCTGAGTG–3’ larm17196Rv 5’–CATGTCAGATCAGACAGTTC–3’Short-arm probe sarm26294Fw 5’–CTGAGCTATGTACTGGATGC–3’ AscI-sarmRv2 5’–TGAAGAGGCGCGCCCAGAGACAGAAAAAGCAC–3’ES cells first screening PGK S1 5’–CCTCCCCTACCCGGTAGAATTGACC–3’ GL1 typing RV2 5’–GAACTGTCAGATTTGGTGACACAGAAAGGC–3’ES cells second screening 3rdlox Fw 5’–CCACCCGACCCCTGCCAGAACATAATGCTCTCTTGCATC–3’ 3rdlox Rv 5’–GCTGTCGCCAGAGGAGAGAGTGGGTGCTTACTTAC–3’Eogt mice genotyping 3rdloxFw 5’–CCACCCGACCCCTGCCAGAACATAATGCTCTCTTGCATC–3’ 3rdloxRv2 5’–GCTGTCGCCAGAGGAGAGAGTGGGTGCTTACTTAC–3’ 25307Rv 5’–CCAAGGCGGTCTTGGCCCAT–3’RT-PCR for Eogt Exon 9 Fw 5’– AGGCTACACGCAGCTCAATT –3’ Exon 11 Rv 5’– AGAAGCCGTGTTTTCGTTGC –3’qRT-PCR for Eogt Exon 1 Fw 5’–AAGCTGCAGGTCCGTGAAAA–3’ Exon 2 Rv 5’–TAGGTTAGGCTACCGCGTCT–3’Underlined sequences are 15 bp homologous overlaps required for In-Fusion cloning . Eogtfloxed/neo mice were crossed with Ayu1-Cre ( Niwa et al . , 1993 ) or FLPe mice ( B6;SJL-Tg ( ACTFLPe ) 9205Dym/J backcrossed more than nine times to C57BL/6J ) to generate mice carrying floxed or exon 10-deleted Eogt alleles , respectively , in collaboration with Satoru Takahashi ( Moriguchi et al . , 2006 ) . Mice with a floxed Rbpj allele or a Tek-Cre transgene were obtained from T . Honjo ( Han et al . , 2002 ) and M . Yanagisawa ( Kisanuki et al . , 2001 ) , respectively , and provided by the RIKEN BRC through the National Bio-Resource Project of the MEXT , Japan . Rbpj-null mice were generated by mating RbpjF/F mice with Ayu1-Cre deleter mice . Tek-Cre: EogtF/F mice were generated in collaboration with Kenji Uchimura ( Nagoya University ) . Notch1 null mice ( Notch1tm1Con/J ) ( Conlon et al . , 1995 ) were obtained from the Jackson laboratory . Mice with a ligand binding domain mutation in Notch1 ( Notch1lbd ) and mice lacking the O-fucose site in EGF12 of Notch1 ( Notch112f ) were previously described ( Ge and Stanley , 2008 , 2010 ) . Notch112f mice were used after backcrossing >10 times to C57Bl/6 . Backcrossed Notch112f/12f progeny die around mid-gestation ( manuscript in preparation ) . All experimental procedures were performed in accordance with the Guidelines for Animal Experimentation in Nagoya University Graduate School of Medicine and Japanese Government Animal Protection and Management Law ( Permit Number: 26397 ) . Animal experiments at the Albert Einstein College of Medicine were performed with the approval of the Institutional Animal Care and Use Committee ( Permit Number: 20140803 ) . Whole mount staining of mouse retina was performed as reported previously ( Tual-Chalot et al . , 2013 ) , unless otherwise noted . Briefly , eyes were fixed with 4% paraformaldehyde ( PFA ) in phosphate buffered saline without divalent cations ( PBS-CMF ) , pH 7 . 4 at room temperature ( RT ) for 15 min , except for preserving filopodia at the vascular front , eyes were fixed in 4% PFA in PBS-CMF at 4°C for 2 hr . After dissection of retinas in PBS-CMF , flat retinas were prepared by dropping cold methanol . After washing with PBS-CMF , retinas were incubated with Perm/Block solution containing PBS-CMF , pH 7 . 4 , 0 . 3% TritonX-100 , and 0 . 2% bovine serum albumin ( BSA ) supplemented with 5% donkey or goat serum for 1 hr at RT . Immunostaining was performed overnight at 4°C by incubation with biotin-IB4 ( 1:100 or 1:250 ) and Cy3- or FITC-conjugated anti-αSMA diluted in Perm/Block solution ( 1:200 ) , followed by four washes with PBS-CMF containing 0 . 3% TritonX-100 ( PBSTX ) for 10 min at RT , followed by incubation overnight at 4°C in SA-488 ( 1:200 ) or Dylight649-conjugated streptavidin ( 1:250 ) in Perm/Block solution . Alternatively , retinas were directly labeled with Dylight594-conjugated IB4 ( 2 . 5 µg/ml in Perm/Block ) . After four washes with PBSTX for 15 min at RT and a rinse with PBS-CMF , retinas were mounted using prolong diamond anti-fade mounting medium ( Life Sciences , [1664835] ) for observation using a TiE-A1R-KT5 microscope ( Nikon ) or a P250 High Capacity Slide Scanner ( Perkin Elmer [1S10OD019961-01] ) . Whole mount staining for NG2+ pericytes was performed following the procedure described above using rabbit anti-NG2 antibody ( 1:200 ) or rat anti-NG2 antibody ( 1:1000 ) , and DyLight488-conjugated anti-rabbit IgG ( 1:200 ) or CF640R-conjugated anti-rat IgG ( 1:200 ) as primary and secondary antibodies , respectively . Whole mount staining of extravasated fibrinogen was performed as described above using rabbit anti-fibrinogen antibody ( 1:200 ) as a primary antibody and DyLight488-conjugated anti-rabbit IgG ( 1:200 ) as a secondary antibody . The length of αSMA+ blood vessels was calculated by determining the average of the 3–4 longest αSMA+ vessels per retina . The number of vascular branch points was measured by analyzing 3–8 fields of 500 × 500 µm per retina which were chosen randomly to cover the region under the vascular front for P5 , or central regions for P15 retinas . AngioTool ( Zudaire et al . , 2011 ) was used to calculate branch points . For counting the number of filopodia , 4–12 random fields of 250 × 250 µm per retina at the vascular front , and the number of filopodia were counted by eye and marked with a red dot in NIH ImageJ ( Schneider et al . , 2012 ) . Eyes from newborn mice were fixed with 4% PFA/PBS for 15 min at RT . Retinas were dissected in 4% PFA in PBS . Flat retinas were prepared by dropping cold methanol onto the retina and stored in methanol at −20°C . In situ hybridization was performed as described previously ( Powner et al . , 2012 ) with modifications . In brief , flat retinas were washed with PBS , incubated with 1 µl of FITC-labeled probe in diluted in 1 ml of hybridization buffer overnight at 65°C , washed with formamide wash buffer for 15 min 3 times at 65°C , followed by PBS containing 0 . 1% Tween-20 for 20 min , twice at RT . Retinas were incubated in 500 µl IHC blocking buffer ( 3% Triton X-100 , 0 . 5% Tween-20 , 1% FBS in 2x PBS ) for 20 min at RT , followed by AP-conjugated anti-FITC antibody ( 1:200 dilution ) in IHC blocking buffer . Counter staining was performed by incubating retinas in Dylight 594-conjugated IB4 ( 2 . 5 µg/ml ) in IHC blocking buffer . After incubation overnight at 4°C on a slow shaker , retinas were washed with PBS containing 0 . 1% Tween-20 ( PTW ) for 20 min twice at RT , and incubated with BCIP/NBT solution ( Roche [11681460001] ) at RT . Light field images were acquired using FSX100 ( Olympus ) . Deeply anesthetized P15 mice were perfused with 0 . 75 µg/g ( mouse weight ) of Sulfo-NHS-LC-biotin ( Thermo Fisher Scientific [21335] ) dissolved in 10 ml PBS-CMF for 10 min , followed immediately by 10% formalin in sodium phosphate , pH7 . 4 . Eyes were post-fixed in 4% PFA in PBS for 15 min on ice , washed with PBS , and soaked in 2xPBS for 15 min on ice . Dissected retina were flattened by dropping cold methanol and kept at −20°C . The flat retinas were washed with PBS , incubated with Perm/Block solution including 5% goat serum prior to immunostaining with CF488A-conjugated streptavidin ( 10 µg/ml ) and Dylight594-conjugated IB4 ( 2 . 5 µg/ml ) for 4 hr at RT or overnight at 4°C . After four washes with PBS , retinas were mounted and observed using confocal microscope A1R-TiE ( Nikon ) . Brain microvascular ECs were isolated as described previously ( Ruck et al . , 2014 ) . In brief , 10 mice at P14 were sacrificed , and brains were isolated and transferred to 5 ml of PBS . Cerebellum and brainstem were removed with forceps . Meninges were detached by rolling the brains on sterile blotting paper . Meninges-free brains were transferred to a 50 ml Falcon tube filled with 13 . 5 ml of DMEM , minced first with a 25 ml pipette , then with a 10 ml pipette until the medium became milky . Tissue homogenates were digested by adding 0 . 6 ml of 10 mg/ml collagenase type 2 ( Worthington [4174] ) in DMEM and 0 . 2 ml of 1 mg/ml DNase I ( Boehringer Mannheim [104159] ) in PBS for 1 hr at 37°C on an orbital shaker at 180 rpm . After digestion , 10 ml of DMEM was added and the tissue suspension was centrifuged at 1000 x g for 10 min at 4°C . The pellet was resuspended using a 25 ml pipette in 25 ml of 20% ( w/v ) BSA in DMEM approximately 25 times , and centrifuged at 1000 x g for 20 min at 4°C . The pellet was resuspended in 9 ml of DMEM and supplemented with 1 ml of 10 mg/ml collagenase type 2 and 0 . 1 ml of 1 mg/ml DNase . After digestion for 1 hr at 37°C , 15 ml of 20% FBS in DMEM containing penicillin/streptomycin was added to stop digestion . After centrifugation at 400x g for 5 min , the pellet was resuspended with 3 ml of 0 . 1% BSA in PBS , mixed with 22 . 5 µl of Dynabeads Biotin Binder ( Veritas [11047] ) precoated with 5 µg of biotin-labeled anti-CD31 antibody ( clone 390; Biolegend [102404] ) . The mixture was incubated with rotation at RT for 15 min . Beads with bound ECs were collected using a magnet , washed with PBS , and subjected to total RNA isolation using PureLink RNA Mini Kit ( Ambion [12183020] ) . Quantitative RT ( qRT ) -PCR was performed using M-MLV Reverse Transcriptase ( Invitrogen [28025–013] ) for reverse transcription , and SsoAdvanced Universal SYBR Green Supermix ( Biorad [172–5271] ) for real-time PCR analysis according to the manufacturer's instructions . Primers used for PCR are shown in Table 2 . 10 . 7554/eLife . 24419 . 025Table 2 . Primers used for qRT-PCR . DOI: http://dx . doi . org/10 . 7554/eLife . 24419 . 025TargetForwardReverseNotch1AGTGTGACCCAGACCTTGTGAAGTGGCTGGAAAGGGACTTGNotch2CCCAAGGACTGAGAGTCAGGGGCAGCGGCAGGAATAGTGANotch3ATTTGAGGGGTGCTGAAGTGGAAGGCTGGGACAGAGAGAANotch4TCCGGACTTTTAAGGCCAAATTCCCATTGCTGTGCATACTCTDll4CCCACAATGGCTGTCGTCATAACCCTTTGGCCCACTGTTGN-cadherinAGGGTGGACGTCATTGTAGCCTGTTGGGGTCTGTCAGGATCD31AGCCAACAGCCATTACGGTTAAGCCTTCCGTTCTCTTGGTGGAPDHGGTGCTGAGTATGTCGTGGACCTTCCACCATGCCAAAGTTJag1TCTCTGACCCCTGCCATAACTTGAATCCATTCACCAGATCCHes1ACACCGGACAAACCAAAGACCGCCTCTTCTCCATGATAGGHey1CATGAAGAGAGCTCACCCAGACGCCGAACTCAAGTTTCCHey1 ( mouse specific ) TGAATCCAGATGACCAGCTACTGTTACTTTCAGACTCCGATCGCTTACVegfr1ACATTGGTGGTGGCTGACTCTCCCTCTCCTTCGGCTGGCATCVegfr2GCGGGCTCCTGACTACACCCAAATGCTCCACCAACTCTGVegfr3CCGCAAGTGCATTCACAGAGTCGGACATAGTCGGGGTCTT Lung microvascular ECs were isolated as described previously ( Ruck et al . , 2014 ) unless otherwise noted . In brief , mice at 3 months were sacrificed , and lungs were minced and digested with 1 mg/ ml of collagenase/dispase ( Roche [10 269 638 001] ) in DMEM for 45 min at 37°C . A single cell suspension was passed through a 70 μm cell strainer , collected by centrifugation , and resuspended in 1 ml 0 . 1% BSA/PBS . ECs were isolated using Dynabeads Biotin Binder precoated with biotin-labeled anti-CD31 antibody ( clone 390; Biolegend [102404] ) . Beads were resuspended in HuMedia-EG2 ( Kurabou , [KE-2150S] ) containing 2% FCS and 10 ng/ml hEGF , 5 ng/ml hFGF-b , 1 . 34 μg/ml hydrocortisone hemisuccinate , 10 μg/ml heparin , 50 μg/ml gentamycin , 50 ng/ml Amphotericin B , and then , plated on CellBIND Surface Culture Dish ( Corning , [3295] ) . After reaching 70–80% confluent , cells were detached using 0 . 05% Trypsin/EDTA , collected by centrifugation , and resuspended in 1 ml 0 . 1% BSA/PBS . ECs were subsequently purified using Dynabeads Biotin Binder precoated with biotin-labeled anti-CD102 antibody ( clone 3C4; Biolegend [105604] ) . Beads were resuspended with complete HuMedia-EG2 media , and cultured onto CellBIND Surface Culture Dish . | As an embryo develops , its cells undergo a series of divisions and transformations until they specialize to form distinct cell types . These transformations are regulated by many molecules and signaling pathways . Often one cell will release signaling molecules that are then recognized by specialized receptor proteins on other cells that then trigger a reaction in the receiving cell . A signaling molecule that binds to a receptor is often referred to as a ligand . Specific chains of sugar molecules known as glycans are attached to receptor proteins and help to regulate the signaling pathways that control how cells develop in an embryo . One of these receptors is called the Notch receptor , which carries several types of glycans . Mutations that prevent the glycans from being attached to this receptor can lead to congenital diseases . For example , EOGT is an enzyme that attaches a sugar molecule onto the Notch receptor , and mutations in the gene that encodes EOGT are found in people with Adams-Oliver syndrome – a rare condition that causes their skin and limbs to fail to develop properly . However , it was not clear what exact role the attachment of the sugar molecule ( referred to as O-GlcNAc ) plays and if it regulates Notch signaling . Sawaguchi et al . studied animal cells grown in the laboratory to investigate ligand binding to Notch receptors and Notch signaling . The experiments showed that if the gene for the EOGT enzyme was deleted , some ligand molecules could not bind well to the Notch signaling receptor . This affected how the receptor became activated and how it triggered a signaling reaction in the receiving cell . In mutant mice that lacked EOGT , the blood vessels of the developing retina had reduced Notch signaling . Blood vessels are important for the normal development of the retina . Together , the results suggest that O-GlcNAc on Notch receptors are important for specific aspects of Notch signaling , including the signals that lead to the normal development of retinal blood vessels . A next step will be to identify other roles of the O-GlcNAc glycans on Notch receptors and to explore how they affect Notch signaling during development . A better understanding of these mechanisms will extend our knowledge of Adams-Oliver syndrome , which at the moment remains a poorly understood condition . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"biochemistry",
"and",
"chemical",
"biology"
] | 2017 | O-GlcNAc on NOTCH1 EGF repeats regulates ligand-induced Notch signaling and vascular development in mammals |
Mitochondrial calcium ( Ca2+ ) import is a well-described phenomenon regulating cell survival and ATP production . Of multiple pathways allowing such entry , the mitochondrial Ca2+ uniporter is a highly Ca2+-selective channel complex encoded by several recently-discovered genes . However , the identity of the pore-forming subunit remains to be established , since knockdown of all the candidate uniporter genes inhibit Ca2+ uptake in imaging assays , and reconstitution experiments have been equivocal . To definitively identify the channel , we use whole-mitoplast voltage-clamping , the technique that originally established the uniporter as a Ca2+ channel . We show that RNAi-mediated knockdown of the mitochondrial calcium uniporter ( MCU ) gene reduces mitochondrial Ca2+ current ( IMiCa ) , whereas overexpression increases it . Additionally , a classic feature of IMiCa , its sensitivity to ruthenium red inhibition , can be abolished by a point mutation in the putative pore domain without altering current magnitude . These analyses establish that MCU encodes the pore-forming subunit of the uniporter channel .
Since the initial demonstration that mitochondria take up substantial amounts of cytoplasmic Ca2+ ( Deluca and Engstrom , 1961 ) , detailed studies have revealed that this uptake can sculpt the cytoplasmic Ca2+ transient ( Wheeler et al . , 2012 ) , enhance ATP synthesis ( Balaban , 2009 ) , and trigger cell death ( Zoratti and Szabo , 1995 ) . Of several pathways for Ca2+ entry , a uniporter found in the inner membrane possesses the largest capacity for uptake and was shown to be a highly Ca2+-selective ion channel ( Kirichok et al . , 2004 ) . However , despite this substantial progress , the identities of the genes encoding the functional uniporter were largely unknown until only recently . In the past several years , investigators from several laboratories have identified mitochondrial calcium uptake 1 ( MICU1 , formerly Cbara1 ) ( Perocchi et al . , 2010 ) , mitochondrial calcium uniporter ( MCU , formerly Ccdc109a ) ( Baughman et al . , 2011; De Stefani et al . , 2011 ) , and mitochondrial calcium uniporter regulator 1 ( MCUR1 , formerly Ccdc90a ) ( Mallilankaraman et al . , 2012 ) as potentially integral components of this channel . These discoveries have spurred an explosion of research into mitochondrial Ca2+ transport , yet an outstanding question in this field is which , if any , of these several genes forms the pore subunit of the channel . So far , the discovery of each of these genes occurred via Ca2+ imaging assays following RNAi-mediated knockdown . Although extremely useful for screening , this technique is ill-suited for separating modulatory effects from altered expression of the channel-forming subunit itself , and each of these genes was found to inhibit Ca2+ uptake via this assay . This difficulty arises because Ca2+ imaging measures Ca2+ entry , which depends critically on ( i ) the number of open portals , ( ii ) the voltage gradient , and ( iii ) the Ca2+ concentration gradient driving flux , and none of these is controlled during imaging experiments . In fact , changes in Ca2+ buffers or efflux pathways can alter matrix Ca2+ , and changes in pH or other divalents can alter indicator fluorescence , without reflecting a true difference in uniporter activity . Rather , the ideal method for assaying electrogenic ion transport is voltage-clamping , which allows precise control of the voltage and concentration gradient and thus accurate measurement of the changes in channel density . Initial attempts to apply voltage-clamping focused on MCU , which possesses two transmembrane domains and highly-conserved acidic residues in a putative pore domain ( Baughman et al . , 2011; De Stefani et al . , 2011; Bick et al . , 2012 ) . However , although MCU peptides reconstituted into lipid bilayers showed some activity in single-channel recordings ( De Stefani et al . , 2011 ) , these experiments failed to recapitulate the typical single-channel behavior of the uniporter , most importantly its characteristic conductance and long-lasting openings ( 99% open probability ) ( Kirichok et al . , 2004 ) . Although such a discrepancy may be explained by the incorporation of the channel into bilayers in the absence of regulatory subunits , lipid bilayer reconstitution is also notorious for false positive findings . Since it detects the activity of single molecules , channel-like behavior can be measured from trace contaminants after even crystallography-grade purification ( Accardi et al . , 2004 ) . Thus , the identity of the pore-forming subunit remains to be definitively established . To overcome the limitations of single-channel recordings , we chose to voltage-clamp whole mitoplasts ( mitochondria stripped of their outer membrane and expanded osmotically to allow access to the inner membrane ) ( Figure 1A ) . This allows electrophysiological interrogation of all the channels assembled in their native milieu in individual mitochondria . Moreover , this technique allows us to express mutated channels and examine mitochondrial Ca2+ currents ( IMiCa ) for altered key features . Using this method , we show here that MCU does recapitulate key features of IMiCa , and show that a single point-mutation can generate resistance to pharmacologic inhibition . 10 . 7554/eLife . 00704 . 003Figure 1 . MCU expression recapitulates IMiCa . ( A ) A HEK-293T cell mitoplast under differential interference contrast ( far left ) , showing a typical figure-eight shape . The lobe bounded only by the mitochondrial inner membrane appears less dense ( white arrow ) than the lobe also bounded by the outer membrane ( black arrow ) . Matrix-targeted mCherry demonstrates that the inner membrane forms a surface on both lobes ( middle left ) . GFP-tagged mitofusin-1 ( an outer membrane GTPase , middle right ) reveals the outer membrane partially encapsulating only one lobe . Far right: merged image . ( B ) Western blots demonstrate reduced MCU expression following short-hairpin RNA-mediated knockdown of MCU but not GFP ( control ) . ATP5B is a loading control . ( C ) Exemplar current traces demonstrate endogenous IMiCa in control knockdown ( black trace ) , which is largely blocked by 100 nM RuR ( red trace ) . Voltage ramps from −160 to +80 mV for 750 ms were delivered every 6 s from a holding potential of 0 mV . ( D ) Significantly reduced IMiCa after MCU knockdown . ( E ) Summary data , n = 6–10 per condition , error bars report SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 00704 . 003
Stable transfection of HEK-293T cells with a short-hairpin RNA targeting MCU produced a substantial reduction in the protein when assayed by Western blot ( Figure 1B ) or quantitative real-time polymerase chain reaction ( 17 ± 5% transcripts remaining compared to shGFP ) . We isolated mitoplasts from these cells using the Kirichok protocol ( Fedorenko et al . , 2012; Fieni et al . , 2012; Figure 1A ) . As expected from Ca2+-imaging experiments ( Baughman et al . , 2011 ) , mitoplasts from control cells showed robust IMiCa during voltage ramps from −160 mV to +80 mV ( Figure 1C ) . Because IMiCa features a half-saturation value ( K0 . 5 ) of 20 mM [Ca2+]bath , we maximized current by recording in a 100 mM Ca2+ gluconate bath solution ( Kirichok et al . , 2004 ) . Utilizing high external Ca2+ allows us to conclude that changes observed after modifying MCU expression is due to altered channel levels rather than modulation of K0 . 5 , which might be set by accessory subunits . Other critical features of IMiCa replicated in HEK-293T cells include its strong inward-rectification and high-affinity blockade by ruthenium red ( RuR , 87 ± 2% inhibition in 100 nM RuR , Figure 1C , E ) . Compared to the control condition , IMiCa in mitoplasts from shMCU-expressing cells was markedly smaller ( Figure 1D ) . The RuR-sensitive component of total Ca2+ current was reduced by 78 ± 14% ( p<0 . 001 , Figure 1E ) , with no significant difference in the RuR-insensitive residual component , suggesting that the knockdown was specific to IMiCa and not a generalized reduction in membrane conductance . Moreover , differences were not due to alterations in mitochondrial structure , as mitoplast capacitance , a surrogate for inner membrane surface area ( 100µm2/pF ) , was consistent across all conditions tested here ( shGFP: 0 . 48 ± 0 . 10 pF , shMCU: 0 . 34 ± 0 . 09 pF , p>0 . 05 ) . Next , we examined if overexpression of wild-type or mutant human MCU proteins substantially changed IMiCa . We focused on an MCU mutated to encode an alanine at a highly-conserved serine in the putative pore-forming loop ( S259A ) ( Baughman et al . , 2011; Bick et al . , 2012 ) . The wild-type and mutant MCU proteins were tagged with a carboxyl-terminal FLAG epitope , and localized appropriately to mitochondria ( Figure 2A , B ) . To confirm targeting to the inner membrane , we treated HEK-293T mitochondrial fractions with increasing concentrations of digitonin , and assayed for enzymatic digestion with proteinase K ( Baughman et al . , 2011; Figure 2C , D ) . Since the FLAG-tagged proteins were protected from digestion to a comparable extent as peptidyl-prolyl cis-trans isomerase F ( PPIF , a matrix protein ) , whereas proteins of the outer- and inter-membrane spaces were not , the MCU constructs are appropriately targeted with their carboxyl-termini facing the mitochondrial matrix . Turning to our whole-mitoplast recordings ( Figure 2E , G ) , we found that overexpression of wild-type MCU-Flag produced a robust increase in IMiCa of approximately 3 . 4-fold compared to endogenous HEK-293T currents ( compare to Figure 1C , E ) . This enhanced IMiCa retained its sensitivity to RuR ( Figure 2E , G ) . 10 . 7554/eLife . 00704 . 004Figure 2 . MCU mutants alter IMiCa sensitivity to RuR . ( A ) Confocal imaging of HEK-293T cells for FLAG-tagged MCU ( left ) and cytochrome C oxidase I ( a mitochondrial marker , center ) . The merged image ( right ) shows colocalization . ( B ) As in ( A ) but for MCU-259A-FLAG . ( C ) Immunoblot analysis of proteinase K digestion after wild-type MCU-FLAG-expressing HEK-293T mitochondrial fractions are digitonin-permeabilized at the specified concentrations , confirming mitochondrial targeting . ( D ) Proteinase K digestion analysis for MCU-S259A-FLAG mutant . ( E ) Significant increase in IMiCa after MCU-FLAG overexpression . ( F ) Loss of RuR block in S259A mutant . ( G ) Summary data , n = 5–9 per condition . DOI: http://dx . doi . org/10 . 7554/eLife . 00704 . 004 Finally , we studied the S259A-MCU mutant to see if it disrupted key features of IMiCa . In Ca2+-imaging experiments , this mutant had preserved Ca2+ uptake but diminished sensitivity to uniporter inhibitors ( Baughman et al . , 2011 ) . At baseline , mitoplasts from HEK-293T cells transfected with the S259A mutant had capacitances similar to mitoplasts after wild-type MCU overexpression ( MCU-Flag: 0 . 32 ± 0 . 06 pF , S259A: 0 . 29 ± 0 . 10 pF , p>0 . 05 ) . Moreover , S259A overexpression mirrored the increase in IMiCa seen after wild-type MCU transfection , confirming a fully-functional channel ( Figure 2F , G ) . However , this variant displayed markedly decreased sensitivity to RuR , with minimal inhibition at 100 nM . Since overexpression occurred on a background of endogenous channels , this mutant appears to act in a dominant-negative fashion . In particular , the S295A RuR-inhibited fraction ( 148 ± 33 pA/pF , Figure 2G ) was much less than the RuR-inhibited fraction in endogenous IMiCa ( 372 ± 42 pA/pF , Figure 1F ) . This suggests that , since the channel is an oligomer of multiple MCU subunits ( Baughman et al . , 2011 ) , mutant S259A-MCU can incorporate with endogenous , wild-type subunits to form a Ca2+-conducting channel , which nonetheless has drastically-reduced RuR sensitivity .
In our study , we use the most direct assay available to measure mitochondrial Ca2+ currents—whole-mitoplast electrophysiology ( Kirichok et al . , 2004 ) . By isolating mitoplasts and recording directly from all the channels embedded in the inner membrane , we measure the ensemble current and thus avoid confounding minor contaminants . This method , moreover , is the same technique that originally demonstrated that the mitochondrial Ca2+ uniporter was an ion channel , allowing us to make a direct comparison to the characteristic features of this current ( IMiCa ) , and to determine if the genetically-modified channels could alter them . We focused on MCU , the leading candidate for the pore-forming subunit , as it possesses two transmembrane domains and has highly-conserved acidic residues in a putative pore-like domain . First , we show that reducing MCU transcripts via knockdown , or enhancing them through overexpression , produce parallel changes in IMiCa . Second , we show that a central feature of IMiCa—its exquisite sensitivity to blockade with ruthenium red—can be abolished via a single point mutation in the pore domain . MCU subunits carrying this mutation act in a dominant-negative fashion , producing functional channels that conduct Ca2+ ions but are largely insensitive to ruthenium red . We conclude that the functional uniporter pore at the mitochondrial inner membrane is formed by MCU multimers . We believe these experiments put to rest the outstanding question of the pore identity , firmly establishing that MCU produces IMiCa .
HEK-293T were obtained from ATCC and grown in Dulbecco’s modified Eagle medium with high glucose supplemented with 10% fetal bovine serum and penicillin/streptomycin ( Life Technologies , Grand Island , NY ) . HEK-293T cells ( 250 , 000 cells per well ) were infected with lentivirus expressing short hairpin RNA ( for knockdown experiments ) or wild-type or mutant FLAG-tagged MCU . 2 days after infection , cells were split and selected with the appropriate antibiotic . Stable HEK-293T cells lines expressing shGFP and shMCU were maintained in 2 μg/ml puromycin ( Sigma , St . Louis , MO ) , while MCU-FLAG and S259A lines were maintained in 200 μg/ml hygromycin B ( Sigma ) . Vectors for expressing shRNA ( pLKO . 1 ) were obtained from the Broad Institute’s RNAi Consortium ( Sigma ) . To silence MCU , we used a hairpin targeting the 3′ UTR , with sequence 5′-GCAAGGAGTTTCTTTCTCTTT-3′ ( TRCN0000133861 ) . The control hairpin was against GFP , with sequence 5′-ACAACAGCCACAACGTCTATA-3′ ( shGFP ) ( TRCN0000072181 ) . The FLAG-tagged full-length human MCU cDNA ( NM_138357 . 1 ) was cloned into the pLJM5 vector ( Sancak et al . , 2010 ) . The MCU mutant S259A was generated with QuikChange according to the manufacturer’s instructions ( Agilent , Santa Clara , CA ) . Human Mitofusin-1 ( NM_033540 ) was cloned out of HEK-293T cell cDNA ( see Quantitative polymerase chain reaction section , below ) and ligated in-frame into the pEGFP-N3 vector ( Clontech , Mountain View , CA ) using KpnI and BamHI restriction sites . Matrix-targeted mCherry was engineered by replacing the GFP construct of the pAcGFP1-Mito plasmid ( Clontech ) , using BamHI and NotI restriction sites . Primers against human MCU were designed using PRIMER-Blast ( NCBI ) . The forward primer was 5′-TTCCTGGGACATCATGGAGC-3′ , and the reverse primer was 5′-TGTCTGTCTCTGGCTTCTGG-3′ . For qPCR , confluent HEK-293T cells were resuspended in TriZOL reagent ( Life Technologies ) , and total RNA was subsequently purified following the manufacturer’s instructions . 2 μg of RNA was used to synthesize cDNA using SuperScript ViLo ( Life Technologies ) and 100 ng of this was used per well for qPCR . Reactions were run in triplicate using SYBR Green ( Agilent ) on an Eppendorf Mastercycler epgradient S Realplex4 qRT-PCR cycler ( Eppendorf , Hauppauge , NY ) . Analysis compared expression in shMCU cells to shGFP cells using the 2−ΔΔCt method , normalizing to glyceraldehyde 3-phosphate dehydrogenase expression . 10 μg of protein from cell lysates or 5 μg of protein from crude mitochondrial preparations were loaded on 12% or 18% Tris-Glycine gels , proteins were transferred to PVDF membranes and then incubated with the indicated antibodies at 4°C overnight . We used antibodies against TOMM20 ( sc-17764; Santa Cruz Biotechnology , Dallas , TX ) , TIMM23 ( No . 611222; BD Biosciences , San Jose , CA ) , PPIF ( ab110324; Abcam , Cambridge , MA ) , ATP5B ( ab14730; Abcam ) , FLAG epitope ( No . 2368; Cell Signaling Technology , Danvers , MA ) , and MCU ( polyclonal chicken antibody raised against residues 41–351 of human MCU tagged with a 6His epitope; Covance , Dedham , MA ) . Two confluent plates of HEK-293T cells were washed with PBS and the cells scraped into 1 . 5-ml microcentrifuge tubes in 1 ml of final volume . The cells were lysed by passaging through a 27 . 5 G needle several times . Samples were spun at 800g for 10 min at 4°C to pellet nuclei and non-lysed cells . The supernatant was removed and spun separately at 8000g for 10 min at 4°C to pellet the mitochondrial fraction . Mitochondria were resuspended in a sucrose buffer ( 200 mM sucrose , 10 mM Tris-MOPS , 1 mM EGTA-Tris ) to a final concentration of 1 mg/ml . 20 μg of mitochondria were treated with digitonin ( 0 . 01–0 . 4% , Sigma ) or 1% Triton-X100 ( Sigma ) and 100 μg/ml of proteinase K for 15 min at room temperature , in a 30 μl final volume . Proteinase K was then inhibited by addition of 5 mM PMSF . 5 μg of protein was loaded to Tris-glycine gels for Western blotting . Stably-transfected HEK-293T cells were washed in PBS , which was used as the base solution in all following steps . Cell were fixed in 4% formaldehyde , permeabilized in 0 . 5% Triton-X100 , and blocked in 10% goat serum , 0 . 1% Tween-20 ( Biorad , Hercules , CA ) . Primary antibodies were a rabbit anti-FLAG and a mouse anti-MTCO1 ( a mitochondrial marker , ab14705; Abcam ) . Secondary antibodies were Alexa-488 conjugated to a goat anti-rabbit , and Alexa-546 conjugated to goat anti-mouse ( Life Technologies ) . Cells were imaged on an Olympus Fluoview 1000 confocal microscope ( Center Valley , PA ) . For mitoplast images , HEK-293T cells were co-transfected with mt-mCherry and MFN1-GFP plasmids using Lipofectamine 2000 reagent ( Life Technologies ) according to the manufacturer’s instructions . Mitoplasts were isolated ( as described in ‘Whole-mitoplast recording’ ) 2 days following transfection , and imaged on the confocal microscope as above . Brightness levels have been slightly adjusted in the images to improve contrast . This adjustment has been applied to each image as a whole and has not obscured or eliminated any particular feature of the image . All reagents were from Sigma unless otherwise stated . We adapted prior methods to HEK-293T cells ( Fedorenko et al . , 2012; Fieni et al . , 2012 ) . Cells were grown to confluency on three 15-cm dishes . All subsequent manipulations were done using ice-cold solutions and equipment . Cells were scraped onto divalent-free PBS and spun down to pellet cells . The cell pellet was subsequently resuspended in a sucrose solution ( 250 mM sucrose , 5 mM HEPES , 1 mM EGTA , 0 . 1% BSA ) . The suspension was homogenized using an overhead stirrer . Centrifugation at 700g pelleted nuclei . The supernatant was transferred to a new tube and centrifuged at 8500g to pellet a crude mitochondrial fraction . This pellet was resuspended in a hypertonic mannitol solution ( 440 mM mannitol , 140 mM sucrose , 5 mM HEPES , 1 mM EGTA ) for 10 min . The suspension was then passed through a French Pressure Cell homogenizer ( Thermo Scientific , Waltham , MA ) at 2000 p . s . i . to disrupt the mitochondrial outer membrane . The eluate was centrifuged at 10 , 500g to pellet mitoplasts . Mitoplasts were stored in a hypertonic solution ( 750 mM KCl , 100 mM HEPES , 1 mM EGTA ) on ice and protected from light . For recordings , mitoplasts were aliquoted into the initial bath solution ( 150 mM KCl , 10 mM HEPES , 1 mM EGTA ) to allow them to expand for 20 min . An individual mitoplast was visualized and attached to the pipette tip with gentle suction . Pipettes were formed from borosilicate glass ( Sutter , Novato , CA ) with tip resistances of 20–60 MOhm . We used sodium gluconate for pipette internal solutions ( 150 mM sodium gluconate , 10 mM HEPES , 2 mM EDTA , with sucrose to achieve an osmolarity of 410–430 mOsmol ) . Electrophysiology was performed using an Axopatch 200B amplifier ( Molecular Devices , Sunnyvale , CA ) . Upon giga-ohm seal formation , pipette capacitance transients were cancelled , and the inner membrane was ruptured by 10–50 ms pulses ( 350–1000 mV ) . After establishing the whole-mitoplast configuration , a series of 4 ms pulses of 20 mV were recorded to allow curve fitting for capacitance measurements . Bath solutions were subsequently exchanged for a high-Ca2+ solution ( 100 mM calcium gluconate , 20 mM HEPES , 2 mM CaCl2 , with sucrose to achieve osmolarity of 295–305 mOsms ) and high-Ca2+ +100 nM RuR solution ( Sigma ) . RuR solutions were prepared fresh for every experiment . Solution exchanges were rapid , as prolonged matrix exposures to Ca2+ caused activation of a significant leak component . Comparisons were made via the Student’s t-test . | Mitochondria are tiny organelles , less than a micrometre across , found inside almost all eukaryotic cells . Their main function is to act as the ‘power plant’ of the cell , generating adenosine triphosphate or ATP , which is the source of chemical energy for cellular processes . Beyond generating ATP , mitochondria perform many other functions: they contribute to various signalling pathways; they influence cellular differentiation; and they are involved in processes related to cell death . Mitochondria are quite distinctive in appearance—they are enclosed by two membranes , a porous outer one and a largely impermeable inner membrane . Most mitochondrial functions involve proteins that control the movement of various molecules and ions across the inner membrane . One particularly important ion that must pass through this membrane is calcium; once inside the mitochondria , these calcium ions regulate cell survival and the generation of ATP . Although several calcium import mechanisms exist , the best-studied pathway involves a pore-forming protein complex called the mitochondrial calcium uniporter . This ion channel has an exquisite selectivity , allowing only calcium into mitochondria even when other ions outnumber it a million-fold . Previously , researchers had identified several genes that are required for the formation of the uniporter , but it had not been established which of these encodes the central pore through which the calcium ions pass . Now , Chaudhuri et al . have shown that one of these—a gene called mitochondrial calcium uniporter ( MCU ) —codes for the protein subunit that creates the pore . Prior studies used optical methods or purified proteins to study genes encoding the uniporter complex , producing controversial results regarding pore identity . This study uses a much more direct assay , namely electrophysiology performed on mitochondrial inner membranes . To access the inner membrane , the authors stripped off the outer membrane from whole mitochondria , and made them expand . By using a technique called voltage-clamping , Chaudhuri et al . were able to precisely measure calcium ion movement through intact or mutated channels . This technique controls confounding factors and minimizes the effect of contaminants that can plague interpretation of data acquired by other methods . They showed that blocking the expression of the MCU gene reduced the flow of calcium ions through the uniporter , whereas increasing MCU expression increased calcium transport . One unique feature of the mitochondrial calcium uniporter is that it can be blocked by a dye called ruthenium red . Chaudhuri et al . used this property to confirm that the MCU gene encodes the pore-forming subunit of the channel complex—they identified a single point mutation in MCU that did not affect the channel’s ability to transport calcium ions , but did abolish its sensitivity to ruthenium red . Together , these results show that the MCU gene encodes the pore of the mitochondrial calcium uniporter , and should lead to further research into the physiology and structure of this channel . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2013 | MCU encodes the pore conducting mitochondrial calcium currents |
Defining mechanisms of direct-acting antivirals facilitates drug development and our understanding of virus function . Heteroaryldihydropyrimidines ( HAPs ) inappropriately activate assembly of hepatitis B virus ( HBV ) core protein ( Cp ) , suppressing formation of virions . We examined a fluorophore-labeled HAP , HAP-TAMRA . HAP-TAMRA induced Cp assembly and also bound pre-assembled capsids . Kinetic and spectroscopic studies imply that HAP-binding sites are usually not available but are bound cooperatively . Using cryo-EM , we observed that HAP-TAMRA asymmetrically deformed capsids , creating a heterogeneous array of sharp angles , flat regions , and outright breaks . To achieve high resolution reconstruction ( <4 Å ) , we introduced a disulfide crosslink that rescued particle symmetry . We deduced that HAP-TAMRA caused quasi-sixfold vertices to become flatter and fivefold more angular . This transition led to asymmetric faceting . That a disordered crosslink could rescue symmetry implies that capsids have tensegrity properties . Capsid distortion and disruption is a new mechanism by which molecules like the HAPs can block HBV infection .
Worldwide , an estimated 240 million people suffer from chronic hepatitis B virus ( HBV ) ; infection is most often acquired in early childhood . Over time , chronic HBV can lead to liver disease including liver failure , cirrhosis , and hepatocellular carcinoma , contributing to more than 700 , 000 deaths each year ( Gish et al . , 2015 ) . A vaccine directed against surface protein is preventative but not therapeutic . Entry inhibitors prevent new infection but are unlikely to affect maintenance of a chronic infection ( Volz et al . , 2013 ) . Reverse transcriptase inhibitors directed against the viral polymerase suppress viremia and improve liver health , but even after years of treatment are rarely curative ( Gish et al . , 2007; Shi et al . , 2015 ) . Interferon α and derivatives can stimulate HBV clearance in a small subset of patients ( Gish et al . , 2015 ) . The viral core protein ( Cp ) has become an important target for developing direct-acting antivirals ( DAAs ) ( Zlotnick et al . , 2015 ) . The HBV Cp is present at virtually every step of the virus lifecycle in an infected cell ( Zlotnick et al . , 2015 ) . In the nucleus , Cp is associated with viral nucleic acid ( Bock et al . , 2001; Li et al . , 2010 ) . In cytoplasm , Cp assembly packages viral polymerase complexed with an RNA copy of the viral genome and has an active role in reverse transcription of the relaxed circular mature genome ( Tan et al . , 2015 ) . Cp also has signals that direct core intracellular trafficking and secretion of enveloped cores ( Kann et al . , 2007 ) . When transported to the nucleus , for new infection and to maintain infection , the core disassembles to release its contents . Curiously , approximately 90% of the capsids from enveloped particles are empty and the rest contain the viral genome ( Ning et al . , 2011 ) . This leads to the hypothesis that assembly can be nucleated by a polymerase-viral RNA complex for virions or spontaneously by Cp . This bifurcation of paths suggests that assembly is susceptible to disruption . The oligomeric nature of Cp function makes it a particularly attractive antiviral target because even when DAA-resistant mutants arise , residual wildtype protein co-assembled into a complex retains DAA sensitivity ( Tanner et al . , 2014 ) . The HBV Cp is a 183-residue , homodimeric protein ( Venkatakrishnan and Zlotnick , 2016 ) . The first 149 residues form the assembly domain . Each monomer contributes two helices to a four-helix bundle that forms the intradimer interface ( Venkatakrishnan and Zlotnick , 2016 ) . The final 34 residues of the full length Cp form the disordered nucleic acid-binding C-terminal domain . This domain is optional for capsid assembly but required for nucleic acid packaging; this region also carries putative nuclear localization and nuclear export signals ( Li et al . , 2010; Kann et al . , 2007 ) . Cp assembles into icosahedral 90-dimer T = 3 capsids and 120-dimer T = 4 capsids . In capsids , dimers are arranged in a pattern of fivefold vertices and quasi-sixfold vertices . Thus , owing to quasi-equivalence , a T = 3 capsid is comprised of 60 AB dimers and 30 CC dimers ( with true twofold symmetry ) and a T = 4 capsid is comprised of 60 AB and 60 CD dimers . In vitro assembly of the assembly domain or full-length Cp yields about 90% T = 4 capsids; a similar ratio is observed in vivo ( Venkatakrishnan and Zlotnick , 2016; Stannard and Hodgkiss , 1979 ) . Studies of in vitro assembly of Cp have provided insights into the biology of HBV , the general problem of capsid assembly and the mechanism of assembly-directed antivirals . It is hypothesized that assembly is initiated by a conformational transition of Cp to an active state , suggesting a basis for regulated assembly in vivo ( Packianathan et al . , 2010 ) . The net association energy is weak , but since each of the many Cp dimers forming a capsid are tetravalent , the result is a globally stable structure ( Zlotnick , 2003 ) . In vitro assembly of HBV Cp is sensitive to ions and temperature , supporting the hypothesis that assembly can be allosterically regulated ( Stray et al . , 2004; Ceres et al . , 2004 ) . Furthermore , there is a close relationship between the familiar phenomenon of allostery and the structural principle of tensegrity in virus capsids ( Domitrovic et al . , 2013a ) . Putative Cp allosteric effectors ( e . g . heteraryldihydropyrmidines or HAPs ) have been discovered that have antiviral effect ( Deres et al . , 2003; Stray et al . , 2005; Stray and Zlotnick , 2006a ) . These small molecules ostensibly act by inducing an assembly-active state and by stabilizing Cp-Cp interactions by filling a pocket at the interdimer interface between dimers of the quasi-sixfold ( Venkatakrishnan and Zlotnick , 2016; Bourne et al . , 2006 ) . We refer to this class of molecules , which now includes a diverse set of chemistries ( Pei et al . , 2017 ) , as Core protein Allosteric Modulators ( CpAMs ) . The well-characterized mechanism for CpAM action is to over-initiate assembly , forming empty capsids and large aberrant structures . Both products act to sequester Cp from its biological functions . We suggest that CpAMs likely have additional effects on the HBV lifecycle and are developing fluorescent CpAMs as tools to better probe their effects both in vitro and in cells . Here , we used cryo-electron microscopy ( cryo-EM ) as a means to look at capsids in complex with the fluorescent CpAM HAP-TAMRA . The fluorescent CpAM provides a useful signal for binding , and we hoped that the extra mass would assist identifying the molecule in structural studies . Cryo-EM provides advantages compared to previous studies using crystalized forms of Cp , in that capsid-wide deformations can be observed , including asymmetric deformations . This establishes a basis for further antiviral development focused on disruption of capsid structure that is distinct from the previously established mechanisms of CpAM activity .
To better understand the relationship between CpAMs and core protein , we synthesized a fluorescent CpAM , HAP-TAMRA ( Figure 1 ) . HAP-TAMRA’s HAP13 core ( Bourne et al . , 2008 ) was extended with a three-carbon linker appended with a tetramethylrhodamine ( TAMRA ) dye . As the HAP13 core used in this synthesis was a racemic mixture , only half the compound is expected to be active ( Deres et al . , 2003 ) . We anticipated the TAMRA moiety to extend into the capsid lumen from the HAP-binding pocket and cause minimal disruption to interactions between the HAP and protein . To verify the activity of HAP-TAMRA , we examined its effect on purified dimeric assembly domain , Cp149 . We observed that assembly reactions of purified core protein in the presence of HAP-TAMRA produced large aberrant products that are characteristic of HAP-induced assembly ( Figure 2b ) . For reference , assembly reactions without CpAMs ( Figure 2b ) led to particles morphologically consistent with an icosahedral shell . In this work , our focus was the binding of HAP-TAMRA to pre-assembled capsids , similar to those in Figure 2a . We first evaluated formation of the capsid HAP-TAMRA complex using size exclusion chromatography with a diode array absorbance detector ( Figure 2c , d ) . Capsids co-eluted with bound HAP-TAMRA in the capsid fraction . Free HAP-TAMRA eluted at the end of the column volume , as expected for a small molecule . TAMRA absorbance spectra associated with the capsid and free probe were distinctly different . The free probe exhibited the familiar absorbance profile of TAMRA , with a major peak near 555 nm , and a shoulder at 520 nm . The absorbance spectrum for capsid-bound HAP-TAMRA was shifted , with a major peak near 520 nm and a shoulder around 550 nm that resembled examples of spectra from π-stacked TAMRA complexes ( Adachi et al . , 2014 ) . Using the change in signal at 520 nm , we monitored the amount of capsid- bound HAP-TAMRA . We also examined the utility of the ratio of 520/555 nm absorbance as a measure of binding ( Figure 2d ) . For both the absorbance at 520 nm and the 520/555 ratio , the signal associated with capsid peaked at about one HAP-TAMRA per dimer . These data also suggest that HAP-TAMRA binds capsid with high affinity compared to the 8 µM Cp dimer used in these experiments . The appearance of the 520 nm peak under these conditions was restricted to the capsid fraction as shown by the absence of a ratiometric change in the free fraction ( Figure 2d ) . This absorbance signal change is also visible in samples that are not purified by size exclusion ( Figure 3a ) . We chose to use the increase in absorbance at 520 nm to study the kinetics of binding , ( Figure 3b–f ) . Using 2 . 5 to 40 µM HAP-TAMRA , we observed that binding was surprisingly slow , reaching half-saturation in approximately 10 min at all concentrations . When the different kinetic trajectories were normalized to the same maximum and minimum values , zero and one , it was evident that they were identical ( Figure 3c ) , indicating that the rate of binding under these conditions was independent of HAP-TAMRA concentration . Therefore , we examined binding kinetics in terms of a first order reaction . Slow kinetics can indicate that the binding pocket is not usually accessible to the ligand , but that a conformational transition of the protein allows the pocket to open transiently . Curve fits using a single exponential did not fit the observed binding kinetics but an association model with two phases of equal magnitude matched the kinetic trajectories . The average for the fast half-life was 0 . 6 ± 0 . 1 min and for the slow half-life was 7 . 1 ± 1 . 4 min ( Figure 3b ) . This HAP-TAMRA concentration independence and the slow binding rate strongly suggest that the primary barrier for HAP binding is a conformational transition at the site of the binding pocket . Because the signal for binding is an absorbance shift consistent with stacking of two or more TAMRA moieties , examination of the absorbance when the reaction was complete should also give an indication of the mechanism of binding . We observed that the change in absorbance increased linearly until capsids were saturated at one HAP-TAMRA per dimer ( Figure 2d ) . The change in absorbance per bound HAP-TAMRA was also evaluated ( i . e . the change in extinction coefficient ) and found to be 11200 ± 4% M−1 ( Figure 3e ) . This result indicates that there is the same degree of TAMRA–TAMRA interaction in saturated capsids and in capsids where there was on average only one HAP-TAMRA per capsid . The change in absorbance spectrum suggested TAMRA-TAMRA interaction that would be expected to cause static quenching; this prediction was confirmed in Figure 3f where the residual fluorescence is largely due to unbound probe . Previous structural studies of HAPs with HBV core protein have depended on crystallography , which selects for regular complexes . Thus , when we looked at cryo-EM 2D class averages of HAP-TAMRA-saturated capsids we anticipated a field of isometric particles similar to apo-capsids ( drug free ) ( Figures 2a and 4b ) . However , in both raw micrographs and 2D class averages , the heterogeneity of particle shapes with bound HAP was evident ( Figure 4a , c ) : many particles had sharp corners , facets , and irregular outlines . To distinguish ‘distorted’ from the normal capsids , we performed 2-D class averaging to compare particles from Cp149-apo ( Figure 4b ) and Cp149+HAP TAMRA ( Figure 4c ) datasets . In this technique , particles are grouped based on dominant features of the microscopic image , rotationally oriented , and presented as an average . Apo-capsids ( Figure 4b ) had a round cross section with obvious spikes and distinct internal features . The capsids with HAP-TAMRA had a broad distribution of asymmetric shapes with ellipsoidal character and distinct faceting . Because of the structural irregularity of these capsids , they would be of limited use in pursuit of a high-resolution reconstruction . Therefore , we examined Cp150 capsids; Cp150 is a variant of Cp149 with an engineered disulfide bond linking the disordered C-terminal tails when in capsid form ( Lee et al . , 2017 ) . Our rationale was that the inter-dimer disulfides would provide a soft positional restraint on the capsid , and prevent it from adopting conformations with extreme asymmetry due to bound probe . The 2D class averages of Cp150 capsids incubated with HAP-TAMRA ( Figure 4c ) showed regular and approximately spherical particles; suitable for further detailed structural analysis . To examine HAP-bound capsids , we reconstructed T = 4 and T = 3 Cp150-HAP-TAMRA capsids to 4 . 0 Å and 3 . 7 Å resolution , respectively , from a single dataset ( Figures 5 and 6; Table 1 ) . The resolution was sufficient to position and rebuild molecular models of protein and HAP-TAMRA ( Video 1 ) . These structures are , to our knowledge , the first instance of using cryo-EM to visualize a drug bound to a virus capsid and to show that T = 3 capsids bind CpAMs ( Video 2 ) . Cryo-EM also allowed high resolution analysis of changes to quaternary structure without the potential bias of crystallographic packing constraints . The T = 4 capsids were slightly larger than crystallographic structures , though each crystal structure has shown unique features ( Venkatakrishnan and Zlotnick , 2016 ) . The comparison to the T = 3 capsid , in particular , shows commonalities and differences in Cp–Cp interactions . Besides the difference in diameter and the differing number of component Cp dimers , the local difference in T = 4 and T = 3 architecture is that quasi-sixfold vertices have twofold and threefold symmetry , respectively . A T = 4 quasi-sixfold has two repeats ( B , C , D , B’ , C’ , and D’ ) subunits and in T = 3 there are three repeats ( B , C , B’ , C’ , B’' , and C’' ) ( Figure 5 ) . In T = 4 capsids , quasi-sixfold hexamers have four HAP-TAMRA molecules , while the T = 3 quasi-sixfold hexamers accommodate three HAP-TAMRA molecules . The HAP13 portion of the probe is nestled in the pocket between two adjacent dimers with the linker extending from the pocket , positioning the TAMRA near the center of the quasi-sixfold pore . In the T = 4 quasi-sixfold , the HAPs are in B , C , B’ , and C’ pockets , which are respectively capped by C , D , C’ and D’ subunits ( Figure 5b ) . In the T = 3 particles , HAPs are in B pockets capped by C subunits ( Figure 5d ) . These results place the HAP moiety of HAP-TAMRA in the same pocket associated with other CpAMs HAP1 , AT130 , HAP18 , and sulfamoyl benzamides . ( Bourne et al . , 2006; Venkatakrishnan et al . , 2016; Zhou et al . , 2017 ) No small molecule density was observed at the corresponding A ( fivefold ) and D sites , which are occluded by A and B subunits , respectively . The pocket of the T = 3 C subunit is occluded by the neighboring B subunit , resembling the interaction of the T = 4 D subunit with its neighboring B subunit . Despite the quaternary differences between T = 3 and T = 4 capsids , the local environment of the HAP-TAMRA in the HAP pocket was remarkably similar ( Figure 7 ) . The HAP13 had extremely well-ordered density in the pocket as do most protein side chains in the region ( visualized in Figure 7e at a contour of 3 . 6 σ , where σ is the number of standard deviations from the mean ) . The quality of HAP-TAMRA density gradually decays in the TAMRA region , which is visible as a loss of density in Figure 7 , and was also seen as a decay in local resolution in Figure 8 . This observation is consistent with fewer positional constraints in the exposed TAMRA moiety , and higher molecular motion . In particular , the T = 4 C subunit the TAMRA density is only contiguous at 2 . 6 σ . For the T = 4C subunit , the loss of linker density makes precise fitting the connection from the HAP to the TAMRA ambiguous . However , linker density clearly points towards a planar mass that is parallel to and ‘stacked’ on the TAMRA from the B’ subunit . By symmetry , there is an equivalent interaction between TAMRAs from C’ and B . This resulting best fit into the available density should not be considered an exclusive model . However , as 80–90% of the capsids have T = 4 symmetry , we suggest that these TAMRA-TAMRA interactions are the cause of the observed change in the optical signal ( Figure 2c , d ) . When comparing T = 3 to T = 4 , and T = 4 CpAM-bound to T = 4 apo structures , we found that local similarities in protein structure do not translate to similarities one or two dimers away . Long range differences in quaternary structure were apparent when overlaying entire hexamers , for example overlaying T = 3 and T = 4 quasi-sixfold hexamers based on one set of AB dimers ( Figure 9 ) . The next subunit around the quasi-sixfold , the D subunit ( T = 4 ) or C subunit ( T = 3 ) , show pronounced differences while the dimer on the far side of quasi-sixfold is completely misaligned . The dramatic effect arises because the small angular differences are amplified by distance . Another relevant example is the assembly-inactive core protein mutant Y132A . ( Packianathan et al . , 2010; Zhou et al . , 2017; Klumpp et al . , 2015; Qiu et al . , 2017 ) The Y132A mutation disrupts normal interdimer interaction allowing crystal contacts that form hexagonal layers reminiscent of but distinct from subunit interactions in capsids . In this cryo-EM structure , density for Y132 is well-resolved ( Figure 7e ) . While the Y132 mutant supports high resolution crystallography ( Packianathan et al . , 2010; Klumpp et al . , 2015 ) , it is impossible to predict CpAM effects to a capsid in that context . Such CpAM-induced quaternary differences between HBV capsids do not appear arise from changes in the binding to the HAP pocket . Rather , they arise from the way the capping subunit overlays the pocket and how these differences propagate across a capsid . The structural differences between the HAP-bound T = 4 particle and an apo-capsid are evident when capsids are aligned based on icosahedral symmetry . It is immediately clear that the HAP-bound particle has expanded , particularly at the fivefold ( Venkatakrishnan et al . , 2016 ) . The pairwise Cα separation between the HAP-TAMRA capsid and an apo-capsid ( 1QGT [Wynne et al . , 1999a] ) shows large-scale changes and regions of constraint that were not obvious when focusing on a single asymmetric unit ( Figure 10 ) . The cryo-EM reconstruction shows a greater expansion than seen in crystallographic structures of other CpAM-capsid complexes ( Bourne et al . , 2006; Venkatakrishnan et al . , 2016; Katen et al . , 2013 ) . In comparison to the apo capsid , the displacement maximum is almost 6 Å around the 5-fold axis , and the displacement minimum is 1 . 7 Å around the 3-fold axis . The two-dimer asymmetric unit is rotated along the long axis of the CD dimer , near the 3-fold , leading the AB dimer to pivot upward , raising at the 5-fold . We suggest that the basis of this quaternary change is a flattening of the quasi-sixfold hexamers . Flattening quasi-sixfold and concomitantly making the 5-fold vertices protrude is likely a subdued version of what we observed for Cp149 ( Figure 4 ) . An important observation is that the structural variation that leads to faceting of T = 4 capsids can also be observed by examining the cryo-EM local resolution maps ( Figure 8 ) . The local resolution is systematically better around the 2-fold symmetry axis , and worse around the 5-fold . This result suggests a greater positional variation of A subunits , equivalent to a crystallographic B factor . In both AB and CD dimers , the highest resolution is observed in the core of the protein , the chassis subdomain ( Packianathan et al . , 2010 ) , and the lowest resolution , falling to about 5 Å , is at the spike tips . Protein–protein interactions around the quasi-sixfold were extremely well-ordered as were the HAP moieties of HAP-TAMRA , about 3 . 6 Å in the overall 4 Å resolution T = 4 structure . The resolution of the TAMRA density was poorer , about 5 Å . This same pattern was also observed in the T = 3 capsid structure . Studies with crosslinked Cp150 suggest the hypothesis that CpAM-induced structural changes and fluctuations would be concentrated at fivefold vertices and substantially greater without the constraint of the interdimer C150 disulfide . For this reason , we performed a 3D reconstruction of the heterogeneous complexes of Cp149 and HAP-TAMRA ( Figure 11a ) , using data exemplified in Figure 4a and c . The Cp149 +HAP TAMRA reconstruction could only reach 22 Å resolution , which is a likely outcome from icosahedral averaging a wide array of structural states . The striking feature of the reconstruction is the absence of density at the fivefold . This may arise because the A subunits are extremely mobile or , more likely , that the individual A subunits are relatively static but in a variety of positions . This second hypothesis is consistent with the heterogeneity of the class averages in Figure 4c . This result suggests that the A subunits are the favored site of capsid failure . A comparator structure calculated to the same resolution shows well-ordered fivefold density ( Figure 11b ) .
CpAMs , exemplified here by HAP-TAMRA , can distort capsid shape . HAP-TAMRA with crosslinked Cp150 dimers yielded capsids that retained icosahedral symmetry and were suitable for relatively high-resolution reconstruction . HAP-TAMRA with the uncrosslinked capsids of Cp149 led to asymmetric faceting , severe distortions , and broken capsids . There are advantages to examining both Cp150 and Cp149 CpAM complexes , one for structural detail and the other for biological relevance ( Wang et al . , 2015 ) . In HBV infections , core protein exerts many of its functions while in the capsid form . Defects in the capsid structure due to CpAMs may alter exposure of signals for cellular trafficking ( Kann et al . , 2007; Chen et al . , 2016 ) or dysregulate capsid dissociation ( Rabe et al . , 2009 ) . Core protein and capsid play active roles in reverse transcription which we would expect to also be compromised in a deformed capsid ( Nassal , 2008; Hu and Seeger , 2015; Liu et al . , 2015 ) . Core protein mutations that fill the HAP site lead to defects in both RNA packaging and transcription of the DNA plus-sense strand suggesting that CpAMs can have similar action ( Tan et al . , 2015 ) . Though there have been reports of HAPs causing capsids to dissociate ( Stray and Zlotnick , 2006b ) , the structural basis for this paradoxical effect of stabilizing Cp-Cp interaction but inducing capsid failure have , with this report , only now begun to be explored . The T = 4 Cp150+HAP TAMRA cryo-EM structure has a larger diameter than crystallographic structures of the apo capsid ( Wynne et al . , 1999a ) and CpAM-bound capsids ( Venkatakrishnan and Zlotnick , 2016 ) . The HAP pockets show little structural difference in quasi-equivalent sites in capsids or in crystal structures of the assembly-incompetent mutant Y132A ( Packianathan et al . , 2010; Zhou et al . , 2017; Klumpp et al . , 2015 ) . This indicates that CpAM is affecting structure primarily by modulating the protein-protein interactions of the capping subunit . In addition , CpAMs can subtly modify the tertiary and quaternary structure of an individual dimer ( Venkatakrishnan and Zlotnick , 2016 ) . In broad terms , CpAMs flatten sixfold; during assembly this can result in non-capsid structures with extended regions of hexagonal sheet ( Stray et al . , 2005 ) . or cylinders with a hexagonal repeat ( Liu et al . , 2017 ) . The flattening of quasi-sixfold , observed in molecular detail in the crosslinked Cp150 capsid , is also evident in uncrosslinked Cp149 capsids as flat regions and sharp angles . The sharp angles are likely to arise at fivefold . This is analogous to phage P22 which is assembled as a rounded procapsids that matures to have flattened sixfold and sharply angled fivefold ( Teschke and Parent , 2010 ) . In the 22 Å resolution image reconstruction ( Figure 10c ) , the absence of density at fivefold could be due to highly dynamic A subunits , or to relatively static structures where the A subunits have multiple orientations . In the 2D class averages of Cp149+HAP TAMRA capsids ( Figure 4c ) most classes are asymmetrically elliptical and/or have a periphery punctuated by sharp angles , suggestive of faceting . Consider the effect of flattening a quasi-sixfold in a T = 4 HBV capsid: the A subunit in the adjacent fivefold would protrude and the C and D subunits in adjacent quasi-sixfold would be flattened . The effect on the adjacent quasi-sixfold could lead to a large flat area , resulting in a faceted or elliptical particle , or a discontinuity where the adjacent quasi-sixfold was folded . In the extreme case , such flattening could rupture a particle as observed in micrographs of Cp149+HAP TAMRA ( Figure 4a ) . This study also has implications for the study of the physics of nanoscale materials . For a flattened quasi-sixfold to disrupt a complete capsid , retaining the flattened region must be energetically more favored than maintaining the global capsid structure . CpAMs are known to strengthen protein-protein interaction energy ( Zlotnick et al . , 2015; Venkatakrishnan and Zlotnick , 2016 ) . Making the capsid surface stiffer is predicted from condensed matter theory to promote buckling transitions that lead to faceting ( Lidmar et al . , 2003; Klug et al . , 2006 ) . We therefore suggest that crosslinking Cp150 preserved capsid symmetry by allowing a compromise where CpAMs partially flattened surfaces but the stress was distributed globally , consistent with the mechanical principles of tensegrity structures . Tensegrity structures can also be designed to undergo structural transitions in response to relatively small stimuli ( Domitrovic et al . , 2013b; Skelton et al . , 2001; Fuller , 1975 ) in the same way that HBV responds to CpAM binding . Indeed , a reoccurring point in literature on the mechanical properties of icosahedral virus capsids , is that their stability is fundamentally related to their symmetry ( Klug et al . , 2006; Zandi et al . , 2004; Zandi and Reguera , 2005 ) . The motivation to construct a fluorescent CpAM was that it would expand our ability to evaluate CpAM-Cp interaction in vitro and in vivo . We used changes in the absorbance profile , as well as fluorescent quenching , to observe CpAM binding to capsid . The photochemical effects we describe has been well-documented for TAMRA and other rhodamine dyes . It is a consequence of changes in electronic excitation due to proximity and dipole alignment of two or more aromatic systems ( Bergström et al . , 2002; Ogawa et al . , 2009 ) . Taking advantage of HAP-TAMRA optical properties , we observed that HAP-TAMRA bound capsid tightly and slowly . Slow binding suggests that sites are only rarely open to CpAMs . The slow rate of binding is reminiscent of a proteolytic analysis of HBV capsid stability where it was observed that Cp149 capsids would undergo a slow partial unfolding transition ( Hilmer et al . , 2008 ) . The half-life of the unfolding transition for cleavage was on the order of 100 min . In contrast , intact capsids can persist in solution for months ( Uetrecht et al . , 2010 ) . All-atom molecular simulations of capsids indicate that HBV capsid structure is highly dynamic and is capable of departure from icosahedral symmetry ( Perilla et al . , 2016 ) . The multidimensional nature of a continuously deforming virus capsid ( composed of 120 interconnected dimers ) frustrates a rigorous treatment of binding . However , we have identified some important constraints by relating HAP-TAMRA conformation to binding cooperativity . The ∆A520 signal we observe depends on formation of a π -stacked dimer ( Adachi et al . , 2014 ) . The absorbance shift we observe ( Figure 3f ) indicates that all of the bound HAP-TAMRA is involved in π-stacked dimers , even when there is on average approximately a single HAP-TAMRA per quasi-sixfold vertex . This would not occur if HAP-TAMRA were randomly distributed over a capsid . In a random distribution at the lowest concentration we tested only about 1/3 of the probe would be in quasi-sixfold vertices with two or more bound molecules . Furthermore , the structure of the T = 4 capsid suggests interaction only between probes in the B and C’ or C and B’ sites ( Figure 5b ) . Of the six possible pairwise interactions for a T = 4 hexamer with two sites filled ( BB’ , BC’ , BC , CB’ , CC , B’C’ ) , only one third is expected to yield a ∆A520 signal . Thus , at the lowest HAP-TAMRA:dimer ratio tested , random binding predicts 1/3*1/3 = 1/9 or 11% of the maximum possible ∆A520 . Nonetheless , we observe approximately 100% of the possible signal . This indicates a high degree of cooperativity of binding , at least at a local level . Binding kinetics also suggests some degree of cooperativity . We would expect to see sigmoidal kinetics if sites filled randomly . Thus , once a B site is filled there must be a preference for filing an adjacent C’ with a much faster rate constant ( or for a C site a preference for an adjacent B’ ) . In a slightly more complicated model , B and C may have different rate constants for the initial binding event . This second model could fit to our data with two first order curves of equal amplitude ( Figure 3b ) . However , this local structural change model does not account for the global structural change we observe . Furthermore , contrary to Occam’s razor , a simple fit may obscure complex , multi-phase kinetics . If we consider that flattening one quasi-sixfold will affect the structure of the neighboring quasi-sixfold , we can see how capsid asymmetry is induced . Asymmetric distortion of quasi-sixfold will affect binding site structures and therefore their kinetics of binding and affinities . There may be a continuum of binding sites from flattened quasi-sixfold , as might be found in a large aberrant structure or the flat areas of an oblate ellipse ( i . e . large radius of curvature ) , to bent quasi-sixfold that might be found where there is a small radius of curvature . The result of this continuum of sites will not fit a single first order rate law , but may be approximated by as the sum of two or more ( Zlotnick et al . , 1994 ) . In a broad sense , our results reaffirm the view that virus capsids are not inert containers . It is likely that some of the most important biological functions of icosahedral capsids may occur when they become non-icosahedral . We have demonstrated that non-icosahedral states of HBV capsids can be induced using CpAMs . Indeed , CpAMs can deform and even disrupt intact capsids , which suggest modes of action beyond the established mechanisms of assembly activation and misdirection . A molecule which simultaneously affects two or more stages of the lifecycle would reduce the impact of escape mutations , and increases the likelihood of clearing an infection .
The Hepatitis B subtype adyw core protein assembly domain and the Cp150 mutant , in which three native cysteine are mutated to alanine and an additional C-terminal cysteine appended , were expressed in E . coli and purified by size exclusion as previously described ( Zlotnick et al . , 2007 ) . Cp149 and Cp150 capsids were prepared for preliminary Cryo-EM from purified dimer by initiating assembly at 10 μM dimer concentration in 300 mM NaCl , 50 mM HEPES , pH 7 . 5 and allowing the reaction to proceed overnight . Residual un-assembled dimer was removed by purifying the fresh capsids via size exclusion . Purified capsids were then incubated with HAP-TAMRA at a molar excess overnight . For cryo-EM , capsids were then concentrated to >10 mg/ml . Synthesis of HAP13 ( Bourne et al . , 2008 ) and HAP-TAMRA are described in the patent literature ( Zlotnick et al . , 2014 ) . Analysis of the product by HPLC at the 280 nm and 555 nm wavelengths showed a product with ~99% purity; ESI-TOF MS: calculated for C51H50ClFN8O7 , m/z 940 . 35; M+ + H , m/z 941 . 36; Found , m/z 941 . 3; calculated for M+ + Na , m/z 963 . 4; Found , m/z 963 . 3 . Samples from Figure 2a , b represent Cp149 assembly reactions at 50 mM HEPES , 300 mM NaCl , and 10 μM dimer . The reaction in Figure 2b was initiated in the presence of 40 μM HAP-TAMRA . Both samples were adsorbed to the surface of EMS carbon film 300 mesh coper grids , washed with water , and stained with 2% uranyl acetate . Samples were imaged with a JEOL 3200-FS electron microscope operated at 300kV . Samples for Cryo-EM were applied to a glow-discharged Quantifoil holey-carbon grids ( R2/2 ) . The grids were blotted with filter paper for 4 s before automated plunging into liquid ethane using an FEI vitrobot . The Cp149+HAP TAMRA data from Figure 11 and Figure 4 panel a were imaged using a JEOL 3200-FS electron microscope operated at 300kV with an in-column energy filter . Images were recorded at a nominal magnification of 80 , 000X , with a pixel size of 1 . 5 Å , maintaining less than 35e- Å2 dose , recorded on a Gatan Ultrascan 4000 4k × 4 k CCD detector . The 2D class averages shown in Figure 4 panels b , c , and d were collected on the same JEOL 3200-FS microscope , but with a Direct Electron DE-12 detector , with an image sampling rate of 1 . 01 Å/pixel . For the high-resolution 3D structure determination , first presented in Figure 5 , samples were imaged on a FEI Titan Krios operated at 300 kV at a nominal magnification of 22 , 500 . Images were recorded on a Gatan K2 Summit detector operating in ‘super-resolution’ mode , resulting in a pixel size of 0 . 65 Å with a dose of ~33 e- Å2 . Each exposure was 8 s long , and was collected as 35 individual frames . The data collection process was semi-automated using the Leginon system . ( Suloway et al . , 2005 ) The final maps are deposited in the EMDB database as 7295 for the T = 3 map and 7294 for the T = 4 map ( Table 1 ) . Cryo-EM classification and reconstruction was implemented using standard protocols of the EMAN2 , Motioncorr2 , and Relion software programs ( Kimanius et al . , 2016; Zheng et al . , 2016; Tang et al . , 2007 ) . Upon convergence of the 3D structures , each map was sharpened by applying a negative B-factor which was obtained using the Guinier fitting procedure implemented in Relion ( Rosenthal and Henderson , 2003 ) . To assess the local variability in the quality of the structure a local resolution analysis was employed , also implemented in Relion . The local resolution procedure determined local FSC with a sampling window of 10 Å , using the same negative B-factor obtained at the end of refinement . The crystal structures of both apo capsid ( 1qgt ) ( Wynne et al . , 1999b ) and HAP18 bound capsid ( 5d7y ) ( Venkatakrishnan et al . , 2016 ) were used as starting points for flexible model refinement imposing icosahedral non-crystallographic symmetry constraints , and using the PHENIX and eLBOW software programs ( Afonine et al . , 2012; Moriarty et al . , 2009; Emsley et al . , 2010 ) . The model validation statistics we report were obtained from the final output of the Phenix real space refinement tool . To refine the pixel size of the map , we fit the model of the asymmetric unit into density and measured cross-correlation across pixel sizes . The optimal pixel size was determined to be 1 . 285 Å , compared to 1 . 30 Å as determined by the microscope calibration . The final maps and models reflect this change . Structural comparisons and figure creation were carried out in UCSF Chimera . The final structure coordinates are deposited in the protein data bank as 6BVN for the T = 3 structure and 6BVF for the T = 4 structure . The final buffer conditions for all binding assays were 300 mM NaCl , 20 mM Tris , 1% DMSO , and pH 7 . 5 with varying protein and HAP-TAMRA . Size exclusion chromatography assays were performed using a superose 6 30/10 column plumbed into a Shimadzu HPLC equipped with a diode array detector . The HAP-TAMRA absorbance eluting in the capsid fractions was attributed to the capsid bound population . All HAP-TAMRA absorbance eluting later was attributed to free ligand . It is important to note that because the HAP-TAMRA probe is synthesized as a racemic mixture , and because only a single enantiomer is known to bind the HAP pocket , we expected to see at least half of the input HAP-TAMRA eluting as free ligand . When plotting the increase in 520 nm signal in the capsid fraction ( Figure 2d ) , the A520 at 7 . 5 ml was used , corresponding to the pre-established center of the capsid elution volume . Absorbance and fluorescence was measured in a 96-well plate using a 200 µl sample volume by a Biotek Synergy H1 plate reader . Measurements based on fluorescence used an excitation wavelength of 520 nm and monitored emission at 580 nm . For plotting the change in 520/555 nm absorbance ratio , we used the maximum value of the capsid peak at each wavelength . Kinetic assays based on absorbance were sampled in 1 min intervals , where each data point is the value of 520 nm absorbance . To account for the varying presence of excess free probe in the titration time course ( Figure 3b ) , the signal is presented as Δ520 nm absorbance compared to a probe-only reference . We defined this as ΔA520 = A520raw - A520inert – A520binding ( t=0 ) where A520inert accounts for the absorbance of dye that does not participate in binding ( because it was either the inactive enantiomer or present in superstoichiometric quantities ) , and A520binding ( t=0 ) is the initial absorbance of the fraction of HAP-TAMRA that will participate in binding . This analysis assumes linear binding until saturation , based on the results in Figure 2d , but could underestimate the signal if binding were weaker . | Viruses are simple structures formed of genetic information wrapped inside a shell . For the hepatitis B virus , this casing looks like a soccer ball . It is composed of 240 copies of the same protein , arranged in a pattern of pentagons and hexagons . These proteins form a protective shield for the virus’ genetic information: they also interact with the cells of the host during key events of the virus’ life cycle . When the hepatitis B virus infects a cell , it hijacks the cellular machinery to replicate . New shell proteins are produced and assemble within the cell . A type of potential antiviral drug called a CpAM disrupts this process: it causes the shell to assemble too early and inaccurately , which impairs the life cycle of the virus . However , a CpAM can bind to the shell even after it has already assembled . How this binding affects the virus is still unclear . Here , Schlicksup et al . attach a fluorescent molecule to a CpAM , and use a cutting-edge microscopy method to look at the structures at the atomic level . This makes it possible to examine in detail how the CpAM attaches to a correctly formed virus shell . Schlicksup et al . show that when the CpAM binds to the shell , it disrupts and sometimes even breaks the soccer-like pattern of the shell: the hexagons flatten , and the pentagons buckle . These misshaped shells could prevent the virus from interacting with the cellular structures necessary for infection or prevent it from releasing the virus’ genetic information . This is a new antiviral mechanism for a CpAM . By acting both before and after the shell has assembled , the CpAM targets the virus at different points of its life cycle . Hepatitis B affects over 240 million people worldwide . While a vaccine exists , there is still no cure for it . A better understanding of the physics of the virus’ shell and the mode of action of CpAMs could lead to better drugs against the disease . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"microbiology",
"and",
"infectious",
"disease"
] | 2018 | Hepatitis B virus core protein allosteric modulators can distort and disrupt intact capsids |
Anxiety results in sub-optimal motor learning , but the precise mechanisms through which this effect occurs remain unknown . Using a motor sequence learning paradigm with separate phases for initial exploration and reward-based learning , we show that anxiety states in humans impair learning by attenuating the update of reward estimates . Further , when such estimates are perceived as unstable over time ( volatility ) , anxiety constrains adaptive behavioral changes . Neurally , anxiety during initial exploration increased the amplitude and the rate of long bursts of sensorimotor and prefrontal beta oscillations ( 13–30 Hz ) . These changes extended to the subsequent learning phase , where phasic increases in beta power and burst rate following reward feedback were linked to smaller updates in reward estimates , with a higher anxiety-related increase explaining the attenuated belief updating . These data suggest that state anxiety alters the dynamics of beta oscillations during reward processing , thereby impairing proper updating of motor predictions when learning in unstable environments .
Anxiety involves anticipatory changes in physiological and psychological responses to an uncertain future threat ( Grupe and Nitschke , 2013; Bishop , 2007 ) . Previous studies have established that trait anxiety interferes with prefrontal control of attention in perceptual tasks , whereas state anxiety modulates the amygdala during detection of threat-related stimuli ( Bishop , 2007; Bishop , 2009 ) . An emerging literature additionally identifies the dorsomedial and dorsolateral prefrontal cortex ( dmPPC and dlPFC ) and the dorsal anterior cingulate cortex ( dACC ) as central brain regions modulating sustained anxiety , both in subclinical and clinical populations ( Robinson et al . , 2019 ) . Computational modeling work has started to examine the mechanisms through which anxiety might impair learning , revealing that individuals with high trait anxiety do not correctly estimate the likelihood of outcomes during aversive or reward learning in uncertain environments ( Browning et al . , 2015; Huang et al . , 2017; Pulcu and Browning , 2019 ) . In the area of motor control , research has shown that stress and anxiety have detrimental effects on performance ( Baumeister , 1984; Beilock and Carr , 2001 ) . These results have been interpreted as anxiety interferring with information-processing resources , and as a shift towards an inward focus of attention and an increase in conscious processing of movement ( Eysenck and Calvo , 1992; Pijpers et al . , 2005 ) . The effects of anxiety on motor learning are , however , often inconsistent , and a mechanistic understanding of these effects is still lacking . Delineating mechanisms through which anxiety influences motor learning is important to ameliorate the impact of anxiety in different settings , including in motor rehabilitation programs . Motor variability could be one component of motor learning that is affected by anxiety; it is defined as the variation of performance across repetitions ( van Beers et al . , 2004 ) , and is affected by various factors including sensory and neuromuscular noise ( He et al . , 2016 ) . As a form of action exploration , movement variability is increasingly recognized to benefit motor learning ( Todorov and Jordan , 2002; Wu et al . , 2014; Pekny et al . , 2015 ) , particularly during reward-based learning , with discrepant effects in motor adaptation paradigms ( He et al . , 2016; Singh et al . , 2016 ) . These findings are consistent with the vast amount of research on reinforcement learning that demonstrates increased learning following initial exploration ( Sutton and Barto , 1998; Olveczky et al . , 2005 ) . Yet contextual factors can reduce variability . For instance , an induced anxiety state leads to ritualistic behavior , characterized by movement redundancy , repetition , and rigidity ( Lang et al . , 2015 ) . This finding resembles the reduction in behavioral variability and exploration that manifests across animal species during phasic fear in reaction to certain imminent threats ( Morgan and Tromborg , 2007 ) . On the basis of these results , we set out to test the hypothesis that state anxiety modulates motor learning through a reduction in motor variability . A second component that could be influenced by anxiety is the flexibility to adapt to changes in the task structure during learning . Individuals who are affected by anxiety disorders exhibit an intolerance of uncertainty , which contributes to excessive worry and emotional dysregulation ( Ouellet et al . , 2019 ) . Turning to non-clinical populations , computational studies have established that highly anxious individuals exhibit difficulties in estimating environmental uncertainty both in aversive and reward-based tasks ( Browning et al . , 2015; Huang et al . , 2017; Pulcu and Browning , 2019 ) . Failure to adapt to volatile or unstable environments thus impairs learning of action-outcome contingencies in these settings . Accordingly , in the context of motor learning , and more specifically , in reward-based motor learning , we proposed that an increase in anxiety would affect individuals’ estimation of uncertainty about the stability of the task structure , such as the rewarded movement . On the neural level , we posited that changes in motor variability are driven by activity in premotor and motor areas . Support for our hypothesis comes from animal studies demonstrating that variability in the primate premotor cortex tracks behavioral variability during motor planning ( Churchland et al . , 2006 ) . Further evidence supports the hypothesis that changes in variability in single-neuron activity in motor cortex drive motor exploration during initial learning , and reduce it following intensive training ( Mandelblat-Cerf et al . , 2009; Santos et al . , 2015 ) . In addition , the basal ganglia are crucial for modulating variability during learning and production , as shown in songbirds and , indirectly , in patients with Parkinson’s disease ( Kao et al . , 2005; Olveczky et al . , 2005; Pekny et al . , 2015 ) . In the present study , we analyzed sensorimotor beta oscillations ( 13–30 Hz ) as a candidate brain rhythm associated with the modulation of motor exploration and variability . Beta oscillations are modulated with different aspects of performance and motor learning ( Herrojo Ruiz et al . , 2014; Bartolo and Merchant , 2015; Tan et al . , 2014 ) , as well as in reward-based learning ( HajiHosseini et al . , 2012 ) . Increases in sensorimotor beta power following movement have been proposed to signal greater reliance on prior information about the optimal movement ( Tan et al . , 2016 ) , which would reduce the impact of new evidence on the update of motor commands . We therefore tested the additional hypothesis that changes in sensorimotor beta oscillations mediate the effect of anxiety on belief updates and the estimation of uncertainty driving reward-based motor learning . Crucially , in addition to assessing sensorimotor brain regions , we were interested in prefrontal areas because of prior work in clinical and subclinical anxiety linking the prefrontal cortex ( dmPFC and dlPFC ) and the dACC to the maintenance of anxiety states , including worry and threat appraisal ( Grupe and Nitschke , 2013; Robinson et al . , 2019 ) . Thus , beta oscillations across sensorimotor and prefrontal electrode regions were evaluated . Traditionally , the primary focus of research on oscillations was on power changes , although there is a renewed interest in assessing dynamic properties of oscillatory activity , such as the presence of brief bursts ( Poil et al . , 2008 ) . Brief oscillation bursts are considered to be a central feature of physiological beta waves in motor-premotor cortex and the basal ganglia ( Feingold et al . , 2015; Tinkhauser et al . , 2017; Little et al . , 2018 ) . Accordingly , we assessed both the power and burst distribution of beta oscillations to capture dynamic changes in neural activity that were induced by anxiety and their link to behavioral effects . To test our hypotheses , we recorded electroencephalography ( EEG ) in three groups of participants while they completed a reward-based motor sequence learning paradigm , with separate phases for motor exploration ( without reinforcement ) and reward-based learning ( using reinforcement ) . We manipulated anxiety by informing participants about an upcoming public speaking task ( Lang et al . , 2015 ) . Using a between-subject design , the anxiety manipulation targeted either the motor exploration or the reward-based learning phase . Analysis of the EEG signals aimed to assess anxiety-related changes in the power and burst distribution in sensorimotor and prefrontal beta oscillations in relation to changes in behavioral variability and reward-based learning .
The analysis of the EEG signals focused on sensorimotor and prefrontal ( anterior ) beta oscillations and aimed to assess separately ( i ) tonic and ( ii ) phasic ( or event-related ) changes in spectral power and burst rate . Tonic changes in average beta activity would be an indication that the anxiety manipulation had an effect on the general modulation of underlying beta oscillatory properties . Complementing this analysis , assessment of the phasic changes in the measures of beta activity during trial performance and following feedback presentation allowed us to investigate the neural processes that drive reward-based motor learning and their alteration by anxiety . These analyses focused either on all channels ( tonic changes ) or on a subset of channels across contralateral sensorimotor cortices and anterior regions ( phasic changes; see statistical analysis details in 'Materials and methods' ) .
The results revealed several interrelated mechanisms through which state anxiety impairs reward-based motor learning . First , state anxiety reduced motor variability during an initial exploration phase . This was associated with limited improvement in scores during subsequent learning . Second , the smaller change in the expectation of reward throughout time led to a decrease in the expectation of volatility . Along with those results , we observed an overestimation of uncertainty about volatility due to state anxiety , which promoted the drop in the volatility estimate . Additional computational results demonstrated that larger precision-weighted prediction errors relating to reward and volatility had the effect of constraining the trial-to-trial behavioral adaptations in state anxiety . This contrasted with the findings for volatility in control participants , where larger pwPE relating to this quantity promoted behavioral exploration . On the neural level , anxiety during initial exploration was associated with elevated sensorimotor beta power and a distribution of bursts of sensorimotor beta oscillations with a longer tail ( smaller scaling exponent ) . The latter result indicated a more frequent presence of longer bursts , resembling recent findings of abnormal burst duration in movement disorders ( Tinkhauser et al . , 2017 ) . The anxiety-induced higher rate of long burst events and higher beta power during initial exploration also manifested in prefrontal electrodes and extended to the following learning phase , where phasic trial-by-trial feedback-locked increases in these measures accounted for the attenuated updating of expectation on reward . These results provide the first evidence that state anxiety induces changes in the distribution of sensorimotor and prefrontal beta bursts , as well as in beta power , which may account for the observed deficits in the update of predictions during reward-based motor learning . Evidence from our main experiment suggested that the finding of anxiety-related reduced motor variability during exploration was associated with the outcome of subsequently impaired learning from reward . These results validate previous accounts on the relationship between motor variability and Bayesian inference ( Wu et al . , 2014 ) . In addition , the association between larger initial task-related variability and higher scores during the following learning phase extends results on the faciliatory effect of exploration on motor learning , at least in tasks that require learning from reinforcement ( Wu et al . , 2014; Pekny et al . , 2015; Dhawale et al . , 2017; see also critical view in He et al . , 2016 . Crucially , state anxiety constrained the total amount of task-related variability only when induced during the initial exploration phase . The lack of between-group differences in cvIKI during learning in both experiments suggests that this measure could not account for the anxiety-related deficits in reward-based learning . Our Bayesian learning model provided additional insight on this aspect . The modelling results suggested that state anxiety can impair learning from reward not only by influencing the posterior distributions of beliefs ( expectations and uncertainty ) but also by altering how pwPE relating to those beliefs affect behavioral variability . The response model consistently demonstrated in experimental and control groups that smaller pwPEs driving reward updates on the previous trial ( leading to decreased expectation of reward ) were followed by an increase in task-related motor variability ( higher exploration ) . On the other hand , trials of larger pwPE relating to reward were followed by reduced task-related behavioral changes . By contrast , the effect of pwPE on volatility differed substantially in control and anxiety groups . Although large pwPEs on volatility promoted subsequent larger task-related behavioral changes in control participants , they constrained behavioral exploration in the anx1 and anx2 groups . Accordingly , state anxiety facilitated the use of task-related variability during reward-based learning only in trials following smaller pwPE reducing volatility estimates . This led participants who were affected by the prior or concurrent state anxiety manipulation to underestimate environmental volatility . Thus , they had the expectation that reward estimates are more stable throughout time . Anx1 and anx2 participants also had larger uncertainty about volatility . This implies that they were less confident about their volatility estimate , and allowed for a greater influence of new information in updating this quantity . This finding is additionally reinforced in anx1 by the result of a larger ω2 , reflecting a different learning style that is characterized by sharper and more pronounced steps of update in μ2 . The results align well with recent computational work in decision-making tasks , showing that high trait anxiety leads to alterations in uncertainty estimates and adaptation to the changing statistical properties of the environment ( Browning et al . , 2015; Huang et al . , 2017; Pulcu and Browning , 2019 . Notwithstanding the similarities in the anx1 and anx2 groups concerning the expectation of volatility and associated uncertainty , the fact that anx2 participants achieved high scores in the task and were not impaired in learning requires further clarification . Our post-hoc analyses revealed that the drop in μ2 in anx2 could be accounted for by the narrower distribution of scores encountered by this group . In addition , these participants introduced smaller trial-to-trial changes in temporal variability when compared to control participants . Thus , anx2 participants had a tendency to exploit the current motor program more than control participants , suggesting a more conservative approach to success . Anx1 participants also introduced smaller trial-to-trial changes in trial-wise temporal variability ( cvIKItrial ) , yet their behavioral changes had a slower benefit on reward . In both groups , however , the more pronounced tendency to exploit the current motor program was associated with alterations in how pwPE relating to volatility influenced behavioral changes . Overall , our findings provide the first evidence that computational mechanisms similar to those described for trait anxiety and decision-making underlie the effect of temporary anxious states on motor learning . This might be particularly the case in the context of learning from rewards , such as feedback about success or failure , which is considered one of the fundamental processes through which motor learning is accomplished ( Wolpert et al . , 2011 ) . Previous studies manipulating psychological stress and anxiety to assess motor learning showed both a deleterious and a faciliatory effect ( Hordacre et al . , 2016; Vine et al . , 2013; Bellomo et al . , 2018 ) . Differences in experimental tasks , which often assess motor learning during or after high-stress situations but not during anxiety induction in anticipation of a stressor , could account for the previous mixed results . Here , we adhered to the neurobiological definition of anxiety as a psychological and physiological response to an upcoming diffuse and unpredictable threat ( Grupe and Nitschke , 2013; Bishop , 2007 ) . Accordingly , anxiety was induced using the threat of an upcoming public speaking task ( Feldman et al . , 2004; Lang et al . , 2015 ) , and was associated with a drop in the HRV and an increase in state anxiety scores during the targeted blocks . Although the average state anxiety scores were not particularly high , they were significantly higher during the targeted phases than during the initial resting state phase . Future studies should use more impactful stressors to study the effect of the full spectrum of state ( and trait ) anxiety on motor learning ( Bellomo et al . , 2018 ) . What is the relationship between the expression of motor variability and state anxiety ? As hypothesized , state anxiety during initial exploration reduced the use of variability across trials . This converges with recent evidence demonstrating that anxiety leads to ritualistic behavior ( repetition , redundancy , and rigidity of movements ) that allow the subject to regain a sense of control ( Lang et al . , 2015 ) . The outcome also aligns well with animal studies in which evidence shows a reduction in motor exploration when the stakes are high ( high-reward situations , social context; Kao et al . , 2008; Dhawale et al . , 2017; Woolley et al . , 2014 ) . These interpretations , however , seem to stand in contrast with our findings in anx2 participants , who were affected by the anxiety manipulation during learning but with no significant effect on the total degree of motor variability expressed during this phase . Similar results were obtained in the second experiment , as anx3 and control participants did not differ in the amount of across-trials variability expressed during learning . Bayesian computational modelling clarified these findings demonstrating that anx2 participants used increased exploitation of their current motor program . Also , their trial-to-trial changes in temporal variability were smaller than those in the control group , particularly following large pwPEs that increased the expectation on volatility . This outcome was also found in both anx1 and anx3 participants in the second experiment . Thus , anxiety consistently constrained dynamic trial-to-trial changes in temporal variability—with these changes negatively influenced by pwPEs on volatility . Notably , however , the strategy in anx2 participants of more extensively exploiting the inferred rewarded solution ( relative to control participants ) was successful , and therefore differs from the learning impairment exhibited by anx1 participants . In the second experiment , removing the initial exploration phase led to impaired reward-based learning in anx3 participants . This group also tended to explore less than controls at the trial level as a function of changes in volatility pwPEs . Thus , the combined evidence suggests that normal use of initial variability in anx2 participants protected their performance from the subsequent impact of the anxiety manipulation . Initial use of variability in anx2 might promote faster learning of the mapping between actions and their asociated outcome , contributing to successful goal-directed exploitation . We interpret these results to indicate that initial unconstrained exploration is important for later subsequent successful motor learning . Some considerations should be taken into account . Task-related motor variability might be pivotal for learning from reinforcement or reward signals ( Sutton and Barto , 1998; Dhawale et al . , 2017; Wu et al . , 2014 ) , whereas in other contexts , such as during motor adaptation , the evidence is conflicting ( He et al . , 2016; Singh et al . , 2016 ) . An additional consideration is that greater levels of motor variability could reflect both an intentional pursuit of an explorative regime and an unintentional higher level of motor noise , in the latter case similar to that observed in previous work ( Wu et al . , 2014; Pekny et al . , 2015 ) . A recent study established that motor learning is improved by the use of intended exploration , not motor noise ( Chen et al . , 2017 ) . Our paradigm cannot dissociate intended and unintended exploration . This limitation will be addressed in future work by using a separate initial phase with regular performance to assess motor noise as a measure of unintended exploration . Another consideration is that our use of an initial exploration phase that did not provide reinforcement or feedback signals was motivated by the work of Wu et al . ( 2014 ) , which demonstrated a correlation between initial variability ( no feedback ) and learning curve steepness in a subsequent reward-based learning phase—a relationship previously observed in the zebra finch ( Kao et al . , 2005; Olveczky et al . , 2005; Ölveczky et al . , 2011 ) . This suggests that higher levels of motor variability do not solely amount to increased noise in the system . Instead , this variability represents a broader action space that can be capitalized upon during subsequent reinforcement learning by searching through previously explored actions ( Herzfeld and Shadmehr , 2014 ) . Accordingly , an implication of our results is that state anxiety could impair the potential benefits of an initial exploratory phase for subsequent learning . Last , we used a reward-based motor learning paradigm in which different performances could provide the same feedback score . The rationale for using this task was to explore the effect of state anxiety on volatility estimates , as recent work demonstrates that anxiety primarily affects learning in volatile conditions ( Browning et al . , 2015; Huang et al . , 2017 ) . This scenario , however , implied that a high expression of task-related motor variability during learning would be associated with a more volatile perception of the task , which is indeed supported by our correlation results . This could be a confounding factor when explaining the group effects . Importantly , however , further analyses revealed that the total degree of motor variability during learning and the mean learned performance did not differ between groups , suggesting that these are not confounding factors that could explain the reward-based-learning group results . Instead , our findings underscore that computational mechanisms related to how pwPE on reward and volatility influence behavioral changes are the main factors driving the effects of concurrent or prior state anxiety on reward-based motor learning . At the neural level , an important finding was that anxiety during initial exploration increased the power of beta oscillations and the rate of long beta bursts ( long-tailed distribution ) . The increases in power and the rate of long-lived bursts manifested after completion of the sequence , reflecting an anxiety-related enhancement of the post-movement beta rebound ( Kilavik et al . , 2012; Kilavik et al . , 2013 ) . This effect was observed in a region of contralateral sensorimotor and right prefrontal channels , and could be explained by anxiety alone , despite a small effect of motor variability on the modulation of these neural changes across sensorimotor electrodes . Further , larger sensorimotor beta power at the termination of the sequence performance was associated with a more constrained use of task-related variability . Our analyses did not provide a detailed anatomical localization of the effect , but the findings in sensorimotor regions that partially contribute to changes in motor variability are consistent with the involvement of premotor and motor cortex in driving motor variability and learning , as previously reported in animal studies ( Churchland et al . , 2006; Mandelblat-Cerf et al . , 2009; Santos et al . , 2015 ) . The results also converge with the representation in the premotor cortex of temporal and sequential aspects of rhythmic performance ( Crowe et al . , 2014; Kornysheva and Diedrichsen , 2014 ) . During learning , an unexpected result was that , in anx2 participants , there was an increase in beta power at the end of the sequence performance but not during feedback processing—and despite the anxiety manipulation successfully affecting the HRV . This outcome , as well as the lack of beta burst effects in this group , seems to be in agreement with the lack of learning impairments when compared with control participants . An additional unexpected result during learning blocks was the presence in anx1 participants of higher rates of long bursts and greater beta power at the end of the trial and during feedback processing , across both sensorimotor and prefrontal electrodes . These phasic changes in beta activity in anx1 participants extended from the previous phase , and the outcome aligns with the finding of prefrontal involvement in the emergence and maintenance of anxiety states ( Davidson , 2002; Grupe and Nitschke , 2013; Bishop , 2007 ) . Thus , our results revealed that , in the context of motor learning , anxious states induce changes in sensorimotor and prefrontal beta power and burst distribution . These changes are maintained after physiological measures of anxiety return to baseline , and thus continue to affect relevant behavioral parameters . Anxiety has been shown to modulate different oscillatory bands depending on the context , such as gamma activity in visual areas and amygdala when processing fearful faces ( Schneider et al . , 2018 ) , alpha activity in response to processing emotional faces ( Knyazev et al . , 2008 ) or theta activity during rumination ( Andersen et al . , 2009 ) . Beta-band oscillations could be particularly relevant to flesh out the effects of anxiety on performance during motor tasks . Mechanistically , phasic trial-by-trial feedback-locked changes in the sensorimotor beta power and burst distribution were related to the computational alterations in updating expectations on reward found in anx1 participants , and thus explained their poorer performance during reward-based learning . Specifically , a higher rate of long beta bursts and increased power following feedback were associated with a reduced update in the expectation of reward . The computational quantity that determines the update of expectations in our Bayesian model is the precision-weighted PEs . Here , pwPE relating to reward were inversely related to the rate of long beta bursts and beta power , and were therefore attenuated in anx1 participants because of their enhanced feedback-related beta activity . We found no significant contribution of pwPE relating to volatility to explaining changes in beta activity , suggesting that additional frequency ranges should be considered when linking hierarchical pwPEs to neural oscillations during learning . In the context of the predictive coding hypothesis , PEs ( or pwPEs ) are hypothesized to be mediated by gamma oscillations , whereas the neuronal signaling of predictions is mediated by lower frequencies ( e . g . , alpha 8–12 Hz , Friston et al . , 2015 ) . Further studies point to beta oscillations as the cortical oscillatory rhythm associated with encoding predictions , although the evidence to date is scarce ( Arnal and Giraud , 2012 ) . More recently , beta oscillations have been associated with the change to predictions rather than with predictions themselves ( Sedley et al . , 2016 ) , which is consistent with our findings as pwPEs were the quantities determining the change to predictions . In line with these results , a post-performance increase in beta power during motor adaptation is considered to index confidence in priors , and thus a reduced tendency to change the ongoing motor command ( Tan et al . , 2016 ) . More generally , beta oscillations along cortico-basal ganglia networks have been proposed to gate incoming information to modulate behavior ( Leventhal et al . , 2012 ) and to maintain the current motor state ( Engel and Fries , 2010 ) . Consequently , the phasic increase in beta power and the rate of beta bursts following feedback presentation could represent neural states that impair the encoding of pwPEs and the update of predictions about lower level quantities—reward here—induced by anxiety states . Notably , the modulation of feedback-locked beta activity was not explained by changes in pwPE relating to volatility . We speculate that the effect of reduced reward estimates on the expectation of volatility in the HGF suggests that abnormal increases in beta activity following feedback presentation indirectly influenced volatility estimates , while it had a direct effect on reward expectation . Our findings show that the assessment of neural activity in sensorimotor regions is crucial to understanding the effects of anxiety on motor learning and to determining mechanisms , above and beyond the role of prefrontal control of attention , in mediating the effects of anxiety on cognitive and perceptual tasks ( Bishop , 2007; Bishop , 2009; Eysenck and Calvo , 1992 ) . Our data imply that the combination of Bayesian learning models and analysis of oscillation properties can help us to better understand the mechanisms through which anxiety modulates motor learning . Future studies should investigate how the brain circuits that are involved in anxiety interact with motor regions to affect motor learning . In addition , assessing burst properties across both beta and gamma frequency ranges would further allow us to delineate and dissociate the neural mechanisms responsible for anxiety biasing decision-making and motor learning .
Sixty right-handed healthy volunteers ( 37 females ) aged 18 to 44 ( mean 27 years , SEM , 1 year ) participated in the main study . In a second , control experiment , 26 right-handed healthy participants ( 16 females , mean age 25 . 8 , SEM 1 , range 19–40 ) took part in the study . Participants gave written informed consent prior to the start of the experiment , which had been approved by the local Ethics Committee at Goldsmiths University . Participants received a base rate of either course credits or money ( 15 GBP; equally distributed across groups ) and were able to earn an additional sum up to 20 GBP during the task depending on their performance . We used pilot data from a behavioral study using the same motor task to estimate the minimum sample sizes for a statistical power of 0 . 95 , with an α of 0 . 05 , using the MATLAB ( The MathWorks , Inc , MA , USA ) function sampsizepwr . In the pilot study , we had one control and one experimental group of 20 participants each . In the experimental group , we manipulated the reward structure during the first reward-based learning block ( in this block , feedback scores did not count towards the final average monetary reward ) . For each behavioral measure ( motor variability and mean score ) , we extracted the standard deviation ( sd ) of the joint distribution from both groups and the mean value of each separate distribution ( e . g . , m1 , control; m2 , experimental ) , which provided the following minimum sample sizes: Between-group comparison of behavioral parameters ( using a two-tailed t-test ) : MinSamplSizeA = sampsizepwr ( 't' , [m1 sd] , m2 , 0 . 95 ) = 18–20 participants . Accordingly , we recruited 20 participants for each group in the main experiment . Next , using the behavioral data from the anxiety and control groups in the current main experiment , we estimated the minimum sample size for the second , behavioral control experiment: Between-group comparison of behavioral parameters ( using a two-tailed t-test ) : MinSamplSizeA = sampsizepwr ( ’t’ , [m1 sd] , m2 , 0 . 95 ) = 13 participants . Therefore , for the second control experiment , we recruited 13 participants for each group . Participants were seated at a digital piano ( Yamaha Digital Piano P-255 , London , UK ) and in front of a PC monitor in a light-dimmed room . They sat comfortably in an arm-chair with their forearms resting on the armrests of the chair . The screen displayed the instructions , feedback and visual cues for the start and end of a trial ( Figure 1A ) . Participants were asked to place four fingers of their right hand ( excluding the thumb ) comfortably on four pre-defined keys on the keyboard . Performance information was transmitted and saved as Musical Instrument Digital Interface ( MIDI ) data , which provided time onsets of keystrokes relative to the previous one ( inter-keystroke-interval—IKI in ms ) , MIDI velocities ( related to the loudness , in arbitrary units , a . u . ) , and MIDI note numbers that corresponded to the pitch . The experiment was run using Visual Basic , an additional parallel port and MIDI libraries . In all blocks , participants initiated the trial by pressing a pre-defined key with their left index finger . After a jittered interval of 1–2 s , a green ellipse appeared in the center of the screen representing the GO signal for task execution ( Figure 1A ) . Participants had 7 s to perform the sequence , which was ample time to complete it before the green circle turned red indicating the end of the execution time . If participants failed to perform the sequence in the correct order or initiated the sequence before the GO signal , the screen turned yellow . In blocks 2 and 3 during learning , performance-based feedback in the form of a score between 0 and 100 was displayed on the screen 2 s after the red ellipse , that is , 9 s from the beginning of the trial . The scores provided participants with information regarding the target performance . The performance measure that was rewarded during learning was the Euclidean norm of the vector corresponding to the pattern of temporal differences between adjacent IKIs for a trial-specific performance . Here , we denote the vector norm by ‖Δz‖ , with 𝚫𝐳 being the vector of differences , 𝚫𝐳= ( z2-z1 , z3-z2 , … , zn-zn-1 ) , and zi representing the IKI at each keystroke ( i=1 , 2 . . , n ) . Note that IKI values themselves represent the difference between the onset of consecutive keystrokes , and therefore 𝚫𝐳 indicates a vector of differences of differences . Specifically , the target value of the performance measure was a vector norm of 1 . 9596 ( e . g . , one of the maximally rewarded performances leading to this vector norm of IKI-differences would consist of IKI values: [0 . 2 , 1 , 0 . 2 , 1 , 0 , 2 , 1 , 0 . 2] s; that is a combination of short and long intervals ) . The score was computed in each trial using a measure of proximity between the target vector norm ‖Δzt‖ and the norm of the performed pattern of IKI differences ‖Δzp‖ , using the following expression: ( 3 ) score=100exp ( −|‖Δzt‖−‖Δzp‖| ) In practice , different temporal patterns leading to the same vector norm ‖Δzp‖ could achieve the same score . Participants were unaware of the existence of various solutions . Higher exploration across trials during learning could thus reveal that several IKI patterns were similarly rewarded . To account for this possibility , the perceived rate of change of the hidden goal ( environmental volatility ) during learning was estimated and incorporated into our mathematical description of the relationship between performance and reward ( see below ) . Anxiety was induced during block1 performance in group anx1 , and during block2 performance in the anx2 group by informing participants about the need to give a 2 min speech to a panel of experts about an unknown art object at the end of that block ( Lang et al . , 2015 ) . We specified that they would first see the object at the end of the block ( it was a copy of Wassily Kandinsky’ Reciprocal Accords [1942] ) and would have 2 min to prepare for the presentation . Participants were told that the panel of experts would take notes during their speech and would be standing in front of the testing room ( due to the EEG setup participants had to remain seated in front of the piano ) . Following the 2 min preparation period , participants were informed that due to the momentary absence of panel members , they instead had to present in front of the lab members . Participants in the control group had the task of describing the artistic object to themselves , and not in front of a panel of experts . They were informed about this secondary task before the beginning of block1 . To assess state anxiety , we acquired two types of data: ( 1 ) the short version of the Spielberger State-Trait Anxiety Inventory ( STAI , state scale X1 , 20 items; Spielberger , 1970 ) and ( 2 ) a continuous electrocardiogram ( ECG , see EEG , ECG and MIDI recording session ) . The STAI X1 subscale was presented four times throughout the experiment . A baseline assessment at the start of the experiment before the resting state recording was followed by an assessment immediately before each experimental block to determine changes in anxiety levels . In addition , a continuous ECG recording was obtained during the resting state and three experimental blocks were used to assess changes in autonomic nervous system responses . The indexes of heart rate variability ( HRV , coefficient of variation of the inter-beat-interval ) and mean heart rate ( HR ) were evaluated , as their reduction has been linked to changes in anxiety state due to a stressor ( Feldman et al . , 2004 ) . Here , we provide details on the computational Bayesian model that we adopted to estimate participant-specific belief trajectories , determined by the mean ( expectation ) and variance ( uncertainty ) of the posterior distribution . The model was implemented using the HGF toolbox for MATLAB ( http://www . translationalneuromodeling . org/tapas/ ) . The model consists of a perceptual and a response model , representing an agent ( a Bayesian observer ) who generates behavioral responses on the basis of a sequence of sensory inputs that it receives . In many implementations of the HGF , the sensory input is replaced with a series of outcomes ( e . g . feedback , reward ) associated with participants’ responses ( de Berker et al . , 2016; Diaconescu et al . , 2017 ) . As general notation , we let lowercase italics denote scalars ( x ) , which can be further characterized by a trial superscript xk and a subscript i denoting the level in the hierarchy xik ( i = 1 , 2 ) . The HGF corresponds to the perceptual model , representing a hierarchical belief-updating process , that is a process that infers hierarchically related environmental states that give rise to sensory inputs ( Stefanics et al . , 2018; Mathys et al . , 2014 ) . In the version for continuous inputs ( see Mathys et al . , 2014; function tapas_hgf . m ) , we used the series of feedback scores as input: uk=score; normalized to range 0–1 . From the series of inputs , the HGF then generates belief trajectories about external states , such as the reward value of an action or a stimulus . Learning occurs in two hierarchically coupled levels ( x1 , x2 ) , one for ‘perceptual’ beliefs ( x1: the reward associated with the current performance ) , and the phasic volatility of those beliefs ( x2 ) . These two levels evolve as coupled Gaussian random walks , with the lower level coupled to the higher level through its variance ( inverse precision ) . The Gaussian random walk at each level xi is determined by its posterior mean ( μi ) and its variance ( σi ) . Further , the variance of the lower level , x1 , depends on x2 through an exponential function:f ( x2 ) =exp ( κx2+ω1 ) where κ was fixed to 1 and ω1 is a model parameter that was estimated for each participant by fitting the HGF model to the experimental data ( scores and responses ) using Variational Bayes . At the top level , the variance is typically fixed to a constant parameter , ϑ=exp ( ω2 ) , where ω2 is also a free paratemer to be estimated in each individual . The specific coupling between levels indicated above has the advantage of allowing simple variational inversion of the model and the derivation of one-step update equations under a mean-field approximation . This is achieved by iteratively integrating out all previous states up to the current trial k ( see appendices in Mathys et al . , 2014 ) . Importantly , the update equations for the posterior mean at level i and for trial k depend on the prediction errors weighted by uncertainty σi ( or its inverse , precision πi=1/σi ) according to the following expression: ( 5 ) Δμik=μik-μik-1∝π^i-1kπikδi-1k The first term in the above expression is the change in the expectation for state xi on trial k , μik , relative to the prediction on trial k-1 , μik-1 . The prediction on trial k-1 is denoted by the ‘hat’ or diacritical mark ^ , μik-1=μ^ik . The term prediction thus refers to the expectation of xi before seeing the feedback score from the current trial: it corresponds with the mean of the posterior distribution of xi up to trial k-1 . By contrast , the term expectation refers to the mean of the posterior distribution of xi up to trial k . The difference term Δμik is proportional to the prediction error of the level below , δi-1k , representing the difference between the expectation μi-1k and the prediction μ^i-1k of the level below . The prediction error is weighted by the ratio between the prediction of the precision of the level below , π^i-1k , and the precision of the current belief , πik . Thus the product of the precision weights and the prediction error constitute the precision-weighed prediction error ( pwPE ) , which therefore regulates the update of expectations on trial k: Δμik=ϵik . The pwPE expressions for level 1 and 2 are defined below in Equation 14 and Equation 15 . Equation 5 illustrates that higher uncertainty in the current level ( σik , lower πik in the denominator ) leads to faster update of beliefs; moreover , smaller uncertainty ( higher precision ) of the prediction of the level below also increases the update of beliefs . For the two-level HGF model for continuous inputs , the generic equation Equation 5 takes the explicit forms shown below ( Equation 6 and Equation 10; equations taken directly from the TAPAS toolbox; see also Mathys et al . , 2011; Mathys et al . , 2014 ) . Updates of expectations for level 1: ( 6 ) μ1k=μ^1k+π^ukπ1kδuk , with π^uk representing the prediction of the precision of the input ( feedback scores; see Table 1 ) and δuk the prediction error about the input: ( 7 ) δuk=uk-μ^1k , Precision updates for level 1: ( 8 ) π1k=π^1k+πuk , where π^1k is defined as ( using ρ=0 , κ=1 , tk=1 ) : ( 9 ) π^1k=1 ( 1π1k-1+exp ( μ2k-1+ω1 ) ) , Update of expectations for level 2: ( 10 ) μ2k=μ^2k+121π2kw1kδ1k , with ( 11 ) w1k=exp ( μ2k-1+ω1 ) π^1k Precision updates for level 2: ( 12 ) π2k=π^2k+12w1k ( w1k+ ( 2w1k-1 ) δ1k ) , and ( 13 ) π^2k=11π2k-1+exp ( ω2 ) . From Equation 6 and Equation 10 , it follows that the pwPEs for level 1 and 2 , ϵ1 and ϵ2 , respectively , are: ( 14 ) ϵ1k=μ1k−μ^1k=π^ukπ1kδuk , ( 15 ) ϵ2k=μ2k-μ^2k=121π2kw1kδ1k . Next , we mapped the expectation on the inferred perceptual beliefs , reward μ1 and volatility μ2 , and the corresponding pwPEs to the performance output that the participant generates during every trial using a separate response model . We adapted the family of response models used by Marshall et al . ( 2016 ) to our task . In that work , the authors explained participant’s observed log ( RT ) responses on a trial-by-trial basis as a linear function of various HGF quantities using a multiple regression . We implemented similar models , but adapted them to our task ( new scripts are available in the Open Science Framework Data Repository: https://osf . io/sg3u7/ ) . The models we tested used two different performance parameters: The coefficient of variation of inter-keystroke intervals , cvIKItrial , as a measure of the extent of timing variability within the trial . The logarithm of the mean performance tempo in a trial , log ( mIKItrial ) , with IKI in milliseconds . We were interested in how HGF quantities on the previous trial explained changes in the performance parameters in the subsequent trial and therefore used these dependent variables:ΔcvIKItrialk=cvIKItrialk-cvIKItrialk-1Δlog ( mIKItrial ) k=log ( mIKItrialk ) -log ( mIKItrialk-1 ) For each of those two performance measures , the corresponding response model was a function of a constant component of the performance measure ( intercept ) and HGF quantities on the previous trial , such as: the expectation on reward ( μ1 ) , the expectation on volatility ( μ2 ) , the precision-weighted PE relating to reward ( ϵ1 ) , or the precision-weighted PE relating to volatility ( ϵ2 ) . In total , we assessed the following two families of four alternative response models HGF11-14 and HGF21-24 . Model HGF11:ΔcvIKItrialk=β0+β1μ1k-1+β2ϵ1k-1+ζ Model HGF12: ( 16 ) ΔcvIKItrialk=β0+β1μ1k-1+β2μ2k-1+ζ Model HGF13: ( 17 ) ΔcvIKItrialk=β0+β1μ2k-1+β2ϵ2k-1+ζ Model HGF14:ΔcvIKItrialk=β0+β1ϵ1k-1+β2ϵ2k-1+ζ Model HGF21: ( 18 ) Δlog ( mIKItrial ) k=β0+β1μ1k-1+β2ϵ1k-1+ζ Model HGF22: ( 19 ) Δlog ( mIKItrial ) k=β0+β1μ1k-1+β2μ2k-1+ζ Model HGF23: ( 20 ) Δlog ( mIKItrial ) k=β0+β1μ2k-1+β2ϵ2k-1+ζ Model HGF24: ( 21 ) Δlog ( mIKItrial ) k=β0+β1ϵ1k-1+β2ϵ2k-1+ζ The priors on the model parameters ( ω1 , ω2 ) , the response model parameters ( β0 , β1 , β2 , ζ ) , the initial expected states ( μ10 , μ20 ) and the precision of the input ( πu ) are provided in Table 1 . All priors are Gaussian distributions in the space in which they are estimated and are therefore determined by their mean and variance . The variance is relatively broad to let the priors be modified by the series of inputs ( feedback scores ) . Quantities that need to be positive ( e . g . , the variance or uncertainty of belief trajectories ) are estimated in the log-space , whereas general unbounded quantities are estimated in their original space . We used Random Effects Bayesian Model Selection ( BMS ) to assess the different models of learning at the group level ( Stephan et al . , 2009; code freely available from the MACS toolbox , Soch and Allefeld , 2018 ) . First , the log-model evidence ( LME ) values for models HGF11-14 were combined to get the log-family evidence ( LFE ) , and similarly for models HGF21-24 . The LFE values were subsequently compared using BMS to assess which family of models provided more evidence . BMS generated ( i ) the estimated model-family frequencies , that is , how frequently each family of models is optimal in the sample of participants; and ( ii ) the exceedance probabilities , reflecting the posterior probability that one family is more frequent than the others ( Soch et al . , 2016 ) . In the winner family , additional BMS determined the final optimal model . EEG and ECG signals were recorded using a 64-channel ( extended international 10–20 system ) EEG system ( ActiveTwo , BioSemi Inc ) placed in an electromagnetically shielded room . During the recording , the data were high-pass filtered at 0 . 16 Hz . The vertical and horizontal eye-movements ( EOG ) were monitored by electrodes above and below the right eye and from the outer canthi of both eyes , respectively . Additional external electrodes were placed on both left and right earlobes as reference . The ECG was recorded using two external channels with a bipolar ECG lead II configuration . The sampling frequency was 512 Hz . Onsets of visual stimuli , key presses and metronome beats were automatically documented with markers in the EEG file . The performance was additionally recorded as MIDI files using the software Visual Basic and a standard MIDI sequencer program on a Windows Computer . We used MATLAB and the FieldTrip toolbox ( Oostenveld et al . , 2011 ) for visualization , filtering and independent component analysis ( ICA; runica ) . The EEG data were highpass-filtered at 0 . 5 Hz ( Hamming windowed sinc finite impulse response [FIR] filter , 3380 points ) and notch-filtered at 50 Hz ( 847 points ) . Artifact components in the EEG data related to eye blinks , eye movements and the cardiac-field artifact were identified using ICA . Following IC inspection , we used the EEGLAB toolbox ( Delorme and Makeig , 2004 ) to interpolate missing or noisy channels using spherical interpolation . Finally , we transformed the data into common average reference . Analysis of the ECG data with FieldTrip focused on detection of the QRS-complex to extract the R-peak latencies of each heartbeat and use them to evaluate the HRV and HR measures in each experimental block . We first assessed the standard power spectral density ( PSD , in mV2/Hz ) of the continuous raw data in each performance block and separately for each group . The PSD was computed with the standard fast Fourier Transform ( Welch method , Hanning window of 1 s with 50% overlap ) . The raw PSD estimation was normalized into decibels ( dB ) with the average PSD from the initial rest recordings ( 3 min ) . Specifically , the normalized PSD during the performance blocks was calculated as ten times the base-10 logarithm of the quotient between the performance-block PSD and the resting state power . In addition , we assessed the time course of the spectral power over time during performance . Trials during sequence performance were extracted from −1 to 11 s locked to the GO signal . This interval included the STOP signal ( red ellipse ) , which was displayed at 7 s , and—exclusively in learning blocks—the score feedback , which was presented at 9 s . Thus , epochs were effectively also locked to the STOP and Score signals . Artifact-free EEG epochs were decomposed into their time-frequency representations using a 7-cycle Morlet wavelet in successive overlapping windows of 100 ms within the total 12s-epoch . The frequency domain was sampled within the beta range from 13 to 30 Hz at 1 Hz intervals . For each trial , we thus obtained the complex wavelet transform , and computed its squared norm to extract the wavelet energy ( Ruiz et al . , 2009 ) . The time-varying spectral power was then simply estimated by averaging the wavelet energy across trials . This measure of spectral power was further averaged within the beta-band frequency bins and normalized by subtracting the mean and dividing by the standard deviation of the power estimate in the pre-movement baseline period ( [−1 , 0] s prior to the GO signal ) . We estimated the distribution , onset and duration of oscillation bursts in the time series of beta-band amplitude envelope . We followed a procedure adapted from previous work to identify oscillation bursts ( Poil et al . , 2008; Tinkhauser et al . , 2017 ) . In brief , we used as threshold the 75% percentile of the amplitude envelope of beta oscillations . Amplitude values above this threshold were considered to be part of an oscillation burst if they extended for at least one cycle ( 50 ms , as a compromise between the duration of one 13 Hz-cycle [76 ms] and 30 Hz-cycle [33 ms] ) . Threshold-crossings that were separated by less than 50 ms were considered to be part of the same oscillation burst . As an additional threshold , the median amplitude was used in a control analysis , which revealed qualitatively similar results , as expected from previous work ( Poil et al . , 2008 ) . Importantly , because threshold crossings are affected by the signal-to-noise ratio in the recording , which could vary between the different performance blocks , we selected a common threshold from the initial rest recordings separately for each participant ( Tinkhauser et al . , 2017 ) . Distributions of the rate of oscillation bursts per duration were estimated using equidistant binning on a logarithmic axis with 20 bins between 50 ms and 2000 ms . General burst properties were assessed during exploration and learning blocks separately , first as averaged values within the full block-related recording , and next as phasic changes over time during trial performance . Trial-based analysis focused on the interval 0–11000 ms following the GO signal , which included the time window following the STOP signal ( at 7000 ms: exploration and learning blocks ) and the score feedback ( at 9000 ms: learning blocks ) . Statistical analysis of behavioral and neural measures focused on the separate comparison between each experimental group and the control group ( contrasts: anx1 – controls , anx2 –controls ) . Differences between experimental groups , anx1 – anx2 , were evaluated exclusively concerning the overall achieved monetary reward . We used non-parametric pair-wise permutation tests to assess differences between conditions or between groups in the statistical analysis of behavioral or computational measures . When multiple testing was performed , we implemented a control of the false discovery rate ( FDR ) at level q = 0 . 05 using an adaptive linear step-up procedure ( Benjamini et al . , 2006 ) . This control provided an adapted threshold p-value ( termed PFDR ) . Further , to evaluate differences between sets of multi-channel EEG signals corresponding to two conditions or groups , we used two approaches: Non-parametric effect size estimators were used in association with our pair-wise nonparametric statistics , following Grissom and Kim , 2012 . In the case of between-subject comparisons , the standard probability of superiority , Δ , was used . Δ is defined as the proportion of greater values in sample B relative to A , when values in samples A and B are not paired: Δ=P ( A>B ) ranges from 0 to 1 . The total number of comparisons is the product of the size of sample A and sample B ( Ntot=sizeA*sizeB ) , and therefore , Δ=N ( A>B ) /Ntot . In the case of ties , Δ is corrected by subtracting in the denominator the number of ties from the total number of comparisons ( Ntot-Nties ) . For within-subject comparisons , we used the probability of superiority for dependent samples , Δdep , which is the proportion of all within-subject ( paired ) comparisons in which the values for condition B are larger than for condition A . 95% confidence intervals ( termed simply CI ) for Δ and Δdep were estimated with bootstrap methods ( Ruscio and Mullen , 2012 ) . Last , associations between parameters were quantified using non-parametric rank correlations ( Spearman ρ ) , which are robust against outliers . However , we used linear correlations in the case of multiple linear regressions for the HGF response model , following Marshall et al . ( 2016 ) . | Feeling anxious can hinder how well someone performs a task , a phenomenon that is sometimes called “choking under pressure” . Anxiety may also impair a person’s ability to learn a new manual task , like juggling or playing the piano; however , it remains unclear exactly how this happens . People learn manual tasks more quickly if they can practice first , and the more someone varies their movements during these trial runs , the faster they learn afterwards . Yet , anxiety can affect movement; for example , anxious people often make repetitive motions like hand-wringing or fidgeting . There is also evidence that very anxious people may learn less from the outcomes of their actions . To understand how anxiety may affect the learning of manual tasks , Sporn et al designed experiments where people learned to play a short sequence of notes on a piano . The main experiment involved 60 participants and was split over two phases . In the first ‘exploration’ phase , participants had to play the piano sequence using any timing they liked and were encouraged to explore different rhythms . In the second ‘learning’ phase , participants were rewarded with a higher score the closer they got to playing the notes with a certain rhythm , without being told that this was their specific goal . To see how anxiety affected performance , the participants were split into three groups . One group were told in the initial exploration phase that they would give a public talk after they completed the piano task , which reliably made them more anxious . A second group were told about the anxiety-inducing public speaking only during the learning phase; while a third group – the controls – were not aware of any public speaking task . People in the second group could learn the rhythm as well as the controls . Participants who were made anxious during the exploration phase , however , scored fewer points and were less likely to learn the piano sequence in the second phase . They also varied their movements less in the first phase . As a follow-up , Sporn et al . repeated the experiment with 26 people but without the initial exploration phase . This time the anxious participants were less able to learn the piano sequence and scored fewer points . This suggests that the initial exploration in the previous experiment had enabled later anxious participants to succeed in the learning phase despite being anxious . Finally , Sporn et al . also used a technique called electroencephalography ( or EEG for short ) to record brain activity and observed differences in participants with and without anxiety , particularly when they received their scores . The EEG signals showed that anxiety altered rhythmic patterns of brain activity called “sensorimotor beta oscillations” , which are known to be involved in both movement and learning . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2020 | Alterations in the amplitude and burst rate of beta oscillations impair reward-dependent motor learning in anxiety |
Heat shock factor 1 ( HSF1 ) , a key regulator of transcriptional responses to proteotoxic stress , was linked to estrogen ( E2 ) signaling through estrogen receptor α ( ERα ) . We found that an HSF1 deficiency may decrease ERα level , attenuate the mitogenic action of E2 , counteract E2-stimulated cell scattering , and reduce adhesion to collagens and cell motility in ER-positive breast cancer cells . The stimulatory effect of E2 on the transcriptome is largely weaker in HSF1-deficient cells , in part due to the higher basal expression of E2-dependent genes , which correlates with the enhanced binding of unliganded ERα to chromatin in such cells . HSF1 and ERα can cooperate directly in E2-stimulated regulation of transcription , and HSF1 potentiates the action of ERα through a mechanism involving chromatin reorganization . Furthermore , HSF1 deficiency may increase the sensitivity to hormonal therapy ( 4-hydroxytamoxifen ) or CDK4/6 inhibitors ( palbociclib ) . Analyses of data from The Cancer Genome Atlas database indicate that HSF1 increases the transcriptome disparity in ER-positive breast cancer and can enhance the genomic action of ERα . Moreover , only in ER-positive cancers an elevated HSF1 level is associated with metastatic disease .
Breast cancer is the most common malignancy in women worldwide . Four clinically relevant molecular types are distinguished based on the expression of estrogen receptors ( ERs ) and HER2 ( ERBB2 ) . Among them , luminal adenocarcinomas , characterized by the expression of estrogen receptors , constitute about 70% of all breast cancer cases . There are two classical nuclear estrogen receptors , ERα and ERβ ( encoded by ESR1 and ESR2 genes , respectively ) , and structurally different GPR30 ( GPER1 ) , which is a member of the rhodopsin-like family of the G protein-coupled and seven-transmembrane receptors . ERα expression is most common in breast cancer , and its evaluation is the basis for determining the ER status . The activity of estrogen receptors is modulated by steroid hormones , mainly estrogens , which are synthesized from cholesterol via androgens in the reaction catalyzed by aromatase ( Fuentes and Silveyra , 2019 ) . According to epidemiological and experimental data , estrogens alongside the mutations in BRCA1 and BRCA2 , CHEK2 , TP53 , STK11 ( LKB1 ) , PIK3CA , PTEN , and other genes , are key etiological factors of breast cancer development ( Yaşar et al . , 2017; Verigos and Magklara , 2015 ) . The mechanism of estrogen-stimulated breast carcinogenesis is not clear . According to the widely accepted hypothesis , estrogens acting through ERα stimulate cell proliferation and can support the growth of cells harboring mutations that then accumulate , ultimately resulting in cancer . Another hypothesis suggests the ERα-independent action of estrogens via their metabolites , which can exert genotoxic effects , contributing to cancer development ( Yager and Davidson , 2006; Pescatori et al . , 2021 ) . Nevertheless , hormonal therapies targeting either estrogen production ( i . e . , aromatase inhibitors ) or the hormone receptor itself such as selective ER modulators ( SERMs; i . e . , tamoxifen ) and selective ER degraders ( SERDs; i . e . , fulvestrant ) are widely used to block the mitogenic action of estrogens in patients with ER-positive breast cancer ( Renoir , 2012; Farcas et al . , 2021 ) , contributing to the decline in mortality from breast cancer in recent decades ( Iwase et al . , 2021 ) . Previously , we have found that the major female sex hormone 17β-estradiol ( E2 ) stimulates activation of heat shock factor 1 ( HSF1 ) in estrogen-dependent breast cancer cells via MAPK signaling ( Vydra et al . , 2019 ) . HSF1 is a well-known regulator of response to cellular stress induced by various environmental stimuli . It mainly regulates the expression of the heat shock proteins ( HSPs ) , which function as molecular chaperones and regulate protein homeostasis ( Ran et al . , 2007 ) . HSF1-regulated chaperones control , among others , the activity of estrogen receptors ( Echeverria and Picard , 2010 ) . ERs remain in an inactive state trapped in multimolecular chaperone complexes organized around HSP90 , containing p23 ( PTGES3 ) , and immunophilins ( FKBP4 or FKPB5 ) ( Segnitz and Gehring , 1995 ) . Upon binding to E2 , ERs dissociate from the chaperone complexes and become competent to dimerize and regulate the transcription . ERs bind DNA directly to the estrogen-response elements ( EREs ) , or indirectly , via tethering factors , and promote the transcription at either nearby promoters or through chromatin loops from distal enhancers . The dynamic action of ERs , which enables the adaptation of cancer cells and impacts the clinical outcome , relies on many transcriptional coactivators and corepressors ( Heldring et al . , 2007; Renoir , 2012; Farcas et al . , 2021 ) . HSP90 is essential for ERα hormone binding ( Fliss et al . , 2000 ) , dimer formation ( Powell et al . , 2010 ) , and binding to the EREs ( Inano et al . , 1994 ) . Also , the passage of the ER to the cell membrane requires association with the HSP27 ( HSPB1 ) oligomers in the cytoplasm ( Razandi et al . , 2010 ) . More than 20 chaperones and co-chaperones associated with ERα in human cells have been identified through a quantitative proteomic approach ( Dhamad et al . , 2016 ) , but their specific contribution in the receptor action still needs to be investigated . Moreover , HSF1 is involved in the regulation of a plethora of non-HSP genes , which support oncogenic processes: cell cycle regulation , signaling , metabolism , adhesion , and translation ( Mendillo et al . , 2012 ) . A high level of HSF1 expression was found in cancer cell lines and many human tumors ( Vydra et al . , 2014; De Thonel et al . , 2011 ) and was shown to be associated with the increased mortality of ER-positive breast cancer patients ( Santagata et al . , 2011; Gökmen-Polar and Badve , 2016 ) . E2-activated HSF1 is transcriptionally potent and takes part in the regulation of several genes essential for breast cancer cell growth ( Vydra et al . , 2019 ) . Furthermore , HSF1-regulated chaperones are necessary for ERα proper function . Thus , a hypothetical positive feedback loop between E2/ERα and HSF1 signaling may exist , which putatively supports the growth of estrogen-dependent tumors . Here , to study the cooperation of HSF1 and ERα in estrogen signaling and the influence of HSF1 on E2-stimulated transcription and cell growth and mobility , we created novel experimental models based on HSF1-deficient cells and performed an in-depth bioinformatics analysis of the relevant genomics data . We also compared the influence of HSF1 on ER-positive and ER-negative breast cancers transcriptomes from The Cancer Genome Atlas ( TCGA ) database .
To study the contribution of HSF1 in E2 signaling , we established MCF7 cell lines with reduced HSF1 expression . Firstly , we tested a few HSF1-targeting shRNAs ( Figure 1—figure supplement 1A ) . Then , the most potent variant that reduced HSF1 level about 10-fold ( termed afterward shHSF1 ) was chosen for further studies . Although the heat shock response was significantly reduced , the expression of HSP genes ( HSPA1A , HSPH1 , HSPB1 , and HSPB8 ) was still induced after this HSF1 knockdown ( Figure 1—figure supplement 1B ) . Thus , we additionally created MCF7 variants with HSF1 functional knockout using the CRISPR/Cas9 gene targeting approach ( clones arisen from two individual cells termed KO#1 and KO#2 afterward ) . Then , considering the slight differences between clones , we created an additional experimental model: six new individual HSF1-negative ( HSF1− ) and six HSF1-positive ( HSF1+ ) MCF7 clones obtained using the DNA-free CRISPR/Cas9 system ( which was more effective ) were pooled before analyses . The complete elimination of HSF1 ( Figure 1A , Figure 1—figure supplement 1A ) was connected with a substantial loss of inducibility of HSP genes ( Figure 1—figure supplement 1B ) and proteins ( HSP105/HSPH1 , HSP90 , HSP70/HSPA1 ) following hyperthermia ( Figure 1B ) . The ability of cells to form colonies in the clonogenic assay was reduced in all MCF7 experimental models of HSF1 depletion ( using shRNA and sgRNA; Figure 1C , Figure 1—figure supplement 1C ) . Moreover , the population size of ALDH-positive ( stem/progenitor ) cells correlated with the HSF1 level and was reduced in HSF1-deficient cells ( Figure 1—figure supplement 1D ) . Also , the increased contribution of cells in the G1 phase was associated with the HSF1 knockout ( Figure 1—figure supplement 1E ) . HSF1 knockdown did not affect the proliferation rate , while the functional HSF1 knockout led to a slight reduction in the proliferation rate under standard conditions ( this effect was not visible under less favorable growing conditions , i . e . , in phenol red-free 5% dextran-activated charcoal-stripped fetal bovine serum ( FBS ) ; Figure 1D , Figure 1—figure supplement 1F ) . To check if HSF1 deficiency would affect the growth of another ERα-positive cell line , we modified T47D cells using the CRISPR/Cas9 method ( Figure 1—figure supplement 2A ) . Under standard conditions , we did not observe differences between HSF1+ and HSF1− T47D cells in the proliferation and clonogenic assay ( not shown ) . Unlike MCF7 cells , HSF1− T47D cells grew slightly faster than HSF1+ cells , but this difference was not statistically significant and no differences were observed in the cell cycle ( Figure 1—figure supplement 2B and C ) . We have previously demonstrated that HSF1 was activated after E2 treatment of ERα-positive cells and it was able to bind to the regulatory sequences of several target genes , which correlated with the upregulation of their transcription ( Vydra et al . , 2019 ) . Since most of these genes code for proteins involved in E2 signaling , we expected that HSF1 downregulation could affect E2-dependent processes , especially cell proliferation . Therefore , we compared E2-stimulated proliferation of HSF1-proficient and HSF1-deficient MCF7 cells in all experimental models . HSF1 deficiency resulted in weaker growth stimulation by E2 ( than in the corresponding control cells ) , but a statistically significant difference was not observed in all experimental conditions/cell variants ( Figure 1D , Figure 1—figure supplement 1G ) . However , E2-stimulated proliferation was not significantly reduced in HSF1-deficient T47D cells ( Figure 1—figure supplement 2B ) . These results indicate that HSF1 may influence the growth of ER-positive breast cancer cells , unstimulated and stimulated by estrogen , although the effect also depends on other factors ( differences between cells , culture conditions ) . We then searched for differences between modified cells in response to longer E2 treatment . We noticed that stimulation of HSF1-proficient MCF7 cells with E2 for 7–14 days resulted in cell-cell dissociation , the acquisition of an ameboid- or mesenchymal-like morphology ( Figure 1E and F , Figure 1—figure supplement 1I ) , and enhanced adhesion to collagens ( I and IV ) but reduced to vitronectin ( Figure 1G; adhesion to fibronectin , laminin , and tenascin was not affected; not shown ) . These changes enabled cells to migrate faster ( Figure 1H , Figure 1—figure supplement 1H ) . HSF1 deficiency counteracted cell scattering after E2 stimulation ( Figure 1E and F , Figure 1—figure supplement 1I ) . This was associated with the reduced adhesion to collagens and cell motility ( Figure 1G and H , Figure 1—figure supplement 1H ) . It is noteworthy that T47D cells differed from MCF7 cells in response to E2 treatment for 14 days , especially in acquired cell morphology . Amoeboid-like morphology was dominant among the scattered MCF7 cells , while mesenchymal-like morphology was dominant in T47D cells ( Figure 1E and F , Figure 1—figure supplement 2D and E ) . Also , adhesion to collagens was not affected by E2 in T47D cells ( Figure 1—figure supplement 2F ) . Nevertheless , E2 treatment enhanced migration of HSF1-proficient but not HSF1-deficient T47D cells ( Figure 1—figure supplement 2G ) . In a search for the mechanism responsible for a distinct response to estrogen in ER-positive cells with different levels of HSF1 , we analyzed global gene expression profiles by RNA-seq in all MCF7 cell variants . At control conditions ( no E2 stimulation ) , we found relatively few genes differentially expressed in HSF1-proficient and HSF1-deficient cells that were common for different models of HSF1 downregulation . These included mainly known HSF1 targets ( e . g . , HSPH1 , HSPE1 , HSPD1 , HSP90AA1 ) slightly repressed in HSF1-deficient cells . Analyzing the response to E2 , we initially compared cell variants from different models: with the normal level of HSF1 ( WT , SCR , and MIX ) and HSF1-deficient cells ( shHSF1 , KO#1 , and KO#2 ) ( Supplementary file 1 , sheet 1 ) . We found 50 genes similarly regulated by E2 ( 47 upregulated and 3 downregulated ) in all HSF1-proficient MCF7 cell variants ( Figure 2—figure supplement 1A and B ) . On the other hand , only 13 genes were similarly upregulated after E2 stimulation in HSF1-deficient MCF7 cell variants ( Figure 2—figure supplement 1A and C ) . The gene set enrichment analyses indicated that HSF1 deficiency negatively affected the processes activated by estrogen ( the early and late estrogen response; Figure 2—figure supplement 1E ) . Moreover , though almost all genes upregulated by E2 in HSF1-proficient cells were also upregulated in HSF1-deficient cells ( except NAPRT ) , the degree of their activation ( measured as a fold change E2 versus Ctr ) was usually weaker in the latter cells ( Figure 2—figure supplement 1F ) , which indicated that the transcriptional response to estrogen was inhibited in the lack of HSF1 . Interestingly , however , several E2-dependent genes revealed slightly higher basal expression ( without E2 stimulation ) in HSF1-deficient cells ( Figure 2—figure supplement 1G ) , which suggested that in the absence of E2 , HSF1 could be involved in the suppression of these genes . Considering differences between KO#1 and KO#2 HSF1 knockout clones derived from individual cells ( Figure 2—figure supplement 1D ) , we performed an additional transcriptomic analysis using a putatively more representative MCF7 cell model obtained by DNA-free CRISPR/Cas9 method ( heterogeneous populations of HSF1+ and HSF1− cells ) ( Supplementary file 1 , sheet 2 ) . The analysis showed that 3715 genes significantly changed the expression ( 2336 upregulated and 1479 downregulated ) in HSF1+ cells after E2 stimulation . On the other hand , only 2969 genes ( 1818 upregulated and 1151 downregulated ) changed the expression in HSF1− cells ( Figure 2A ) . Thus , approximately 20% of genes responding to E2 treatment in HSF1+ cells did not respond similarly in HSF1− cells . Moreover , among genes up- or downregulated in both cell variants , approximately 68% or 81% , respectively , responded less effectively ( fold change E2 versus Ctr ) in HSF1− cells than HSF1+ cells ( Figure 2A , bottom panel ) . The gene set enrichment analyses revealed the slight differences in the early and late estrogen response pathways ( Hallmark gene sets M5906 and M5907 ) but also in genes defining epithelial-mesenchymal transition ( M5930 ) . Interestingly , the expression of genes from these pathways already differentiated untreated HSF1− and HSF1+ cells . Genes encoding cell cycle-related targets of E2F transcription factors ( M5925 ) , involved in the G2/M checkpoint ( M5901 ) as well as ECM proteoglycans ( M27219 ) and collagen formation ( M631 ) , also discriminated HSF1− and HSF1+ cells ( Figure 2—figure supplement 2A ) . The analysis confirmed that the transcriptional response to estrogen was inhibited in the lack of HSF1 . In addition , signaling pathways related to proliferation , migration , and collagen adhesion were identified as primarily affected , which was consistent with the results of functional tests . Among E2-responding genes that were common for all MCF7 cell models , 46 were upregulated and 2 were downregulated ( Figure 2B ) . Though a fraction of genes with higher basal expression in HSF1− cells than in HSF1+ cells ( potentially repressed by HSF1 ) was smaller compared to other models of HSF1 deficiency ( Figure 2—figure supplement 1 ) , it remained relevant ( Figure 2C ) . To validate the RNA-seq results , we selected 13 estrogen-induced genes for RT-qPCR analyses using nascent RNA ( Figure 2D ) . In the case of nine genes , the degree of activation was substantially lower in HSF1− than in HSF1+ cells . When the basal expression in E2-untreated cells was compared , there were 12 genes expressed at a higher level in untreated HSF1− cells in comparison to HSF1+ cells ( Figure 2D ) . Additional RT-qPCR analyses using total RNA showed that of the 15 genes tested 12 were less activated after E2 treatment in HSF1− than in HSF1+ cells . When the basal expression in E2-untreated cells was compared , six genes were expressed at a significantly higher level and one at a lower level in HSF1− than HSF1+ cells ( Figure 2—figure supplement 2B ) . Therefore , although the response to E2 was highly variable ( differences were observed between cell models ) , RT-qPCR-based validation generally confirmed differences between HSF1-proficient and HSF1-deficient MCF7 cells revealed by the RNA-seq . These changes at the transcriptional level might have direct functional consequences in HSF1− cells ( reduced level of E2-stimulated lcnRNAs , e . g . , LINC01016 ) but also were connected with the reduced protein level of E2-stimulated genes ( HSPB8 , PHLDA1 , and EGR3 are shown as examples; Figure 2E ) . To further study the influence of HSF1 on estrogen signaling , we analyzed ERα binding to chromatin in HSF1-proficient and HSF1-deficient MCF7 cells . We performed ChIP-seq analyses using the first functional knockout model ( KO#2 and MIX cells ) and validation by ChIP-qPCR using the model obtained by the DNA-free CRISPR/Cas9 system . A list of all ERα-binding sites detected by ChIP-seq in unstimulated cells and after 30 or 60 min of E2 treatment is presented in Supplementary file 2 . These analyses revealed that in unstimulated cells ERα binding was more efficient ( more binding sites and increased number of tags per peak ) in HSF1-deficient cell variant ( KO#2 ) than in the corresponding HSF1-proficient control ( MIX cells ) ( Figure 3A and B ) ( it is worth noting that the MIX cell variant was also different from wild-type cells , indicating that the genome organization was affected by the CRISPR/Cas9 procedure itself , possibly due to off-targets ) . ERα target sequences in IGFBP4 or GREB1 are examples of such increased binding efficiency in unstimulated HSF1-deficient cells ( Figure 3D ) . Estrogen treatment for 30 or 60 min resulted in enhanced ERα binding in all cell variants . However , fold enrichment ( E2 versus Ctr ) was lower in HSF1-deficient cells than in HSF1-proficient cells ( Figure 3C ) . Moreover , the number of detected peaks in the E2-treated HSF1-deficient cells was only slightly higher than in unstimulated cells ( Figure 3A ) and enhanced ERα binding was primarily manifested in sites already existing in unstimulated cells ( Figure 3C and D ) . We additionally searched for ERα- binding preferences in HSF1-proficient and HSF1-deficient cells . After estrogen treatment , ERβ ( ESR2 ) and ERα ( ESR1 ) motifs were centrally enriched in ERα-binding regions in all cell variants ( Figure 3—figure supplement 1 ) . Moreover , in untreated cells , the motif for PBX1 ( not centrally enriched in peak regions ) , which is a pioneer factor known to bind to the chromatin before ERα recruitment ( Magnani et al . , 2011 ) , was identified by MEME-ChIP analysis in all cell variants ( not shown ) . This indicates that ERα chromatin-binding preferences were not substantially changed in HSF1-deficient cells . Validation of ChIP-seq results revealed that in the case of IGFBP4 and GREB1 ( i . e . , sequences highly enriched with ERα after E2 stimulation ) the binding efficiency of ERα was higher in unstimulated HSF1− cells than in the corresponding HSF1+ cells . On the other hand , although estrogen treatment strongly induced ERα binding , this induction was considerably lower in HSF1− cells ( Figure 3E ) . Therefore , we confirmed that in this experimental system the deficiency of HSF1 may result in enhanced binding of unliganded ERα ( in particular at strongly responsive ERα-binding sites ) and weaker subsequent enrichment of ERα binding upon estrogen stimulation . However , other patterns of the response to E2 treatment are also possible , especially in sequences that were weakly enriched in ERα after stimulation , as exemplified by AMZ1 , SDK2 , SMPD3 , and SMTNL2 ( Figure 3D and E ) . Observed differences in response to E2 between cells with different levels of HSF1 may result from altered expression of ERα in HSF1-deficient cells . We found that although the kinetics of ERα activation ( as assessed by S118 phosphorylation ) in response to E2 treatment was similar in HSF1+ and HSF1− MCF7 cells , ERα and pS118 ERα levels were lower in HSF1− cells ( Figure 3F ) . ERα is known to be kept in an inactive state by HSP90 ( Pratt and Toft , 1997 ) , in particular by HSP90AA1 ( Dhamad et al . , 2016 ) , that is , the HSF1 transcriptional target . Thus , looking for a reason for the decreased ERα level and its dysregulated binding to DNA in HSF1-deficient cells , we focused on ERα and HSP90 interactions . Analyses of the proximity of both proteins by PLA revealed that the number of ERα /HSP90 complexes decreased after estrogen treatment in HSF1+ MCF7 cells ( Figure 3—figure supplement 2A ) . This indicates that liganded ( and transcriptionally active ) ERα is indeed released from the inhibitory complex with HSP90 . HSP90AA1 expression was substantially reduced in HSF1-deficient cells ( RNA-seq analyses ) , which correlated with the reduced HSP90 protein level ( Figure 3—figure supplement 2B ) . Also , the ERα level was considerably decreased in most HSF1-deficient cell variants ( except KO#1 cells; Figure 3—figure supplement 2C ) , especially in cells cultured in phenol-free media ( Figure 3F ) . Therefore , we hypothesized that the number of ERα/HSP90 complexes could be reduced in HSF1-deficient cells , which would result in enhanced basal transcriptional activity of ERα in untreated cells . However , we observed an increased number of such complexes both in untreated and E2-stimulated HSF1− cells when compared to HSF1+ cells ( Figure 3—figure supplement 2A ) . This indicates that the response to estrogen could be dysregulated in HSF1-deficient cells , also at the level of ERα/HSP90 interactions , in a mechanism not related directly to the HSP90 and ERα downregulation mediated by the HSF1 deficiency . Since estrogen-activated HSF1 was shown to bind to chromatin , we compared the binding patterns of ERα and HSF1 in wild-type MCF7 cells ( using our ChIP-seq data deposited in the NCBI GEO database; accession no . GSE137558; Vydra et al . , 2019 ) . Although in untreated cells ( Ctr ) there were 1535 and 2248 annotated peaks for ERα and HSF1 respectively ( compared to the input ) , only a few ( below 50 ) binding sites with overlapped peaks for both transcription factors were identified . Moreover , these common binding regions were characterized by a small number of tags ( smaller in the case of ERα ) ( Figure 4A; Supplementary file 3 , sheet 1 ) . On the other hand , the search for ERα and HSF1 common binding regions created after estrogen treatment ( E2 versus Ctr ) returned more than 200 peaks ( Supplementary file 3 , sheet 2 ) . They represented only a small fraction of the total number of ERα-binding sites ( ~2 . 6% from 8320 peaks; in the case of HSF1 , this represents 35% of 571 peaks ) ( Figure 4B ) . Numbers of tags per peak and fold enrichment increased after E2 stimulation for both factors , yet more for ERα than HSF1 binding in such regions ( Figure 4C ) . These results suggest that although there is a significant overlap between two sets of peaks ( p-value=0 . 0099 , ChIPpeakAnno , peakPermTest ) , the cobinding of both factors in the same DNA region may not be critical in the regulation of the ERα transcriptional activity . Instead , we postulate that HSF1 may influence the organization of the chromatin loops created after estrogen stimulation . When we combined ERα and HSF1 ChIP-seq peaks with data from chromatin interaction analysis by paired-end tag sequencing ( ChIA-PET ) performed by Fullwood et al . , 2009 , it was evident that the HSF1-binding sites mapped to ERα-interacting loci ( ERα anchor regions ) ( Figure 4D , Figure 4—figure supplement 1 ) even if actual ERα binding was not detected in the same locus ( examples of such anchors in FAM102A , HSPB8 , PRKCE , and WWC1 regulatory sequences are shown in Figure 4D ) . HSF1 peaks unrelated to ERα anchoring were also existing ( Figure 4—figure supplement 1B ) . Further analyses of the spatial organization of chromatin by chromosome conformation capture ( 3C ) technique revealed that some interactions between different ERα anchor regions were dependent on the presence of HSF1 . This is exemplified by HSPB8 and WWC1 loci analyzed in HSF1-proficient and HSF1-deficient cells ( Figure 4E ) , which confirms the role of HSF1 in the formation of ERα-mediated chromatin loops . Though the cobinding of HSF1 and ERα to DNA was rare and relatively weak , particularly in untreated cells , the proximity of both factors was easily detected . In general , both transcription factors colocalized in the nucleus when assessed by immunofluorescence ( Figure 4—figure supplement 2A ) . Thus , PLA spots indicating putative HSF1/ERα interactions were mainly located in the nucleus and their number increased after E2 treatment ( Figure 4F , Figure 4—figure supplement 2B ) . However , large diversity was observed between individual cells , which suggests that also HSF1 binding to DNA may be differentiated at the single-cell level . Nevertheless , we concluded that the proximity of HSF1 and ERα putatively reflecting their interactions frequently happens in the cell nucleus . The PLA results showed that interactions between ERα and HSF1 are possible , while the binding patterns observed in ChIP-seq combined with the ChIA-PET results suggest that different modes of these interactions are possible . To distinguish between cobinding , tethering , and canonical binding , regions of HSF1 and ERα ChIP-seq peaks ( and ChIA-PET reads ) were analyzed whether each sequence contained a motif match of HSF1 ( heat shock element [HSE] ) and/or ERα ( ERE ) ( Figure 5A; Supplementary file 4 ) . Analyses of all peaks/reads existing after E2 treatment showed that most of them reflected canonical binding through the corresponding motif . We found 569 genes that could be independently co-regulated by both transcription factors . In addition , ERα and HSF1 may directly cooperate in the regulation of 275 genes: cobinding was found for 65 genes and possible tethering ( i . e . , the presence of both transcription factors in a given chromatin region containing only one motif ) for 220 genes ( Figure 5B and C ) . In the last group , it is also possible that ERα and HSF1 bound to chromatin in different sites can interact , leading to the formation of a chromatin loop . Moreover , various binding patterns were found in the regulatory region annotated to one gene ( Figure 5D ) . This analysis showed that ERα and HSF1 can interact more frequently through tethering ( or when each is bound to a different region ) than cobinding . Our in vitro analyses indicated that HSF1 could support the transcriptional action of ERα upon estrogen treatment . On the other hand , HSF1-regulated chaperones are necessary to keep estrogen receptors in an inactive state in the absence of ligands , which collectively indicated important functional crosstalk between both factors . Therefore , to further study the significance of the interaction between ERα and HSF1 in actual breast cancer , we utilized RNA-seq data deposited in TCGA database . The analysis revealed that the transcript level of HSF1 negatively correlated with the ESR1 transcript level , although this tendency was relatively weak ( Figure 6A ) . Neither ESR1 nor HSF1 transcript levels had a significant prognostic value ( Figure 6—figure supplement 1A ) . Therefore , out of all breast cancer cases , we selected four groups ( numbered from I to IV ) characterized by significantly different levels of ERα ( mRNA and protein level ) and HSF1 ( mRNA ) expression: ER−/HSF1low , ER−/HSF1high , ER+/HSF1low , and ER+/HSF1high ( Figure 6B ) . These groups varied in molecular subtypes composition . In ER+ cancers ( luminal A , luminal B , and normal-like ) , the HSF1low group was more homogenous ( mostly luminal A ) than the HSF1high group . In ER– cases ( basal-like and HER2-enriched ) , the HSF1high group was more homogenous ( mostly basal-like ) ( Figure 6C ) . Importantly , the exclusion of cases with moderate/intermediate expression of ESR1 or HSF1 enabled us to observe the effect of both transcription factors on the survival of breast cancer patients , although the expression of HSF1 alone had no significant effect on the survival in either ER– or ER+ group analyzed separately ( Figure 6—figure supplement 1B ) . Nevertheless , the most divergent groups were ER+/HSF1low and ER−/HSF1high ( better and worse prognosis , respectively; p=0 . 0044 ) , which represented luminal A and basal-like enriched groups ( Figure 6D , Figure 6—figure supplement 1B ) . The difference between ER+/HSF1low and ER−/HSF1high cancers was also clearly visible in the multidimensional scaling ( MDS ) plots where the cancer cases belonging to these groups were separated . MDS plotting generally separated ER+ cases from ER−/HSF1high cases , while ER−/HSF1low cases were scattered between them ( Figure 6E ) . On the other hand , HSF1high and HSF1low cases were not separated , although they were slightly shifted against each other . When looking at molecular subtypes , it became apparent that ER−/HER2-positive cancers were separated from ER−/basal-like cancers and slightly overlapped with ER+ cancers . These analyses indicate collectively that HSF1 and ERα may affect survival and have stronger prognostic value if analyzed together but only when extreme expression values are taken into account . Since in vitro analyses showed an effect of HSF1 on E2-stimulated cell migration that may facilitate metastasis formation , we checked HSF1 levels in metastatic ( defined as all cases with a nonzero number of positive lymph nodes or with distant metastases; 418 cases ) and nonmetastatic ( 399 cases ) breast cancers ( data deposited in TCGA database ) . ER+ ( defined by our criteria , Figure 6B ) was the only group in which HSF1 expression level was higher in metastatic cases than in nonmetastatic ones ( logFC = 0 . 32 , p-value=0 . 0005 ) ( Figure 6F ) . When groups of patients defined by ER status were analyzed for overrepresentation of metastatic tumors , we found that they might be more common among ER+ tumors ( 51 . 3% versus 39 . 6% in the ER− group; p-value=0 . 018 , Fisher’s exact test ) ( Figure 6G , upper panel ) . Furthermore , only in the ER+ group a proportional increase in metastatic disease was observed with the increase in HSF1 expression ( Figure 6H ) and metastatic tumors were overrepresented in ER+/HSF1high ( 62% versus 44 . 3% in ER+/HSF1low ) ( p-value=0 . 059 , Pearson’s chi-squared test , verified by chi-squared posthoc test ) ( Figure 6G , bottom ) . When all patients ( split into groups by ER status from TCGA clinical data and HSF1 expression split by median value or intervals ) were analyzed , among-groups differences were also present ( p-value = 0 . 001 ) ( Figure 6—figure supplement 2 ) . These analyses suggest that the action of HSF1 and its effect on metastasis formation may differ in ER+ versus ER− breast cancers . Furthermore , we analyzed global gene expression profiles in breast cancers with different ERα and HSF1 statuses . Differential expression tests between the above-selected groups of patients ( Supplementary file 5 ) revealed that generally ERα had a much stronger influence on the transcriptome ( i . e . , ER+ versus ER− ) than HSF1 ( i . e . , HSF1high versus HSF1low ) . Nevertheless , differences between ER+ and ER− cases were higher in the presence of high levels of HSF1 , which implicates that HSF1 increases the disparity of the transcriptome of ER+ cancers . Also , the differences in the transcript levels between HSF1high and HSF1low cancers were higher in ER+ than ER− cases ( Figure 7A ) . Remarkably , the most divergent were ER+/HSF1low and ER−/HSF1high cancers , which resembled the most significant differences in the survival probability ( Figure 6D ) . Then , we looked at differences in numbers of differently expressed genes ( DEGs ) between patients’ groups . To eliminate the possible influence of the group size on DEGs , we repeated each test 10 times , randomly subsampling groups to an equal number of cases and averaging the number of DEGs . Furthermore , to check whether heterogeneity of selected groups regarding molecular subtypes could affect observed differences in gene expression profiles , only basal-like ( ER− ) and luminal A ( ER+ ) cancers were included in these tests ( Figure 7B ) . In general , these analyses also revealed that the number of genes differentiating ER+ and ER− cases was higher in HSF1high cancers , while the number of genes differentiating HSF1high and HSF1low cases was higher in ER+ cancers . The most divergent were again ER+/HSF1low and ER−/HSF1high cases while the most similar , ER−/HSF1low and ER−/HSF1high ( Figure 7C ) . This tendency was maintained when groups with mixed molecular subtypes composition were analyzed as well as more homogenous cancer groups ( i . e . , only basal-like and luminal A ) . Furthermore , the prognostic value of both ESR1 and HSF1 was visible in such homogenous groups ( Figure 6—figure supplement 1C ) , which may simply reflect the prognostic difference between the basal-like and luminal A ( i . e . , ER-negative and ER-positive ) breast cancer subtypes . Also , HSF1high cases were dominant in basal-like cases , while HSF1low were dominant in luminal A cases . Further analyses showed that the level of HSF1 did not affect the survival of ER-positive luminal A cancers but may slightly worsen the prognosis of basal-like cancers ( Figure 6—figure supplement 1C ) . Differences in gene expression profiles between pairwise compared groups of cancer were further illustrated on volcano plots that additionally separated upregulated and downregulated genes ( Figure 7—figure supplement 1 ) . Then we searched for the hypothetical influence of the HSF1 status on functions of ERα-related genes in actual cancer tissue . The gene set enrichment analysis identified terms related to estrogen response among the most significant ones associated with transcripts differentiating between ER+ and ER− cancers . It is noteworthy that terms related to spliceosomal complex assembly , especially the formation of a quadruple snRNP complex , were differentiating HSF1high and HSF1low cancers ( Figure 7—figure supplement 2 ) . The more detailed analysis focused on terms related to hormone signaling and metabolism showed differences between HSF1high and HSF1low cases when ER+ and ER− cancers were compared . These analyses indicate that HSF1 may enhance estrogen signaling . On the other hand , the analysis focused on terms related to response to stimulus and protein processing ( i . e . , functions presumed to be dependent on HSF1 action via the HSPs expression ) revealed that most of them reached the statistical significance of differences between ER+/HSF1high and ER−/HSF1high cases ( Figure 7D ) . We additionally compared the expression of E2-regulated genes ( the set identified in MCF7 cells by RNA-seq , i . e . , 46 upregulated and 2 downregulated genes; Figure 2 ) in selected groups of breast cancers with different levels of ESR1 and HSF1 . The analysis revealed the highest upregulation of PGR and LINC01016 genes in ER+ compared to ER− cancers ( regardless of HSF1 status ) ( Figure 7E ) . It is noteworthy , however , that not all genes upregulated by E2 in MCF7 cells revealed an increased expression level in ER+ compared to ER− cancers . Especially , FOXC1 and LINC00511 were expressed at a higher level in ER− cancers . Moreover , regardless of ER status , cancers with high HSF1 levels revealed a higher expression of MYBL1 than cancers with low HSF1 levels . Furthermore , expression of a few genes systematically differentiated cancers with high levels of both factors ( ER+/HSF1high ) compared to cancers with the low level of at least one factor ( including RAPGEFL1 , AMZ1 , KCNF1 , HSPB8 upregulated , and CYP24A1 , SIM1 downregulated in ER+/HSF1high cancers ) , which was consistent in all relevant comparisons ( marked with green boxes in Figure 7E ) . Nevertheless , the observed features of gene expression profiles confirmed collectively that HSF1 affects the genomic action of ERα in breast cancer . ER-positive breast cancers are frequently treated with tamoxifen , a selective estrogen receptor modulator . More recent therapeutic options include palbociclib , a selective inhibitor of the cyclin-dependent kinases CDK4 and CDK6 , approved for women with advanced metastatic cancer . Thus , we studied the influence of HSF1 on the response of MCF7 and T47D cells to these drugs . Treatment of HSF1+ cells with 4-hydroxytamoxifen ( 4-OHT ) resulted in slightly enhanced proliferation ( Figure 8A ) . This may be a consequence of ERα activation ( estimated by its phosphorylation at S118; Figure 8B ) and induction of ERα-regulated genes ( not shown ) and is consistent with previous reports ( Ali et al . , 1993 ) . Although HSF1 functional knockout by itself had different effects in both cell lines ( MCF7 growth was inhibited , while T47D growth was enhanced after HSF1 knockout; see Figure 1D , Figure 1—figure supplement 1F and G , Figure 1—figure supplement 2B ) , treatment with 4-OHT did not result in increased proliferation , thus it gave better results than in HSF1+ cells ( Figure 8A ) . 4-OHT slightly inhibited E2-stimulated cell proliferation , and the differences between HSF1+ and HSF1− cells reflected differences in response to E2 . T47D cells were more resistant to palbociclib than MCF7 cells . The difference between HSF1+ and HSF1− was not significant in T47D cells while in MCF7 cells inhibitory concentration 50 ( IC50 ) of palbociclib was more than twofold lower in HSF1− than in HSF1+ cells ( Figure 8C ) . Palbociclib also inhibited E2-stimulated cell proliferation , yet only in MCF7 cells , it was slightly more effective in the absence of HSF1 ( Figure 8A ) . These results show only some tendencies ( the statistical significance depends on the tests used ) but suggest that ER-positive breast tumors with low HSF1 expression may be more sensitive to treatment with 4-OHT and palbociclib than cases with high HSF1 levels .
The precise mechanisms by which estrogens stimulate the proliferation of breast cancer cells are still unclear . We found that estrogen action may be supported by HSF1 , a deficiency of which in ER-positive MCF7 breast cancer cells slows down the mitogenic effect of estrogen . This may be a consequence of a reduced level of ERα and transcriptional response to estrogen in these cells . In addition , analyses of the transcriptome of breast cancers from TCGA database showed the importance of HSF1 as evidenced by higher transcriptome disparity in ER-positive cases with a high expression of HSF1 rather than with low HSF1 levels . The effect of E2 and ERα on cell migration and metastasis also is unclear and published data are inconsistent . E2 was shown to suppress the invasion of ER-positive breast cancer cells ( Padilla-Rodriguez et al . , 2018 ) or to enhance breast cancer cell motility and invasion ( Sanchez et al . , 2010; Zheng et al . , 2011; Ho et al . , 2016; Vazquez Rodriguez et al . , 2017 ) . Correspondingly , ERα silencing or inhibition ( by fulvestrant , a selective estrogen receptor degrader ) was shown to enhance cell migration and invasion ( Bouris et al . , 2015; Gao et al . , 2017 ) or to reduce motility ( Bischoff et al . , 2020 ) . We showed that longer exposure to E2 induced cell scattering and increased mobility in ER-positive breast cancer cells and HSF1 deficiency could counteract these processes . It is noteworthy that responses to E2 and the effects of HSF1 are slightly different in MCF7 and T47D cells . Ameboid-like morphology and enhanced adhesion to collagens are induced by E2 in MCF7 , while mesenchymal-like morphology is induced in T47D cells . Generally , T47D cells differ from MCF7 cells in response to estrogen , partially due to a lower level of ERα in T47D ( Vydra et al . , 2019 ) . Moreover , T47D cells harbor a p53 missense mutation ( L194F ) , which causes p53 stabilization ( Lim et al . , 2009 ) . The mutant p53 exhibits gain-of-function activities in mediating cell survival , and this is likely the reason for the differences between T47D and MCF7 cells . Nevertheless , the data from TCGA showing a correlation between increasing levels of HSF1 and metastatic disease in ER-positive breast cancers support the observations from the in vitro model that HSF1 may affect migration . The mechanism of supportive action of HSF1 in ER-positive cells was already proposed , by which upon E2 treatment HSF1 is phosphorylated via ERα/MAPK signaling , gains transcriptional competence , and activates several genes essential for breast cancer cell growth and/or ERα action ( Vydra et al . , 2019 ) . Here , we found that HSF1 deficiency results in a weaker response to estrogen stimulus of many estrogen-induced genes . It is noteworthy that the reduced transcriptional response to estrogen could at least partially result from the enhanced binding of unliganded ERα to chromatin and higher basal expression of ERα-regulated genes . This suggests that HSF1-dependent mechanisms may amplify ERα action upon estrogen stimulation while inhibiting it in the absence of ligands . The proper action of ERs depends on HSF1-regulated chaperones , especially HSP90 . As expected , the number of HSP90/ERα complexes decreased after ligand ( E2 ) binding in cells with normal levels of HSF1 . However , although HSP90 was downregulated in HSF1-deficient cells , more HSP90/ERα complexes were found both in untreated and estrogen-stimulated cells . Hence , increased activity of ERα in HSF1-deficient cells could not be explained by the reduced sequestration of unliganded ERα by HSP90 . Accordingly , additional HSF1-dependent factors may influence the formation of these complexes . Nevertheless , because it is known that HSP90 inhibitors affected the ERα level ( Fliss et al . , 2000; Nonclercq et al . , 2004; Fiskus et al . , 2007; Wong and Chen , 2009 ) , a decreased level of ERα observed in our experimental model may be a consequence of the decreased level of HSP90 . Ligand-independent genomic actions of ERα are also regulated by growth factors that activate protein-kinase cascades , leading to phosphorylation and activation of nuclear ERs at EREs ( Stellato et al . , 2016 ) . The involvement of HSF1 in the repression of estrogen-dependent transcription was reported in MCF7 cells treated with neuregulin ( NRG1 ) , the ligand for the HER2 ( NEU/ERBB2 ) receptor tyrosine kinase ( Khaleque et al . , 2008 ) . Interactions of HSF1 with the corepressor metastasis-associated protein 1 ( MTA1 ) and several additional chromatin-modulating proteins were implicated in that process . Therefore , since the lack of HSF1 can alter the cellular context , it cannot be ruled out that HSF1 influences unliganded and liganded ERα by various mechanisms that have to be further investigated . Our observation from cell culture models that silencing or knockout of HSF1 has a different effect on ERα-regulated genes in the absence or presence of estrogen implicates that the consequences in real cancer may depend on the hormonal status of the patient , which is connected with age ( pre-/postmenopausal ) or use of contraceptive and hormone replacement therapies . Transcriptional activation by ERα is a multistep process modulated by coactivators and corepressors . Cofactors interact with the receptor in a ligand-dependent manner and are often part of large multiprotein complexes that control transcription by recruiting components of the basal transcription machinery , regulating chromatin structure , and/or modifying histones ( Welboren et al . , 2009; Kovács et al . , 2020; Pescatori et al . , 2021 ) . Liganded ERα may bind directly to DNA ( to ERE ) , and indirectly via tethering to other transcription factors such as FOS/JUN ( AP1 ) , STATs , ATF2/JUN , SP1 , and NFκB ( Björnström and Sjöberg , 2005; Welboren et al . , 2009; Heldring et al . , 2011 ) . It was established that direct ERE binding is required for most ( 75% ) of the estrogen-dependent gene regulation and 90% of the hormone-dependent recruitment of ERα to genomic binding sites ( Stender et al . , 2010 ) . Therefore , 10% of ERα binding occurs through tethering factors . Here , we found that HSF1 can potentially be an additional factor tethering liganded ERα to DNA . ERα has been shown to function via extensive chromatin looping to bring genes together for coordinated transcriptional regulation ( Fullwood et al . , 2009 ) . Since ERα anchor sites were identified also in sites bound by HSF1 but not ERα , we propose that HSF1 may be a part of this ‘looping’ machinery . Other components in the same anchoring center are also possible . According to the data from ENCODE , the HSF1-binding sites may coincide with NR2F2 , JUND , FOSL2 , CEBPB , GATA3 , MAX , HDAC2 , etc . It is consistent with the finding that transcription factor binding in human cells occurs in dense clusters ( Yan et al . , 2013 ) . In general , estrogen-induced HSF1 binding was weaker than ERα binding . However , PLA analyses indicated a large heterogeneity in a cell population regarding ERα and HSF1 interactions . The final transcriptional activity of ERα is modulated by interactions with various tethering factors , including HSF1 . Therefore , we hypothesize that it can be modulated differently at the single-cell level by different cofactors and chromatin remodeling factors . Thus , the response measured on the whole-cell population is heterogeneous , while stochastic when a single cell is considered . Some premises indicate that high levels of HSF1 may be associated with resistance of estrogen-dependent breast cancers to hormonal therapies based on antiestrogens . A significant association between high HSF1 expression and increased mortality among the ER-positive breast cancer patients receiving hormonal therapy was first noticed by Santagata et al . , 2011 , then confirmed by Gökmen-Polar and Badve , 2016 . It was proposed that overexpression of HSF1 in ER-positive breast cancers was associated with a decreased dependency on the ERα-controlled transcriptional program for cancer growth ( Silveira et al . , 2021 ) . However , this conclusion was based on experiments performed without estrogen stimulation . Our in vitro studies indicate that the influence of HSF1 on ERα action depends on the presence of the estrogen and HSF1 may repress the ERα-controlled transcriptional program only in the absence of the ligand . Nevertheless , we confirmed that HSF1-deficient cells responded better to 4-hydroxytamoxifen treatment than cells with normal HSF1 levels . Also , palbociclib , an inhibitor of CDK4/6 , was more effective in these cells . Enhanced resistance to hormonal therapies could be mediated by HSF1-regulated genes . HSPs themselves can be prognostic factors in breast cancer and especially oncogenic properties of HSP90AA1 correlated with aggressive clinicopathological features and resistance to the treatment ( Whitesell et al . , 2014; Klimczak et al . , 2019 ) . Here , we have proposed a novel mechanism of the HSF1 action in ER-positive breast cancers , which is independent of typical HSF1-regulated genes . This mechanism assumes that HSF1 influences the transcriptional response to estrogen via the reorganization of chromatin structure in estrogen-responsive genes . This mode of HSF1 action may be important in all ERα-expressing cells . For example , ERα is a critical transcription factor that regulates epithelial cell proliferation and ductal morphogenesis during postnatal mammary gland development . It is noteworthy that HSF1 has been shown to promote mammary gland morphogenesis by protecting mammary epithelial cells from apoptosis and increasing their proliferative capacity ( Xi et al . , 2012 ) . Experimental data indicate that the level of HSF1 can be used to predict response to treatment , while HSF1 targeting may improve the efficacy of breast cancer treatment and prevent the development of metastases . High HSF1 nuclear levels ( estimated by immunohistochemistry in patients with invasive breast cancer at diagnosis; in situ carcinomas and stage IV cancers were excluded from the outcome analysis ) were previously associated with decreased survival specifically in ER-positive breast cancer patients ( Santagata et al . , 2011 ) . However , in another study performed on samples from patients with ER-positive tumors , only a weak association was found between the HSF1 protein expression and poor prognosis ( Gökmen-Polar and Badve , 2016 ) . Nevertheless , both studies showed a significant correlation between HSF1 transcript levels and the survival in ER-positive breast cancer patients . In our analysis , using data from TCGA gene expression database , we did not observe such a correlation . ESR1 and HSF1 levels may be prognostic only if analyzed groups of patients are preselected regarding the high/low expression , and may reflect the differences between luminal and basal cancers . We found that in cancers with high expression of estrogen receptor ( ER+ ) , the HSF1low group consisted mainly of luminal A cases , which are known to have the best prognosis . Although high HSF1 levels slightly reduced the survival in ER+ cancer patients , they had a greater negative outcome on survival in ER-negative patients . In that group , HSF1high cases consisted mainly of the basal-like subtype , known to have a worse prognosis . In conclusion , HSF1 and ERα cooperate in response to estrogen stimulation . The regulation is known to be mediated by HSF1-dependent chaperones , which are important for the proper ERα action ( Echeverria and Picard , 2010 ) . Additionally , however , estrogen via ERα and MAPK activates HSF1 ( Vydra et al . , 2019 ) , which together with ERα forms new chromatin loops that enhance estrogen-stimulated transcription ( Figure 9 ) . This may be affected by other factors ( acting differently in individual cells ) . Moreover , HSF1 may be involved in the repression of unliganded ERα . Furthermore , genes activated by ERα and HSF1 play an important role in regulating the growth and spread of estrogen-dependent tumors .
Human ERα-positive MCF7 and T47D breast cancer cell lines were purchased from the American Type Culture Collection ( ATCC , Manassas , VA ) and the European Collection of Authenticated Cell Cultures ( ECACC , Porton Down , UK ) , respectively . Cells were cultured in DMEM/F12 medium ( Merck KGaA , Darmstadt , Germany ) supplemented with 10% FBS ( EURx , Gdansk , Poland ) and routinely tested for mycoplasma contamination . For heat shock , logarithmically growing cells were placed in a water bath at a temperature of 43°C for 1 hr . The cells were allowed to recover for the indicated time in a CO2 incubator at 37°C . For estrogen treatment , cells were seeded on plates and the next day the medium was replaced into a phenol-free medium supplemented with 5% or 10% dextran-activated charcoal-stripped FBS ( PAN-Biotech GmbH , Aidenbach , Germany ) . 17β-estradiol ( E2; #E4389 , Merck KGaA ) was added 48 hr later to a final concentration of 10 nM unless otherwise stated for the indicated time . For longer E2 treatments , the medium was changed every two days . In the case of treatments with palbociclib ( hydrochloride salt , #P-7788 , LC Laboratories , Woburn , MA ) and 4-hydroxytamoxifen ( #T176 , Merck KGaA ) , an equal volume of DMSO was added as vehicle control . The growth media were not replaced either before or after treatments . Working solutions were prepared fresh before each experiment in a culture medium ( without antibiotics ) . The shRNA target sequence for human HSF1 ( NM_005526 . 4 ) was selected using the RNAi Target Sequence Selector ( Clontech , Mountain View , CA ) . The target sequences were shHSF1 - 5′GCA GGT TGT TCA TAG TCA GAA-3′ ( 1994–2013 in NM_005526 . 4 ) , shHSF1 . 2–5′CCT GAA GAG TGA AGA CAT A ( 526–544 ) , and shHSF1 . 3–5′ CAG TGA CCA CTT GGA TGC TAT ( 1306–1326 ) . The negative control sequence was 5′-ATG TAG ATA GGC GTT ACG ACT . Sense and antisense oligonucleotides were annealed and inserted into the pLVX-shRNA vector ( Clontech ) at BamHI/EcoRI site . Infectious lentiviruses were generated by transfecting DNA into HEK293T cells and virus-containing supernatant was collected . Human MCF7 cells were transduced with lentiviruses following the manufacturer’s instructions and selected using a medium supplemented with 1 μg/ml puromycin ( Life Technologies/Thermo Fisher Scientific , Waltham , MA , USA ) . To remove the human HSF1 gene , Edit-R Human HSF1 ( 3297 ) crRNA , Edit-R tracrRNA , and Edit-R hCMV-PuroR-Cas9 Expression Plasmid ( Dharmacon , Lafayette , CO ) were introduced into MCF7 cells using DharmaFECT Duo ( 6 µg/ml ) ( Dharmacon ) according to producer’s instruction . Transfected cells were enriched by puromycin ( 2 µg/ml ) selection for 4 days . Afterward , single clones were obtained by limiting dilution on a 96-well plate . The efficiency of the HSF1 knockout was monitored by western blot . Out of 81 tested clones , 2 individual clones with the HSF1 knockout ( KO#1 and KO#2 ) and 6 pooled control clones ( MIX ) were chosen for the next experiments . Among individually tested HSF1-targeting crRNAs , only two were effective ( target sequences: GTGGTCCACATCGAGCAGGG and AAAGTGGTCCACATCGAGCA , both in exon 3 on the plus strand ) . For validation experiments , a new model was created using DNA-free system: Edit-R Human HSF1 ( 3297 ) crRNAs ( GGTGTCCGGGTCGCTCACGA in exon 1 on the minus strand and AAAGTGGTCCACATCGAGCA in exon 3 on the plus strand ) , Edit-R tracrRNA ( Dharmacon ) , and eSpCas9-GFP protein ( #ECAS9GFPPR , Merck KGaA ) were introduced into MCF7 and T47D cells using Viromer CRISPR ( Lipocalyx GmbH , Halle [Saale] , Germany ) according to the manual provided by the producer . Single clones were obtained by limiting dilution on a 96-well plate . The efficiency of the HSF1 knockout was monitored by western blot and confirmed by sequencing ( Genomed , Warszaw , Poland ) . Five ( T47D ) or six ( MCF7 ) individual unaffected clones ( HSF1+ ) or with the HSF1 functional knockout ( HSF1− ) were pooled each time before analyses . Whole-cell extracts were prepared using RIPA buffer supplemented with cOmplete protease inhibitors cocktail ( Roche ) and phosphatase inhibitors PhosSTOP ( Roche , Indianapolis , IN ) . Proteins ( 20–30 μg ) were separated on 10% SDS-PAGE gels and blotted to a 0 . 45 μm pore nitrocellulose filter ( GE Healthcare , Europe GmbH , Freiburg , Germany ) using Trans Blot Turbo system ( Bio-Rad , Hercules , CA ) for 10 min . Primary antibodies against HSF1 ( 1:4000 , ADI-SPA-901 ) , HSP90 ( 1:2000 , ADI-SPA-836 ) , and HSP70 ( 1:2000 , ADI-SPA-810 ) , all from Enzo Life Sciences ( Farmingdale , NY ) , HSP105 ( 1:600 , #3390-100 , BioVision , Milpitas , CA ) , ERα ( 1:2000 , #8644 ) , phosphoERα ( S118 ) ( 1:2000 , #2511 ) , HSPB8 ( 1:1000 , #3059 ) , all from Cell Signaling Technology ( Danvers , MA ) , PHLDA1 ( 1:1000 , #sc-23866 ) , EGR3 ( 1:1000 , #sc-390967 ) , HSPA8/HSC70 ( 1:5000 , #sc-7298 ) , all from Santa Cruz Biotechnology ( Dallas , TX ) , and ACTB ( 1:25 , 000 , #A3854 , Merck KGaA ) were used . The primary antibody was detected by an appropriate secondary antibody conjugated with horseradish peroxidase ( Thermo Fisher Scientific ) and visualized by ECL kit ( Thermo Fisher Scientific ) or WesternBright Sirius kits ( Advansta , Menlo Park , CA ) . Imaging was performed on x-ray film or in a G:BOX chemiluminescence imaging system ( Syngene , Frederick , MD ) . The experiments were repeated in triplicate , and blots were subjected to densitometric analyses using ImageJ software to calculate relative protein expression after normalization with loading controls ( statistical significance of differences was calculated using t-test ) . For nascent RNA labeling , 500 μM of 4-thiouridine ( Cayman Chemical , Ann Arbor , MI ) was added to control and E2-treated cells for the duration of the treatment ( 4 hr ) . Next , total RNA was isolated using the Direct-Zol RNA MiniPrep Kit ( Zymo Research , Irvine , CA ) , digested with DNase I ( Worthington Biochemical Corporation , Lakewood , NJ ) , and cleaned with RNAClean XP beads ( Beckman Coulter Life Science , Indianapolis , IN ) . 5 µg of total RNA from each sample were taken for nascent RNA fraction isolation using methane thiosulfonate ( MTS ) chemistry according to Duffy and Simon , 2016 . After the biotinylation step using MTSEA-biotin-XX ( Biotium , Fremont , CA ) , s4U-RNA was cleaned with RNAClean XP beads and isolated using μMacs Streptavidin Kit ( Miltenyi Biotec , Bergisch Gladbach , Germany ) as described ( Garibaldi et al . , 2017 ) . Total RNA ( 1 μg ) and nascent RNA ( isolated from 5 μg of total RNA ) from each sample were converted into cDNA as described ( Kus-Liskiewicz et al . , 2013 ) . Quantitative PCR was performed using a Bio-Rad C1000 Touch thermocycler connected to the head CFX-96 . Each reaction was performed at least in triplicates using PCR Master Mix SYBRGreen ( A&A Biotechnology , Gdynia , Poland ) . Expression levels were normalized against GAPDH , ACTB , HNRNPK , HPRT1 , if not stated otherwise . The set of delta-Cq replicates ( Cq values for each sample normalized against the geometric mean of four reference genes ) for control and tested samples were used for statistical tests and estimation of the p-value . Shown are median , maximum , and minimum values of a fold change versus untreated control . The primers used in these assays are described in Supplementary file 6 . Cells were plated onto 6-well dishes ( 1 × 103 cells per well ) and cultured for 14 days . Afterward , cells were washed with the phosphate-buffered solution ( PBS ) and fixed with methanol . Colonies were stained with 0 . 2% crystal violet , washed , and air-dried . Colonies were counted manually . Cells ( 2 × 104 cells per well ) were seeded and cultured in 12-well plates . At the indicated time , cells were washed with PBS , fixed in cold methanol , and rinsed with distilled water . Cells were stained with 0 . 1% crystal violet for 30 min , rinsed with distilled water extensively , and dried . Cell-associated dye was extracted with 1 ml of 10% acetic acid . Aliquots ( 200 μl ) were transferred to a 96-well plate and the absorbance was measured at 595 nm ( Synergy2 microtiter plate reader , BioTek Instruments , Winooski , VT ) . Grow curves are shown as the ratio of the absorbance on days 2 , 4 , and 6 against day 0 and were calculated from 3 to 6 independent experiments , each in 2–3 technical replicates . The effect of drug treatment on cell viability was determined colorimetrically using the CellTiter 96 AQueous One Solution Cell Proliferation Assay ( Promega , Madison , WI ) according to the manufacturer’s protocol . Cells seeded into 96-well plates ( 4 × 103 MCF7 cells and 1 × 104 T47D cells per well ) were incubated with palbociclib in concentrations ranging from 0 to 100 µM in DMSO for the next 72 hr ( max . DMSO concentration <0 . 5% ) . The experiment was performed 3–5 times with three replicates for each concentration of the tested compound . IC50 values were determined using the Quest Graph IC50 Calculator ( AAT Bioquest , Inc , 28 September 2021 , https://www . aatbio . com/tools/ic50-calculator ) with the option ‘Set minimum response to zero’ . Cells were treated with E2 for 14 days , then 5 × 104 cells per well were plated on fibronectin-coated Nunc Lab-Tek II chambered coverglass ( #155383 , Nalge Nunc International , Rochester , NY ) and allowed to grow for 24 hr . Cells were briefly washed with PBS , fixed for 10 min with 4% paraformaldehyde ( PFA ) solution in PBS , washed with PBS ( 3 × 5 min ) , treated with 0 . 1% Triton-X100 in PBS for 5 min , and washed again in PBS ( 3 × 5 min ) . F-actin was visualized by incubation with phalloidin–tetramethylrhodamine B isothiocyanate ( 1:800 , #P1951 , Merck KGaA ) while nuclei were counterstained with DAPI . Images were taken using Carl Zeiss LSM 710 confocal microscope with ZEN navigation software . The experiments were performed in triplicate . Cells were counted in 10 randomly selected fields for each replicate . Cells were cultured for 14 days with or without E2 ( 10 nM ) . The ECM Cell Adhesion Array kit ( #ECM 540 , Merck KGaA ) was used according to the manufacturer’s instructions . Adhesion to collagen I and IV ( #804592 and #C5533 , Merck KGaA ) was independently validated on collagen-coated ( 1 µg/cm2 ) 96-well plates . Cells ( 7 × 104 cells per well ) were allowed to adhere for 30 min . The wells were washed three times with PBS , fixed with cold methanol , and stained with 0 . 1% crystal violet . The adsorbed dye was extracted with a 10% acetic acid solution for 5 min , and measurement was performed at 595 nm on the Synergy2 microtiter plate reader ( BioTek Instruments ) . Transwell chambers ( with 8 μm pore size membrane , Becton Dickinson ) were coated with fibronectin ( 10 μg/ml , Becton Dickinson ) . Cells ( 8 × 104 ) were suspended in a HEPES-buffered serum-free medium containing 0 . 1% BSA , seeded in the top of the chambers , and placed in the wells containing medium supplemented with 10% FBS . After 24 hr , the inserts were washed with PBS , fixed with cold methanol , rinsed with distilled water , and stained with 0 . 1% crystal violet for 30 min . The cells on the upper surface of the inserts were gently removed with a cotton swab . Cells that migrated onto the lower surface were counted under a microscope in five random fields; all experiments were performed three times ( three technical repetitions each ) . The assay was performed using a kit from STEMCELL Technologies ( Vancouver , Canada , #01700 ) according to the protocol . Cells ( 6 × 105 ) were harvested by trypsinization and resuspended in 1 ml ALDEFLUOR Buffer . After the addition of 5 μl of BODIPY-aminoacetaldehyd e ( BAAA ) , the substrate for aldehyde dehydrogenase ( ALDH ) , and a brief mixing , 500 μl of the cell suspension ( 3 × 105 ) was immediately transferred to another tube supplemented with 5 μl of diethylaminobenzaldehyde ( DEAB ) , a specific inhibitor of ALDH , and pipetted to mix evenly . Tubes were incubated at 5% CO2 , 37°C for 60 min . Cells were collected by centrifugation and resuspended in ALDEFLUOR Buffer . Analyses were performed using the BD Canto III cytometer ( Becton Dickinson , Franklin Lakes , NJ ) . Cells ( 3 × 105 per well ) were plated onto 6-well plates . The next day medium was replaced and cells were grown for an additional 48 hr . Afterward , cells were harvested by trypsinization , rinsed with PBS , and fixed with ice-cold 70% ethanol at −20°C overnight . Cells were collected by centrifugation , resuspended in PBS containing RNase A ( 100 µg/ml ) , and stained with 100 µg/ml propidium iodide solution . DNA content was analyzed using flow cytometry to monitor the cell cycle changes . Cells were plated onto Nunc Lab-Tek II chambered coverglass ( #155383 , Nalge Nunc International , Rochester , NY ) and fixed for 15 min with 4% PFA solution in PBS , washed , treated with 0 . 1% Triton-X100 in PBS for 5 min , and washed again in PBS ( 3 × 5 min ) . IF imaging was performed using primary antibodies: anti-HSP90 ( 1:200; ADI-SPA-836 , Enzo Life Science ) , anti-HSF1 ( 1:300; ADI-SPA-901 , Enzo Life Sciences ) , or anti-ESR1 ( 1:200; C15100066 , Diagenode ) and secondary Alexa Fluor ( 488 or 594 ) conjugated antibodies ( Abcam ) . Finally , the DNA was stained with DAPI . Images were taken using Carl Zeiss LSM 710 confocal microscope with ZEN navigation software . To detect the ERα/HSP90 and ERα/HSF1 interactions , the Duolink In Situ Proximity Ligation Assay ( PLA ) ( Merck KGaA ) was used according to the manufacturer’s protocol . Cells were plated onto Nunc Lab-Tek II chambered coverglass ( #155383 , Nalge Nunc International ) 1 day before the experiment . Cells were fixed for 15 min with 4% PFA solution in PBS , washed in PBS , and treated with 0 . 1% Triton-X100 in PBS for 5 min . After washing , slides were incubated in Blocking Solution and immunolabeled ( overnight , 4°C ) with primary antibodies diluted in the Duolink Antibody Diluent: rabbit anti-HSP90 ( 1:200; #ADI-SPA-836 , Enzo Life Science ) and mouse anti-ERalpha ( 1:200; #C15100066 , Diagenode , Liège , Belgium ) or mouse anti-HSF1 ( 1:200; #sc-17757 , Santa Cruz Biotechnology ) and rabbit anti-ERα ( 1:200; #8644 , Cell Signaling Technology ) , as well as rabbit anti-HSF1 ( 1:300; ADI-SPA-901 , Enzo Life Sciences ) and mouse anti-ERalpha; negative controls were proceeded without one primary antibody or both . Then the secondary antibodies with attached PLA probes were used . Signals of analyzed complexes were observed using Carl Zeiss LSM 710 confocal microscope with ZEN navigation software; red fluorescence signal indicated proximity ( <40 nm ) of proteins recognized by both antibodies ( Fredriksson et al . , 2002 ) . Z-stacks images ( 12 slices; 5 . 5 μm ) were taken at ×630 magnification . From each experimental condition , spots from 10 to 15 images were identified using Photoshop ( Red Channel → Select → Color Range ) and counted ( Picture → Analysis → Record the measurements ) . Next , the mean number of spots per cell ( nucleus , cytoplasm ) in each image was calculated . Experiments were repeated three times . Outliers were determined using the Grubbs , Tukey criterion , and QQ plot . For each dataset , the normality of distribution was assessed by the Shapiro–Wilk test and , depending on data distribution homogeneity of variances , was verified by the Levene test or Brown–Forsythe test . For analysis of differences between compared groups with normal distribution , the quality of mean values was verified by the ANOVA test with a pairwise comparison done with the HSD Tukey test or Games–Howell test and Tamhane test depending on the homogeneity of variance . In the case of non-Gaussian distribution , the Kruskal–Wallis ANOVA was applied for the verification of the hypothesis on the equality of medians with Conover–Iman’s test for pairwise comparisons . p=0 . 05 was selected as a statistical significance threshold . Total RNA was isolated from all MCF7 cell variants ( untreated , treated with 10 nM E2 for 4 hr , conditions based on Vydra et al . , 2019 ) using the Direct-Zol RNA MiniPrep Kit ( Zymo Research ) and digested with DNase I ( Worthington Biochemical Corporation ) . For each experimental point , RNAs from three biological replicates were first tested by RT-qPCR for the efficiency of treatments . They were sequenced separately for HSF1+ and HSF1− cell variants or pooled before sequencing for WT , SCR , shHSF1 , MIX , KO#1 , and KO#2 cell variants . cDNA libraries were sequenced by Illumina HighSeq 1500 ( run type: paired-end , read length: 2 × 76 bp ) . Raw RNA-seq reads were aligned to the human genome hg38 in a Bash environment using hisat2 v 2 . 0 . 5 . ( Kim et al . , 2015 ) with Ensembl genes transcriptome reference . Aligned files were processed using Samtools ( v . 1 . 13 ) ( Li et al . , 2009 ) . Furthermore , reads aligned in the coding regions of the genome were counted using FeatureCounts ( v . 1 . 6 . 5 ) ( Liao et al . , 2014 ) . Further data analyses were carried out using the R software package ( v . 3 . 6 . 2; R Foundation for Statistical Computing; http://www . r-project . org ) . Read counts were normalized using DESeq2 ( v . 1 . 32 . 0 ) ( Lowe et al . , 2014 ) , then normalized expression values were subjected to differential analysis using NOISeq package ( v . 3 . 12 ) ( Tarazona et al . , 2015 ) ( E2 versus Ctr in all cell variants separately ) . To find common genes between samples , lists of differentiating genes were compared and Venn diagrams were performed ( package VennDiagram v . 1 . 6 . 20 from CRAN ) . Heatmaps of normalized read counts or log2 fold changes ( E2 versus Ctr ) for genes shared between samples were generated ( package pheatmap v . 1 . 0 . 12 from CRAN ) . The hierarchical clustering of genes was based on Euclidean distance . Colors are scaled per row . Gene set enrichment analysis was performed as follows: from the count matrices , we filtered out all the genes with less than 10 reads in each of the libraries . Then , we analyzed the gene-level effects of E2 stimulation of cells with normal/decreased HSF1 levels , performing the DESeq2 test for paired samples , with pairs defined by the cell variant ( HSF1+ and HSF1− , and separate analysis of three cell variants with normal-HSF1 level: WT , SRC , MIX , and three cell variants with decreased HSF1 level: shHSF1 , KO#1 , and KO#2 ) . Finally , we performed the gene set enrichment analysis in the same way as for TCGA data ( see below for details ) – for each test , genes were ranked according to their minimum significant difference ( MSD ) , CERNO test from tmod package was used to find enriched terms , and tmodPanelPlot function was used to visualize the results . The raw RNA-seq data were deposited in the NCBI GEO database; accession nos . GSE159802 and GSE186004 . The ChIP assay was performed according to the protocol from the iDeal ChIP-seq Kit for Transcription Factors ( Diagenode ) as described in detail in Vydra et al . , 2019 . For each IP reaction , 30 µg of chromatin and 4 μl of mouse anti-ERalpha monoclonal antibody ( C15100066 , Diagenode ) was used . For negative controls , chromatin samples were processed without antibody ( mock-IP ) . Obtained DNA fragments were used for global profiling of chromatin-binding sites or gene-specific ChIP-qPCR analysis using specific primers covering the known EREs . The set of delta-Cq replicates ( difference of Cq values for each ChIP-ed sample and corresponding input DNA ) for control and test sample were used for ERα-binding calculation ( as a percent of input DNA ) and estimation of the p-values . ERE motifs in individual peaks were identified using MAST software from the MEME Suite package ( v . 5 . 1 . 1 ) ( Bailey et al . , 2015 ) . The sequences of used primers are presented in Supplementary file 7 . In each experimental point , four ChIP biological replicates ( each from 30 μg of input chromatin ) were collected and combined in one sample before DNA sequencing . Immunoprecipitated DNA fragments and input DNA were sequenced using the Illumina HiSeq 1500 system and QIAseq Ultralow Input Library Kit ( run type: single read , read length: 1 × 65 bp ) . Raw sequencing reads were analyzed according to standards of ChIP-seq data analysis as described below . Quality control of reads was performed with FastQC software ( https://www . bioinformatics . babraham . ac . uk/projects/fastqc ) and low-quality sequences ( average phred <30 ) were filtered out . Remained reads were aligned to the reference human genome sequence ( hg19 ) using the Bowtie2 . 2 . 9 program ( Langmead and Salzberg , 2012 ) . Individual peaks ( Ab-ChIPed samples versus input DNA ) and differential peaks ( 17β-estradiol-treated versus untreated cells ) were detected using MACS software ( v . 1 . 4 . 2 ) ( Feng et al . , 2012 ) , whereas the outcome was annotated with Homer package ( v . 4 . 11 ) ( Heinz et al . , 2010 ) . Peak intersections and their genomic coordinates were found using Bedtools software ( Quinlan and Hall , 2010 ) . The input DNA was used as a reference because no sequences were obtained using a mock-IP probe . The locations of identified ERα-binding sites were compared to genomic coordinates of E2-induced HSF1 peaks from our previous ChIP-seq analysis ( NCBI GEO database; accession no . GSE137558 ) . We defined ERα/HSF1-binding sites as ‘common’ if at least the center of one peak was within the corresponding peak . Dot plots showing peak size distribution were generated using MedCalc Statistical Software ( v . 19 . 2 . 1; MedCalc Software Ltd , Ostend , Belgium; https://www . medcalc . org; 2020 ) . Coverage of bam files was normalized using deepTools ( v . 3 . 5 . 0; bamCoverage and bamCompare functions ) ( Ramírez et al . , 2016 ) , with scaling factors normalizing the coverage to 1 million reads . ChIP-seq heatmaps were prepared using peakHeatmap function from ChIPseeker Bioconductor package ( v . 1 . 26 . 2 ) , with margins of 3000 nucleotides upstream and downstream from the promoter . Venn diagrams peak overlap statistics and permutation tests were generated using the ChIPpeakAnno package ( v . 3 . 24 . 2 ) ( Zhu et al . , 2010 ) . The raw ChIP-seq data were deposited in the NCBI GEO database; accession no . GSE159724 . The consensus DNA sequences for ERα binding were identified in silico by Motif Analysis of Large Nucleotide Datasets ( MEME-ChIP , v . 5 . 1 . 1 ) ( Bailey et al . , 2015 ) using a 150 bp region centered on the summit point and visualized by CentriMo ( Local Motif Enrichment Analysis ) ( Bailey and Machanick , 2012 ) . Data from ERα-related chromatin interaction analysis by paired-end tag sequencing ( ChIA-PET ) ( Fullwood et al . , 2009; GSE18046 ) were processed as follows: raw paired-end tags detected in the same genome localizations were combined to individual peaks using GenomicRanges 1 . 42 . 0R package ( ‘reduce’ function to merge overlapping ranges ) , obtaining genomic coordinates of the ERα binding/anchor sites , then identified peaks were annotated with Homer package ( v . 4 . 11 ) ( Heinz et al . , 2010 ) . The locations of detected anchors were compared to genomic coordinates of HSF1 and ERα peaks ( versus input ) from ChIP-seq analysis of E2-treated MCF7 cells ( GSE137558 and GSE159724 , respectively ) . We defined ERα_ChIA-PET , HSF1_ChIP-seq , and ERα_ChIP-seq binding sites as overlapping if the center of each peak was within the two corresponding peaks . Additionally , for the sequence of all peaks from each analysis , a search for ERE and HSE motifs was performed using MAST software from the MEME Suite package ( v . 5 . 4 . 1 ) ( Bailey et al . , 2015 ) with position weight matrix ( PWM ) of ERα and HSF1 from the JASPAR database ( Fornes et al . , 2020 ) . Based on the presence of HSF1/ERα in genomic locus and/or motif matching , three types of binding regions , including canonical binding , cobinding , and tethered binding regions , were identified within ChIP-seq/ChIA-PET peaks . A procedure is illustrated in Figure 5A ( lists of all peaks with annotation and information about the presence of motifs are presented in Supplementary file 4 ) . The procedure was carried out according to the protocol from Deng and Blobel , 2017 . In brief , 1 × 107 cells per sample were trypsinized and fixed with 1% formaldehyde in 1× PBS . Crosslinking was quenched by 0 . 125 M glycine and cells were lysed ( 10 mM Tris pH 8 . 0 , 10 mM NaCl , 0 . 2% NP-40 , protease inhibitors ) . Cell nuclei were resuspended in HindIII RE buffer ( 10 mM Tris pH 8 . 0 , 50 mM NaCl , 10 mM MgCl2 , 100 μg/ml BSA ) and incubated sequentially with 0 . 3% SDS ( 1 . 5 hr ) and 1 . 8% Triton X-100 ( 1 . 5 hr ) at 37°C with rotation . Chromatin was cleaved using 450U HindIII restriction enzyme ( BioLabs , Ipswich , MA ) at 37°C overnight and diluted 15-fold in ligation buffer ( 50 mM Tris pH 7 . 5 , 10 mM MgCl2 , 10 mM DTT , 1% Triton X-100 , 100 μg/ml BSA ) . Ligation was carried out using 4000U T4-DNA ligase ( EURx ) at 16°C overnight , in the presence of 1 mM ATP . All samples were de-crosslinked ( 65°C , overnight with mixing ) , RNase A and Proteinase K treated , and DNA was isolated using standard Phenol/Chloroform/Isoamyl alcohol purification method . Precipitated DNA was dissolved in 10 mM Tris pH 8 . 0 and used as a template in PCR analyses . The primers used are listed in Supplementary file 8 . | About 70% of breast cancers rely on supplies of a hormone called estrogen – which is the main hormone responsible for female physical characteristics – to grow . Breast cancer cells that are sensitive to estrogen possess proteins known as estrogen receptors and are classified as estrogen-receptor positive . When estrogen interacts with its receptor in a cancer cell , it stimulates the cell to grow and migrate to other parts of the body . Therefore , therapies that decrease the amount of estrogen the body produces , or inhibit the receptor itself , are widely used to treat patients with estrogen receptor-positive breast cancers . When estrogen interacts with an estrogen receptor known as ERα it can also activate a protein called HSF1 , which helps cells to survive under stress . In turn , HSF1 regulates several other proteins that are necessary for ERα and other estrogen receptors to work properly . Previous studies have suggested that high levels of HSF1 may worsen the outcomes for patients with estrogen receptor-positive breast cancers , but it remains unclear how HSF1 acts in breast cancer cells . Vydra , Janus , Kuś et al . used genetics and bioinformatics approaches to study HSF1 in human breast cancer cells . The experiments revealed that breast cancer cells with lower levels of HSF1 also had lower levels of ERα and responded less well to estrogen than cells with higher levels of HSF1 . Further experiments suggested that in the absence of estrogen , HSF1 helps to keep ERα inactive . However , when estrogen is present , HSF1 cooperates with ERα and enhances its activity to help cells grow and migrate . Vydra , Janus , Kuś et al . also found that cells with higher levels of HSF1 were less sensitive to two drug therapies that are commonly used to treat estrogen receptor-positive breast cancers . These findings reveal that the effect HSF1 has on ERα activity depends on the presence of estrogen . Therefore , cancer therapies that decrease the amount of estrogen a patient produces may have a different effect on estrogen receptor-positive tumors with high HSF1 levels than tumors with low HSF1 levels . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"cancer",
"biology"
] | 2021 | Heat shock factor 1 (HSF1) cooperates with estrogen receptor α (ERα) in the regulation of estrogen action in breast cancer cells |
Predicting neurological recovery after spinal cord injury ( SCI ) is challenging . Using topological data analysis , we have previously shown that mean arterial pressure ( MAP ) during SCI surgery predicts long-term functional recovery in rodent models , motivating the present multicenter study in patients . Intra-operative monitoring records and neurological outcome data were extracted ( n = 118 patients ) . We built a similarity network of patients from a low-dimensional space embedded using a non-linear algorithm , Isomap , and ensured topological extraction using persistent homology metrics . Confirmatory analysis was conducted through regression methods . Network analysis suggested that time outside of an optimum MAP range ( hypotension or hypertension ) during surgery was associated with lower likelihood of neurological recovery at hospital discharge . Logistic and LASSO ( least absolute shrinkage and selection operator ) regression confirmed these findings , revealing an optimal MAP range of 76–[104-117] mmHg associated with neurological recovery . We show that deviation from this optimal MAP range during SCI surgery predicts lower probability of neurological recovery and suggest new targets for therapeutic intervention . NIH/NINDS: R01NS088475 ( ARF ) ; R01NS122888 ( ARF ) ; UH3NS106899 ( ARF ) ; Department of Veterans Affairs: 1I01RX002245 ( ARF ) , I01RX002787 ( ARF ) ; Wings for Life Foundation ( ATE , ARF ) ; Craig H . Neilsen Foundation ( ARF ) ; and DOD: SC150198 ( MSB ) ; SC190233 ( MSB ) ; DOE: DE-AC02-05CH11231 ( DM ) .
Spinal cord injury ( SCI ) may result in motor , sensory , and autonomic dysfunction in various degrees , depending on the injury severity and location . In the USA , the incidence of SCI is around 18 , 000 new cases per year , with a total prevalence ranging from 250 , 000 to 368 , 000 cases ( National Spinal Cor Injury Statistical Center , 2021 ) . The dramatic life event of SCI imposes a high socioeconomic burden , with an estimated lifetime cost between $1 . 2 and $5 . 1 million per patient ( National Spinal Cor Injury Statistical Center , 2021 ) . Prior retrospective observational single-center studies in humans suggest that lower post-surgery mean arterial pressure ( MAP ) predicts poor outcome ( Cohn et al . , 2010; Hawryluk et al . , 2015; Saadeh et al . , 2017; Ehsanian , 2020 ) , which have resulted in clinical guidelines focused on avoidance of hypotension by maintaining MAP >85 mmHg during the first 7 days of injury ( Aarabi et al . , 2013 ) . The rational for MAP augmentation to avoid hypotension is based on the hypothesis that decreased spinal cord prefusion leads to ischemia and additional tissue lost ( Mautes et al . , 2000; Soubeyrand et al . , 2014 ) . Importantly , experimentally raising MAP during acute SCI in animals by using vasopressors increases the risk for hemorrhage and consequent tissue damage ( Soubeyrand et al . , 2014; Streijger et al . , 2018; Guha et al . , 1987 ) . In acute cervical patients with SCI , spinal cord hemorrhage correlates with poor prognosis for neurological recovery ( Miyanji et al . , 2007 ) . These findings collectively suggest that hypo- and hypertension must be accounted for in MAP management , but there remains a gap in evidence from clinical studies that definitively informs MAP guidelines ( Saadeh et al . , 2017 ) . One of the challenges resulting in the lack of strong evidence for MAP management in patients with acute SCI is the heterogeneity of injury . The heterogeneity of SCIs results in data complexity that benefit from modern analytic tools . Using the machine intelligence approach of topological data analysis , we have previously shown that hypertension during SCI surgery ( ultra-acute phase ) predicts long-term functional recovery in rodent models ( Nielson et al . , 2015 ) . These prior cross-species findings motivated the present multicenter study where we apply a data-driven workflow in patients with ultra-acute SCI surgical records from two different Level 1 trauma centers . By employing machine intelligence tools , we show that deviation from an optimal MAP range during surgery predicts lower likelihood of neurological recovery and suggest new targets for therapeutic intervention .
Operating room ( OR ) records from n = 94 patients ( 98 surgical records , 3 patients with multiple surgeries ) from the Zuckerberg San Francisco General Hospital ( ZSFG , from 2005 to 2013 ) and n = 33 patients ( 33 surgical records ) from the Santa Clara Valley Medical Center ( SCVMC , from 2013 to 2015 ) that underwent spinal surgery were collected retrospectively . For ZSFG , monitoring data was extracted from the values manually recorded by the anesthesiologist at 5 min intervals ( Q5 ) . For SCVMC , monitoring data was extracted from the Surgical Information Systems ( SIS ) ( Alpharetta , GA ) at 1 min intervals ( Q1 ) . Demographics and outcome variables were extracted from an existing retrospective registry . AIS ( American Spinal Injury Association [ASIA] Impairment Scale ) grade at admission ( first complete AIS upon admission to the hospital before surgery ) and discharge ( latest complete AIS grade after surgery before discharge from hospital ) were estimated using the available ISNCSCI exams ( International Standards for Neurological Classification of SCI ) and the neurosurgery , trauma surgery , emergency department , and physical medicine and rehabilitation physical exam notes . To ensure compatibility between centers on the estimated AIS grades , one independent physician conducted the estimates for all the patients in each center ( SM for SCVMC and JT for ZSFG ) and one independent ISNCSCI certified physician ( WW ) extracted the AIS grades for all the patients ( across centers ) . In case of conflict between grades , both physicians established a consensus . From the total 131 surgical records , three records were excluded for not having monitoring recorded for both MAP and HR , three were excluded because surgeries were not related to SCI , and seven surgeries were excluded from three patients that were submitted to more than one surgery . The final cohort for exploratory topological data analysis included 118 patients with complete MAP and heart rate ( HR ) monitoring . For confirmatory regression analysis , 15 patients were excluded because AIS grade could not be extracted either at admission and/or discharge ( Figure 2—figure supplement 1 ) . AIS improvement was defined as an increase of at least one AIS grade from admission to discharge . The final list of extracted variables included: MAP and HR continuous monitoring , age , length of surgery in minutes , days from surgery to hospital discharge , estimated AIS grade at admission , estimated AIS grade at discharge and AIS improvement ( ‘yes’ , ‘no’ ) . All data was de-identified before pre-processing and analysis . Protocols for retrospective data extraction were approved by Institutional Research Board ( IRB ) . The differences in the AIS improvers/non-improvers population characteristics were tested at the univariate level using R ( see software below ) . For continuous numerical variables ( age , length of surgery , and days from surgery to discharge ) , the group mean differences were tested using unpaired Student’s t-test ( two-sided test ) . For categorical variables ( AIS admission , AIS at discharge , and dichotomized neurological level of injury [NLI] ) , their levels were compared using Fisher’s exact test ( two-sided test ) . p-Values are presented in Table 1 . We sought to determine a range of MAP in which time outside that range might predict improvement . To consider the time at which MAP was outside a range , we performed an increasing window of MAP for either a symmetric range or an asymmetric one . For the symmetric range , a 1 mmHg range increment at each site of the center ( 90 mmHg , the mean MAP for improvers ) was created . For the asymmetric range , the lower limit was fixed at 76 mmHg and the upper limit was incremented 1 mmHg at the time . The time of MAP ( in min ) being outside each range was estimated for each patient . Logistic regression ( see above ) was used to build prediction models for three binary outcome metrics: AIS improvement of at least one grade from admission to discharge , whether patient was AIS grade A at discharge , or whether the patient was AIS grade D at discharge . For each one of the classification tasks , the following predictors were considered: quadratic aMAP ( both linear and quadratic terms ) , aHR , length of surgery ( min ) , days from surgery to discharge , age , AIS grade at admission , and dichotomized NLI . We performed model selection ( a . k . a . feature selection ) through an exhaustive search of all potential combinations of at least one of the predictors using the glmulti R package ( Calcagno , 2020 ) . The most parsimonious models were selected to be the one minimizing the small-sample corrected Akaike information criteria ( AIC ) for each task . We then investigated the performance of each one of the most parsimonious models using LOOCV and adjusting the classification threshold to balance prediction sensitivity and specificity . Briefly , each model was trained n ( patient ) times with an n−1 training sample and tested the performance in the remaining sample . A vector of n probabilities of predictions was then used to measure the LOOCV model performance . Model fitting and prediction performance were conducted using the caret R package ( Kuhn et al . , 2019 ) . Receiving operating curves ( ROC ) and area under the curve ( AUC ) for the LOOCV prediction were obtained using the ROCR R package ( Sing et al . , 2005 ) . All data wrangling , processing , visualization , and analysis was performed using the R programming language ( R version 3 . 5 . 1 ) ( R core team , 2019 ) and RStudio ( RStudio version 1 . 2 . 1335 ) ( Team , 2018 ) in Windows 10 operating system , with the exception of the Q1 OR measures form SCVMC that were preprocessed in MATLAB before downsampling to Q5 in R . The most relevant R functions and packages ( beyond the installed with R ) used and the references for each function/package and methods are reported in the following table . For more details , see the source code available ( Supplementary file 1 and Source code 1 ) . The final de-identified datasets for analysis are deposited and accessible at the Open Data Commons for SCI ( odc-sci . org , RRID:SCR_016673 ) under DOIs 10 . 34945/F5R59J and 10 . 34945/F5MG68 . The R code to run all the analysis present in this publication , including visualizations , is available as supplementary material . The code would reproduce the entire analysis and plots when run using the same versions of R , RStudio , and packages specified in this publication . Otherwise results might change .
Intra-operative monitoring records ( MAP , HR ) and neurological outcome data were extracted and curated from two Level 1 trauma centers . A final cohort of 118 patients was included ( Figure 1a and Table 1 ) . The cohort represents a varied dataset of intra-operative MAP and HR patterns and respective aMAP across time in surgery and aHR across time in surgery values ( Figure 1b–c ) . Using a machine intelligence analytical pipeline ( Figure 2a ) , we extracted a similarity network of patients ( Pai and Bader , 2018; Parimbelli et al . , 2018 ) from a low-dimensional space embedded using a non-linear algorithm , Isomap ( Tenenbaum et al . , 2000 ) , on a distance matrix derived from the MAP and HR records and then performed topological network extraction using persistent homology metrics ( Rieck and Leitte , 2015; Figure 2a and Figure 2—figure supplement 1 ) . The results of this dimensionality reduction suggested that four dimensions are enough to capture most of the variance and the topological structures of the original data ( Figure 2—figure supplement 1c-e ) . Clustering the network of patients through a random-walk algorithm , Walktrap ( Pons and Latapy , 2005 ) , revealed 11 different clusters where patients were regarded to share intra-operative hemodynamic phenotypes ( Figure 2 andFigure 2—figure supplement 1f-h ) . Importantly , this workflow was unsupervised: only the OR hemodynamic time-series was used to derive patient clustering , and therefore any association captured by the network must be dependent on hemodynamic patterns . Exploratory network analysis showed a gradient distribution of patients by their aMAP ( Figure 2b–d ) and aHR ( Figure 2e–g ) , confirming that the network captured a valid representation of the raw high-dimensional dataset . We then investigated the association of the clusters to patient recovery as defined by whether the patient improved at least one AIS grade A–D ( Roberts et al . , 2017 ) between time before surgery and time of discharge from the hospital . Mapping the proportion of patients with AIS improvement onto the similarity network ( Figure 2h–j ) revealed that patients with recovery localized to clusters associated with a middle range of MAP ( Figure 2k ) . Those clusters also showed a higher proportion of less severe AIS grades at discharge ( AIS C , D , and E ) than other clusters ( Figure 2—figure supplement 2 ) . In contrast , clusters of patients showing an extreme variation of MAP were highly enriched with patients with no AIS recovery and patients with more severe AIS grades at discharge ( AIS A and B , Figure 2—figure supplement 2 ) . This analysis suggested that there is a limited range of MAP during surgery associated with neurological recovery . The exploratory network analysis revealed that clusters with higher proportion of patients that increased AIS of at least one grade were associated with having a middle range aMAP ( Figure 2 and Figure 2—figure supplements 1 and 2 ) and that clusters of patients with aMAP on the extremes contained fewer improvers . We hypothesized that there might be a non-linear relationship between intra-operative MAP and the probability of AIS grade improvement . To confirm this hypothesis , logistic regression models with LOOCV were used ( Figure 3 , Table 2 ) . We fitted a null model ( no predictors ) as well as linear , polynomial , and cubic models of aMAP ( Figure 3 , Table 2 ) to test the non-linearity of the hypothesis . The linear model showed a significant improvement over the null model with a positive association , suggesting that the higher the aMAP , the higher the probability of AIS grade improvement . However , polynomial logistic regression demonstrated a significant quadratic fit ( Table 2 ) with lower LOOCV error than the linear model , indicating that a quadratic form of aMAP better predicts the probability of improvement . Notably , the cubic model did not show significant improvement over the quadratic one . Exploratory network analysis suggested an asymmetrical function of AIS improvement with respect to aMAP ( Figure 2k ) ; we therefore also tested spline models to relax the symmetry of polynomial models . A spline model of degree 2 resulted in a significant fit over the linear model ( Table 2 ) while a spline model of degree 3 resulted in a similar fit as compared to the cubic model . There was no evidence from which to choose between the spline model of degree 2 and the quadratic model . Accordingly , we did not pursue the asymmetric model further , although we explore an asymmetric MAP range below . We sought to explore additional patient characteristics that might explain or affect MAP association with recovery . To test whether other factors could be responsible for the observed non-linear association , we first compared the quadratic model with aMAP as a predictor alone , a model that also includes several covariates ( aHR , length of surgery , days from surgery to discharge , age , and AIS grade at admission ) , and a model with only the covariates . The significance of the quadratic fit holds even after accounting for the covariates ( Table 3 , 4 ) , and none of the terms in the covariates-only model had a significant coefficient ( Table 5 ) . These results indicate that even in the presence of the other factors , aMAP is still non-linearly associated with AIS grade improvement at discharge . Patients with more severe injuries are more likely to suffer hemodynamic dysregulations ( Lehmann et al . , 1987 ) . Hence , we studied whether the relationship of MAP and AIS improvement was maintained in the subcohort of patients with an AIS grade of A at admission . We first filtered the data for the subcohort and then fitted a full model as above but without the AIS grade at admission covariate . The resulting model showed the linear aMAP coefficient to be significant and the quadratic term close to significant ( p = 0 . 14 ) with the second biggest coefficient ( Table 6 ) . A likelihood ratio test between a linear model with covariates and a quadratic model with covariates resulted in p-values = 0 . 07 . On the other hand , in the full model with covariates fitted to the entire cohort , none of the AIS grades at admission had significant coefficients , which suggested that the non-linear relationship of MAP with neurological recovery was sustained across injury severity in that model . This apparent divergence in results might be explained by the reduction in power for the AIS A cohort model . Next , given that the level of the cord injury can be related to different degrees of hemodynamic dysregulation ( Lehmann et al . , 1987 ) , we studied the effect of the NLI at admission on the association of MAP and patient recovery . Our cohort was very heterogeneous on the NLI , with most patients having cervical injuries and the rest distributed along the mid and lower segments of the cord ( Table 7 ) . Thus , we divided the population into two categories: cervical and non-cervical patients . Running the same full model with just the cervical patients resulted in similar results as compared to the full model on the entire cohort , maintaining the quadratic aMAP significance ( Table 8 ) . In the non-cervical cohort , we did not find a significant association of the quadratic aMAP to recovery ( Table 9 ) . We then performed additional analyses to determine whether this difference in aMAP relationship to recovery between cervical and non-cervical patients was due to differences in the likelihood of recovery between the two NLI populations . A univariate analysis suggested that the proportion of improvers and not improvers in the cervical and non-cervical population were marginally different ( Table 1 ) . Moreover , a logistic regression predicting AIS grade improvement by NLI categorization indicated that non-cervical patients were significantly less likely to recover ( β = –0 . 93 , p = 0 . 041 ) . While these results suggest that a quadratic aMAP is important for predicting AIS grade recovery in cervical patients , the lack of significant results in the non-cervical patients must be interpreted with caution due to the reduced number of cases , the heterogeneous distribution , and the low number of improvers in the group . Finally , we sought to determine whether the probability of recovery associated to MAP could be influenced by the time the patient is in the hospital . For that , we break down the potential causal pathway between MAP dysregulation , AIS improvement , and days from surgery to discharge . We first fitted a logistic regression model with AIS improvement as response and days to discharge as the only predictor . This resulted in a non-significant p-value of p = 0 . 32 , suggesting that days to discharge does not associate with probability of improvement . Second , we fitted a linear model with days to discharge as a response and quadratic aMAP ( both linear and quadratic terms ) as predictors . This resulted as a significant coefficient of the quadratic term ( p = 0 . 047 ) , although the model was not significant ( p = 0 . 13 for the F statistic ) and the adjusted R2 was small ( 0 . 0217 ) . We also investigated whether days to discharge interacts with MAP and quadratic MAP to predict AIS improvement , with no significant results on the interaction ( interaction days to discharge with aMAP: linear term p = 0 . 61; quadratic term p = 0 . 91 ) . These suggest that these two factors do not moderate each other . Finally , eliminating days to discharge from the full covariate model predicting AIS improvement does not have a major effect on the model fit . A likelihood ratio test between both models shows a non-significant change in variance explained ( p = 0 . 729 ) with a deviance difference of ~0 . 1% . All together indicates that the non-linear relationship between aMAP and AIS improvement is independent of the days from surgery to discharge . Since aMAP can obscure episodes of high deviation from average ( Hawryluk et al . , 2015 ) and has a non-linear relationship with recovery , we hypothesized that there might be a range of intra-operative MAP that better predicts AIS grade improvers . To test this hypothesis , we analyzed the amount of time MAP was out of a specific range ( Figure 4 ) . Since our modeling suggested both a symmetric and asymmetric range , we performed two different analyses . First , starting at a MAP of 90 mmHg , we implemented an algorithm to iteratively expand the MAP range symmetrically 1 mmHg higher and lower and calculate the time MAP was outside the range ( Figure 4a ) . Exploratory analysis of the similarity network indicated a high association between the time out of a MAP range of 73–107 mmHg with the topological distribution of patients ( Figure 4b and Figure 4—figure supplement 1 ) . To validate this range and the associated lower and upper MAP thresholds , we used a logistic model with LASSO regularization with the predictors being the time outside of each MAP range as in Figure 4b . This allowed us to systematically reduce the number of relevant predictors until only one remained ( non-zero coefficient ) . Interestingly , the logistic LASSO regression with LOOCV revealed that a MAP range from 76 to 104 mmHg was optimal in our dataset since it produced the most reproducible prediction of recovery ( average LOOCV prediction accuracy of 61 . 16%; Figure 4c and Figure 2—figure supplement 2 , Table 9 ) . Next , we studied the possibility of an asymmetric range by fixing the lower limit to 76 mmHg and increasing the upper limit by 1 mmHg at the time ( Figure 4d ) . The association of the patient distribution in the network plateau at a range of 76–116 mmHg ( Figure 4e and Figure 4—figure supplement 1 ) and the logistic LASSO found the range 76–117 mmHg be the most predictive of recovery ( average cross-validation prediction accuracy of 57 . 28%; Figure 4f and Figure 4—figure supplement 2a , Table 10 ) . While both the exploratory analysis and the logistic LASSO produced similar ranges , the later analysis is performed through statistical modeling rather than descriptive associations , and therefore we further discuss the results of the LASSO . Altogether , the findings indicate that the time of MAP outside a measurable normotensive range during surgery is associated with lower odds of recovering at least one AIS grade . Our analysis suggests the optimal range for recovery is between 76–104 and 76–117 mmHg . Notice that while range 76–104 mmHg has higher predictive utility than 76–117 mmHg ( mean LOOCV accuracy of 61 . 16 % vs . 57 . 28% ) , the difference in variance of the probability of AIS improvement explained by these two predictors is minimal ( <4% difference in RV ) . Therefore , from a modeling perspective , we broadly conclude that the upper limit of the MAP range is probably anywhere between 104 and 117 mmHg . Finally , we wanted to study the prediction utility of a model including the analyzed features together with other patient characteristics . We focused on three classification tasks: a model predicting AIS improvement of at least one grade at discharge , a model predicting AIS A at discharge , and a model predicting AIS D at discharge . We chose to predict AIS A and D instead of a multiclass prediction of the AIS at discharge in concordance to our previous studies ( Kyritsis et al . , 2021 ) and because of the low representation of other grades in our dataset ( Table 1 ) . For each of the three classification tasks , we performed an exhaustive search of all possible additive models with at least one of the predictors of interest: quadratic aMAP , aHR , length of surgery , days from surgery to discharge , age , AIS grade at admission , dichotomized NLI ( cervical , non-cervical ) , time of MAP out of range 76–104 , and time of MAP out of range 76–117 . We selected the parsimonious model as the model that minimized the small-sample corrected AIC ( Table 12 ) . Next , for the selected best model for each task , we performed LOOCV performance evaluation and prediction threshold calibration balancing prediction sensitivity and specificity ( Figure 5 ) . The model predicting AIS improvement had a cross-validated AUC of 0 . 74 , the model predicting AIS A at discharge had a cross-validated AUC of 0 . 88 , and the model predicting AIS D at discharge had a cross-validation AUC of 0 . 84 . Other metrics of classification performance can be seen in Table 12 . Both the parsimonious model predicting AIS improvement and the one predicting AIS A at discharge included quadratic aMAP as an important predictor . The model predicting AIS A also included the time of MAP out of range 76–117 mmHg . The model predicting AIS D did not include any of the MAP associated terms , suggesting that patients discharged with AIS D can be predicted without considering their MAP during OR . In fact , training the same model but with the inclusion of the quadratic aMAP term resulted in slightly worse prediction performance ( AUC 0 . 84 vs . 0 . 83 ) . Training the models predicting AIS improvement and AIS A at discharge but without a MAP component ( quadratic MAP term or time of MAP out of range ) reduced the model performance considerably ( AUC , AIS improvement: 0 . 74 vs . 0 . 52; AIS A discharge: 0 . 88 vs . 0 . 78 ) . Altogether , this suggests that models can be built for predicting AIS improvement or AIS A at discharge and that such the model performance critically depends on MAP during OR . Conversely , we did not find evidence that predicting AIS D at hospital discharge is dependent on intra-operative MAP .
Acute hypotension is common in patients with SCI due to neurogenic shock ( Lehmann et al . , 1987; Krassioukov et al . , 2007 ) and autonomic dysregulation ( Lehmann et al . , 1987 ) , probably contributing to post-traumatic spinal ischemia ( Streijger et al . , 2018; Hall and Wolf , 1987 ) , which is known to cause impaired neurological recovery in animal models ( Fehlings et al . , 1989 ) . Level 4 evidence from a small single-center case series study in the 1990s suggested that MAP augmentation to 85–90 mmHg during the first 5–7 days after injury was linked to neurological recovery in acute SCI ( Levi et al . , 1993; Vale et al . , 1997 ) . These results are the basis of clinical guidelines for avoidance of hypotension in acute SCI management ( Aarabi et al . , 2013 ) . However , while numerous clinical studies support MAP augmentation , the arbitrary , recommended MAP goal has been controversial ( Cohn et al . , 2010; Hawryluk et al . , 2015; Saadeh et al . , 2017 ) . Recent analysis of high-frequency ICU monitoring data ( Hawryluk et al . , 2015 ) and systematic meta-analysis of post-surgery management ( Saadeh et al . , 2017 ) suggest that the MAP threshold to avoid is actually lower ( ~75 mmHg ) than the current recommendation of 85 mmHg , and that MAP management might be effective at shorter duration ( < 5 days post-injury ) than the 7-day goal ( Saadeh et al . , 2017 ) . The present study represents a multicenter , data-driven , and cross-validated re-evaluation in a different setting ( during surgery as compared with prior ICU studies ) . Our analysis support that there must be a MAP range during surgery at which neurological recovery is maximized , providing further evidence that MAP management for maintaining normotension might be more beneficial for patient outcome than MAP augmentation for hypotension avoidance alone ( Ehsanian , 2020; Nielson et al . , 2015 ) . The low boundary of 76 mmHg found in our ultra-early analysis further supports previous suggestions for lowering the intervention threshold ( Cohn et al . , 2010; Hawryluk et al . , 2015; Saadeh et al . , 2017 ) . On the other side , we find an upper boundary to MAP management between 104 and 117 mmHg , above which the probability of improvement is reduced . Thus , the proposal for MAP augmentation with vasopressors to increase spinal cord perfusion ( Saadoun and Papadopoulos , 2016 ) has a limit since spinal hyper-perfusion pressure can be detrimental ( Saadoun and Papadopoulos , 2016 ) . The physiological rational is that high blood pressure induced by vasopressors can translate to increased risk of hemorrhage in the injured spinal cord , exacerbating tissue damage ( Soubeyrand et al . , 2014; Streijger et al . , 2018; Guha et al . , 1987 ) . Moreover , the use of some vasopressors might cause more complications in patients ( Inoue et al . , 2014 ) while also potentially contributing to intra-spinal hemorrhage . In fact , recent results in acute experimental SCI suggest controlling for hemodynamic dysregulation through a cardiac-focused treatment instead of using standard vasopressors such as norepinephrine ( Williams et al . , 2020 ) . Specifically , the authors demonstrated that dobutamine can correct for hemodynamic anomalies and increase blood flow to the spinal cord while reducing the risk of hemorrhage compared to norepinephrine . Furthermore , hypertension during surgery in rodent SCI has been associated with lower probability of recovery ( Nielson et al . , 2015 ) , probably related to higher tissue damage . Our findings together with previous work ( Ehsanian , 2020 ) also translate these animal study results to humans , indicating that prolonged periods of hypertension early after injury can be a predictor of poor neurological recovery in patients with SCI . An important finding of our study is the indication that level of injury and injury severity modify the association of MAP with neurological recovery . We observed that normotensive MAP during surgery predicts AIS improvement in patients with cervical SCI but not in patients with lower injuries ( thoracic , lumbar , and sacral ) . While the heterogeneity of our population and low sample size for patients with non-cervical SCI sets limitations on interpreting the results , our finding raises a relevant question regarding precision management of patients with SCI . Patients with cervical SCI present more frequently with hemodynamic and cardiac abnormalities than patients with thoracolumbar SCI , increasing the need for treatment ( Lehmann et al . , 1987 ) . This is due to sympathetic dysregulation in upper cord injuries , which reduces sympathetic tone likely causing reduced heart contractility , bradycardia , and hypotension ( Lehmann et al . , 1987; Myers et al . , 2007; Teasell et al . , 2000 ) . This is particularly true for individuals with severe cervical injury ( Lehmann et al . , 1987 ) . In that context , our results may indicate that those patients with cervical SCI that are more difficult to maintain within a normotensive MAP are probably less likely to improve in neurological function . Alternatively , it could also be the case that more aggressive MAP management treatment is performed in these patients during their course in the hospital , which could increase the chances of aggravating secondary cord injury . Hemodynamic instability early after injury could serve as a prognostic physiology-based biomarker in a subset of the population , providing a potential tool for precision medicine in SCI . Hence , we have established basic prediction models around non-linear features of MAP that could serve as a benchmark for future machine learning development . Another relevant contribution of this work is the analytical workflow . First , we demonstrate that topology-based analytics can undercover associations for hypothesis generation during exploratory analysis in a cross-species validation . Our group has previously used a similar workflow in data from animal models ( Nielson et al . , 2015 ) suggesting that hypertension is a predictor of neurological recovery and providing rational for the present study . Hence , our work constitutes a successful story of translating machine intelligence analytical tools from animals to humans . Second , we provide further illustration that patient similarity networks are useful and interpretable representations of multidimensional datasets that capture associations during exploratory analysis that can then be validated by network-independent confirmatory analysis . Third , we successfully combine Isomap , a non-linear dimensionality reduction method , with topology-based metrics to evaluate embedding solutions . Fourth , our method for finding the MAP range could be expanded and deployed in other settings . Lastly , our workflow captures representations of multidimensional time-series of different lengths into a network that is actionable . Limitations of this study include the retrospective nature of the analysis , the relatively small sample size ( although large for SCI ) , and the use of an estimated ordinal scale ( AIS grade ) as an indicator of neurological recovery . An important consideration is the difficulty of determining AIS grade early after injury . Moreover , other factors not considered in this analysis such as MAP levels before or after surgery or the use of vasopressors might influence the results . Future research with more granular data should address these and other important questions . | Spinal cord injury is a devastating condition that involves damage to the nerve fibers connecting the brain with the spinal cord , often leading to permanent changes in strength , sensation and body functions , and in severe cases paralysis . Scientists around the world work hard to find ways to treat or even repair spinal cord injuries but few patients with complete immediate paralysis recover fully . Immediate paralysis is caused by direct damage to neurons and their extension in the spinal cord . Previous research has shown that blood pressure regulation may be key in saving these damaged neurons , as spinal cord injuries can break the communication between nerves that is involved in controlling blood pressure . This can lead to a vicious cycle of dysregulation of blood pressure and limit the supply of blood and oxygen to the damaged spinal cord tissue , exacerbating the death of spinal neurons . Management of blood pressure is therefore a key target for spinal cord injury care , but so far , the precise thresholds to enable neurons to recover are poorly understood . To find out more , Torres-Espin , Haefeli et al . used machine learning software to analyze previously recorded blood pressure and heart rate data obtained from 118 patients that underwent spinal cord surgery after acute spinal cord injury . The analyses revealed that patients who suffered from either low or high blood pressure during surgery had poorer prospects of recovery . Statistical models confirming these findings showed that the optimal blood pressure range to ensure recovery lies between 76 to 104-117 mmHg . Any deviation from this narrow window would dramatically worsen the ability to recover . These findings suggests that dysregulated blood pressure during surgery affects to odds of recovery in patients with a spinal cord injury . Torres-Espin , Haefeli et al . provide specific information that could improve current clinical practice in trauma centers . In the future , such machine learning tools and models could help develop real-time models that could predict the likelihood of a patient’s recovery following spinal cord injury and related neurological conditions . | [
"Abstract",
"Introduction",
"Methods",
"Results",
"Discussion"
] | [
"medicine",
"computational",
"and",
"systems",
"biology"
] | 2021 | Topological network analysis of patient similarity for precision management of acute blood pressure in spinal cord injury |
The transport receptor Crm1 mediates the export of diverse cargos containing leucine-rich nuclear export signals ( NESs ) through complex formation with RanGTP . To ensure efficient cargo release in the cytoplasm , NESs have evolved to display low affinity for Crm1 . However , mechanisms that overcome low affinity to assemble Crm1-export complexes in the nucleus remain poorly understood . In this study , we reveal a new type of RanGTP-binding protein , Slx9 , which facilitates Crm1 recruitment to the 40S pre-ribosome-associated NES-containing adaptor Rio2 . In vitro , Slx9 binds Rio2 and RanGTP , forming a complex . This complex directly loads Crm1 , unveiling a non-canonical stepwise mechanism to assemble a Crm1-export complex . A mutation in Slx9 that impairs Crm1-export complex assembly inhibits 40S pre-ribosome export . Thus , Slx9 functions as a scaffold to optimally present RanGTP and the NES to Crm1 , therefore , triggering 40S pre-ribosome export . This mechanism could represent one solution to the paradox of weak binding events underlying rapid Crm1-mediated export .
In all eukaryotes , transport between the nucleus and the cytoplasm is channeled through nuclear pore complexes ( NPCs ) embedded within the nuclear envelope ( Tran and Wente , 2006; D'Angelo and Hetzer , 2008 ) . Yeast NPCs are approximately 60 MDa ( Fernandez-Martinez et al . , 2012 ) and are composed of multiple copies of about 30 nucleoporins ( Rout et al . , 2000 ) . Cargos of different sizes and charges pass through the central transport channel of the NPC , which is filled with a meshwork of natively unfolded Phe-Gly ( FG ) -repeats present in FG-nucleoporins ( Frey and Görlich , 2007; Terry and Wente , 2009 ) . These nucleoporins generate a permeability barrier that allows passive diffusion of small molecules , such as ions and metabolites ( Frey and Görlich , 2007; Patel et al . , 2007 ) . Macromolecules ( >40 kDa ) require the assistance of nuclear transport receptors , including members of the importin-β-like family , to efficiently overcome this selectivity barrier ( Macara , 2001; Ribbeck and Görlich , 2001; Rout and Aitchison , 2001 ) . These transport receptors , also termed importins and exportins , mediate the majority of molecular exchange between the nucleus and cytoplasm ( Cook et al . , 2007 ) . Transport receptors recognize their cargo via specific signal sequences ( Cook and Conti , 2010; Xu et al . , 2010; Güttler and Görlich , 2011 ) and translocate them through the NPC by transiently interacting with FG-repeats . The small GTPase Ran coordinates the movement of importins and exportins between the nucleus and the cytoplasm , and directs the compartment-specific binding and release of transported cargos ( Fried and Kutay , 2003; Pemberton and Paschal , 2005; Cook et al . , 2007 ) . Ran exists in both GDP- and GTP-bound forms , and the two states are asymmetrically distributed , with RanGTP significantly enriched in the nucleus ( Nakielny and Dreyfuss , 1999; reviewed in Görlich and Kutay , 1999 ) . This gradient of RanGTP is established through the spatial separation of regulators of the Ran-cycle ( Izaurralde et al . , 1997 ) . Whereas the Ran guanine nucleotide exchange factor , RCC1 ( Prp20 in yeast ) ( Bischoff and Ponstingl , 1991; Fleischmann et al . , 1991 ) , localizes to the nucleus , the Ran GTPase-activating protein , RanGAP1 ( Rna1 in yeast ) , is found in the cytoplasm ( Hopper et al . , 1990; Matunis et al . , 1996 ) . Interactions between RanGTP and transport receptors are crucial for the directionality of nucleocytoplasmic exchange ( Nachury and Weis , 1999 ) . In the nucleus , RanGTP induces the release of imported cargos from importins ( Rexach and Blobel , 1995; Görlich et al . , 1996 ) . In addition , RanGTP promotes the interaction of cargos with exportins for their transport to the cytoplasm ( Fornerod et al . , 1997; Kutay et al . , 1997; Stade et al . , 1997; Solsbacher et al . , 1998 ) . In budding yeast , ribosome assembly accounts for a major proportion of the nucleocytoplasmic transport ( Rout et al . , 1997; Schlenstedt et al . , 1997; Sydorskyy et al . , 2003; Kressler et al . , 2012; Schütz et al . , 2014 ) . mRNAs encoding ribosomal proteins ( r-proteins ) are exported into the cytoplasm . Newly synthesized r-proteins are imported into the nucleus and then targeted to the nucleolus for incorporation into nascent pre-ribosomes ( Schütz et al . , 2014 ) . Additionally , >300 transiently interacting non-ribosomal assembly factors aid the construction and maturation of ribosomes ( Bassler et al . , 2001; Dragon et al . , 2002; Grandi et al . , 2002; Schäfer et al . , 2003; Gerhardy et al . , 2014 ) . Correctly assembled pre-ribosomal particles are transported through NPCs into the cytoplasm ( Tschochner and Hurt , 2003; Panse and Johnson , 2010 ) . In addition to other cargos , it is estimated that each yeast NPC facilitates the export of ∼25 pre-ribosomal particles every minute ( Warner , 1999 ) . Transporting pre-ribosomal cargos from the nucleus through the NPC into the cytoplasm , therefore , represents a major task for the export machinery . The Ran-cycle-dependent exportin Crm1 plays an essential role in exporting pre-ribosomal particles to the cytoplasm ( Hurt et al . , 1999; Moy and Silver , 1999; Gadal et al . , 2001; Johnson et al . , 2002 ) . Crm1 recognizes and directly binds leucine-rich nuclear export signals ( NESs ) on cargos in the presence of RanGTP to form a Crm1-export complex ( Fornerod et al . , 1997 ) . Although an essential NES-containing adaptor , Nmd3 , has been identified for 60S pre-ribosome export ( Ho and Johnson , 1999; Ho et al . , 2000; Gadal et al . , 2001 ) , a similar adaptor to recruit Crm1 to the 40S pre-ribosome remains elusive . It has been suggested that multiple NES-containing adaptors such as Ltv1 and Rio2 recruit Crm1 ( Seiser et al . , 2006; Zemp et al . , 2009; Merwin et al . , 2014 ) , thereby guaranteeing efficient 40S pre-ribosome export . The essential mRNA and 60S pre-ribosome transport receptor Mex67-Mtr2 , which does not directly utilize the RanGTP gradient , also facilitates nuclear export of the 40S pre-ribosomal cargo ( Faza et al . , 2012 ) . Despite the identification of several components of the export machinery , assembly steps and mechanisms that prepare the pre-ribosomal cargo for transport through the NPC remain largely unexplored . To initiate export , Crm1 must cooperatively bind RanGTP and its NES-containing cargo in the nucleus , to form a trimeric export complex ( Petosa et al . , 2004; Dong et al . , 2009; Monecke et al . , 2013 ) . NESs have evolved to maintain relatively low affinity to Crm1 to avoid defects in disassembly of the export complex in the cytoplasm ( Engelsma et al . , 2004; Kutay and Güttinger , 2005 ) . Mechanisms that promote Crm1-export complex assembly and thereby ensure export of the NES-containing cargo at a reasonable rate remain poorly understood . Here , we identify yeast Slx9 as a new type of RanGTP-binding protein that promotes assembly of a Crm1-export complex on the 40S pre-ribosome-associated NES-containing adaptor Rio2 . Our data raise the possibility of a yet-unidentified family of RanGTP-binding proteins that act as scaffolds to optimally present RanGTP and NES-containing cargos to Crm1 , orchestrating a non-cooperative stepwise assembly that drives fast and efficient Crm1-mediated export .
Slx9 is a 24-kDa basic protein that co-enriches with pre-ribosomal particles in the 40S maturation pathway ( Gavin et al . , 2002; Faza et al . , 2012 ) and is required for efficient nuclear export of 40S pre-ribosomes ( Li et al . , 2009; Faza et al . , 2012 ) . However , the precise contribution of Slx9 to 40S pre-ribosome export has remained unclear . To investigate the function of yeast Slx9 , we generated slx9 variants by random mutagenesis and analyzed the growth of the resulting strains at different temperatures . One allele , slx9L108P , hereafter termed slx9-1 , caused slow growth at temperatures between 20°C and 30°C , indistinguishable from slx9∆ cells ( Figure 1A , top panel ) . Like slx9∆ , slx9-1 cells were not impaired in growth at 37°C ( Figure 1A ) . Western analysis of whole cell lysates revealed that Slx9 and Slx9-1 were present at similar levels ( Figure 1A , bottom panel ) , indicating that impaired growth of the slx9-1 strain is not due to reduced levels of the mutant protein . As previously observed , Slx9-GFP localized primarily to the nucleolus , where it co-localized with the nucleolar marker Gar1-mCherry , as well as to the nucleoplasm ( Faza et al . , 2012 and Figure 1B ) . Slx9-1-GFP displayed an identical localization ( Figure 1B ) , indicating that the mutant protein is correctly targeted to the nucleolus and nucleoplasm . Slx9 maximally co-enriched with Enp1-TAP that purifies both the 90S and 40S pre-ribosomes ( Faza et al . , 2012 ) . A similar purification from slx9-1 cells revealed that Enp1-TAP co-enriched at least as much Slx9-1 mutant protein as Slx9 ( Figure 1C ) . Together , these data show that Slx9-1 is correctly expressed , localized , and recruited to 40S pre-ribosomes . 10 . 7554/eLife . 05745 . 003Figure 1 . slx9-1 phenocopies the slx9∆ mutation . ( A ) The slx9-1 allele does not complement the slow growth of slx9∆ cells . Top: SLX9 , slx9∆ , and slx9-1 cells were spotted in 10-fold dilutions on SD-plates and grown at the indicated temperatures for 3–6 days . Bottom: Slx9 protein levels from whole cell extracts derived from the indicated strains were determined by Western analysis using antibodies directed against Slx9 . Levels of the protein Arc1 served as a loading control . ( B ) Slx9-1 localizes to the nucleolus/nucleoplasm . Cells expressing Gar1-mCherry and Slx9-GFP or Slx9-1-GFP were grown until mid-log phase . Localization of the indicated fusion proteins was analyzed by fluorescence microscopy . Gar1-mCherry served as a nucleolar marker . Scale bar = 5 µm . ( C ) Slx9-1 is recruited to the early 40S pre-ribosome . Enp1-TAP was isolated by tandem affinity purification ( TAP ) from the indicated strains . Calmodulin-eluates were separated on a 4–12% gradient gel and analyzed by either silver staining or Western using the indicated antibodies . The ribosomal protein uS7 served as a loading control . ( D ) slx9-1 cells are impaired in nuclear export of 40S pre-ribosomes . Top: localization of uS5-GFP was monitored by fluorescence microscopy . Bottom: localization of 20S pre-rRNA was analyzed by FISH using a Cy3-labeled oligonucleotide complementary to the 5′ portion of ITS1 ( red ) . Nuclear and mitochondrial DNA was stained by DAPI ( blue ) . Scale bar = 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05745 . 003 Previous studies showed that slx9∆ cells accumulate the small subunit reporter uS5-GFP ( yeast Rps2 , nomenclature according to Ban et al . , 2014 ) and 20S pre-rRNA in the nucleoplasm ( Li et al . , 2009; Faza et al . , 2012 ) , indicating a defect in 40S pre-ribosome export . Using these reporters , we tested whether slx9-1 cells have defects in 40S pre-ribosome export . Whereas WT cells displayed cytoplasmic uS5-GFP localization , slx9-1 cells showed a strong nuclear accumulation of this reporter , similar to that observed in slx9∆ cells ( Faza et al . , 2012 and Figure 1D , top panel ) . As expected , fluorescence in situ hybridization ( FISH ) of 20S pre-rRNA in WT cells showed a strong nucleolar Cy3-ITS1 signal ( red ) with virtually no nucleoplasmic staining . In contrast , slx9-1 cells displayed a nucleoplasmic signal of Cy3-ITS1 localization , which co-localized with the DAPI signal ( Figure 1D , bottom panel ) . These data indicate that slx9-1 cells , like slx9∆ cells ( Faza et al . , 2012 ) , are impaired in 40S pre-ribosome export . Therefore , we conclude that Slx9-1 is recruited to the 40S pre-ribosome but fails to fulfill its function in nuclear export of the pre-ribosomal cargo . Mutations in MEX67 and MTR2 ( mex67Δloop and mtr2Δloop116-137 ) , which encode the essential transport receptor Mex67-Mtr2 , are synthetically lethal when combined with the slx9Δ mutant ( Faza et al . , 2012 ) . In addition , we found that slx9∆ displayed a synthetic growth defect with a strain expressing Rrp12-GFP ( Figure 2A ) . Rrp12 is a 40S pre-ribosome export factor that directly interacts with FG-rich nucleoporins ( Oeffinger et al . , 2004 ) . Based on these genetic interactions , we asked whether Slx9 functions as a novel export factor for the 40S pre-ribosome . A salient feature of an export factor is that it rapidly shuttles between the nucleus and the cytoplasm . To test this , we employed the established heterokaryon assay ( Altvater et al . , 2014 ) . WT cells expressing Slx9-GFP were mated to kar1-1 cells , which are deficient in nuclear fusion after cell conjugation , leading to heterokaryon formation . In order to distinguish between the two nuclei , kar1-1 cells also contained Nup82-mCherry as a marker for nuclear pores . As controls , we used the shuttling 40S assembly factor Enp1 and the non-shuttling nucleolar protein Gar1 fused to GFP . Whereas Gar1-GFP was never seen in the nucleus of kar1-1 cells ( red signal ) , both Enp1-GFP and Slx9-GFP localized to both nuclei ( Figure 2B ) . These data are consistent with the shuttling of Slx9 between the nuclear and the cytoplasmic compartments . 10 . 7554/eLife . 05745 . 004Figure 2 . Slx9 is a RanGTP binding protein . ( A ) slx9-1 genetically interacts with factors involved in 40S pre-ribosome export . slx9-1 is synthetically lethal with mex67∆loop , mtr2∆loop116-137 , or yrb2∆ and strongly synthetically enhanced with rrp12-GFP . Strains containing the indicated WT and mutant alleles were spotted in 10-fold serial dilutions on 5-FOA-SD or SD and grown at 20–30°C for 3–6 days . ( B ) Slx9 shuttles between the nucleus and the cytoplasm . Cells expressing Enp1-GFP , Gar1-GFP , or Slx9-GFP were mated with kar1-1 cells expressing Nup82-mCherry . The resulting heterokaryons were analyzed by fluorescence microscopy . Scale bar = 5 µm . ( C ) Slx9 directly binds to RanGTP . GST-Slx9 or GST-Ssb1C was immobilized on GSH-Sepharose before incubating with either buffer alone or buffer containing 2 µM His6-RanQLGTP , 50 nM Crm1-His6 or 2 µM His6-RanQLGTP , and 50 nM Crm1-His6 . After washing , bound proteins were eluted in LDS sample buffer , separated by SDS-PAGE and visualized by Coomassie staining or Western blotting using the indicated antibodies . L = input . ( D ) Slx9 specifically interacts with the GTP-bound form of Ran . GST-Slx9 , GST-Yrb1 , or GST-Ntf2 was immobilized on GSH-Sepharose and incubated with buffer alone or 2 µM His6-Ran loaded with GDP or GTP . Analysis of the eluted proteins was carried out as described in ( C ) . L = input . ( E ) Slx9-1 binding to RanGTP is impaired . Top: GST-Slx9 or GST-Slx9-1 immobilized on GSH-Sepharose was incubated with buffer alone or 2 µM His6-RanQLGTP . Analysis of the eluted proteins was carried out as described in ( C ) . L = input . Bottom: bar graph depicts the bound His6-RanQLGTP Western blot signal normalized to GST-Slx9 and GST-Slx9-1 levels , respectively . Four independent experiments were performed and Western blots were quantified by software ImageJ ( Version 1 . 44o ) . Error bars ( S . D . ) are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 05745 . 004 We wondered whether Slx9 functions directly in 40S pre-ribosome export as a NES-containing adaptor for the exportin Xpo1 ( hereafter termed Crm1 ) . We , therefore , investigated whether Slx9 binds Crm1 in the presence of Gsp1Q71L ( Maurer et al . , 2001 ) in vitro ( equivalent to the human RanQ69L GTP-stabilized mutant , hereafter termed RanQLGTP; Bischoff et al . , 1994 ) . The C-terminal domain of Ssb1 ( Ssb1C ) that contains a functional NES ( Shulga et al . , 1999 ) served as a positive control . Unlike Ssb1C ( Maurer et al . , 2001 and Figure 2C , lane 8 ) , Slx9 was unable to form a trimeric export complex with Crm1 and RanQLGTP ( Figure 2C , lane 4 ) . Surprisingly , these studies revealed that Slx9 directly bound RanQLGTP ( Figure 2C , lane 2 and 4 ) . Therefore , although Slx9 does not contain a functional NES , it is a Ran-binding protein . Since Slx9 is a shuttling protein , we tested whether Slx9 interacts with both RanGTP and RanGDP in vitro . As controls , we used Ntf2 , the import factor for RanGDP ( Ribbeck et al . , 1998; Smith et al . , 1998 and Figure 2D , lane 8 ) and the yeast RanBP1 homolog Yrb1 that binds to RanGTP ( Schlenstedt et al . , 1995 and Figure 2D , lane 6 ) . We found that , like Yrb1 , Slx9 interacted exclusively with RanGTP ( Figure 2D , lanes 2 and 3 ) . Based on these data , we conclude that Slx9 is a shuttling RanGTP-binding protein . The slx9-1 mutant did not rescue the slow growth and impaired 40S pre-ribosome export of slx9∆ cells ( Figure 1A ) . Furthermore , like slx9∆ cells , slx9-1 cells genetically interacted with mex67 and mtr2 mutants ( mex67∆loop and mtr2∆116-137 ) and rrp12-GFP ( Figure 2A ) . These findings prompted us to test whether Slx9-1 binds to RanQLGTP in vitro , using the assay described above . We found a decrease of approximately 50% in the levels of RanQLGTP bound to Slx9-1 as compared to Slx9 ( Figure 2E ) . Based on these data , we conclude that Slx9-1 is modestly impaired in binding RanQLGTP . A conserved basic patch on Ran is involved in the interaction with known Ran-binding proteins ( Nilsson et al . , 2001 ) . Based on homology to human Ran , arginine 142 , and lysine 143 in yeast Ran were mutated to alanine ( RanQLGTPRKAA ) or glutamate ( RanQLGTPRKEE ) and the contribution of this basic patch to Slx9:RanGTP complex formation was analyzed in vitro . In agreement with previous studies ( Nilsson et al . , 2001 ) , these RanQLGTP mutants bound weakly to the importin β-like transport receptor , Kap123 ( Figure 3A , compare lane 10 with lanes 11 and 12 ) , and interacted more strongly with the RanBP1 homolog Yrb1 ( Figure 3A , compare lane 6 with lanes 7 and 8 ) . Pull down studies of Slx9 and these Ran mutants showed that the interactions between Slx9 and these two Ran mutants were impaired , with the charge reversal mutant having a more severe effect than the alanine mutant ( Figure 3A , compare lane 2 with lanes 3 and 4 ) . Altogether , these results suggest that , similar to Kap123 , Slx9 binding to RanQLGTP involves the basic patch . 10 . 7554/eLife . 05745 . 005Figure 3 . The basic patch and acidic tail of Ran modulates interactions with Slx9 . ( A ) The basic patch of RanQLGTP contributes to Slx9 binding . GST-Slx9 , GST-Yrb1 , or GST-Kap123 immobilized on GSH-Sepharose was incubated with buffer alone or 2 µM Ran ( His6-RanQLGTP , His6-RanQLGTPRKAA , or His6-RanQLGTPRKEE ) . After washing , bound proteins were eluted in LDS sample buffer , separated by SDS-PAGE and visualized by Coomassie staining or Western blotting using the indicated antibody . L = input . ( B ) The acidic tail of RanQLGTP negatively regulates interactions with Slx9 . GST-Slx9 , GST-Yrb1 , or GST-Kap123 immobilized on GSH-Sepharose was incubated with buffer alone or 2 µM Ran ( His6-RanQLGTP or His6-Ran∆CQLGTP ) . Analysis of the eluted proteins was carried out as described in ( A ) . L = input . DOI: http://dx . doi . org/10 . 7554/eLife . 05745 . 005 The C-terminal acidic tail of Ran ( -DEDDAL ) plays a crucial role in the interaction with small RanGTP-binding proteins such as Yrb1 . As expected , RanQLGTP lacking the C-terminal acidic tail ( Ran∆CQL GTP ) failed to interact with Yrb1 in vitro ( Maurer et al . , 2001 and Figure 3B , lane 6 ) . We tested whether the acidic tail contributes to the interaction between RanGTP and Slx9 . In contrast to Yrb1 , Ran∆CQLGTP bound stronger to Slx9 compared to RanQLGTP ( Figure 3B , compare lane 2 and 3 ) . This enhanced interaction was specific , since Kap123 bound Ran∆CQLGTP and RanQLGTP to a similar extent ( Figure 3B , compare lanes 8 and 9 ) . These data suggest that the C-terminal acidic tail of Ran negatively regulates RanGTP:Slx9 interactions . Cells lacking the RanGTP- and Crm1-binding protein Yrb2 ( yrb2Δ ) exhibit strong nucleoplasmic accumulation of uS5-GFP and 20S pre-rRNA as well as reduced abundance of 40S subunits ( Moy and Silver , 2002; Altvater et al . , 2012 ) . In addition , mex67Δloop and mtr2Δloop116-137 are synthetically lethal with yrb2Δ ( Faza et al . , 2012 ) . These findings led us to test whether SLX9 genetically interacts with YRB2 . Both slx9∆ and slx9-1 were synthetically lethal with yrb2∆ ( Figure 2A ) , suggesting that Slx9 and Yrb2 functionally overlap to ensure proper nuclear export of 40S pre-ribosomes . Yrb2 and its human homolog , RanBP3 , stimulate the assembly of Crm1-export complexes on certain NES-containing cargos by cooperatively binding Crm1 and RanGTP ( Englmeier et al . , 2001; Lindsay et al . , 2001; Koyama et al . , 2014 ) . The strong genetic interaction between SLX9 and YRB2 raised the possibility that Slx9 also functions in Crm1-complex assembly . However , our interaction studies showed that Slx9 binds RanGTP , but not Crm1 ( Figure 2C , lane 4 ) . We , therefore , wondered whether Slx9 instead facilitates the assembly of a Crm1-export complex by bringing together RanGTP and the NES-containing adaptor . Two 40S pre-ribosome-associated factors , hLtv1 and hRio2 , have been shown to bind Crm1 in the presence of RanGTP ( Zemp et al . , 2009 ) . In agreement with these studies , we found that yeast Ltv1 and yeast Rio2 formed trimeric complexes with Crm1 and RanQLGTP via a cooperative mechanism in vitro ( Figure 4A , lanes 4 and 12 ) . Moreover , the C-terminal regions of these proteins are predicted to contain a NES ( Zemp et al . , 2009; Merwin et al . , 2014 and Figure 4A , top panel ) , and indeed , Rio2 and Ltv1 mutant proteins lacking NESs were unable to form trimeric export complexes ( Figure 4A , lanes 8 and 16 ) . Western analyses revealed that Rio2 , but not Ltv1 , interacted weakly with RanQLGTP , independent of Crm1 ( Figure 4A , lane 2 ) . 10 . 7554/eLife . 05745 . 006Figure 4 . Slx9 directly binds the 40S pre-ribosome nuclear export signal ( NES ) -containing adaptor Rio2 and RanQLGTP . ( A ) Rio2 and Ltv1 export complex formation requires their C-terminal NESs . Top: the positions of the Rio2 and Ltv1 NESs are shown . Hydrophobic residues in these NESs are highlighted in red . Bottom: GST-Rio2 , GST-Rio2∆NES , GST-Ltv1 , or GST-Ltv1∆NES was immobilized on GSH-Sepharose , and complex formation was analyzed as in Figure 2C . L = input . ( B ) Slx9 directly interacts with Rio2 . Immobilized GST-Rio2 or GST-Ltv1 was incubated with buffer alone or 0 . 5 µM Slx9 . Conversely , immobilized GST-Slx9 was incubated with buffer alone or with lysate containing His6-Nmd3 or His6-Rio2 . Analysis of the eluted proteins was carried out as described in Figure 2C . L = input . DOI: http://dx . doi . org/10 . 7554/eLife . 05745 . 006 Ltv1 and Rio2 co-enrich with the Enp1-TAP particle that also maximally co-purifies Slx9 ( Figure 1C; Schütz et al . , 2014 ) suggesting that these NES-containing proteins might interact with Slx9 . To test this , we incubated immobilized GST-Rio2 or GST-Ltv1 with Slx9 . Slx9 directly bound to Rio2 but not to Ltv1 ( Figure 4B , left panel ) . Conversely , GST-Slx9 interacted with Rio2 but not with the essential 60S pre-ribosome-associated NES-containing adaptor Nmd3 ( Figure 4B , right panel ) . Moreover , the GST-Rio2:Slx9 complex efficiently recruited RanQLGTP ( Figure 5A ) . The level of recruitment was not affected by the prior presence or absence of Slx9 , since GST-Rio2 saturated with Slx9 recruited similar amounts of RanQLGTP ( Figure 5B , compare top and bottom panels ) . Also , the levels of RanQLGTP recruitment to GST-Slx9 were not affected by the presence or absence of Rio2 ( Figure 5C , compare top and bottom panels ) . Altogether , these data suggest that Rio2 binds Slx9 and RanQLGTP using distinct surfaces . 10 . 7554/eLife . 05745 . 007Figure 5 . Slx9 binds to Rio2 and RanGTP using distinct binding surfaces . ( A ) GST-Rio2 was immobilized on GSH-Sepharose and incubated with buffer , 2 µM His6-RanQLGTP , or 0 . 5 µM Slx9 ( +1 ) . After washing , the GST-Rio2:Slx9 complex was incubated with 2 µM His6-RanQLGTP ( +2 ) . Analysis of the eluted proteins was carried out as described in Figure 2C . L = input . ( B ) RanGTP does not displace Slx9 from a preformed GST-Rio2:Slx9 complex . Top: immobilized GST-Rio2 was incubated with buffer or increasing concentrations of His6-RanQLGTP ( 62 . 5 nM–32 µM ) . Bottom: immobilized GST-Rio2 was incubated with either buffer or 1 µM Slx9 . The unbound Slx9 was washed away , and the resulting GST-Rio2:Slx9 complex was incubated with increasing concentrations of His6-RanQLGTP ( 62 . 5 nM–32 µM ) . Analysis of the eluted proteins was carried out as described in Figure 2C . L = input . ( C ) RanQLGTP does not displace Rio2 from a preformed GST-Slx9:Rio2 complex . Top: immobilized GST-Slx9 was incubated with buffer or increasing concentrations of His6-RanQLGTP ( 62 . 5 nM–32 µM ) . Bottom: immobilized GST-Slx9 was incubated with excess of Rio2 . The unbound Rio2 was washed away , and the resulting complex GST-Slx9:Rio2 complex was incubated with increasing concentrations of His6-RanQLGTP ( 62 . 5 nM–32 µM ) . Analysis of the eluted proteins was carried out as described in Figure 2C . L = input . DOI: http://dx . doi . org/10 . 7554/eLife . 05745 . 007 We next investigated whether the GST-Rio2:Slx9:RanQLGTP complex could directly recruit Crm1 . To this end , pre-formed GST-Rio2:RanQLGTP or GST-Rio2:Slx9:RanQLGTP complexes ( summarized in Figure 6A ) were incubated with buffer alone or Crm1 ( Figure 6B , lanes 3 , 4 and 7 , 8 ) . Only the GST-Rio2:Slx9:RanQLGTP complex efficiently recruited Crm1 ( Figure 6B , compare lanes 4 and 8 ) . The Crm1-recruitment to a GST-Rio2:Slx9:RanQLGTP complex was dependent on the NES of Rio2 , since a GST-Rio2∆NES:Slx9:RanQLGTP complex was unable to bind Crm1 ( Figure 6B , compare lanes 8 and 10 ) . Moreover , Crm1 recruitment was also dependent on the RanQLGTP bound to Rio2 , since a GST-Rio2:Slx9 complex was unable to bind Crm1 ( Figure 6C , lane 4 ) . These studies indicate that , in order to recruit Crm1 in a non-cooperative manner , Rio2 must bind to both Slx9 and RanGTP . 10 . 7554/eLife . 05745 . 008Figure 6 . Slx9 promotes stepwise assembly of a Crm1-export complex on the NES of Rio2 . ( A ) Flow chart depicting the experimental setup to assemble a Rio2:Slx9:RanQLGTP:Crm1 complex . Immobilized GST-Rio2 was sequentially incubated with Slx9 ( red ) , RanQLGTP ( purple ) , and Crm1 ( green ) . Unbound protein was washed away after each incubation step . ( B ) Crm1 is recruited to the GST-Rio2:Slx9:RanGTP complex in a NES-dependent manner . Immobilized GST-Rio2 or GST-Rio2∆NES was incubated with buffer alone or 0 . 5 µM Slx9 , followed by the stepwise addition of 0 . 2 µM His6-RanQLGTP and 50 nM Crm1-His6 , as depicted in ( A ) . After a final washing step , bound proteins were analyzed as in Figure 2C . L = input . ( C ) Crm1 is not recruited to the GST-Rio2:Slx9 complex . Immobilized GST-Rio2 was incubated with buffer alone or 0 . 5 µM Slx9 , followed by addition of buffer , 50 nM Crm1-His6 , or the stepwise addition of 0 . 2 µM His6-RanQLGTP and 50 nM Crm1-His6 as depicted in ( A ) . Analysis of the bound proteins was carried out as described in Figure 2C . L = input . ( D ) Recruitment of Crm1 to a Rio2:Slx9-1:RanQLGTP complex is impaired . Immobilized GST-Rio2 was incubated with buffer alone , 0 . 5 µM Slx9 or 0 . 5 µM Slx9-1 , followed by the stepwise addition of 0 . 2 µM His6-RanQLGTP and 50 nM Crm1-His6 as depicted in ( A ) . Analysis of the bound proteins was carried out as described in Figure 2C . L = input . DOI: http://dx . doi . org/10 . 7554/eLife . 05745 . 008 We next investigated whether Slx9-1 could assemble a GST-Rio2:Slx9-1:RanQLGTP complex . In vitro binding studies revealed that Rio2 binds Slx9-1 ( Figure 6D , lane 6 ) . In comparison to the Slx9 , Slx9-1 bound at least as well to Rio2 ( Figure 6D , compare lane 2 and 6 ) . Further , the GST-Rio2:Slx9-1 complex was also able to recruit RanQLGTP ( Figure 6D , lane 7 ) , indicating that Slx9-1 is still able to assemble a GST-Rio2:Slx9-1:RanQLGTP complex . However , this complex was impaired in loading Crm1 ( Figure 6D , compare lane 5 and 9 ) . Therefore , we suggest that Slx9 promotes assembly of a Crm1-export complex on Rio2 , and that this function relies on a proper interaction between Slx9 and RanGTP . Crm1 recognizes and binds cargos that contain diverse leucine-rich NESs . Structural analyses of the RanGTP:Crm1 complex bound to prototypic NESs suggest that any peptide can function as a NES as long as its backbone conformation permits its side chains to access the rigid hydrophobic pockets of Crm1 ( Güttler et al . , 2010 ) . To test whether conformational rigidity of the Rio2-NES is critical to recruit Crm1 in the presence of RanQLGTP , three consecutive residues ( 399-EEN-401 ) proximal to the NES were mutated to glycines ( Rio23G ) ( Figure 7A , top panel ) . Because glycine residues lack a side chain , they allow greater conformational flexibility for the polypeptide backbone of these residues ( Ramachandran and Sasisekharan , 1968 ) as well as to the neighboring NES , thus destabilizing it . We found that , like the Rio2∆NES ( Figure 4A , lane 8 ) , Rio23G was unable to cooperatively recruit Crm1 in the presence of RanQLGTP in vitro ( Figure 7A , bottom panel , lane 4 ) . In parallel , we made a Rio2 mutant in which residues 399–401 were replaced by alanines ( Rio23A ) ( Figure 7—figure supplement 1 , top panel ) . Unlike Rio23G ( Figure 7A , bottom panel , lane 4 ) , Rio23A was able to efficiently cooperatively recruit Crm1 in the presence of RanQLGTP ( Figure 7—figure supplement 1 , bottom panel ) , suggesting that the glycine mutations destabilize the NES . 10 . 7554/eLife . 05745 . 009Figure 7 . Slx9 provides a scaffold to load Crm1 onto Rio2-NES . ( A ) Rio23G does not interact with Crm1 in the presence of RanGTP . Top: schematic depicts the positions of mutations proximal to the NES ( 399-EEN-401-GGG ) in the Rio23G . Hydrophobic amino acids of the NES are red and mutated amino acids are orange . Bottom: GST-Rio23G was immobilized on GSH-Sepharose and binding reactions were carried out and analyzed as in Figure 2C . L = input . ( B ) rio2∆NES , but not rio23G , is synthetically lethal with mex67∆loop and mtr2∆loop116-137 . Strains were spotted in 10-fold serial dilutions on 5-FOA ( SD ) plates and grown at 30°C for 2–4 days . ( C ) Slx9 restores Crm1 binding to the Rio23G:Slx9:RanQLGTP complex . GST-Rio2:Slx9:RanQLGTP or GST-Rio23G:Slx9:RanQLGTP was incubated with buffer alone or 50 nM Crm1-His6 . Bound proteins were analyzed as in Figure 2C . L = input . ( D ) Crm1 is impaired in binding a Rio23G:Slx9-1:RanQLGTP complex . Immobilized GST-Rio23G was incubated with buffer alone , 0 . 5 µM Slx9 or 0 . 5 µM Slx9-1 , followed by the stepwise addition of 0 . 2 µM His6-RanQLGTP and 50 nM Crm1-His6 as depicted in ( A ) . Analysis of the bound proteins was carried out as described in Figure 2C . L = input . DOI: http://dx . doi . org/10 . 7554/eLife . 05745 . 00910 . 7554/eLife . 05745 . 010Figure 7—figure supplement 1 . The ‘flexibility’ of the NES region in Rio2 contributes to its interaction with Crm1 in the presence of RanGTP . Top: schematic of Rio2 highlighting the triple A mutation ( 399-EEN-401-AAA , brown ) proximal to the NES . Hydrophobic amino acids of the NES are red and mutated amino acids are brown . Bottom: immobilized GST-Rio23A was incubated with buffer alone or buffer containing 2 µM His6-RanQLGTP , 50 nM Crm1-His6 or 2 µM His6-RanQLGTP , and 50 nM Crm1-His6 . After washing , eluted proteins were separated by SDS-PAGE and visualized by Coomassie staining or Western blotting using indicated antibodies . L = input . DOI: http://dx . doi . org/10 . 7554/eLife . 05745 . 01010 . 7554/eLife . 05745 . 011Figure 7—figure supplement 2 . Genetic interactions between Rio2 alleles and RanGTP-binding proteins Slx9 and Yrb2 . ( A ) slx9∆ and slx9-1 do not genetically interact with rio2∆NES . A RIO2 shuffle slx9∆ strain was transformed with the indicated combinations of empty , WT , or mutant plasmids and spotted in 10-fold dilutions on SD-plates containing 5-FOA and grown at 25°C for 2–4 days . ( B ) rio2∆NES weakly genetically interacts with yrb2∆ . RIO2 shuffle yrb2∆ strain transformed with the indicated combinations of empty , WT or mutant plasmids were spotted in 10-fold dilutions on SD-plates containing 5-FOA and grown at 25°C for 2–4 days . DOI: http://dx . doi . org/10 . 7554/eLife . 05745 . 011 We next assessed the functionality of Rio2∆NES and Rio23G in yeast . Both rio2∆NES and rio23G rescued the lethality of the rio2∆ strain ( Figure 7B ) . However , rio2∆NES was synthetically lethal mex67∆loop and mtr2∆116-137 ( Figure 7B ) , consistent with the model that the transport receptor Mex67-Mtr2 has a redundant function in 40S pre-ribosome nuclear export . Curiously , the rio23G allele that did not recruit Crm1 in the presence of RanQLGTP in vitro ( Figure 7A , bottom panel , lane 4 ) , rescued the synthetic lethality ( Figure 7B ) , indicating that Rio23G is still functional in vivo . Importantly , neither rio2 alleles ( rio2∆NES and rio23G ) were synthetic lethal when combined with slx9∆ and slx9-1 mutant strains ( Figure 7—figure supplement 2A ) . These genetic interactions led us to ask whether Slx9 could stabilize the NES conformer of Rio23G to facilitate Crm1 recruitment . To test this , a GST-Rio23G:Slx9:RanQLGTP complex was incubated with Crm1 . Remarkably , this complex was able to recruit Crm1 similar to the Rio2:Slx9:RanQLGTP complex ( Figure 7C , compare lane 2 and 4 ) . Notably , we found that the GST-Rio23G:Slx9-1:RanQLGTP complex was impaired in loading Crm1 ( Figure 7D , compare lane 1 and 2 ) . Altogether , these data suggest that , within the GST-Rio23G:Slx9:RanQLGTP complex , Slx9 promotes Crm1 loading by stabilizing the region surrounding the Rio2-NES . Our data so far show that a specific mutation within Slx9 impaired Crm1-export complex assembly in vitro and 40S pre-ribosome export in vivo . We , therefore , wondered whether replacing the Rio2-NES with a set of strong NESs could bypass the requirement of Slx9 in 40S pre-ribosome export . The strength of a specific NES is based on its resemblance to the consensus sequence and has been shown to strongly correlate with its affinity for Crm1 in vitro ( Engelsma et al . , 2004; Kutay and Güttinger , 2005 ) . To this end , we replaced the Rio2-NES with the strong NESs of Nmd3 ( hereafter termed as Rio2Nmd3NES ) ( Engelsma et al . , 2004; Kutay and Güttinger , 2005 ) . Functional analyses revealed that the expression of Rio2Nmd3NES complemented the lethality of the rio2∆ strain . Moreover , rio2-nmd3NES was not synthetic lethal with mex67∆loop and mtr2∆116-137 ( Figure 8A ) . Since Rio2Nmd3NES bound to Crm1 in the presence of RanGTP ( Figure 8B , lane 4 ) , we assessed whether Rio2Nmd3NES expression rescued the 40S pre-ribosome export defect seen in slx9∆ cells . >95% of slx9∆ cells expressing Rio2 and Rio2∆NES accumulated uS5-GFP in the nucleoplasm . However , slx9∆ cells expressing Rio2Nmd3NES did not accumulate uS5-GFP in the nucleoplasm ( Figure 8C ) , indicating no apparent impairment in 40S pre-ribosome export . A Rio2 variant containing only the first NES of Nmd3 ( rio2-nmd3NES∆1 ) was unable to rescue the 40S pre-ribosome export defect of slx9∆ cells ( Figure 8C ) , suggesting that both NESs are required to bypass Slx9 function in 40S pre-ribosome export . Notably , the expression of Rio2Nmd3NES in yrb2∆ cells did not rescue the nucleoplasmic accumulation of uS5-GFP ( Figure 8D ) , indicating that the heterologous NESs specifically bypasses Slx9 function but not other steps that drive 40S pre-ribosome export . 10 . 7554/eLife . 05745 . 012Figure 8 . Strong NESs of Nmd3 on Rio2 bypass requirement for Slx9 but not Yrb2 in 40S pre-ribosome export . ( A ) rio2-nmd3NES is not synthetic lethal with mex67∆loop or mtr2∆loop116-137 . Strains were spotted in 10-fold serial dilutions on 5-FOA ( SD ) plates and grown at 30°C for 2–4 days ( B ) The Nmd3-NES ( amino acids 440–518 ) fused to Rio2∆NES bypasses the requirement of the Rio2-NES in export complex formation in vitro . GST-Rio2Nmd3NES was immobilized on GSH-Sepharose and complex formation was carried out and analyzed as in Figure 2C . L = input . ( C ) rio2-nmd3NES rescues the impaired pre40S export of slx9∆ cells . Localization of uS5-GFP in the indicated strains was monitored by fluorescence microscopy . Scale bar = 5 µm . ( D ) The rescue of impaired pre40S ribosome export by rio2-nmd3NES is specific for slx9∆ . yrb2∆ cells transformed with the indicated plasmids was monitored by fluorescence microscopy for the localization of uS5-GFP . Scale bar = 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05745 . 012
Slx9 was first proposed to function in ribosome biogenesis pathway through promoting ITS1 processing , which is required to separate pre-rRNAs of the large and small pre-ribosomal subunits ( Bax et al . , 2006 ) . In support to this early assembly role , slx9∆ interacts genetically with mutations in RRP5 , which encodes an assembly factor required for pre-rRNA processing . However , low-copy overexpression of the transport receptor MEX67-MTR2 , which is involved in 40S pre-ribosome export , suppresses the growth defect and pre-40S export defect of slx9∆ cells . Further , SLX9 genetically interacts with several factors directly involved in 40S pre-ribosome export ( Faza et al . , 2012 ) . These functional studies indicate a role for Slx9 in the nuclear export of the 40S pre-ribosomal cargo ( Faza et al . , 2012 ) . Thus , the pre-rRNA processing defects observed in the slx9∆ mutant could be a consequence of a primary defect in the nuclear export of 40S pre-ribosomes . Surprisingly , although Slx9 did not interact with Crm1 in vitro , it interacts specifically with the GTP-bound form of Ran . This interaction depends on the basic patch of RanGTP . Removal of the RanGTP acidic C-terminal tail strengthened Slx9:RanGTP interactions , suggesting that these proteins interact in a manner distinct from previously identified RanGTP-binding proteins , such as Yrb1 ( Maurer et al . , 2001; Nilsson et al . , 2001 ) . Sequence analyses did not reveal any apparent homology between Slx9 and known Ran-binding proteins . Based on all these data , we propose that Slx9 is a new type of RanGTP-binding protein . To assemble a Crm1-export complex , Crm1 requires binding to both RanGTP and a NES-containing cargo in a cooperative manner ( Dong et al . , 2009; Güttler and Görlich , 2011; Monecke et al . , 2013 ) . This canonical assembly pathway requires Crm1 to switch from a relaxed , low affinity ‘cytoplasmic’ conformation to a strained , high-affinity ‘nuclear’ conformation . In the nuclear conformation , RanGTP is enclosed within the toroid-like fold of Crm1 , away from the cargo-binding site . Structural analyses of Crm1-complexes suggest that RanGTP promotes NES-binding solely by stabilizing the strained nuclear conformation of Crm1 ( Güttler and Görlich , 2011; Monecke et al . , 2013 ) . In contrast , Slx9 mediates a non-canonical stepwise assembly of a Crm1-export complex . First , Slx9 binds RanGTP and the NES-containing 40S pre-ribosome adaptor Rio2 . Interactions between Slx9 , Rio2 , and RanGTP do not rely on cooperative binding , since the pairwise complexes between these proteins were stable . However , all three proteins were required for efficient Crm1 loading to the Rio2-NES . This recruitment also does not rely on cooperative binding between Rio2 , Crm1 , and RanGTP , since the Slx9:Rio2:RanGTP complex was stable in the absence of Crm1 . Mutational and functional studies identified a point mutation , L108P , which impaired the ability of Slx9 to bind RanGTP in vitro . Although this mutant protein was stably recruited to pre-ribosomal particles in vivo and to Rio2 in vitro , it was unable to assemble a Crm1-export complex . Both slx9∆ and slx9-1 cells displayed impaired 40S pre-ribosome export and synthetic phenotypes with factors involved in 40S pre-ribosome export . These data suggest that the 40S pre-ribosome nuclear export defect observed in slx9∆ and slx9-1 cells is due to their failure to efficiently assemble Crm1-export complexes on 40S pre-ribosomal particles . A stepwise formation of a Crm1-export complex could allow currently uncharacterized quality control surveillance steps to monitor pre-ribosome assembly and ensure that only correctly assembled particles are chosen for export . In this respect , Slx9-dependent ribosome export would mirror the assembly process of the 40S pre-ribosomal cargo itself . Crm1-mediated nuclear export is modulated by RanGTP-binding proteins , which promote specific steps along the export pathway . For example , RanBP3/Yrb2 increases the rate of cooperative Crm1-complex assembly in the nucleus ( Koyama et al . , 2014 ) . Yrb2/RanBP3 also modulates Crm1 substrate recognition , promoting interactions with certain NESs and preventing the strong Crm1-Snurportin1 interaction ( Englmeier et al . , 2001; Langer et al . , 2011 ) . At the other end of the export cycle , RanBP1/Yrb1 and RanBP2 interact strongly with RanGTP in the cytoplasm to stimulate Crm1-export complex disassembly , thereby indirectly contributing to export efficiency ( Bischoff and Görlich , 1997; Kehlenbach et al . , 1999; Maurer et al . , 2001 ) . Notably , despite the fact that these RanGTP-binding proteins act at different stages to stimulate export , they typically influence cooperative interaction between cargo , Crm1 , and RanGTP . Slx9 utilized a distinct mechanism from that proposed for Yrb2/RanBP3 for Crm1-export complex assembly ( Englmeier et al . , 2001; Lindsay et al . , 2001; Koyama et al . , 2014 ) . Yrb2/RanBP3 increases the affinity of Crm1 for RanGTP by stabilizing a conformation of Crm1 that promotes canonical assembly of a Crm1-export complex ( Langer et al . , 2011; Koyama et al . , 2014 ) . In contrast , Slx9 does not interact with RanGTP and Crm1 to form a complex . Instead , it interacts with RanGTP and the NES-containing adaptor Rio2 . Finally , biochemical and structural studies show that Yrb2 competes with NES-containing cargo for binding to Crm1 and suggest that NES-binding causes partial dissociation of Yrb2 from the complex ( Koyama et al . , 2014 ) . In contrast , Slx9 remains stably bound to the GST-Rio2:Crm1:RanGTP complex , suggesting that it does not compete with the Rio2-NES for its interaction . Despite these clear differences , both Slx9 and RanBP3/Yrb2 work to overcome the low binding affinity between NESs and Crm1 , representing distinct solutions to the paradox of low affinity interactions driving fast and efficient Crm1-mediated cargo export . Notably , SLX9 strongly genetically interacted with YRB2 , suggesting that both mechanisms are employed to ensure rapid 40S pre-ribosome export . Does Slx9 target other NES-containing cargos ? Although slx9∆ showed a strong 40S pre-ribosome export defect , rio2∆NES grew indistinguishably from WT cells . Notably , rio2∆NESyrb2∆ cells grew slowly ( Figure 7—figure supplement 2B ) but are not synthetically lethal as the slx9∆yrb2∆ strain . These genetic interactions argue that Rio2 is not the sole target of Slx9 . Genetic approaches will uncover additional NES-containing adaptors that employ Slx9 to prepare the 40S pre-ribosome for nuclear export . NESs contain variability in the spacing between key hydrophobic residues , yet are recognized by the same rigid hydrophobic pockets on Crm1 ( Güttler et al . , 2010 ) . The structure of RanGTP-Crm1 bound to prototypic NESs showed that the backbone of NESs adopts different conformations , which permits the efficient insertion of key NES residues ( Güttler et al . , 2010 ) . In support of the proposed NES-conformer selection model , a destabilized NES of Rio23G was unable to bind Crm1 in the presence of RanGTP in vitro . However , Crm1 recruitment to Rio23G was restored in the presence of Slx9 , suggesting a stabilizing function . This stabilization provides a mechanism to overcome the weak binding of NESs to Crm1 , driving Crm1-complex assembly and guaranteeing efficient cargo export into the cytoplasm . One prediction of the NES-conformer selection model would be that NESs with favorable conformations exhibit improved affinity to Crm1 . Slx9 may belong to a family of yet-unidentified RanGTP-binding proteins that induce and stabilize NES-conformers upon binding export cargos . Thus , these RanGTP-binding proteins would allow greater NES variability and potentially regulate the efficiency by which specific NESs are recognized by Crm1 .
Yeast strains used in this study are listed in Supplementary file 1 . Genomic disruptions , insertion of C-terminal tags , and promotor switches at genomic loci were performed as previously described ( Longtine et al . , 1998; Puig et al . , 2001; Janke et al . , 2004 ) . Preparation of media , yeast transformations , and genetic manipulations were performed accordingly to established procedures . Genetic analyses were performed as previously described ( Faza et al . , 2012 ) . Plasmids used in this study are listed in Supplementary file 2 . All recombinant DNA techniques were performed accordingly to established procedures using Escherichia coli XL1 blue cells for cloning and plasmid propagation . rio2∆NES was created by deletion of the DNA sequence encoding the last 12 amino acids of Rio2 . ltv1∆NES was created by removing the last 13 amino acids . The rio2-nmd3NES was created by fusion of the C-terminal region of NMD3 encoding amino acids 440–518 to rio2∆NES . rio2-nmd3NES∆1 was created by fusing NMD3 lacking amino acids 440–487 . The C-terminal DNA-fragment of SSB1 ( SSB1C ) encodes the amino acids 524–613 ( Maurer et al . , 2001 ) . Point mutations in SLX9 , RIO2 , or GSP1 were generated using the QuikChange site-directed mutagenesis kit ( Agilent Technologies , Switzerland ) . All cloned DNA fragments generated by PCR amplification and mutagenized plasmids were verified by sequencing . Tandem affinity purifications ( TAPs ) of pre-ribosomal particles were carried out as previously described ( Faza et al . , 2012; Altvater et al . , 2014 ) . Calmodulin eluates were separated on NuPAGE 4–12% Bis-Tris gradient gels ( Invitrogen , Zug , Switzerland ) . Separated proteins were visualized by either silver staining or Western analysis using indicated antibodies . Whole cell lysates were prepared by a modified post-alkaline extraction protocol as previously described ( Kemmler et al . , 2009 ) . Extracted proteins were separated by SDS-PAGE . Proteins were visualized by Western blotting using the indicated antibodies . Western analyses were performed as previously described ( Kemmler et al . , 2009 ) . The following primary antibodies were used: α-Slx9 ( 1:3000; Faza et al . , 2012 ) , α-Arc1 ( 1:4000; E Hurt , University of Heidelberg , Heidelberg , Germany ) , α-Xpo1 ( Crm1 ) ( 1:3000; this study ) , α-His ( 1:2000; Sigma–Aldrich , USA ) , α-uS7 ( yeast Rps5; 1:4000; Proteintech Group Inc . , Chicago , IL , USA ) , α-TAP ( CBP ) ( 1:4000; Thermo Scientific , Rockford , IL , USA ) , α-Ltv1 ( 1:5000; K Karbstein , Scripps Research Institute , Jupiter , FL , USA ) , α-Rio2 ( 1:1000; Proteintech Group Inc . ) , α-Tsr1 ( 1:5000; K Karbstein , Scripps Research Institute ) , and α-Gsp1 ( Ran ) ( 1:3000; this study ) . For detection , HRP-conjugated α-rabbit ( 1:2000–1:4000; Sigma–Aldrich ) or α-mouse secondary antibodies ( 1:2000–1:4000; Sigma–Aldrich ) were applied . Signals were visualized using the Immun-Star HRP chemiluminescence kit ( Bio-Rad Laboratories , Hercules , CA , USA ) and captured by Fuji Super RX X-ray films ( Fujifilm , Japan ) . Recombinant Slx9 and Slx9-1 , Rio2 , Rio2 variants , and Xpo1 ( yeast Crm1 ) were expressed in E . coli BL21 upon IPTG induction ( final concentration 0 . 3 mM ) . His6-tagged proteins were affinity purified in purification buffer ( 50 mM Tris-HCl , 200 mM NaCl , 10 mM Imidazole , 1 mM β-mercaptoethanol , pH 8 ) , using Ni-NTA agarose ( GE Healthcare , Uppsala , Sweden ) . The GB1-His-domains of these proteins were removed by TEV cleavage . All proteins were stored in PBS-KMT ( 150 mM NaCl , 25 mM sodium phosphate , 3 mM KCl , 1 mM MgCl2 , 0 . 1% Tween , pH 7 . 3 ) after buffer exchange . GST-fusion proteins were purified in PBSKMT using GSH-Sepharose ( GE Healthcare ) . His6-Gsp1 ( Ran ) WT and mutants were expressed and purified as previously described ( Solsbacher et al . , 1998; Maurer et al . , 2001 ) . For in vitro binding studies , recombinant GST-tagged proteins were immobilized on GSH-Sepharose ( GE Healthcare ) in PBSKMT and incubated with the indicated proteins for 1 hr at 4°C . Binding studies between GST-tagged Slx9 , Yrb1 , or Kap123 and Ran variants were performed as described before ( Solsbacher et al . , 1998 ) . To prevent nonspecific interactions , all binding assays were carried out in the presence of competing E . coli lysates . Trimeric export complex formation between GST-tagged Slx9 , Rio2 , or Ltv1 variants and His6-RanQLGTP and/or Crm1-His6 was adapted and modified from Rothenbusch et al . ( 2012 ) and Solsbacher et al . ( 1998 ) . Immobilized GST-fusion proteins were incubated with buffer alone or buffer containing 2 µM His6-RanQLGTP or 50 nM Crm1-His6 or 2 µM His6-RanQLGTP and 50 nM Crm1-His6 . The interactions between GST-fusions of Slx9 , Ntf2 , and Yrb1 and RanGTP or RanGDP were analyzed as follows: first , His6-Ran was incubated with GDP or GTP ( 50× molar excess of the protein concentration ) in the presence of 6 mM EDTA in KPi buffer ( 24 . 8 mM KH2PO4 , 25 mM , K2HPO4 , 20 mM KCl , 5 mM MgCl2 , 2 mM Imidazole , 5 mM β-mercaptoethanol , pH 6 . 8 ) on ice for 40 min . This incubation step was terminated by the addition of 0 . 3 mM MgCl2 . In the second step , 2 µM Ran ( GTP or GTP ) was incubated with immobilized GST-fusion proteins for 1 hr at 4°C . To form a Rio2:Slx9:Ran:Crm1 complex , GST-Rio2 variants were immobilized on GSH-Sepharose and incubated with buffer alone or 2 µM purified Slx9 variants for 1 hr at 4°C . Then , samples were incubated with 0 . 2 µM purified His6-RanQLGTP for 1 hr at 4°C . In the last step , samples were incubated with buffer alone or 50 nM purified Crm1-His6 . After each incubation step , unbound proteins were removed by three times washing with PBSKMT . All bound proteins were eluted in LDS-sample buffer ( Invitrogen ) and separated by SDS-PAGE . Separated proteins were visualized by Coomassie staining or by Western analysis using antibodies against Slx9 , Crm1 , or Ran . 1/3 of bound proteins and 1/3–3× of input were analyzed on a Coomassie gel . 1/6 of bound proteins and 1/24 ( Slx9 ) or 1/50 ( Ran , Crm1 , Nmd3 , and Rio2 ) of the input was used for Western analysis . Cells were visualized using a DM6000B microscope ( Leica , Germany ) equipped with a HCX PL Fluotar 63×/1 . 25 NA oil immersion objective ( Leica ) . Images were acquired with a fitted digital camera ( ORCA-ER; Hamamatsu Photonics , Japan ) and Openlab software ( Perkin–Elmer , USA ) . Localization of pre-40S subunits was monitored employing the uS5-GFP reporter construct as previously described ( Faza et al . , 2012; Altvater et al . , 2014 ) . Co-localization of Slx9-GFP and Slx9-1-GFP with Gar1-mCherry was done as previously described ( Faza et al . , 2012 ) . The heterokaryon assay was adapted and modified from ( Belaya et al . , 2006; Altvater et al . , 2012 ) . Briefly , equal amounts of cells expressing Enp1-GFP , Gar1-GFP , or Slx9-GFP were mated with kar1-1 cells expressing Nup82-mCherry and concentrated onto 0 . 45-µM nitrocellulose filter . Mixtures were placed on YPD plates containing 50 µM cycloheximide . After 1 hr incubation at 30°C , cells were analyzed by fluorescence microscopy . Localization of 20S pre-rRNA in the different strains was analyzed using a Cy3-labeled oligonucleotide probe ( 5′-Cy3-ATG CTC TTG CCA AAA CAA AAA AAT CCA TTT TCA AAA TTA TTA AAT TTC TT-3′ ) that is complementary to the 5′ portion of ITS1 as described ( Faza et al . , 2012; Altvater et al . , 2014 ) . | Plants , fungi , and animals store their genetic material within the nucleus of each of their cells . This structure is surrounded by a double layer of membrane that prevents the contents of the nucleus from mixing with the contents of the rest of the cell ( namely the cytoplasm ) . Exchange of material between the nucleus and cytoplasm occurs through pores embedded within the nuclear membrane . To travel through one of these pores , large molecules ( also called cargos ) require the assistance of so-called ‘transport receptors’ such as the Crm1 protein . This protein recognizes and binds to the part of a cargo molecule called a ‘nuclear export signal’ , and the Crm1 protein also binds to another protein called RanGTP . Nuclear export signals bind weakly to Crm1 , which in turn ensures that these cargos are easily released in the cytoplasm once transport is completed . However , this weak binding means that it has remained a mystery how Crm1 is able to efficiently transport cargos out of the nucleus to begin with . Now , Fischer et al . have analyzed how one cargo that contains a nuclear export signal , namely molecules called ribosome precursors , assembles with Crm1 . The experiments identified another protein called Slx9 that shuttles rapidly between the inside and the outside of the nucleus . Fischer et al . observed that Slx9 binds directly to RanGTP and brings it together with the ribosome precursor cargo . When a complex of Slx9 , RanGTP , and cargo is assembled , it further recruits Crm1 to the cargo . Thus , Slx9 acts as a scaffold to bring cargo into contact with Crm1 and RanGTP , and a tight complex is formed that enables the export of the cargo out of the nucleus . Yeast cells lacking Slx9 delay export of ribosome precursors out of the nucleus . These findings imply the existence of yet-unidentified proteins like Slx9 that help Crm1 to rapidly transport diverse cargos out from the nucleus and into the cytoplasm . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2015 | A non-canonical mechanism for Crm1-export cargo complex assembly |
Herbivore-induced defenses are widespread , rapidly evolving and relevant for plant fitness . Such induced defenses are often mediated by early defense signaling ( EDS ) rapidly activated by the perception of herbivore associated elicitors ( HAE ) that includes transient accumulations of jasmonic acid ( JA ) . Analyzing 60 HAE-induced leaf transcriptomes from closely-related Nicotiana species revealed a key gene co-expression network ( M4 module ) which is co-activated with the HAE-induced JA accumulations but is elicited independently of JA , as revealed in plants silenced in JA signaling . Functional annotations of the M4 module were consistent with roles in EDS and a newly identified hub gene of the M4 module ( NaLRRK1 ) mediates a negative feedback loop with JA signaling . Phylogenomic analysis revealed preferential gene retention after genome-wide duplications shaped the evolution of HAE-induced EDS in Nicotiana . These results highlight the importance of genome-wide duplications in the evolution of adaptive traits in plants .
Induced defense is widespread in plants and can improve the fitness of plants under herbivore attack ( Baldwin , 1998; Kessler et al . , 2004 ) . Many plants recognize and distinguish the damage caused by feeding insects from mechanical damage by perceiving herbivore-associated elicitors ( HAE ) to induce rapid early defense signaling ( EDS ) that includes the accumulation of jasmonic acid ( JA ) and its derivatives , phytohormones that play a central role in the activation of induced defenses ( Erb et al . , 2012; Howe and Jander , 2008; Wu and Baldwin , 2010 ) . Increases or decreases in leaf JA concentrations can directly activate or impair induced anti-herbivore defenses , respectively ( Farmer and Ryan , 1992; Kessler et al . , 2004; Wu and Baldwin , 2010 ) , highlighting the importance of JA accumulation for induced defenses . However , increased JA levels can also reduce plant fitness due to the physiological and ecological costs of defense elicitation when defenses are not needed ( Baldwin and Hamilton , 2000; Glawe et al . , 2003; Heil and Baldwin , 2002; van Dam and Baldwin , 1998 ) . For example , in Nicotiana attenuata , an increase in endogenous JA levels by supplying methyl-jasmonic acid ( MeJA ) reduced plant fitness by 26% when plants were protected from herbivore attack ( Baldwin , 1998 ) . Thus induced JA accumulations can result in net fitness gains or losses depending on the cost/benefit ratio of induced defenses , which varies among attacking herbivore species and environmental conditions . Therefore , a robust and complex signaling network that regulates and fine-tunes induced JA biosynthesis , metabolism and JA-dependent induced downstream defenses is essential for plants to realize their fitness optima . Using reverse genetics , such as RNA-inference ( RNAi ) and virus induced gene silencing ( VIGS ) , several genes that are rapidly induced by HAE were found to regulate JA biosynthesis and metabolism in plants , particularly in the wild tobacco Nicotiana attenuata which has been established as an ecological model system for plant-herbivore interactions ( Wu and Baldwin , 2010 ) . The HAE-regulated signaling network includes: protein kinases , such as the wounding induced protein kinase ( NaWIPK ) ( Wu and Baldwin , 2010; Wu et al . , 2007 ) and calcium-dependent protein kinases ( NaCDPK4/5 ) ( Yang et al . , 2012 ) , which positively and negatively regulate HAE-induced JA accumulations , respectively; transcription factors , such as NaWRKY3/6 , which positively regulate HAE-induced JA accumulations ( Skibbe et al . , 2008 ) ; and ethylene ( ET ) biosynthesis and perception genes ( NaETR1 , NaACO and NaACS ) ( von Dahl et al . , 2007 ) , which crosstalk with JA-regulated downstream defense responses ( Onkokesung et al . , 2010; Voelckel et al . , 2001 ) , such as nicotine biosynthesis ( Kahl et al . , 2000; Shoji et al . , 2000 ) . While these studies have provided mechanistic insights into induced defenses , they also revealed the complexity of the HAE-induced EDS network . A systematic investigation of its complete genetic architecture is essential to understanding the molecular mechanisms and evolution of HAE-induced EDS . Gene duplications play a key role in network evolution ( Pastor-Satorras et al . , 2003; Teichmann and Babu , 2004 ) . Duplicated genes can either be retained in the same network to increase network complexity and robustness or evolve to function in new networks through subfunctionalization and/or neofunctionalization processes ( De Smet and Van de Peer , 2012; Duarte et al . , 2006 ) that can be detected from changes in the spatiotemporal expression or protein interaction patterns of the duplicated genes . Although both gene expression and protein-protein interaction divergences between duplicated genes increase over time ( Arabidopsis Interactome Mapping , 2011 ) , several factors , such as the type of duplication and the functionality of the genes , affect the rate and extent of those divergences ( Hanada et al . , 2008; Rizzon et al . , 2006 ) . For instance , expression divergences between duplicated genes involved in stress responses tend to be greater than those of duplicated genes involved in developmental processes ( Ha et al . , 2007 ) . While studies based on the analysis of gene ontologies and genome-wide duplications suggest that linage-specific duplication ( LD ) followed by expression divergence are important for the evolution of stress responses in plants ( Hanada et al . , 2008; Rizzon et al . , 2006 ) , whole genome duplications ( WGD ) events , which are prominent in the plant kingdom , provide a major source of duplicated genes and contribute significantly to the evolution of cellular networks , such as gene regulatory ( Blanc and Wolfe , 2004 ) , protein-protein interaction ( Arabidopsis Interactome Mapping , 2011 ) and metabolic networks ( Gachon et al . , 2005; Hofberger et al . , 2013 ) . Furthermore , duplicated copies from WGD events are more likely to be retained in a network than those from LD , especially for genes that are dosage-sensitive , such as transcription factors , and protein kinases ( Arabidopsis Interactome Mapping , 2011; Birchler and Veitia , 2007; Casneuf et al . , 2006; Edger and Pires , 2009; Freeling , 2009 ) . However , the relative contribution of WGD and LD to the evolution of HAE-induced EDS networks and the patterns of expression divergence between duplicated genes in these networks have not been studied . Understanding the molecular mechanisms and evolution of HAE-induced EDS requires the identification of the genome-wide HAE-induced EDS networks . Because functionally related genes tend to be transcriptionally coordinated ( Persson et al . , 2005; Stuart et al . , 2003 ) , co-expression network analysis has been widely used to infer the function of genes and uncover biological pathways ( Klie et al . , 2014; Usadel et al . , 2009; Yonekura-Sakakibara et al . , 2008 ) . Distinct from ‘classical’ gene expression analysis using genome-wide expression profiling of control and treated samples to identify ‘up’ or ‘down’ regulated genes , co-expression network analysis uses expression measurements from a large number of samples that vary in their genotype , treatment , tissue or sampling time to enhance the statistical power of the analysis ( Zhang and Horvath , 2005 ) . However , due to the high specificity among tissues and treatments and the speed of the HAE-elicited responses ( within 30 min ) ( Gulati et al . , 2014 , 2013; Kim et al . , 2011 ) , the general co-expression network approach that uses gene expression data from different tissues or time course experiments are not particularly useful . One solution to overcome this specificity issue is to use natural variation , such as occurs amongst closely related species or different genotypes within species , to identify co-expressed gene networks ( Ardlie et al . , 2015; Delker et al . , 2010 ) . In the identification of HAE-induced EDS networks , the comparison of closely related species has at least two advantages over the use of different genotypes within a species: ( 1 ) their greater genetic and phenotypic diversity ( Xu et al . , 2015 ) which increases the power of detecting co-expressed genes; ( 2 ) their divergence times are over several millions of years which allows for the identification of evolutionarily conserved co-expression networks that are likely functionally important . Closely-related Nicotiana species within the clade of Petunioides show highly specific HAE-induced defenses and thus provide an ideal system for identifying HAE-induced EDS networks ( Xu et al . , 2015 ) . Our previous study revealed that a single HAE , such as the fatty acid-amino acid conjugate C18:3-Glu ( FAC ) – the most active elicitor found in the oral secretions of the Solanaceae specialist herbivore Manduca sexta ( OSMs ) larvae – elicits diverse defense responses among closely related Nicotiana species when added to standardized puncture wounds . In addition , a single Nicotiana species , such as N . pauciflora , showed distinct defense responses to the FAC , OSMs and oral secretions from the generalist herbivore Spodoptera littoralis ( OSSl ) ( Xu et al . , 2015 ) . Here we sequenced the leaf transcriptomes of six closely-related Nicotiana species from the Petunioides clade ( N . obtusifolia , N . linearis , N . acuminata , N . pauciflora , N . miersii and N . attenuata ) that had been induced by three different HAEs or simply wounded ( induced by wounding plus water ) to characterize the HAE-induced EDS networks in Nicotiana . We compared HAE-induced transcriptomic responses among the six species and identified a co-expression gene network that represents the HAE-induced EDS in Nicotiana based on three independent lines of evidence: ( 1 ) the induction of the network correlates with variation in JA accumulations both among species treated with the same HAE and within species treated with different HAEs; ( 2 ) the induction of genes in this network that are largely not dependent on induced JA accumulations; ( 3 ) the consequences of silencing a hub gene in this network for HAE-induced JA metabolism and defenses . Analysis of the evolutionary history of all genes in the EDS network revealed that preferential gene retention after the Solanaceae whole genome triplication ( WGT ) event shaped the evolution of HAE-induced EDS in Nicotiana .
Closely related Nicotiana species showed highly divergent early transcriptomic responses within 30 min of FAC elicitation ( Figure 1 ) , consistent with observations from metabolomic and insect performance studies ( Xu et al . , 2015 ) . Two species , N . obtusifolia and N . miersii , which did not respond to FAC-treatments by amplifying their wound-induced accumulations of jasmonic acid ( JA ) within 2 hr , showed overall little induced transcriptomic responses ( Figure 1 and Figure 1—figure supplement 1 ) . In the other four species , FAC elicitation induced both high levels of JA and significant transcriptomic changes . The variation in FAC-induced transcriptomic responses largely resulted from the up-regulation of genes . Among all the species , FAC elicitation up-regulated more genes ( 1149 . 3 ± 667 . 8; mean ± standard deviation ) than it down-regulated ( 353 . 2 ± 278 . 9 ) . To validate the expression changes observed from the RNA-seq analysis , we quantified the transcript abundance of 12 N . attenuata genes that were found to be up-regulated by FAC using qPCR . Although two genes only showed marginally significant increases in their transcript levels ( p=0 . 08 , likely due to their large variation in expression among replicates ) , all 12 genes showed consistent FAC-induced up-regulation ( Figure 1—figure supplement 2 ) , suggesting an overall high reliability of the RNA-seq results . Interestingly , despite the large variation in FAC-induced transcriptomic responses among the species , 34 genes were induced by FAC in all six species ( Supplementary file 1A ) which likely represent the conserved FAC-induced stress response genes . Among them , eight are transcription factors from the WRKY ( 3 ) , AP2/ERF ( 2 ) TT2 ( 1 ) and PLATZ ( 1 ) families , and two are E3 ubiquitin-protein ligase-like proteins ( Supplementary file 1A ) . Taken together , the data suggest that while the induction of only a few genes are conserved after FAC elicitation , the overall FAC-induced early transcriptomic responses are highly variable among closely related Nicotiana species . 10 . 7554/eLife . 19531 . 003Figure 1 . FAC elicits divergent transcriptome responses among closely related Nicotiana species . ( a ) FAC-induced JA responses among six Nicotiana species . Phylogenetic tree was constructed based on orthologous genes and numbers on each branch indicates bootstrap values . X-axis indicates time after elicitation and Y-axis denotes JA concentrations . FM= fresh mass . Gray and black colored lines refer to control ( wounding and water ) and FAC-induced samples , respectively . Different letters indicate significance between two treatments ( Student’s-t test , p<0 . 05 ) . ( b ) transcriptomic similarity between control and FAC-induced samples ( 30 min after elicitation ) in the six species ( order is same as panel a ) . The color gradients indicate the Pearson correlation coefficients among samples . ( c ) number of differentially up- and down-regulated genes after FAC elicitation in the six species ( order is same as panel a ) . Y-axis depicts the number of genes . Each colored bar indicates a different species . Blue: N . obtusifolia , light green: N . linearis , dark green: N . attenuata , light blue: N . miersii , orange: N . acuminata , pink: N . pauciflora . d and e , Venn diagrams of up- ( d ) and down- ( e ) regulated genes in each of the six Nicotiana species . Circle size indicates the relative number of up/down regulated genes in each species . Each filled circle indicates a different species , with color code as in panel c . DOI: http://dx . doi . org/10 . 7554/eLife . 19531 . 00310 . 7554/eLife . 19531 . 004Figure 1—figure supplement 1 . Z-score of FAC-induced gene expression changes in six species . Each heatmap depicts z-scores of the up and down-regulated genes after FAC elicitation in the six species . Z-scores were calculated based on log2 transformed FPKM values . The total number of genes that were up- or down-regulated by FAC in each species is shown at the bottom . DOI: http://dx . doi . org/10 . 7554/eLife . 19531 . 00410 . 7554/eLife . 19531 . 005Figure 1—figure supplement 2 . Validations of 12 selected FAC-induced genes in N . attenuata . The N . attenuata leaf samples were harvested at 30 min after the treatments of wounding + water ( WW ) or wounding + FAC ( FAC ) for the quantification of transcript abundance of the selected 12 genes . P-values were calculated using the Students’ t-test . Data are presented as means ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 19531 . 005 The highly divergent FAC-induced transcriptomic responses provide an excellent opportunity to identify co-expression networks . We identified FAC-induced gene co-expression networks using the weighted gene co-expression network analysis ( WGCNA ) method ( see details in Materials and Methods ) . In total , five gene modules ( M1-M5 ) were identified using control ( wounding + water ) and FAC-induced gene expression profiles from all six species ( Figure 2a ) . Among these five modules , module M4 showed the highest correlation with HAE-induced JA accumulations , a marker of induced defense signal ( Figure 2b ) . In all four species that showed FAC-induced JA accumulations ( Figure 2c ) , the majority of M4 module genes were also significantly induced by FAC ( p<0 . 05 and fold change greater than 1 . 5 , exact negative binomial test ) . In contrast , in the two species , N . obtusifolia and N . miersii , which did not show FAC-induced JA accumulations ( Figure 2c ) , less than 22% of the M4 module genes were induced . The intra-modular connectivity of the M4 module , a parameter that indicates the degree of co-expression among 3genes in a network , was significantly higher in FAC-induced samples than in control samples ( p=0 . 0002 , Kruskal-Wallis rank sum test ) , consistent with the observation that FAC elicits co-expression among genes in the M4 module ( Figure 2d ) . Furthermore , the expression kinetic analysis of the M4 module genes using a previously published microarray dataset ( Kim et al . , 2011 ) revealed that most of the M4 module genes were largely transiently expressed ( Figure 2—figure supplement 1 ) after HAE-elicitation . These data suggest that the identified M4 module is likely associated with FAC-induced EDS . 10 . 7554/eLife . 19531 . 006Figure 2 . The M4 co-expression module is correlated with induced defense and is induced by FAC elicitation . ( a ) the cluster dendrogram of five modules ( M1-M5 ) . Each color indicates a different co-expression module . Y-axis indicates the height of the clustering tree . Yellow: M1 , green: M2 , brown: M3 , turquoise: M4 , blue: M5 . ( b ) the average correlation coefficient between each module and the maximum induced JA level within 2 hr . Y-axis indicates the value of average correlation coefficients . Each color represents one identified module . Mean and standard error are shown for each bar . The M4 module has a significantly higher correlation coefficient than the other modules ( *** , p<0 . 001 , Wilcoxon–Mann–Whitney test ) . ( c ) percentage of genes in module M4 induced by FAC-elicitation across six species . ( d ) the intra-modular connectivity of the M4 module in control and FAC-induced samples . WW indicates control samples ( wounding and water ) and FAC indicates the FAC-elicited samples . DOI: http://dx . doi . org/10 . 7554/eLife . 19531 . 00610 . 7554/eLife . 19531 . 007Figure 2—figure supplement 1 . Expression kinetics of the M4 module genes based on previous microarray data . Each line indicates the expression levels of a gene and the blue line indicates the average values of all genes . DOI: http://dx . doi . org/10 . 7554/eLife . 19531 . 007 More than 53% of the M4 module genes were induced in N . pauciflora at 30 min after FAC-elicitation ( Figure 2b ) , the time point when JA was not yet induced in this species , indicating that the induction of gene module M4 is independent of or precedes that of JA . To further test this hypothesis , using the N . attenuata genome-wide microarray , we measured FAC-induced gene expression changes in JA deficient N . attenuata plants ( irAOC ) , in which a key JA biosynthesis gene was silenced and the induced JA levels were reduced to basal levels ( Kallenbach et al . , 2012 ) . For the comparison , we also performed genome-wide microarray analysis for the same N . attenuata wild type ( WT ) RNA samples that were used for the RNA-seq analysis . In the WT samples , fewer FAC-induced genes were detected by the microarray ( 771 ) than by the RNA-seq analysis ( 1752 ) ; however more than 81 . 2% of the FAC up-regulated genes identified using the microarray were also found from the RNA-seq analysis , indicating an overall consistency between RNA-seq and microarray data , and the expected higher power and sensitivity of RNA-seq in detecting differentially expressed genes . More than 87% of the M4 module genes that were induced by FAC in WT plants , both from the RNA-seq and microarray experiments , were also up-regulated in the JA-deficient irAOC plants ( Figure 3 ) . Likewise , based on the microarray data of samples that were collected at 30 min after FAC-elicitation , the majority ( 85 . 1% ) of up-regulated genes in WT plants were also up-regulated ( FDR adjusted p<0 . 05 , fold change >1 . 5 ) in the JA-deficient plants , suggesting that the FAC-induced early expression changes are largely not dependent on FAC-induced JA accumulations . Together , these data suggest that the genes of the M4 module are largely induced by FAC but not by JA , which places their regulation down-stream of HAE perception but upstream or parallel of the activation of JA signaling . 10 . 7554/eLife . 19531 . 008Figure 3 . The majority of M4 module genes were induced by FAC in both WT and JA deficient plants . ( a ) a heatmap representing the expression of M4 module genes in WT and JA deficient plants ( irAOC ) . The color gradient represents the relative expression level . ( b ) a simplified induced defense signaling pathway , which indicates that accumulation of JA is not required for the induction of the majority of the M4 module ( orange color in the pie chart ) . EDS: early defense signaling . Dashed and solid arrows indicate known and putative regulations , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 19531 . 008 A previous study revealed that different HAE can induce distinct defense responses within the same species ( Xu et al . , 2015 ) . To understand the underlying molecular mechanisms , we additionally sequenced the transcriptomes of leaves that were induced by the oral secretions ( OS ) of the Solanaceae specialist herbivore M . sexta ( OSMs ) and the generalist herbivore S . littoralis ( OSSl ) in four different Nicotiana species that showed specific responses to different HAE . This analysis revealed that the level of M4 module gene inductions correlate with the specificity of HAE-induced defense responses within a species . In N . attenuata , FAC , OSMs and OSSl induced similar levels of induced defense responses , and consistently , the majority of the M4 module genes were induced by all three elicitors ( Figure 4 a–d ) . In N . pauciflora , while both FAC and OSMs up-regulated a large fraction of the M4 module genes ( Figure 4b ) and downstream induced defense responses , OSSl only up-regulated 14 . 8% of the M4 module genes and failed to activate the downstream defense responses ( Xu et al . , 2015 ) . Furthermore , while in N . obtusifolia and N . miersii , both FAC and OSMs only up-regulated less than 13 . 9% of the M4 module genes ( Figure 4b ) and did not activate the downstream defenses ( Xu et al . , 2015 ) , OSSl up-regulated 53 . 5% of the M4 module genes and induced downstream defense responses in these two species . These data suggest that the induction of the M4 module genes correlates with the variation of different HAE induced defense responses within species . 10 . 7554/eLife . 19531 . 009Figure 4 . The induction of module M4 is associated with the specificity of different HAE-induced defense responses within species . a and b , the relative JA induction ( a ) and proportion of genes in the M4 module ( b ) induced by different HAEs in four Nicotiana species . The JA induction was scaled between 0 and 1 , to indicate the lowest and highest JA level induced by three different HAEs and control ( WW ) . Each colored bar represents elicitations from different HAEs . Dark green: FAC , purple: M . sexta oral secretion ( OSMs ) , light blue: S . littoralis oral secretion ( OSSl ) . ( c ) Venn diagrams showing the overlap among upregulated genes induced by three different HAEs within each species . Each color represents one HAE with the same color code as in panels a and b . The sizes of the circles represent the total number of genes in each group . ( d–g ) heatmaps showing the expression of M4 gene module members in four species as induced by the three HAEs and control . The color gradient represents the relative expression value . ( h ) OSMs induced lower expression level of JAR1 . 1 than did FAC in three Nicotiana species . Each bar presents the average expression ( TMM normalized FPKM ) of JAR1 . 1 in each species . Each color indicates different treatments . Gray: control ( wounding and water ) , dark green: FAC , purple: OSMs . ( i ) the phylogenetic tree showing the relationship among three paralogs of JAR1 in N . attenuata and orthologues of JAR1 in Arabidopsis thaliana ( At ) , Vitis vinifera ( Vi ) , and Solanum lycopersicum ( Sl ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19531 . 009 Although at a global level , the FAC and OSMs induced similar levels of phytohormones and transcriptomic defense responses ( Figure 4a–c ) , the resulting downstream defenses , such as effects on caterpillar growth rates can be different . For example , our previous study showed that larvae grew faster on leaves induced by OSMs than by FAC in both N . pauciflora and N . attenuata ( Xu et al . , 2015 ) . Because FAC is a subset of the elicitors in OSMs , we reasoned that OSMs might contain other elicitors that suppress the downstream responses of JA accumulations ( Xu et al . , 2015 ) . Consistent with this hypothesis , at a transcriptomic level , we found that OSMs induced a smaller number of M4 genes than did FAC in both N . attenuata and N . pauciflora ( Figure 4b ) . In N . miersii , in which both FAC and OSMs did not induce defense responses . We further identified eight genes ( Supplementary file 1B ) from the M4 module that showed lower expression in response to OSMs than to FAC in both N . attenuata and N . pauciflora . Among these eight genes , one gene , NaJAR1 . 1 ( NIATv7_g23173 , Figure 4e ) , is a member of the jasmonic acid-amido synthetase ( JAR1 ) gene family ( Figure 4f ) . JAR1 catalyzes the formation of jasmonyl-isoleucine ( JA-Ile ) , a conjugate of JA that activates downstream defense responses ( Kang et al . , 2006; Staswick et al . , 2002 ) . In the N . attenuata genome , there are three JAR1 copies that resulted from duplication events , and two of these ( NaJAR4 and NaJAR6 ) are induced by both FAC and OSMs and are involved in the conjugation of JA to amino acids and anti-herbivore defense responses ( Kang et al . , 2006; Wang et al . , 2007 ) . Although the exact functions of NaJAR1 . 1 remain unknown , it shares more than 85% of protein sequence identity to NaJAR4 and NaJAR6 and has the conserved amino acid conjugation domain shared by all JAR1 family members , suggesting that NaJAR1 . 1 is also likely involved in the metabolism of JA . We hypothesize that an unknown component in OSMs , which might be used by the specialist herbivore M . sexta to suppress the expression of NaJAR1 . 1 in order to regulate JA metabolism and thus suppress downstream defense responses in Nicotiana . In summary , the induction of genes in the co-expression module M4 is associated with the specificity of HAE-induced early defense responses within species , and is consistent with the notion that induction of the M4 module is important for HAE-induced defense responses in the genus Nicotiana . The specific responses induced by different HAEs within a species also provided an opportunity to further examine the conservation of the M4 module in Nicotiana . We analyzed the preservation of the M4 module in N . attenuata , N . miersii , N . pauciflora and N . obtusifolia , of which we sequenced the transcriptomes of leaves induced by different HAEs to characterize transcriptional responses . The results revealed that the Z-summary scores of the M4 module , which indicate the level of module preservation , are all above 20 ( values above 10 indicates that the module is highly conserved ) for all pair-wise species comparisons , suggesting the M4 module has been retained among different species ( Table 1 ) . The statistical significance of the module preservation is further supported by permutation tests ( in all comparisons , p<2 . 2E-16 ) . We further analyzed the sequence divergence of M4 module genes by calculating the ω ( Ka/Ks ratio ) of M4 module genes shared between N . attenuata and N . obtusifolia , the two most divergent species in the dataset . The results revealed that the ω value of most M4 module genes ( 94 . 5% ) are significantly less than 1 ( p<0 . 05 , Fisher’s exact test , median ω=0 . 19 ) , indicating they were under strong purifying selection . The distribution of ω from M4 module genes was not different from all leaf expressed genes ( median ω=0 . 20 , p=0 . 18 , Wilcoxon–Mann–Whitney test ) , suggesting that the M4 module genes were not subject to strong divergent selection between N . attenuata and N . obtusifolia . Together , these results are consistent with the hypothesis that the identified M4 module is conserved among the different Nicotiana species . 10 . 7554/eLife . 19531 . 010Table 1 . M4 module is highly preserved among four studied Nicotiana species . The number in each cell refers to the z-summary score calculated using 'modulePreservation' function from WGCNA package . Species in row and column indicate the reference and testing datasets , respectively . The score above 10 indicates the co-expression module is preserved , whereas the score bellow 2 indicate the module is not preserved . For all comparisons , p-values based on permutation tests are smaller than 2 . 2E-16 . DOI: http://dx . doi . org/10 . 7554/eLife . 19531 . 010N . obtusifoliaN . attenuataN . miersiiN . paucifloraN . obtusifolia-20 . 518 . 221 . 0N . attenuata29 . 4-34 . 931 . 2N . miersii22 . 838 . 8-27 . 8N . pauciflora23 . 426 . 227 . 4- The M4 co-expression module contains 1274 genes , which were enriched for gene ontology terms with ‘regulation of defense responses’ , ‘jasmonic acid metabolism’ , ‘response to insects’ , and ‘protein modification’ among others ( Figure 5—figure supplement 1 ) . A majority of the JA biosynthetic genes were found in this module and their expressions were positively correlated with each other ( Figure 5 ) , indicating that the identified co-expressed genes reflect their functional relationships . Among all M4 module genes , 782 were significantly induced by FAC in the three species ( N . attenuata , N . acuminata and N . linearis ) which all showed JA accumulations 30 min after FAC elicitation . We infer that these 782 genes represent the core HAE-induced EDS network , which includes 75 protein kinases and 96 transcription factors ( Figure 5 ) . Previous research has shown that silencing genes in this conserved signaling network can directly affect herbivore-induced JA biosynthesis , metabolism and downstream defenses in N . attenuata . This includes the following protein kinase-encoding genes: NaWIPK ( Wu et al . , 2007 ) , NaMPK4 ( Hettenhausen et al . , 2013 ) , NaBAK1 ( Yang et al . , 2011 ) and NaCDPK4/5 ( Wu et al . , 2007; Yang et al . , 2012 ) , which are positive or negative regulators of JA biosynthesis and induced defense in N . attenuata; as well as the transcription factor , NaWRKY6 , which is involved in differentiating mechanical wounding from herbivore attack and mediates plants’ herbivore-specific defenses ( Skibbe and Galis , 2008 ) . Furthermore , several genes in this network have also been shown to be involved in phytohormone crosstalk and regulate JA-induced downstream defense responses , including NaACO2 , NaACS3a and NaETR1 which are involved in ET biosynthesis and perception ( von Dahl et al . , 2007 ) ; NaLecRK ( Gilardoni et al . , 2011; von Dahl et al . , 2007 ) that inhibits SA accumulation during herbivory and NaHER1 that suppresses abscisic acid ( ABA ) metabolism after herbivore attack , which , in turn , activates JA accumulation and defenses against insect herbivores ( Dinh et al . , 2013 ) . 10 . 7554/eLife . 19531 . 011Figure 5 . The co-expression network of module M4 . ( a ) the network view of the M4 module . Each node represents a gene in the M4 module , except the filled orange node , which represents a collapsed node from a cluster . The shape of the node represents the property of the gene . Transcription factor: triangle , round rectangle: protein kinases , ellipse: other genes . The size of each node indicates their log2 fold-change after FAC induction . The color of each node represents its Mapman functional annotation . Green: signaling , yellow: transcriptional regulation , red: post translational modification , gray: programmed cell death , purple: biotic and abiotic stress responses , dark blue: transport activity , light blue: hormone metabolism , orange: others . Edges represent the connections between two genes , estimated based on their co-expression coefficient . The genes that were shown to regulate HAE-induced anti-herbivore defenses are also shown in the network . ( b ) the correlation among genes involved in JA biosynthesis and metabolism . The left side shows biosynthesis and metabolism of JA , right side shows the correlation among each other . Each circle indicates the pairwise correlation coefficient between two genes . The size of the circle indicates the coefficient value . Only statistically significant correlations were shown ( p<0 . 05 , Pearson's product moment correlation test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19531 . 01110 . 7554/eLife . 19531 . 012Figure 5—figure supplement 1 . Gene ontology ( GO ) enrichment analysis of the M4 module genes . Each node represents a GO term . The most frequent enriched term is defense signaling . DOI: http://dx . doi . org/10 . 7554/eLife . 19531 . 012 Hub genes , which are defined as highly connected genes in the network , are often functionally important . Based on intra-modular connectivity , we identified 64 hub genes ( top 5% ) in the FAC-induced co-expression network , which include NaWIPK , a key positive regulator of JA biosynthesis and induced defense in N . attenuata ( Meldau et al . , 2009 ) . To provide further mechanistic understanding of these hub genes in regulating induced defenses , we characterised an additional unknown hub gene encoding a putative leucine-rich repeat receptor kinase ( NaLRRK1 ) . The plasma membrane and nuclei localized NaLRRK1 ( Figure 6a ) has an N-terminal extracellular region , a single transmembrane domain , and a C-terminal cytoplasmic region . The expression of NaLRRK1 was co-upregulated with induced JA signaling among the six Nicotiana species ( Figure 6b ) . Measuring NaLRRK1 transcripts in leaves treated with different HAEs and one pathogen-associated elicitor , flg22 , revealed that NaLRRK1 is specifically induced by HAE ( Figure 6c ) . We investigated whether HAE-induced JA signaling regulates the expression of NaLRRK1 using two different jasmonate deficient transgenic plants , in which steps in JA signaling and perception were individually silenced ( Figure 6d ) . Consistent with the microarray results , NaLRRK1 was still significantly induced by FAC 30 min after elicitation in both of the JA-signaling deficient genotypes , revealing that JA is not required for the up-regulation of NaLRRK1 . Interestingly , compared to WT plants at 1 hr , NaLRRK1 transcript levels were higher in irAOC plants – in which JA-Ile levels remain at basal levels - but were lower in irCOI plants , in which JA-Ile levels are constitutively high ( Paschold et al . , 2008 ) ( Figure 6e and f ) . This indicates that JA-Ile levels may suppress the accumulation of NaLRRK1 transcripts . To test this hypothesis , we compared NaLRRK1 transcript accumulations in leaves in which the levels of JA-Ile were elevated by adding different amounts of JA-Ile to wounded leaves together with FAC . The results revealed that increased JA-Ile levels indeed decreased the levels of NaLRRK1 transcripts ( Figure 6g ) . Furthermore , in the transgenic plants 35S-jmt/ir-mje , in which endogenous JA levels are redirected to MeJA resulting in lower levels of induced JA-Ile and abrogated JA-signaling compared to WT plants ( Stitz et al . , 2011 ) , HAE-induced NaLRRK1 transcript accumulation was higher than in WT plants ( Figure 6—figure supplement 1 ) . These results are consistent with the hypothesis that JA-Ile negatively regulates NaLRRK1 transcript levels . 10 . 7554/eLife . 19531 . 013Figure 6 . Jasmonate signaling suppresses the expression of NaLRRK1 . ( a ) Subcellular localization of NaLRRK1 . Nicotiana attenuata leaves were transformed with PM:CFP and NaLRRK1:YFP . After incubation for 48 hr , the transformed leaves were observed under a confocal microscope . The photographs were taken in UV light , visible light ( bright field ) and in combination ( merged signals ) . Scale bar , 20 μm . ( b ) the transcript accumulation of LRRK1 gene in the leaves of six Nicotiana species elicited by wounding + water ( W + water ) and wounding + FAC ( W + FAC ) , estimated from RNA-seq data ( n=3 ) . Asterisk indicates FDR-adjusted p value <0 . 05 and fold change greater than 2 . c , the kinetics of NaLRRK1 transcript accumulation in N . attenuata leaves at 0 , 0 . 5 , 1 and 2 hr after treatments with different elicitors . For each treatment , 20 μL water or elicitors: FAC , oral secretions from M . sexta ( OSMs ) , S . littoralis ( OSSl ) , or flg22 were applied to the wounded leaves . Triple asterisks indicate the significant difference ( p<0 . 01 , n=5 , except flg22 treatment was with 3 replicates ) between treatment and control ( W + water ) . d , a simplified model of JA biosynthesis and metabolism . The two transformed lines , in which AOC and COI were silenced respectively , are indicated . e and f , the transcript accumulation of NaLRRK1 in the two transformed lines in comparison to WT after elicitation with water ( e ) or FAC ( f ) . g , the FAC induced NaLRRK1 transcript accumulation in N . attenuata leaves was suppressed by JA-Ile . N . attenuata leaves were collected at 1 hr after the induction . JA-Ile was applied in two different concentrations . For panel e , f and g , four biological replicates were used . In panel b , c , e f and g , data are presented as means ± SEM . Asterisk indicates significant difference ( *: p<0 . 05; ***: p<0 . 01 , Student’s-t test ) between treatments . DOI: http://dx . doi . org/10 . 7554/eLife . 19531 . 01310 . 7554/eLife . 19531 . 014Figure 6—figure supplement 1 . In N . attenuata , OSMs induced higher NaLRRK1 transcript levels in 35S-jmt/ir-mje plants than in WT plants . ( a ) diagram shows the JA metabolic flux in WT ( left ) and 35S-jmt/ir-mje plants ( right ) . ( b ) in comparison to WT plants , 35S-jmt/ir-mje plants have reduced level of OSMs-induced JA-Ile . ( c ) the transcript levels of NaLRRK1 are higher in 35S-jmt/ir-mje plants than in WT plants . In both panel b and c , filled and unfilled bars indicate 35S jmt/ir-mje and WT plants , respectively . Data are presented as means ± SEM . Rosette stage N . attenuata plants were wounded and 20 µL of OSMs was immediately applied to the wounds . Leaf samples were collected at 1 hr after elicitations . Asterisk indicates significant differences between WT and 35S-jmt/ir-mje plants ( * , p<0 . 05 , Student’s-t test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19531 . 014 We further investigated the roles of NaLRRK1 in regulating HAE-induced defenses in N . attenuata using virus induced gene silencing ( VIGS ) , which reduced HAE-induced NaLRRK1 transcript abundance by more than 88% in comparison to empty vector ( EV ) plants ( Figure 7—figure supplement 1 ) . The levels of a precursor of JA , OPDA , were significantly increased in VIGS-NaLRRK1 plants compared to EV ( Figure 7g ) . Consistently , the transcript levels of genes involved in OPDA biosynthesis , such as NaLOX3 and NaAOS , were all significantly increased in VIGS-NaLRRK1 plants in comparison to EV plants ( Figure 7 a–d ) , suggesting that NaLRRK1 negatively regulates OPDA biosynthesis . Interestingly , the levels of JA and JA-Ile were not significantly different ( Figure 7 h and i ) . However , both the levels of hydroxylated JA-Ile ( 12OH-JA-Ile ) and transcripts of NaCYP94B3-like1/2 - the homologue of AtCYP94B3 that mediates hydroxylation of JA-Ile in N . attenuata ( Luo et al . , 2016 ) - were significantly increased in FAC-induced VIGS-NaLRRK1 plants in comparison to VIGS-EV plants ( Figure 7e , f , j and k ) . Since reduced expression of NaCYP94B3-like1/2 results in lower levels of 12OH-JA-Ile and higher levels of JA-Ile ( Luo et al . , 2016 ) , it is likely that the increased NaCYP94B3-like1/2 transcript accumulations enhanced the hydroxylation of JA-Ile . These results suggest that NaLRRK1 negatively regulates both JA biosynthesis and the hydroxylation of JA-Ile , and potentially suppress the effects the defense responses elicited by JA-signaling . 10 . 7554/eLife . 19531 . 015Figure 7 . Silencing NaLRRK1 increases FAC induced JA biosynthesis and metabolism and downstream defenses . ( a–f ) the VIGS-NaLRRK1 plants have enhanced FAC-induced transcript accumulations of genes involved in JA biosynthesis and metabolism compared to EV plants ( n=5 ) . FAC elicitation significantly increased transcripts of: NaLOX3 ( b ) , NaAOS ( c ) , NaAOC ( d ) and NaCYP94B3-like1/2 . Transcripts levels were measured at 1h after FAC-elicitation . Due to high sequence similarity between NaCYP94B3-like1 and NaCYP94B3-like2 , qPCR primers we used were not able to distinguish these two copies . g-k , the VIGS-NaLRRK1 plants have enhanced JA biosynthesis and metabolism . FAC elicitation induces significantly higher levels of OPDA ( f ) , OH-JA-Ile ( j ) and COOH-JA-Ile ( k ) in VIGS-EV plants than EV plants , but only marginally higher levels of JA ( h ) and JA-Ile ( i ) ( n=7 ) , l , the VIGS-NaLRRK1 plants accumulated higher transcript levels of the transcription factor NaMyb8 than did VIGS-EV plants . m-o , the VIGS-NaLRRK1 plants accumulated higher transcript levels for the defense genes NaTD2 ( m ) and NaTPI ( n ) and higher levels of TPI activity ( o ) than did VIGS-EV plants . ( p ) M . sexta gained significantly less mass when fed on VIGS-NaLRRK1 plants than on VIGS-EV plants ( n=24 ) . The wounding + FAC treated leaf samples were collected at 1 hr after the treatment for gene expression analysis and at 24 hr after the treatment for TPI activity analysis . In all panels , data are presented as means ± SEM . Asterisk indicates significant difference ( * , p<0 . 05; ** , <0 . 01 , *** , p<0 . 001 , Student’s-t test ) between wounding + FAC treatment and control ( wounding + water ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19531 . 01510 . 7554/eLife . 19531 . 016Figure 7—figure supplement 1 . NaLRRK1 transcript abundance was successfully reduced in VIGS-NaLRRK1 plants in comparison to controls . a and b NaLRRK1 transcript levels in leaves that were undamaged ( a ) and or treated with wounding + FAC ( b ) For treated leaves , samples were collected 0 . 5 hr after elicitation . For each treatment , 8 biological replicates were analyzed . No morphological differences between VIGS-EV and VIGS-NaLRRK1 plants at the rosette-stage of growth were observed . The VIGS-NaPDS was used as a positive control for the VIGS process . Asterisks indicate significant differences between VIGS-EV and VIGS-NaLRRK1 plants ( *** , p<0 . 001 , Student’s-t test ) . Data are presented as means ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 19531 . 01610 . 7554/eLife . 19531 . 017Figure 7—figure supplement 2 . VIGS-NaLRRK1 plants accumulated higher levels of FAC induced soluble sugars and invertase activity in comparison to control . ( a–d ) concentrations of different soluble sugars in VIGS-EV and VIGS-NaLRRK1 plants . ( e ) the activity of soluble invertase in VIGS-EV and VIGS-NaLRRK1 plants . Leaves were collected at 24 hr after wounding and FAC elicitations . Asterisks indicate significant differences between VIGS-EV and VIGS-NaLRRK1 plants ( * , p<0 . 05; ** , p<0 . 01; *** , p<0 . 001 , Student’s-t test ) . Data are presented as means ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 19531 . 017 The examination of the downstream JA-dependent defensive traits in VIGS-NaLRRK1 plants revealed that the net effect was a negative regulation of JA-signaling . In N . attenuata , the transcription factor NaMYB8 , whose expression is activated by increased levels of endogenous jasmonates ( Kaur et al . , 2010 ) , upregulates the expression of NaTD2 and NaTPI , two key anti-herbivore defensive enzymes ( Kang et al . , 2006b ) . In comparison to VIGS-EV plants , VIGS-NaLRRK1 plants accumulated significantly higher transcript levels of FAC-induced NaMYB8 , NaTD2 and NaTPI ( Figure 7 i-o ) , consistent with the observed effects on JA signaling . In addition to NaMYB8 regulated genes , FAC induced JA signaling also induces changes in primary metabolism , in particular soluble sugars and soluble invertases activity in N . attenuata ( Machado et al . , 2015 ) . Here , we also found that VIGS-NaLRRK1 plants showed higher levels of induced soluble sugars and invertases activity in comparison to those of VIGS-EV plants ( Figure 7—figure supplement 2 ) . Consistently , these higher levels of defensive responses in VIGS-NaLRRK1 plants resulted in lower growth rates of M . sexta larvae in comparison to those feeding on VIGS-EV plants ( Figure 7 p ) . Taken together , the data suggest that the HAE-induced M4 gene NaLRRK1 and jasmonate signaling form negative feedback loops that regulate and fine tune the induced defenses in N . attenuata . Having identified the M4 module , we were interested in exploring the evolution of HAE-induced EDS networks in Nicotiana by analyzing the evolutionary history of genes in the M4 module . For this , we first analyzed the most recent duplication event for each N . attenuata gene using the species reconciliation approach ( Materials and Methods ) . Among all M4 module genes , 79 . 5% were retained in the genome of Nicotiana and Solanum after at least one round of duplication since the divergence of eudicots from monocots; this percentage retention is significantly higher than the genome-wide average ( odd ratio = 1 . 96 , p<2 . 2E-16 , exact binomial test , Table 2 ) . This suggests that gene duplications played a significant role in the evolution of the HAE-induced EDS network . Because Solanaceae taxa experienced a whole genome triplication ( WGT ) event ( Sato et al . , 2012 ) , we compared the contributions of the Solanaceae WGT event with Nicotiana-specific lineage duplications ( LD ) to the evolution of the M4 module genes . A majority of the most recent duplication events of the M4 module genes occurred in the Solanaceae branch ( 51 . 5% ) , likely due to the WGT ( Sato et al . , 2012 ) . This percentage is significantly higher than the genome-wide average ( odd ratio = 1 . 46 , p=1 . 79E-10 , exact binomial test , Table 2 ) ( Figure 8 ) . Because our phylogenomic approach can not specifically distinguish the ancient segmental duplications from the WGT events , we further identified a subset of genes that is located in the syntenic blocks that resulted from the Solanaceae WGT . These genes are consistently significantly enriched in the EDS network ( odd ratio = 1 . 40 , p=2 . 4E-7 ) . In contrast , only ~8 . 0% of the genes originated from Nicotiana lineage species duplications , which is not different from the genome-wide level ( odd ratio=0 . 84 , p=0 . 12 , exact binomial test , Table 2 ) . In addition , when considering genes that were significantly induced by FAC in all three species , N . attenuata , N . acuminata and N . linearis ( 'conserved EDS' ) , similar patterns were found ( Table 2 ) . These results suggest that Solanaceae WGT contributed more than lineage specific duplication events to the evolution of the HAE-induced EDS network . 10 . 7554/eLife . 19531 . 018Table 2 . Genes from multiple copy gene families and genes containing DTT-NIC1 TE insertions within 1kb upstream region are significantly enriched in the Nicotiana EDS network . The total number of genes used to test gene duplications and the DTT-NIC1 insertions analyses differed due to the additional filtering processes used in the former analysis . For the gene duplication analysis , we excluded all genes whose most recent duplication event was uncertain . WGT: whole genome triplication; NLD refers to Nicotiana lineage specific duplications; complete EDS refers to all of genes identified in the M4 module; conserved EDS refers to M4 genes that were significantly induced by FAC in all three species , N . attenuata , N . acuminata and N . linearis; genome-wide patterns were calculated based on all of genes that were expressed in Nicotiana leaves ( normalized FPKM greater than 5 in at least three samples ) . Bold font color highlights the statistically significant values . Odd ratios were calculated by the following formula: Odd = ( p1/[1 – p1] ) / ( p2/[1 – p2] ) , where p1 is the percentage of genes that are part of the EDS network among testing group , e . g . , genes from multiple gene families or genes retained from Solanaceae WGT , and p2 is the percentage of genes that are part of EDS network among all leaf expressed genes . DOI: http://dx . doi . org/10 . 7554/eLife . 19531 . 018# genes from multiple copy families# genes retained from Solanaceae WGT# genes retained NLDTotal number of genes after filteringGenome wide# genes96916181124914 , 642Complete EDS# genes906587871140Odd ratio1 . 971 . 450 . 86p value< 2 . 2E-163 . 30E-100 . 17Conserved EDS# genes56135565692Odd ratio2 . 181 . 451 . 07p value< 2 . 2E-161 . 71E-060 . 4110 . 7554/eLife . 19531 . 019Figure 8 . Solanaceae WGT contributed to the evolution of HAE-induced defense signaling in Nicotiana . ( a ) the gene duplication history of Nicotiana attenuata genes after the divergence of eudicots and monocots . Phylogenetic tree constructed based on one-to-one orthologue genes . The bars under each branch depict the percentage of duplications that occurred at a given branch . The green bar indicates the genome-wide ( all genes expressed in leaves ) pattern; red bar indicates the duplication events found in module M4 . ( b ) WGT contributed to the evolution of genes that are involved in HAE-induced EDS . All genes are shown as ellipses , and phytohormones as circles . The color of every ellipse shows the most recent duplication events for each gene: blue and gray indicate the Solanaceae WGT , and ancient ( shared with Arabidopsis ) duplications , respectively . The dots underneath each gene represent the number of homologues found in the N . attenuata genome , and the color indicates whether the homologue was induced by FAC in N . attenuata . Red: significantly induced ( FDR adjusted p<0 . 05 , fold change greater than 1 . 5 ) , black: not induced . Dashed lines indicate the indirect functional interactions . TD: THREONINE DEAMINASE; HGL-DTGs: 17-hydroxygeranyllinalool diterpene glycosides; TPI: trypsin proteinase inhibitor . ( c–e ) the co-expression patterns between the two homologous genes that likely resulted from the Solanaceae WGT; The expression values of each gene were from control and FAC-induced samples from the six Nicotiana species . ( f–h ) phylogenetic trees showing the duplication history of NaCDPK4/5 ( f ) , NaHER1/2 ( g ) and NaACO1/2 ( h ) . The blue dot on the phylogenetic tree indicates the duplication events shared among Solanaceae species . The node colors indicate which species the homolog sequences belong to . At ( lilac ) : Arabidopsis thaliana; Vi ( turquoise ) : Vitis vinifera; Pt ( light blue ) : Populus trichocarpa; Na ( yellow ) : N . attenuata; Sl ( green ) : Solanum lycopersicum . DOI: http://dx . doi . org/10 . 7554/eLife . 19531 . 019 Preferential gene retention followed by genome-wide duplications has been suggested as one of the major mechanisms for network evolution and expansion . Because the complete EDS network before the Solanaceae WGT is unknown , it is difficult to directly test the preferential gene retention hypothesis . Instead , we examined a prediction that would result from the preferential gene retention scenario: for a given gene pair , the observed number of gene pairs that are both found in the EDS is higher than the number of two genes found in the EDS by chance . To test this prediction , we identified a subset of gene pairs ( 4292 pairs including 8584 genes ) that resulted from the Solanaceae WGT ( duplications on the Solanaceae branch ) in which at least two of the three original copies were retained in the N . attenuata genome and expressed in leaves . Because both members of these gene pairs were both retained in the genome , we specifically examined the process of preferential gene recruitment to the M4 module . Among this subset of genes , 428 were found in module M4 , and of those , both members of 120 gene pairs were in the M4 module . This is significantly higher than the expected number from the null model which assumes independent recruitment of two duplicated copies in the M4 module ( p<2 . 2E-16 , χ2-test ) , consistent with the prediction that duplicated genes were preferentially retained in the M4 module . Annotations of 120 gene pairs showed that more than 30 . 8% were either transcription factors or protein kinases , the proportion of which was significantly higher than by chance ( 2 . 56% among 4292 pairs , p<2 . 2E-16 , binomial test ) . Consistent with the genome-wide analysis , most of the M4 module genes that were previously functionally characterized in N . attenuata , including NaHER1 , NaCDPK4/5 , NaMEK2 , NaETR1 and NaJAR4/6 , NaACO1/2 evolved from the Solanaceae WGT event ( Figure 8 ) , and their homologs were also known to be involved in EDS in Arabidopsis ( Dong et al . , 2004; Jung et al . , 2007; Ludwig et al . , 2004; Mewis et al . , 2005; Romeis and Herde , 2014; Staswick and Tiryaki , 2004 ) suggesting that their ancestral copies were likely already involved in EDS . Among these genes , the gene pairs of NaHER1/2 , NaCDPK4/5 , NaJAR4/6 and NaACO1/2 were highly co-expressed and all members were retained in the M4 module . Taken together , these results showed that gene duplications , and likely preferential gene retention followed by WGT , shaped the evolution of Nicotiana HAE-induced EDS networks .
Regulation of herbivore induced defenses requires a complex and fine-tuned network ( Bonaventure et al . , 2011; Erb et al . , 2012; Wu and Baldwin , 2010 ) , as its fitness cost/benefit ratio depends on both the type of herbivore attacking the plant and the environmental context of the attack ( Baldwin and Hamilton , 2000; Glawe et al . , 2003; Heil and Baldwin , 2002; Baldwin , 1998 ) . Therefore HAE-induced early defenses signaling that allows a plant to distinguish herbivore attack from wounding plays an important role in this process ( Bonaventure et al . , 2011; Howe and Jander , 2008; Wu and Baldwin , 2010 ) . However , identifying herbivore induced EDS is challenging , due to its specificity among different tissues , time points , treatments and overall complexity ( Howe and Jander , 2008; Wu and Baldwin , 2010 ) . In this study , we took a novel comparative transcriptomic and co-expression network analysis approach using the leaf transcriptome data from six closely related species that were treated with different HAE , which resulted in the identification of a co-expression network that represents the HAE-induced EDS in Nicotiana . This approach assumes that if two genes are functionally connected ( co-expressed ) , the expression changes of one gene will also affect the other one during evolution , thus increasing the statistical power for detecting co-expressed gene modules . Although this assumption cannot be applied to species specific co-expression modules , it can be used to identify gene co-expression modules that are conserved among the studied species ( Table 1 ) , which likely are functionally important . Using comparative transcriptomic and network analysis , we identified a co-expressed gene network ( module M4 ) , in which 782 genes represent the conserved HAE-induced EDS in Nicotiana . Large numbers of transcription factors and protein kinases were found in this network , suggesting rapid transcriptional and post-transcriptional regulations induced by HAE , which then likely led to the re-configuration of whole-plant metabolism to allow for the production of defense responses ( Gulati et al . , 2013 ) . Interestingly , among these 782 genes in the HAE-induced EDS , of which only 28 . 2% and 11 . 7% in N . obtusifolia and N . miersii were elicited by FAC , respectively , at least 67 . 6% were induced by OSSl in each of these two species ( Figure 4 ) . These results suggest that the signaling network , while not elicited by FAC , remains intact in these two species , likely due to changes in FAC-perception . These results are also consistent with the analysis of module conservation which suggested that the M4 module is highly conserved among different Nicotiana species ( Table 1 ) . Molecular signaling cascades often involve negative and positive feedback loops and form circuits . The M4 module is always co-activated and likely upstream or parallel to the activation of JA signaling , indicating that M4 module and JA signaling are likely involved in such circuits . We found NaLRRK1 , a FAC-induced hub gene from the M4 module , and jasmonate signaling form negative feedback loops and are co-activated by HAE elicitation among different species that diverged several millions of years ago . The conserved co-activation between jasmonate signaling and NaLRRK1 by HAE and its negative effect on insect performance suggests that the identified feedback loops are important for plant fitness in Nicotiana . Although increased expression of NaLRRK1 after HAE elicitation may lower anti-herbivore defenses , it may increase the net fitness by reducing fitness costs associated with induced jasmonate signaling , such as the changes in primary metabolites that are important for a plant’s tolerance of tissue removal and regrowth ( Machado et al . , 2013b ) . Clearly more components are involved in the NaLRRK1 and jasmonate signaling negative feedback loops , such as transcription factors and other protein kinases , which are also likely present in the M4 modules . Future molecular studies that identify direct interacting components with NaLRRK1 will shed light on the mechanisms of the NaLRRK1- JA feedback loops . The challenge will be to quantitatively analyze both transcriptional and post-transcriptional regulation of candidate genes at different time points after elicitation , since their interactions might be highly specific . Analyzing the gene duplication history of the genes in the M4 module suggested that preferential gene retention after the WGT shared among Nicotiana spp . and Solanum spp . likely have contributed to the evolution of HAE-induced EDS in Nicotiana . We found that more than 30 . 8% of duplicated pairs that resulted from the Solanaceae WGT , of which both copies were retained in the M4 module , are either transcription factors or protein kinases . This proportion is significantly higher than by chance ( p<2 . 2E-16 ) , consistent with the dosage compensation hypothesis , which predicts that dosage-sensitive genes , of which transcription factors and protein kinases are examples , are more likely to be retained in the signaling network after genome-wide duplications ( Edger and Pires , 2009; Freeling , 2009; Hakes et al . , 2007; Maere et al . , 2005; Wapinski et al . , 2007 ) . Duplicated genes retained in the same network were often considered as evidence of functional redundancy ( De Smet and Van de Peer , 2012; Veron et al . , 2007 ) ; however , genetic redundancy is often evolutionarily unstable and is unlikely to be maintained over long timescales ( De Smet and Van de Peer , 2012 ) . The Solanaceae WGT event can be dated to 91–52 million years ago ( Sato et al . , 2012 ) , yet many of the duplicated gene pairs remain co-expressed after HAE-elicitation . This suggests that retaining these duplicated copies in the same network has been beneficial to plants , likely as a result of increased network complexity and robustness ( De Smet and Van de Peer , 2012 ) . This is consistent with the results of studies on examining the function of NaCDPK4/5 and NaJAR4/6 by simultaneously silencing either member of both copies in N . attenuata ( Kang et al . , 2006; Wang et al . , 2007; Yang et al . , 2012 ) . When NaCDPK4 and NaCDPK5 were individually silenced , HAE-induced JA accumulations were not affected; silencing both copies increased JA accumulation upwards of 3-fold , which resulted in significant negative fitness effects when plants were not attacked ( Yang et al . , 2012 ) . Therefore , retaining both NaCDPK4 and NaCDPK5 in the network may increase the robustness against the negative fitness effects resulted from null mutations in the gene itself or its regulatory systems ( Gu et al . , 2003 ) . Interestingly , the duplicated gene pair , NaJAR4/6 , which catalyzes the formation of JA-Ile , showed additive effects on HAE-induced JA-Ile accumulations , since silencing each individual copy both resulted in reduced JA-Ile levels ( Staswick et al . , 2002 ) and reductions in the activation of downstream defense responses . Thus , retaining both copies in the network resulted in higher level of HAE-induced JA-Ile and defense responses . In addition to gene duplications , expansions of transposable elements ( TEs ) can also contribute to the evolution of induced signaling networks . For example , in rice , the mPing , a miniature inverted-repeat transposable elements ( MITE ) family , rapidly expanded in specific strains and its insertions into the 5’ flanking region rendered adjacent genes inducible by abiotic stresses by introducing cis-regulatory elements and/or epigenetic markers ( Naito et al . , 2009; Yang et al . , 2005; Yasuda et al . , 2013 ) . Similarly , we observed that insertions of the DTT-NIC1 , a Solanaceae specific MITEs family that contains a stress inducible cis-regulatory element , the W-box , into 5’ regulatory regions of genes are significantly enriched among genes of the HAE-induced EDS network in Nicotiana ( p=0 . 00049 , Appendix 1 ) . We speculate that the genome-wide expansions of DTT-NIC1 may have facilitated gene recruitment into the Nicotiana EDS network by introducing cis-regulatory elements into the 5’ flanking regions of Nicotiana genes . However , several other mechanisms could also result in the same observation; for example , biotic stresses may mobilize TEs , which are more likely to insert into genes with open chromatin under stressed conditions . Future studies that measure the contribution of DTT-NIC1 insertions in the inducibility of the identified EDS genes by manipulating the DTT-NIC1 insertions sites are needed to examine these hypotheses .
Plant material was collected as previously reported ( Xu et al . , 2015 ) . In brief , the seeds of six Nicotiana species were germinated and grown in a York chamber under a 16/8 hr light/dark , 26°C and 65% relative humidity regime until the rosette stage . Manduca sexta and Spodoptera littoralis oral secretions ( OS ) were collected on ice from larvae reared on N . attenuata plants until the 3rd-5th instar as previously described ( Halitschke et al . , 2001 ) . To simulate herbivore attack , one leaf of each plant was wounded with a pattern wheel and 20 μL of 1:5 diluted OSMs , OSSl or FAC ( 138 ng μL-1 C18:3-Glu ) or water ( as control ) was added to the puncture wounds . All leaves were collected 30 min after elicitation , their mid-veins rapidly excised , flash frozen in liquid nitrogen and stored at −80°C until analysis . For each species and treatment , three biological replicates were used based on common practice of RNA-seq experiments . The phytohormone data for all samples were analyzed and published in ( Xu et al . , 2015 ) . Total RNA was extracted from ~100 mg aliquots of homogenized leaves that were used for phytohormone analysis ( Xu et al . , 2015 ) using Trizol ( Thermo Fisher Scientifc , Germany ) according to the manufacturer's protocol . All RNA samples were subsequently treated with RNase-free DNase-I ( Thermo Fisher Scientifc ) to remove all genomic DNA contamination . The mRNA was enriched using the mRNA-seq sample preparation kit ( Illumina ) , and ~200 bp insertion size libraries were constructed using the Illumina whole transcriptome analysis Kit following the manufacturer’s standard protocol ( Illumina , HiSeq system ) . All libraries were then sequenced on the Illumina HiSeq 2000 at the sequencing core facility at the Max Planck Institute for Molecular Genetics , Berlin . On average , more than 35 million 50-nt paired-end raw reads for each sample were obtained . All raw reads are deposited in the NCBI short reads archive ( SRA ) under the project number PRJNA301787 . All raw sequence reads were trimmed using AdapterRemoval ( v1 . 1 ) ( Lindgreen et al . , 2012 ) with parameters '--collapse -trimns -trimqualities 2 -minlength 36' before being used for transcriptome assembly . We mapped all reads to the N . attenuata genome ( release v2 ) using Tophat2 ( v . 2 . 0 . 6 ) ( Trapnell et al . , 2009 ) . We used parameter '--segment-mismatches 2 -read-gap-length 2 -m 0 -N 2' for N . attenuata RNA-seq reads and allowed more mismatches for the other five species with parameters '--segment-mismatches 3 -read-gap-length 5 -m 1 -N 7' . The mapping statistics are shown in Supplementary file 1C . The reads count matrix was then extracted from the bam files using HTseq with parameters ‘-a 1 -t exon’ ( Anders et al . , 2015 ) . For both mapping and reads counting , N . attenuata gene models were used . We further simulated the sequence divergence and estimated the expression levels to evaluate the effects of sequence divergence on reads mapping and gene expression estimation . The analysis revealed that sequence divergence did not affect the overall gene expression estimations using our mapping protocol . We constructed the co-expression modules using the R package weighted gene co-expression network analysis ( WGCNA ) ( Langfelder and Horvath , 2008 ) based on the trimmed mean of M-values ( TMM ) normalized FPKM ( Fragments per kilobase of transcript per million mapped reads ) values . Because the expression values among samples clustered by species , we applied a parametric normalization to reduce the effects that resulted from the background expression differences among species using the ‘Combat’ function from the sva package ( Johnson et al . , 2007 ) . The normalized wounding and FAC-induced samples were first used to calculate the soft connectivity and the top 5000 connected genes were selected for module construction . The power selection , module significance and intra-module connectivity analysis were performed according to the WGCNA tutorial ( http://goo . gl/twg20K ) . We only selected the genes in module M4 with membership greater than 0 . 75 for downstream analysis . FAC-induced hub genes were defined as the top 5% most connected genes in FAC-induced transcriptomes . In total , 67 genes were selected as hub genes . For visualization , the M4 module was exported to Cytoscape ( 3 . 1 . 0 ) with the edge weight greater than 0 . 15 as a cutoff . Preservation of the M4 module at co-expression levels among four different species was analyzed using the 'modulePreservation' function from WGCNA . All pairs-wise comparisons were performed based on RNA-seq data from samples treated with WW , OSMs , FAC and OSSl . The M4 module genes that were not expressed in a given species were excluded . In total , 99 . 4% of genes were used for the module preservation tests . The ratio of Ka/Ks ( ω ) was calculated using one-to-one orthologue genes between N . attenuata and N . obtusifolia using KaKs calculator ( Zhang et al . , 2006 ) with 'YN' method . Because the calculations of ω is unreliable for gene pairs with extremely low Ks values , all gene pairs with Ks value less than 0 . 02 were excluded . In total 1182 genes pairs ( 88 . 9% ) were used for the analysis . We identified differentially expressed genes using the edgeR ( Robinson et al . , 2010 ) package based on the raw count data . Genes with greater than 1 . 5-fold change and FDR-adjusted p-values less than 0 . 05 were considered as differentially expressed . For both the gene co-expression network and differential expression analyses , we only considered genes that had a FPKM greater than 5 in at least three samples . The venn diagram analyses for differentially expressed genes among species were performed using the R package venneuler ( Wilkinson , 2011 ) . To identify gene duplication events , we first assigned homologous groups ( HG ) using a similarity-based method . To do so , we used all genes that were predicted from 11 eudicot genomes ( Xu et al , submitted ) . In brief , all-vs-all BLASTP was used to compare the sequence similarity of all protein coding genes , and the results were filtered based on the following criteria: E-value less than 1E-20; match length greater than 60 amino acids; sequence coverage greater than 60% and identity greater than 50% . All BLASTP results that remained after filtering were clustered into HGs using the Markov cluster algorithm ( mcl ) ( Enright et al . , 2002 ) . For each of the identified HGs , we constructed a phylogenetic tree using an in-house developed pipeline . In brief , we aligned all coding sequences for each HG using MUSCLE ( v . 3 . 8 . 31 ) ( Edgar , 2004 ) based on translated protein sequences with TranslatorX ( v . 1 . 1 ) ( Abascal et al . , 2010 ) . For all aligned sequences , all non-informational sites ( gaps in more than 20% of sequences ) were removed using trimAL ( v1 . 4 ) ( Capella-Gutierrez et al . , 2009 ) . Then , for each HG , PhyML ( v . 20140206 ) ( Guindon et al . , 2009 ) was used to construct the gene tree with the best nucleotide substitution model estimated based on jModeltest2 ( v . 2 . 1 . 5 ) ( Darriba et al . , 2012 ) with the following parameters: -f -i -g 4 -s 3 -AIC -a . The support for each branch was calculated using the approximate Bayes method ( PhyML ) . Duplication events within each HG were predicted based on the reconstructed gene trees using a tree reconciliation algorithm , which compares the structure of species and gene trees to infer duplication events ( Page and Charleston , 1997 ) . This approach allows one to predict the history of gene duplication events at each branch of the species tree . To reduce the false positives , we only considered tree structures with approximated Bayes support greater than 0 . 9 at all three closest branches for assignment of gene duplication events . We measured the FAC-induced gene expression changes in WT and JA-deficient plants ( irAOC ) using microarray analyses . The WT samples were the same as the samples used for the RNA-seq analysis . The germination and growth conditions , FAC elicitation , sample collection and RNA extraction for the analysis of the irAOC plants were the same as those described for the analysis of the WT plants . cDNA preparation and hybridizations were performed as described in Kallenbach et al . ( Kallenbach et al . , 2011 ) . Quantile normalization and log2 transformation was performed for all raw microarray data using the R package 'Agi4x44PreProcess' ( http://goo . gl/TJnA6Q ) . Probes with 1 . 5-fold change and adjusted p-values less than 0 . 05 were considered differentially expressed . The sequences of all probes were mapped to the N . attenuata draft genome ( v1 . 0 ) , and only the probes that uniquely mapped to annotated gene regions were considered for downstream analysis . All microarray data were deposited in the public GEO ( Gene Expression Omnibus ) repository ( GSE75006 ) . The gene functional annotation process was part of the N . attenuata genome sequencing effort ( Xu et al , submitted ) . Multiple annotation tools were used . In brief , BLAST2GO ( Gotz et al . , 2008 ) was used to annotate the GO terms for all predicted genes , and the GO enrichment analysis was performed using the ClueGO ( v2 . 1 . 1 ) ( Bindea et al . , 2009 ) plugin in Cytoscape . In addition , all N . attenuata genes were annotated using MapMan ( Thimm et al . , 2004 ) with annotation information from Arabidopsis , tomato , potato and cultivated tobacco . The transcription factors and protein kinase containing genes were identified based on the identified domains in each gene according to the criteria described in Pérez-Rodríguez et al . ( Riano-Pachon et al . , 2007 ) using the iTAK tool ( http://bioinfo . bti . cornell . edu/cgi-bin/itak/index . cgi ) . All N . attenuata genes from NCBI were retrieved and compared to the predicted N . attenuata genes using BLAST and the best hits were annotated accordingly . The functional annotation of all N . attenuata genes and N . attenuata genome data are available from the N . attenuata database server ( http://nadh . ice . mpg . de/NaDH/ ) . The R scripts used for this study and original data are available as source code file and source data . Total RNA was extracted from ~50 mg leaves using Trizol ( Thermo Fisher Scientifc ) according to the manufacturer's protocol . In brief , all RNA samples were subsequently treated with RNase-free DNase-I ( Fermentas ) to remove all genomic DNA contamination . All cDNA samples were synthesized from ~1 µg total RNA using SuperScript II reverse transcriptase ( Thermo Fisher Scientifc ) . The relative transcript accumulation levels of selected genes were measured using qPCR on a Stratagene MX3005P PCR cycler ( Stratagene ) . For all qPCRs , the elongation factor-1A gene ( NaEF1a , accession number: D63396 ) was used as the internal standard for normalization as previously described ( Johnson et al . , 2007 ) . The primer pairs for qPCR are listed in Supplementary file 1D . All qPCR reactions were performed using qPCR core kit for SYBR Green I ( Eurogentec ) in a 20 µL reaction system . At least four biological replicates were used for all qPCR measurements . To characterize the expression of NaLRRK1 , WT plants and three transformed lines were used: irAOC ( line A-457 ) ( Kallenbach et al . , 2012 ) irCOI ( line A-249 ) ( Paschold et al . , 2007 ) and 35S-jmt/ir-mje ( line A-204 ) ( Stitz et al . , 2011 ) . Seed germination procedure was the same as described above . Seedlings were transferred to Teku pots ten days after germination and then were planted into 1L pots in the glasshouse , which was maintained at 26–28°C under 16 hr of light as described in ( Krügel et al . , 2002 ) . The FAC elicitations were the same as described above . For the flg22 treatment ( W+flg22 ) , 20 µL of 100 nM flg22 in water was immediately applied to standardized puncture wounds produced by the fabric pattern wheel . To test the effects of JA-Ile on the expression of NaLRRK1 , 0 . 25 μM or 0 . 125 μM JA-Ile in 20 μL FAC ( containing 12 . 5% ethanol ) , was immediately applied to the puncture wounds in leaves . The leaf samples ( n=5 ) were collected 1 hr after treatment . All leaf samples were flash frozen in liquid nitrogen , and stored at −80°C until analyzed . The construction of NaLRRK1-YFP reporter fusion was carried out as described by Earley et al . and Ran et al . ( Earley et al . , 2006; Li et al . , 2014 ) , and a reporter fusion was also constructed for the A . thaliana plasma membrane ( PM ) intrinsic protein 2a ( accession number: X75883 ) which was previously characterized as a marker for membrane associations ( Shibata et al . , 2016 ) . The open reading frame ( ORF ) of NaLRRK1 and PM were firstly amplified with Phusion Green High-Fidelity DNA polymerase ( Thermo ) by primer pairs listed in the Supplementary file 1D; an additional sequence ( CACC ) was then introduced into the forward primers to facilitate directional cloning into the pENTR/D-TOPO vector ( Thermo ) . The reconstructed plasmids were transformed into E . coli TOP10 competent cells , then amplified and isolated as the 'entry vector' for the Gateway cloning . The 'entry vector' containing the ORF of NaLRRK1 or PM was recombined into destination vectors using LR clonase ( Invitrogen ) to form a C-terminal NaLRRK1-YFP and C-terminal PM-CFP . Recombined plasmids were transformed into E . coli TOP10 competent cells , and then transformed into A . tumefaciens strain GV3101 for subsequent plant transformation . The transformation was performed using A . tumefaciens strain GV3101 following the protocol by Green et al . ( Green et al . , 2012 ) . Fluorescence was visualized 48 hr following the inoculation with a Zeiss LSM 510 Meta confocal microscope ( Carl Zeiss , Jena , Germany ) . The images were analyzed using LSM 2 . 5 image analysis software ( Carl Zeiss , Inc . ) . VIGS based on the tobacco rattle virus ( TRV ) was used to transiently knock down the expression of the candidate gene NaLRRK1 in N . attenuata as previously described ( Galis et al . , 2013 ) . In brief , fragments of ~300 bp of target genes were amplified by PCR with primers listed in Supplementary file 1D . PCR fragments were recovered by agarose gel electrophoresis and purified using a gel band purification kit ( Amersham Biosciences ) according to the manufacturer’s instructions , and subsequently digested with BamHI and SalI and inserted into plasmid , pTV00 ( RNA1 ) . After sequencing to validate the constructs , pTV-fragment-VIGS constructs and pTV00 ( empty vector ) , together with RNA2 , were transformed into Agrobacterium for the VIGS procedure . At 21 days after Agrobacterium inoculation , rosette-stage plants were wounded with a pattern wheel and 20 μL of 1:5 diluted FAC ( 138 ng μL−1 C18:3-Glu before dilution ) or water was added to the puncture wounds . All samples were collected at 1 hr after elicitation with mid-veins excised , flash frozen in liquid nitrogen , and stored at −80°C until analysis . Silencing efficiency was quantified by qPCR . Overall , more than 88% of the target transcripts were silenced by VIGS . Phytohormones were analyzed as described previously ( Wang et al . , 2007 ) . In brief , ~100 mg frozen leaf was homogenized in a Genogrinder with 0 . 8 mL ethylacetate spiked with [9 , 10-2H2]-dihydro-JA and [13C6]-JA-Ile . Homogenates were centrifuged for 30 min at 4°C and the organic phase was collected and evaporated to dryness , which were subsequently reconstituted in 300 mL of 70% ( v/v ) methanol/water for analysis on an advance UPLC ( Bruker ) , equipped with column ZORBAX eclipse XDB ( Agilent ) and quantified on an EVOQ triple quadrupole mass spectrometer ( Bruker ) using the MRM transitions described in ( Schäfer et al . , 2016 ) . To quantify secondary metabolites that were known to function defensively in N . attenuata , leaves of VIGS-EV and VIGS-NaLRRK1 ( n=8 ) plants were treated with W+FAC for 24 hr , harvested and ground in liquid nitrogen and stored at −80°C until analysis . Trypsin proteinase inhibitor ( TPI ) assay was carried out as previously described ( van Dam et al . , 2001 ) . Briefly , 100 mg of ground powder ( n=6 ) was extracted in a protein extraction buffer . The protein content was determined using the Bradford method and PI activity was analyzed with the radial diffusion assay , using soybean trypsin inhibitor ( STI ) as the external standard . Soluble sugars ( glucose , fructose and sucrose ) and starch concentrations were quantified as described by Machado et al . ( Machado et al . , 2013a ) . Briefly , soluble sugars were extracted from plant tissue ( n=6 ) using 80% ( v/v ) ethanol , followed by an incubation step ( 20 min at 80°C ) . The precipitate was collected by centrifugation ( 15 min , 11 , 000 g , °C ) . Pellets were re-extracted twice with 50% ( v/v ) ethanol . Supernatants from all extraction steps were pooled and enzymatically quantified for sucrose , glucose and fructose . The remaining pellets were used for an enzymatic determination of starch . To evaluate the performance of the specialist herbivore M . sexta on transformed plants , neonates were allowed to feed on EV and transformed plants ( n=28 ) and their masses were measured at 0 , 6 , 10 and 14 d after transfer to experimental plants . To ensure that all larvae were at a similar developmental stage and had similar body mass at the start of the bioassay , newly hatched neonates were placed on untreated WT leaves for 48 hr and weighed . The neonates with similar size were selected for the bioassays . 35S-jmt/ir-mje: N . attenuata transgenic plants ectopically expressing Arabidopsis ( Arabidopsis thaliana ) jasmonic acid O-methyltransferase ( 35S-jmt ) and with N . attenuata methyl jasmonate esterase silenced with RNAi . | A variety of different insects feed on plants and these insects often produce molecules known as elicitors that the plants can recognize . This triggers a sophisticated suite of defenses in the plant that can either deter feeding by the insects , or help the plants endure the attack . The elicitors stimulate the rapid accumulation of a plant hormone called jasmonic acid , which in turn activates the defense responses . However , high levels of jasmonic acid can also reduce the ability of the plants to survive and reproduce by activating plant defenses when they are not needed . Therefore , plants need to regulate the signaling networks that control defense so that jasmonic acid only accumulates when the benefit of fighting the insect outweighs the cost of producing the defenses . The costs and benefits of defense responses vary among different insects and environmental conditions , which has made it difficult to study how plants regulate defense signaling networks . To address this question , Zhou et al . investigated the activities of genes in six species of tobacco plant after they have been exposed to different insect elicitors . The experiments identified a network of genes that is activated in response to elicitors and acts largely independent of jasmonic acid signaling . A newly identified gene in this network called NaLRRK1 and jasmonic acid suppress each other , suggesting that NaLRRK1 helps to regulate jasmonic acid levels . Further analysis shows that a process called genome duplication , in which all the genes in an organism are copied , has shaped the evolution of early defense signaling in Nicotiana . Many of the duplicated genes have adopted new roles and been retained in the plants . This highlights the importance of genome duplications in helping plants to adapt to their environment . The next challenge following on from this work would be to identify what specific roles these genes play in the plants , and how they affect the ability of plants to survive insect attacks in their native habitats . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology",
"evolutionary",
"biology"
] | 2016 | Evolution of herbivore-induced early defense signaling was shaped by genome-wide duplications in Nicotiana |
The dynamics of host-parasite interactions are highly temperature-dependent and may be modified by increasing frequency and intensity of climate-driven heat events . Here , we show that altered patterns of temperature variance lead to an almost order-of-magnitude shift in thermal performance of host and pathogen life-history traits over and above the effects of mean temperature and , moreover , that different temperature regimes affect these traits differently . We found that diurnal fluctuations of ±3°C lowered infection rates and reduced spore burden compared to constant temperatures in our focal host Daphnia magna exposed to the microsporidium parasite Ordospora colligata . In contrast , a 3-day heatwave ( +6°C ) did not affect infection rates , but increased spore burden ( relative to constant temperatures with the same mean ) at 16°C , while reducing burden at higher temperatures . We conclude that changing patterns of climate variation , superimposed on shifts in mean temperatures due to global warming , may have profound and unanticipated effects on disease dynamics .
One of the major challenges of the 21st century is understanding how infectious diseases , which have profound ecological and epidemiological impacts on human ( Hotez et al . , 2014 ) , agricultural ( Chakraborty and Newton , 2011 ) , and wildlife ( Harvell et al . , 2019 ) populations , will be affected by climate change . It is now well-established that the interaction between hosts and their pathogens is sensitive to temperature ( Kirk et al . , 2020; Rohr et al . , 2013 ) . For example , disease transmission ( Ben-Horin et al . , 2013 ) , host immunity ( Dittmar et al . , 2014; Rohr and Raffel , 2010 ) , and pathogen growth ( Gehman et al . , 2018; Kirk et al . , 2018 ) can increase with temperature , while other host-pathogen life-history traits such as lifespan and fecundity can decrease ( Altizer et al . , 2013 ) . The interaction between temperature and multiple host and pathogen life-history traits highlights the inherent complexity of temperature effects on infectious diseases . Indeed , each host or pathogen trait may have a unique dependency on temperature and it is their combined effect ( i . e . , R0 , disease outbreak , virulence ) that is often of interest . However , while a growing body of theoretical ( Kirk et al . , 2020; Rohr et al . , 2013 ) and empirical ( Ben-Horin et al . , 2013; Dallas and Drake , 2016; Gehman et al . , 2018; Kirk et al . , 2020; Zhang et al . , 2019 ) studies have quantified the effect of rising mean temperatures on host and pathogen traits ( such as , e . g . , within-host growth [Kirk et al . , 2018] , pathogen transmission [Kirk et al . , 2019] , and epidemiology [Gehman et al . , 2018; Shocket et al . , 2018] ) , the influence of variable temperature regimes such as heatwaves and temperature fluctuations remains unresolved ( Claar and Wood , 2020; Rohr et al . , 2013 ) . Climate change is predicted to increase not only mean temperatures , but also temperature fluctuations and the frequency and intensity of extreme weather events ( Schär et al . , 2004; Vasseur et al . , 2014 ) . Such changes in temperature variance have the potential to modify host-pathogen dynamics ( Franke et al . , 2019; Rohr et al . , 2013 ) . For instance , diurnal temperature fluctuations have been shown to increase malaria transmission at the lower end of the thermal range ( Paaijmans et al . , 2010 ) , while short-term temperature fluctuations led to reduced transmission success due to lower filtration rates in a Daphnia-pathogen system ( Dallas and Drake , 2016 ) . The effect of extreme heat events on host and pathogen traits is also highly variable and may depend on the magnitude , duration , and intensity of the applied heatwave ( Landis et al . , 2012; Schreven et al . , 2017; Zhang et al . , 2019 ) . In a parasitoid-insect interaction , a heatwave of 5°C resulted in greater parasitoid development while a 10°C increase reduced parasitoid growth ( Schreven et al . , 2017 ) . These apparent contrasting results in response to variation in temperature ( here used to refer both to fluctuating temperature regimes and extreme heat events ) imply that alternate temperature regimes or exposure to temperature shifts of different magnitudes will have distinct impacts on host-pathogen interactions . Indeed , whether all temperature variation acts in the same way or leads to different disease outcomes has been identified as a key open question in the field ( Rohr et al . , 2013 ) . Here , we examine the effect of different types of temperature variation on host-pathogen interactions across a broad range of mean temperatures . Specifically , we used the Daphnia magna—Odospora colligata host-pathogen system to test experimentally how temperature variation alters the thermal performance of both the host and the pathogen across their natural temperature range . Daphnia are a well-established ecological model system ( Miner et al . , 2012 ) used frequently in climate change studies ( e . g . , Dallas and Drake , 2016; Hector et al . , 2019; Kirk et al . , 2020 ) , while Ordospora transmission is representative of a classical environmentally transmitted pathogen ( i . e . , it mimics diseases such as SARS-CoV-2 and Vibrio cholerae ) and meets the assumptions of conventional epidemiological models ( e . g . , infection following mass action [Kirk et al . , 2019] , continuous shedding of infectious particles [Ebert , 2005] and little or no spatial structure within host populations ) . Our microcosm experiment comprised three distinct temperature regimes: constant temperatures and two variable temperature regimes with diurnal fluctuations of ±3°C and 3-day heatwaves of 6°C above ambient , all replicated over the natural temperature range of the model system ( i . e . , 10–28°C , Figure 1 ) . These variable temperature regimes were selected to mimic naturally occurring temperature events in habitats our study organisms encounter naturally ( i . e . , small ponds and rock pools ) ( Jacobs et al . , 2008; Kuha et al . , 2018 ) . During the experiment , we measured host longevity , fecundity , infection status , and the number of O . colligate spores within the host gut ( see Materials and methods for details ) . All measurements were conducted on individually kept Daphnia with up to 18 replicates per measurement . To compare the three different temperature regimes ( i . e . , constant , diurnal fluctuations , and heatwave; Figure 1 ) , we fitted a Beta Function using a Bayesian framework . While there are numerous non-linear functions that can be used to fit thermal performance curves ( e . g . , Briére , Ratkowsky equation ) , the advantage of using the Beta Function is that it provides realistic predictions when extrapolating beyond the measured thermal range and that each of its parameters has a clear a biological meaning ( see Shi et al . , 2016 , for a comparison of thermal performance equations ) , where Fm is the fitness at optimal performance for the fitted host or parasite trait , Topt is the temperature at optimal performance , and Tmin and Tmax are , respectively , the critical minimum and maximum temperatures over which fitness of the trait becomes unviable .
Diurnal temperature fluctuations narrowed the thermal performance curve for infectivity compared with constant temperatures ( Figure 2A ) . The estimated ( using the Beta Function ) maximum temperature at which spores were able to cause infections was 5°C lower under fluctuating temperatures than under constant temperatures ( Figure 2A; Tmax=25°C for fluctuating vs . 30°C for constant; confidence intervals for Tmax do not overlap ) . The thermal performance curve for infectivity under the heatwave , where temperatures were raised by 6°C for 3 days and then returned to constant temperature ( Figure 1 ) , was almost identical to that under constant temperature ( all confidence intervals overlap , Figure 2 and Supplementary file 2 ) . However , unlike the constant temperature regime , the heatwave did not differ from the fluctuating regime , as estimates for the maximum temperature had broad confidence intervals , likely caused by lack of data at the higher temperatures . Remaining parameter estimates of the Beta Function were similar for the three temperature regimes , with the highest rate of infection at 19°C , a maximum infection rate of ~95% infection and no infections under 10°C ( Figure 2A and Supplementary file 2; confidence intervals overlap for Topt , Fm , and Tmin ) . Thus , while diurnal fluctuations led to less infection at higher temperatures , a heatwave did not alter infection rates . Spore burden of the two variable temperature regimes deviated from both the constant temperature regime and from each other ( Figure 2B ) . Consistent with infection rates , daily temperature fluctuation led to a lower maximum temperature ( by ~3°C ) for parasite growth within the host , resulting in a narrowed thermal performance curve for burden compared with the other temperature regimes ( Figure 2B , Supplementary file 3 , non-overlapping confidence intervals for Tmax ) . This is supported further by the consistently lower spore burden for the fluctuating regime when compared with the constant temperature regime except near the optimum temperature of 19°C , where spore burdens of both temperature regimes were similar ( confidence intervals for Topt and Fm overlap ) . While infection rates and burden showed a similar thermal performance for diurnal fluctuations ( both narrowing ) , the response to the heatwave differed between infection and burden ( Figure 2 ) . Compared to the constant temperature regime , spore burden in the heatwave showed a shift in the optimum temperature ( from 19 . 4°C to 15 . 7°C ) , and an increase in the number of spore clusters ( Figure 2B , confidence intervals for Topt and Fm do not overlap ) . However , while spore burden was different at ~16°C , spore burden at ~19°C was nearly identical for all three temperature regimes . Moreover , due to the opposite effects at 16°C for both variable temperature regimes ( i . e . , a narrowing of performance under fluctuating temperatures , exacerbation under heatwave ) , spore burden at this temperature differed by almost an order of magnitude ( i . e . , 86 vs . 737 spore clusters ) . Host fitness was generally reduced when exposed to Ordospora spores or when experiencing variable temperature regimes . Daphnia exposed to the parasite had lower reproductive success near the optimum temperature ( ~20°C ) compared to control animals that were not exposed to Ordospora ( non-overlapping confidence intervals for Fm ) and lost between 8% ( constant ) and 24% ( diurnal fluctuation ) of reproductive output ( Figure 3 ) . Comparing host performance among the different temperature regimes shows that animals exposed to Ordospora in variable temperatures had lower reproductive success ( Figure 2C , Supplementary file 4 , non-overlapping 95% confidence intervals for Fm ) , with a small shift ( 1 . 1°C ) in their thermal optimum under the heatwave regime ( Figure 2C , non-overlapping 95% confidence intervals for Topt ) . Control animals that were not exposed to Ordospora also had lower fitness after the heatwave ( Figure 3 , non-overlapping 95% confidence intervals for Fm ) and , while reproduction at the optimal temperature of the control animals experiencing diurnal fluctuations was lower , confidence intervals overlapped with the constant temperature regime ( Figure 3 ) . The host response to the variable temperature regimes differed from that of the pathogen ( compare thermal performance curves for the heatwave and diurnal fluctuating regimes between Figure 2A and B & C ) . While host performance was reduced ( lower Fm ) under both variable temperature regimes , parasite traits showed either a narrowing of the performance curve ( for diurnal fluctuations ) or no effect and greatly increased performance ( for infection and burden under the heatwave ) .
We show that , not only does temperature variation alter the thermal performance of host and pathogen life-history traits in a unique way—driving a shift in performance up to order-of-magnitude over and above the effect of mean temperature—but that the type of variation and the mean temperature at which it occurs are also critical . Indeed , each of the life-history traits we measured was affected differently by thermal variation . With global warming altering the mean and variance of temperature around the world , how this affects diseases and their dynamics is a critical outstanding question ( Claar and Wood , 2020; Rohr et al . , 2013 ) . Our results demonstrate that the combined effect of changing temperature mean and variance can be highly complex , and may alter the vulnerability of host populations ( Harvell et al . , 2019 ) , affect the evolution of host and parasites ( Buckley and Huey , 2016 ) , and , therefore , impede our ability to accurately predict future disease outbreaks . Infection rates were reduced at higher temperatures when animals experienced diurnal fluctuations but not after experiencing a heatwave . Our estimates of maximum temperature for the heatwave , however , have broad confidence intervals , likely due to the lack of data at high temperatures , and extrapolations of our results beyond our highest heatwave temperature ( which was 22°C and reached 28°C during the 3-day heatwave which was close to the estimated maximum thermal tolerance of the parasite ) should be interpreted with caution . In Daphnia , filtration rates determine the contact rate between host and pathogen , and a reduction in filtration can thus lead to reduced levels of infection ( Hall et al . , 2010 ) . As filtration rates of D . magna decline at higher temperatures ( Kirk et al . , 2019 ) , average infection rates under diurnal temperature fluctuations would also be expected to be lower due to the non-linear nature of the thermal performance curve ( i . e . , Jensen’s inequality; Dowd et al . , 2015 ) . In addition , infection probability in our study system decreases sharply when temperatures surpass 22°C ( Kirk et al . , 2019 ) , reducing infection rates under fluctuating temperatures that exceed this temperature ( again , due to Jensen’s inequality ) . In systems where immune function depends on temperature ( e . g . , insects , mosquitos , and ectotherms in general; Paaijmans et al . , 2013 ) , heatwaves may interact with the immune system in complex ways ( Murdock et al . , 2012 ) , particularly when the heatwave occurs early in the infection process . However , given that our heatwave occurred 20 days post-infection and that Daphnia are not known to recover from infection ( Ebert , 2005 ) , the effect of the heatwave on established infections may have been limited . Absence of an effect of a heatwave on infection rates has also been found for a pipefish-trematode host-parasite system ( Landis et al . , 2012 ) . However , though the heatwave did not affect infection rates in our experiment , it did affect parasite burden . Our results show that different types of temperature variation can alter parasite burden and thus affect pathogen growth within the host . While diurnal temperature fluctuations and heatwaves brought about an almost order of magnitude difference in spore burden at a mean temperature of 16°C , no differences were observed at ~19°C . Generally , similar to infection rates , the thermal performance curve for spore burden narrowed under fluctuating temperatures , as predicted by averaging over the non-linear thermal performance curve ( Denny , 2017; Dowd et al . , 2015 ) . The impact of diurnal temperature fluctuations on parasite fitness has been studied previously , with multiple studies suggesting a shift in the thermal performance of parasite fitness under fluctuating temperatures ( Dallas and Drake , 2016; Duncan et al . , 2011; Greenspan et al . , 2017a; Paaijmans et al . , 2010 ) . Indeed , our findings that Ordospora has a narrower thermal performance for spore burden and infectivity under fluctuating temperatures adds to a growing body of evidence ( Dallas and Drake , 2016; Greenspan et al . , 2017b; Hector et al . , 2019; Roth et al . , 2010 ) suggesting that estimates and predictions that ignore temperature variation may over- or underestimate disease burden and prevalence ( Greenspan et al . , 2017a; Raffel et al . , 2012; Rohr et al . , 2013 ) . Moreover , with almost an order-of-magnitude difference between both our two variable temperature regimes at some , though not all , temperatures , our results highlight that both the context and type of temperature variance need to be considered when trying to understand how pathogen performance may be affected by climate change . Spore burden increased following heatwaves , but the effect depended on the mean temperature to which the heatwave was applied . Indeed , the heatwave had either higher , similar or lower spore burden compared to the equivalent constant temperature regime . It was shown recently in a fish-tapeworm host-parasite system that parasite growth , egg production , and the number of first-stage larvae increased after a 1-week exposure to higher temperatures ( increase up to 7 . 5°C ) ( Franke et al . , 2019 ) . Our findings corroborate that heatwaves associated with climate change may , under some conditions , increase disease burden . Indeed , we found a considerable increase in spore burden and a shift in the optimum temperature following a 3-day increase in temperature of 6°C at 16°C . Although some studies have reported increased disease susceptibility following heatwaves ( Dittmar et al . , 2014; Roth et al . , 2010 ) , others found no effect on immune function ( Stahlschmidt et al . , 2017 ) or reduced disease performance after exposure to high temperatures ( Fayer et al . , 1998 ) . Our results may explain these conflicting findings—we found that the effects of a heatwave on spore burden are contingent on the mean temperature to which the heatwave is applied . That is , our results show that the heatwave has either lower or higher burden than equivalent constant temperatures . This context-dependency of heatwaves is supported further by studies in both plant-endoparasite ( Schreven et al . , 2017 ) and herbivore-parasitoid ( Zhang et al . , 2019 ) systems , which showed that the effect of a heatwave on parasite traits depended on the amplitude of the extreme event . As highlighted by a recent review ( Claar and Wood , 2020 ) , effects of warming events on disease traits remain difficult to generalize , and more studies and insight into underlying principles and mechanisms are needed to forecast the effect of extreme heat events on disease dynamics . Indeed , while it is clear from our experiment that a short , 3-day increase in temperature can drastically alter the thermal performance curve for parasite burden , the exact mechanism ( s ) underlying this change remains unidentified . Differences in acclimatization speeds between hosts and pathogens may explain the observed increase in burden of Ordospora at 16°C following a heatwave . According to the temperature variability hypothesis ( Raffel et al . , 2012; Rohr et al . , 2013 ) , parasites , which have faster metabolic rates due to their smaller size , should acclimatise more rapidly to changing temperatures than their larger hosts . In unpredictable variable environments , such as our heatwave regime , parasites should thus have an advantage over their hosts . Moreover , host resistance may also decrease as a result of the trade-off between the energy demand for acclimatization and immunity ( Nelson and Demas , 1996 ) . That varying temperature can lead to higher infection prevalence has been established in Cuban tree frogs , red-spotted newts , and abalone ( Ben-Horin et al . , 2013; Raffel et al . , 2012 ) . While this hypothesis may explain our observation of high burden for the heatwave near 16°C , it does not , however , explain why the response depends on the mean ( i . e . , lower performance at higher temperatures ) . Though Ordospora should have an overall advantage under the temperature variability hypothesis , the realized advantage may be smaller as its thermal range is more restricted than its host ( Kirk et al . , 2018 ) . The heatwave may thus cause proportionally more stress in the parasite than the host at high temperatures , consistent with the thermal stress hypothesis , which suggests that a shift in temperature may reduce the performance of either host or parasite ( Paull et al . , 2015 ) . Indeed , that thermal stress can affect host and pathogen performance has been well supported ( Gehman et al . , 2018; Kirk et al . , 2019; Schreven et al . , 2017; Zhang et al . , 2019 ) . Alternatively , the observed increase in parasite burden due to heatwaves may be system-specific and not explained by differences in acclimatization speed . Estimates show that growth rates of Ordospora increase by a factor of 5 between 20°C and 24°C before declining again ( Kirk et al . , 2018 ) . While the optimal performance of Ordospora occurs around 19°C , due to the balance of thermal performance curves of other host and pathogen traits ( e . g . , mortality , infectivity , etc . ) , a temporary increase to 22°C , as occurred under our heatwave at 16°C , may thus have exacerbated pathogen growth , particularly if different traits react differently to a temperature disturbance , which may have disrupted the balance between host and parasite . Changes in host fecundity in response to temperature variation differed to the response of both parasite traits ( i . e . , infectivity and spore burden ) we measured . While infectivity and burden had either a narrower thermal performance curve or showed a heightened and shifted peak , temperature variation lowered the reproductive output of the host near the thermal optimum . A reduction in reproductive output of the host under variable temperatures is consistent with previous work both on Daphnia ( Schwartz et al . , 2016 ) and in other systems ( Craig and Kipling , 1983; Uvarov et al . , 2011 ) . Similarly , a reduction in host fecundity due to parasitism is well established ( Ebert , 2005 ) . Infection may also reduce the thermal tolerance of the host ( Hector et al . , 2019 ) , which would explain the small shift of the thermal optimum for host reproduction under the heatwave regime . While host responses are thus consistent with expectations , the distinct responses to the different temperature regimes of the different life-history traits we measured ( i . e . , host fecundity , parasite infectively , and parasite burden ) highlight that the effects of temperature variation on host-pathogen systems are complex . When trying to model disease dynamics and outbreaks , we often include a multitude of host and pathogen traits , each with their own thermal dependencies . Recent studies have made advances in predicting disease growth and spread under rising mean temperatures , integrating approaches , and identifying mechanisms that can capture and predict the thermal performance of host and pathogen traits within epidemiological models ( e . g . , metabolic theory ) ( Kirk et al . , 2020 ) . It remains to be seen , however , whether such modeling frameworks can be extended to incorporate temperature variation , especially considering the distinct responses for the life-history traits we measured to each of our variable temperature regimes . Our study shows that temperature variation alters the outcome of host-pathogen interactions in complex ways . Not only does temperature variation affect different host and pathogen life-history traits in a distinct way , but the type of variation and the mean temperature to which it is applied also matters , with up to an order of magnitude change between diurnal fluctuations in temperature and extreme heat events . With global warming altering both the mean and variance of temperature around the world , we can thus expect to see unanticipated changes in disease dynamics of host-pathogen systems . Indeed , extreme temperature events such as El Niño have been linked to disease-driven collapses of keystone predators ( Harvell et al . , 2019 ) , increases in diseases such as dengue and cholera ( Anyamba et al . , 2019 ) , and shifts in the geographic distribution of pathogens ( Claar and Wood , 2020 ) . While temperature variation can thus affect disease dynamics in human , wildlife , and livestock populations—with potentially devastating economic and health consequences ( Altizer et al . , 2013 ) —the complexity of the effects of temperature and its variation currently limits our ability to move beyond system-specific predictions , in particular for extreme temperature events ( Claar and Wood , 2020 ) . We conclude that improving our mechanistic understanding of the role of temperature variation on disease dynamics , and exploring the generality of its effects and how it affects thermal performance curves of both hosts and parasites ( Claar and Wood , 2020 ) , are critical to predicting disease dynamics in a warming world .
The crustacean D . magna plays a key role in ecosystem functioning . Daphnia are filter feeders that consume planktonic algae and other microorganisms , thus promoting water transparency and helping to prevent algal blooms ( Miner et al . , 2012 ) . They are a key food source for planktivorous fish , constitute a major part in the food web ( Ebert , 2005 ) , and play a key role in nutrient cycling ( Elser et al . , 2000 ) . Across its range , Daphnia is affected by a broad variety of pathogens . Here , we use O . colligata , a widely distributed microsporidium parasite that is only known to infect D . magna . This gut parasite has been recently used as a model to understand how changes in mean temperatures under global warming may affect host-parasite systems ( Kirk et al . , 2020 ) . However , the effects of temperature variance remain unstudied . Daphnia become infected when they accidentally ingest water borne spores of Ordospora while filter feeding . After successful establishment , spores divide intracellularly in the gut epithelium of D . magna ( Larsson et al . , 1997 ) until they form a cluster of 32–64 spores . Spores are then released either to the environment or go on to infecting neighboring cells after O . colligata lyses the cell . In the laboratory , we established water baths with temperatures ranging from 10°C to 28°C . Each bath was regulated with a temperature controller ( Inkbird ITC-308 ) that interfaced with cooling ( Hailea HC300A ) and heating ( EHEIM JÄGER 300 W ) units . Pumps ( Micro-Jet Oxy ) were used to create constant flow , which ensured equal temperature distribution within the water baths . Each bath held up to 99 microcosms and was kept under natural lighting conditions ( 16:8 light:dark ) . Temperature and light intensity were recorded using HOBO loggers housed in spare microcosms . Each microcosm was filled with up to 80 ml of Artificial Daphnia Medium ( ADaM , modified to use only 5% of the recommended selenium dioxide concentration; Klüttgen et al . , 1994 ) . To test for the effect of changing both mean temperature and patterns of temperature variation in our host-parasite system , we created three different temperature regimes: one constant and two variable temperature regimes , the latter comprising diurnal temperature fluctuations and a heatwave ( Figure 1 ) . In the constant temperature regime , individual Daphnia were kept at one of seven temperatures for the whole experimental period ( i . e . , 10 , 13 , 16 , 19 , 22 , 25 , and 28°C ) . The diurnal fluctuation regime comprised five temperature levels , which experienced the same mean temperature as the constant regimes but with a fluctuation of ±3°C every 12 hr ( i . e . , 10–16°C , 13–19°C , 16–22°C , 19–25°C , and 22–28°C ) , mimicking diurnal fluctuations in small rock pools ( Jacobs et al . , 2008 ) . The heatwave was performed at four different temperature levels ( 13 , 16 , 19 , and 22°C ) , with conditions identical to the constant regime except for an increase of 6°C for 72 hr , 20 days after animals were exposed to the parasite , mimicking a short heatwave ( Kuha et al . , 2018 ) . We chose these temperature levels because of their relevance for our host and pathogen system , as no infection occurs below 12°C and hosts have high mortality above 30°C ( Kirk et al . , 2018; Kirk et al . , 2019 ) . Animals were kept individually in microcosms , organized into trays , and repositioned daily to avoid positioning effects . In each temperature regime , half of the microcosms were exposed to the parasite while the other half served as controls . For each of the constant temperature levels , we used 12 replicates for both animals exposed to Ordospora and control animals that received a placebo exposure . However , as we expected greater mortality in the variable temperature regimes ( Régnière et al . , 2012 ) , we increased the number of replicates of these regimes to 18 . We based this number of replicates on experience with previous temperature experiments with the Daphnia-Ordospora system ( Kirk et al . , 2019 ) . The Daphnia genotype ( clone FI-OER-3-3 ) , we used was previously isolated from a rock pool at Tvärminne archipelago , Finland and propagated clonally in the laboratory . To generate sufficient animals for the experiment , we grew Daphnia asexually under standardized conditions for 3 weeks . Animals were raised in small populations ( 20 400 ml microcosms , 12 animals per microcosm ) under continuous light at 20°C . The medium ( ADaM ) was replaced at least twice a week and Daphnia were fed ad libitum with Scenedesmus algae ( Scenedesmus sp . ) , which was grown in batch cultures at 20°C in WC Medium ( Kilham et al . , 1998 ) under nutrient- and light-saturated conditions . The experiment was initiated by collecting a cohort of female juveniles ( ~600 females up to 72 hr old ) from the small population microcosms . Individual juveniles were then randomly transferred into 100 ml glass microcosms filled with 40 ml ADaM . These glass microcosms were placed into their assigned water baths and , after an acclimation period of 24 hr , the animals were exposed to the parasite by adding 1 ml medium containing ~10 , 000 spores of O . colligata . This spore solution was prepared by crushing 3560 infected D . magna individuals with known average burden ( determined by using phase-contrast microscopy on a sub-sample ) , using mortar and pestle and diluting down the resulting spore slurry . The unexposed controls received a placebo exposure consisting of crushed uninfected animals diluted in medium . Animals were exposed either to the parasite or placebo for 6 days and were transferred subsequently to clean microcosms with fresh medium ( 80 ml of ADaM ) twice a week until the end of the experiment . Animals were fed four times a week with an increasing amount of algae to accommodate the increased food demand of the growing animals ( from 4 million algae ml–1 at the start of the experiment to 10 million algae ml–1 by day 10 , which was maintained until the end of the experiment ) . Between transfers , evaporation of the medium was offset by refilling microcosms daily with 50-50 ADaM-distilled water . To obtain fitness estimates for the host , we counted the offspring produced and checked mortality of all animals daily . Infection status and spore burden ( i . e . , the number of spores inside the host ) were assessed upon death by dissecting individuals and counting the number of spore clusters ( each cluster holds up to 64 parasite spores ) in the gut with phase-contrast microscopy ( 400× magnification ) . Any animals that remained alive until the end of the experiment ( day 27 ) were terminated within 3 days , dissected and their infection status and burden determined without the observer being aware of the identity of the sample . Because infections cannot be diagnosed accurately in early infection stages , animals that died before day 11 were not considered in analyses . Any male Daphnia that were misidentified as female at the start of the experiment were also excluded . In addition , to prevent potentially confounding effects of animals that died early ( where the parasite had less time to grow ) as having lower spore burden , we included only animals from the last day of the experiment in the analysis of spore burden . Note that , to facilitate good estimates of spore burden , we terminated most hosts before natural death occurred , which limits our ability to assess the effects of virulence ( host mortality , reduced fecundity ) . Analyses were performed using R version 3 . 6 . 1 ( R Development Core Team , 2018 ) interfacing with JAGS ( Lunn et al . , 2009; Plummer et al . , 2006 ) . Datafiles and code are available at https://github . com/charlyknz/HostParasite ( Kunze , 2022; copy archived at swh:1:rev:5f2604fe866f547dd80d5a77f99ef8887b9f10e1 ) . A Beta Function was fitted to each of our different fitness estimates ( i . e . , host fecundity , parasite infectivity , and burden ) for each of the three temperature regimes , as:f=FmTmax-TTmax-ToptT-TminTopt-TminTopt-TminTmax-Topt where f is fitness at temperature T , Fm is estimated fitness at optimal performance for the fitted host or parasite trait , Topt is temperature at optimal performance , and Tmin and Tmax are , respectively , the critical minimum and maximum temperatures over which fitness of the trait becomes unviable . This non-linear function has been shown to capture thermal performance accurately ( Niehaus et al . , 2012 ) and has the advantage that all four parameters in the equation have clear biological meaning . To determine the effect of both mean and variation in temperature on host and pathogen traits , we used a Poisson distribution for reproductive output ( number of offspring per individual ) and spore burden ( number of spore clusters produced by the parasite ) . For pathogen infectivity , we used a binomial distribution . Models were fitted using the MCMC fitting algorithm called from R . All models were fitted in a Bayesian framework with JAGS ( Lunn et al . , 2009; Plummer et al . , 2006 ) , while allowing for separate parameter values for each of the different temperature regimes . Priors for temperature effects were specified in order to satisfy the necessary condition Tmin≤Topt≤Tmax and informed by previous work ( see Supplementary file 1 for the priors ) ( Kirk et al . , 2018; Kirk et al . , 2019; Kirk et al . , 2020 ) . The posterior distribution of all parameters was estimated using three chains , 10 , 000 posterior draws which were then thinned by 5 to yield 6000 samples ( 3∗10 , 000/5 ) . Model convergence was checked using the Gelman-Rubin diagnostic . | Global warming is increasing average temperatures and causing extreme temperature fluctuations and heatwaves . These changes may affect when , where , and how often infectious disease outbreaks occur . This could have profound impacts on agriculture , human health , and wildlife . Studying how extreme temperatures or temperature fluctuations alter infections in laboratory animals may help scientists to better understand the impact of climate change on disease . A small aquatic invertebrate , such as a water flea , is one good candidate for such studies . These tiny creatures can be grown in small glass jars in temperature-controlled aquariums . Kunze , Luijckx et al . show that temperature fluctuations and heat waves have complex effects on parasitic infections in water fleas . In the experiments , water fleas housed with a parasite that infects them were exposed to constant temperatures , fluctuating temperatures , or three-day heatwaves , while being kept at a broad range of mean water temperatures . Then , Kunze , Luijckx et al . measured how these conditions affected the water fleas’ longevity , reproduction , and parasite infections . This revealed that temperature variations had a unique effect on the life span , and reproduction and infection rates of the water fleas , depending on the average water temperature the animals were kept at . Heatwaves drastically increased the number of parasites in the water fleas at an average water temperature of 16 °C but had no effect at all or decreased the number of parasites at 19 °C and 22 °C , respectively . Similarly , at high average water temperatures ( >24 °C ) , temperature fluctuations reduced the number of water fleas infected with parasites and the number of parasites in each infected flea . Moreover , the maximum temperature at which parasites were able to cause infections was 5 °C lower under fluctuating temperatures than under constant temperatures . Kunze and Luijckx et al . show that consistent high temperatures , temperature changes , extreme weather events , and mean water temperature affect disease outcomes in water fleas . More studies are needed to assess how temperature variations change the course of diseases in other organisms and to understand the underlying mechanisms . Learning more about disease-temperature interactions will help scientists predict climate change-driven disease outbreaks . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology",
"short",
"report",
"microbiology",
"and",
"infectious",
"disease"
] | 2022 | Alternate patterns of temperature variation bring about very different disease outcomes at different mean temperatures |
While a first pregnancy before age 22 lowers breast cancer risk , a pregnancy after age 35 significantly increases life-long breast cancer risk . Pregnancy causes several changes to the normal breast that raise barriers to transformation , but how pregnancy can also increase cancer risk remains unclear . We show in mice that pregnancy has different effects on the few early lesions that have already developed in the otherwise normal breast—it causes apoptosis evasion and accelerated progression to cancer . The apoptosis evasion is due to the normally tightly controlled STAT5 signaling going astray—these precancerous cells activate STAT5 in response to pregnancy/lactation hormones and maintain STAT5 activation even during involution , thus preventing the apoptosis normally initiated by oncoprotein and involution . Short-term anti-STAT5 treatment of lactation-completed mice bearing early lesions eliminates the increased risk after a pregnancy . This chemoprevention strategy has important implications for preventing increased human breast cancer risk caused by pregnancy .
Epidemiological studies have demonstrated that a first full-term pregnancy before age 22 greatly reduces breast cancer risk , but a pregnancy after age 35 is stimulatory ( MacMahon et al . , 1970; Polyak , 2006; Schedin , 2006 ) . In the global context of increasing age at first pregnancy , it has become critical to identify the molecular mechanism underlying increased long-term breast cancer risk in parous women who had a late-age first pregnancy , so that prevention strategies may be developed to reduce this risk . When a very young woman becomes pregnant , her breast epithelia are unlikely to have accumulated cells with oncogenic mutations ( Crowley and Curtis , 1963; Nielsen et al . , 1987; Lynch , 2010 ) . Normal breast epithelia , following extensive remodeling by a pregnancy , have been reported to become more differentiated , less proliferative , and indolent to tumorigenesis ( Medina , 2004 ) . However , why a first pregnancy at an older age stimulates life-long breast cancer incidence remains mysterious . As a woman reaches the age of 35 or older , her breast epithelia are more likely to have accumulated cells with oncogenic mutations and to harbor precancerous early lesions than a woman in her 20s , based on autopsy studies ( Bartow et al . , 1987; Nielsen et al . , 1987; Welch and Black , 1997 ) . The effects of a pregnancy on these preexisting cancer-precursor cells have not been rigorously tested . If pregnancy instigates cancer evolution from these premalignant cells , the result may explain the age-dependent impact of pregnancy on breast cancer . Using conventional transgenic mouse models to address the effect of pregnancy on preexisting mammary cells has been problematic because the transgenic promoter , and thus the oncogene it regulates , is usually dramatically induced by pregnancy and lactation hormones and also because transgenic oncogenes often impair the normal development of the mammary gland ( Vargo-Gogola and Rosen , 2007 ) . We have reported intraductal injection of a Rous sarcoma virus-based vector , RCAS , as a means of introducing oncogenes into a small subset of cells in the normally developed mammary gland that has been made susceptible to infection by transgenic expression of the gene encoding the RCAS receptor TVA from the MMTV promoter ( MMTV-tva ) ( Du et al . , 2006 ) . The transgenic tva is only required for the initial infection , while the oncogene is transcriptionally controlled by the proviral RCAS LTR , which is constitutive and not influenced by reproductive hormones ( Toneff et al . , 2010; Li et al . , 2011 ) . Using this method , we tested here the effect of a single pregnancy on carcinogenesis from a small number of oncogene-activated mammary cells in the context of normal mammary epithelia and explored the underlying mechanism of , and means to prevent , increased long-term breast cancer risk caused by a pregnancy .
We chose Erbb2 and Wnt1 as the initiating oncogenes , because their gene products activate two distinct pathways that are frequently altered in human breast cancer ( Klaus and Birchmeier , 2008; Baselga and Swain , 2009 ) . We injected 5–7-week-old MMTV-tva mice ( MA ) intraductally with RCAS ( 108 IUs per gland ) expressing either a constitutively activated version of Erbb2 ( RCAS-caErbb2 ) ( Reddy et al . , 2010 ) or Wnt1 ( RCAS-Wnt1 ) ( Dunn et al . , 2000 ) . This dosage leads to infection of approximately 0 . 3% of the luminal epithelial cells ( Du et al . , 2006 ) . 4 to 7 days later , half of the mice were impregnated and then allowed to lactate for 3 weeks . By introducing the oncogene before exposing the mice to a full-term pregnancy , we ensured that an equal number of cells expressed the oncogene in both parous and control virgin mammary glands . This approach allowed us to examine the effects of pregnancy on preexisting precancerous cells and on breast cancer risk while keeping aging-associated variables constant . While caErbB2 induced tumors much more rapidly than Wnt1 , each parous group developed tumors significantly faster than its corresponding virgin control cohort ( Figure 1A; Figure 2A ) and with higher tumor multiplicity ( Figure 1B; Figure 2B , C ) . Of note , this experimental design was intended to test the effect of pregnancy on precancerous lesion progression and long-term breast cancer risk , and not on pregnancy-associated breast cancer ( PABC ) . The great majority of tumors initiated by Erbb2 appeared at least 7 weeks after the completion of pregnancy , beyond the equivalent window of time considered to be PABC in women ( Borges and Schedin , 2012 ) , and all tumors initiated by Wnt1 appeared more than 36 weeks after the completion of pregnancy . In accord with their longer latency , the great majority of tumors arising in our mouse models did not display any of the cardinal features of PABC , including aggressive growth rate , high Ki67 , and stromal involvement ( data not shown ) ( Schedin , 2006 ) . Therefore , we conclude that pregnancy stimulates the long-term cancer risk from mammary cells that have already acquired an oncogenic mutation . 10 . 7554/eLife . 00996 . 003Figure 1 . Pregnancy promotes survival and carcinogenesis of mammary cells that have already activated ErbB2 . ( A and B ) Kaplan–Meier tumor-free survival curves ( A ) and tumor multiplicity ( B ) . ( C–E ) Identification of early lesions by immunohistochemical staining for the HA tag of RCAS-caErbb2 ( C ) , and quantification of the number ( D ) and area ( E ) of lesions . n ≥ 3 mice for each group . Scale bar = 50 μm . Inset scale bar = 20 μm . ( F–H ) Immunofluorescence for cleaved caspase 3 ( CC3 , white arrows ) in lesions . Insets show CC3 in normal ducts for comparison . Percentages of CC3+ cells in lesions ( G ) and normal ducts ( H ) are shown . Scale bar = 20 μm . n ≥3 mice for each group . The generalized Gehan–Wilcoxon and Fisher’s Exact tests generated the p values for ( A ) and ( B ) , respectively . Student’s t test derived all other p values . All columns indicate the mean , and error bars represent SEM except in ( B ) . Results from experiments showing similar oncogene expression levels in parous and virgin mice , as well as from comparison of cell proliferation are presented in Figure 1—figure supplements 1–3 . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 00310 . 7554/eLife . 00996 . 004Figure 1—figure supplement 1 . Promotion of breast cancer by pregnancy is not caused by increased oncogene expression . ( A ) ImageJ quantification of anti-HA ( a tag for caErbB2 ) fluorescent intensity in premalignant lesions . n = 3 mice . ( B ) Protein was extracted from RCAS-caErbb2-induced tumors ( Figure 1A ) , and probed for the HA tag on RCAS-caErbb2 . Error bars indicate SEM . Student's t test determined all p values . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 00410 . 7554/eLife . 00996 . 005Figure 1—figure supplement 2 . Promotion of breast cancer by pregnancy is not caused by the vector system used to induce oncogene expression . Mammary glands from mice injected with FUCGW-caErbb2-GFP and followed through a pregnancy , 3 weeks of lactation , and 3 weeks of involution ( I21 ) were analyzed by FACS along with age-matched virgin controls ( V ) to determine relative intensity of GFP ( A ) . A second cohort of infected mice ( n = 8 ) was palpated for tumors until day 77 , when 50% of the parous group ( P ) acquired tumors ( B ) . Horizontal bar represents the mean ( A ) . The black fraction represents tumor incidence ( B ) . Fisher's one-sided exact test determined p value for ( B ) , and Student's t test determined all other p values . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 00510 . 7554/eLife . 00996 . 006Figure 1—figure supplement 3 . Promotion of breast cancer by pregnancy is not caused by increased premalignant cell proliferation . Immunofluorescence detected Ki67 in early lesions ( A ) at the indicated times post-viral injection , as well as in uninfected mammary ducts ( B ) . n ≥ 3 . Error bars indicate SEM . Student's t test determined all p values . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 00610 . 7554/eLife . 00996 . 007Figure 2 . Pregnancy promotes development of early lesions and tumors from preexisting mammary cells that overexpress Wnt1 . ( A and B ) Kaplan–Meier survival curves ( A ) and bar graph showing tumor multiplicity ( B ) . Generalized Gehan–Wilcoxon test determined p value for ( A ) , and Pearson’s Chi-square test derived p value for ( B ) . ( C ) Area and incidence of sub-palpable tumors ( indicated by arrows ) were calculated at 18-months post infection using ImageJ . n = 24 virgin and 30 parous mice . Student’s t test defined p values . LN , lymph node . ( D ) Immunofluorescence for the HA tag ( top panel ) located the lesions initiated by RCAS-Wnt1 , and TUNEL assay ( bottom panel ) performed on the consecutive section identified apoptotic cells in lesions . n = 5 mice . Student’s t test determined p values . ( E ) Immunofluorescence for pSTAT5 . Graph indicates the proportion of pSTAT5+ lesions ( >5% pSTAT5+ cells ) . Horizontal bars represent the mean . n ≥3 mice for each group . Student’s t test determined p values . Percentage of pSTAT5+ cells in lesions is shown in associated Figure 2—figure supplement 1B . For all bar graphs except ( B ) , columns represent mean , and error bars represent SEM . All scale bars = 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 00710 . 7554/eLife . 00996 . 008Figure 2—figure supplement 1 . RCAS-Wnt1-induced early lesions of parous mice have fewer Ki67+ cells but more pSTAT5+ cells than the lesions of control virgin mice . ( A ) Immunofluorescence for Ki67 detected proliferation in lesions . n = 5 mice . Columns represent mean , and error bars represent SEM . Scale bar = 20 μm . ( B ) Immunofluorescence detected nuclear pSTAT5 in lesions ( Figure 2E ) . Graph indicates percentage of pSTAT5+ cells in lesions . Horizontal bars represent the mean . Student's t test determined all p values . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 00810 . 7554/eLife . 00996 . 009Figure 2—figure supplement 2 . Early age pregnancy promotes STAT5 activation in precancerous mammary epithelial cells of MMTV-Erbb2 mice . Representative photomicrographs depicting immunohistochemical detection of nuclear pSTAT5 in mammary glands of age-matched virgin ( V ) and parous ( I10 ) transgenic MMTV-Erbb2 mice . n = 5 mice . Scale bar represents 50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 009 We next investigated the underlying mechanism of increased breast cancer risk in these parous mice . Tumor growth rates and angiogenesis were comparable between the parous and virgin groups ( data not shown ) , indicating that the impact of pregnancy on tumorigenesis occurs primarily prior to the formation of frank tumors . To identify the effect of pregnancy on premalignant lesions , we compared mammary glands of RCAS-caErbb2-infected mice at 1 . 5 , 5 . 5 , and 8 . 5 weeks post infection , which are equivalent to pregnancy day 7 . 5 ( P7 . 5 ) , lactation day 12 ( L12 ) , and involution day 10 ( I10 ) in the parous group ( Figure 1C ) . At 1 . 5 weeks post injection , the number of total early lesions ( defined as any hyperplastic ductal foci comprised of more than three layers of epithelial cells that stained positive for the provirus-encoded oncogene product ) was similar between the virgin and P7 . 5 mice ( Figure 1D ) , consistent with their similar rates of initial infection . By 5 . 5 and 8 . 5 weeks post-viral injection , however , the total number of early lesions was greater in L12 and I10 groups than in the respective age-matched virgin mice ( Figure 1D ) . By 8 . 5 weeks post infection , the average lesion area was also significantly larger in I10 mice than in the age-matched virgin animals ( Figure 1C , E ) . Therefore , we conclude that pregnancy accelerates carcinogenesis by promoting the advancement of early lesions from cells that have already activated oncogenic signaling . We confirmed that these parity-induced effects were not due to incidental increase of RCAS-expressed oncogene ( Figure 1—figure supplement 1 ) . We also demonstrated that this observation was not caused incidentally by a particular vectoring system: using intraductal injection of a lentiviral vector ( Bu et al . , 2009 ) to introduce caErbb2 to ductal epithelia , we also observed accelerated tumorigenesis in the parous group relative to the virgin group ( Figure 1—figure supplement 2 ) . We then studied the underlying mechanism of this pregnancy-mediated acceleration of early lesion development . Early lesions in P7 . 5 and virgin control mice had similar proliferative indices based on staining for Ki67 ( Figure 1—figure supplement 3 ) and pHistone3 ( data not shown ) , suggesting that the already high proliferation in oncogene-activated cells cannot be further elevated by pregnancy . Proliferation remained comparable between the early lesions of L12 mice and those of age-matched virgin mice . It was even lower in the early lesions of I10 mice than in those of age-matched virgin mice , suggesting that the molecular network activated in normal epithelia at involution to arrest the cell cycle ( Figure 1—figure supplement 3B ) also operates in these premalignant cells . Collectively , these results suggest that pregnancy-induced acceleration of carcinogenesis from mutated cells does not result from increased proliferation . In response to an oncogenic insult , normal cells usually rapidly activate apoptosis , thus erecting a ‘barrier’ to carcinogenesis ( Lowe et al . , 2004; Halazonetis et al . , 2008 ) . We have reported potent apoptosis induction in mammary early lesions of virgin mice initiated by RCAS-caErB2 ( Reddy et al . , 2010 ) . Therefore , we asked whether this apoptotic reaction was impaired as a result of pregnancy . Early lesions from the above cohorts of RCAS-caErbb2-infected mice were stained for cleaved caspase 3 ( Figure 1F ) and also examined with the TUNEL assay ( data not shown ) . As expected , apoptosis was activated in the early lesions of virgin mice at all three time points ( 1 . 5 , 5 . 5 , and 8 . 5 weeks post infection ) compared to normal ducts ( Figure 1G , H ) . However , the level of apoptosis was significantly lower in early lesions of the parous group at all three time points than those in early lesions of the respective virgin mice ( Figure 1G ) ; it became either comparable to the baseline level in normal ducts during pregnancy and lactation or even significantly lower than that during involution ( Figure 1H ) . These observations demonstrate that with a pregnancy preexisting premalignant mammary cells overcome both the apoptotic response to an oncoprotein and the robust cell death at involution , which normally clears the large mass of cells gained during pregnancy/lactation . Apoptosis in oncogene-activated cells is typically initiated by p53 that transcriptionally activates genes encoding apoptosis effectors including Bax , Bak , PUMA and NOXA ( Green and Kroemer , 2009 ) . However , despite their very low apoptotic rate , the early lesions of I10 mice harbored abundant numbers of cells positive for p53 , Bax , and Bak ( data not shown ) , suggesting that as in virgin mice oncogene-activated cells in parous glands are fully capable of activating p53 and its pro-apoptosis targets , but these precancerous cells nevertheless evade apoptosis . We next asked whether evasion of apoptosis by these precancerous cells specifically in the parous group is due to activation of prosurvival machinery that antagonizes the pro-apoptosis targets of p53 . The STAT5 ( signal transducer and activator of transcription ) transcriptional factor plays a crucial role in normal mammary cell survival as well as proliferation and differentiation ( Liu et al . , 1997; Iavnilovitch et al . , 2002 ) . While found in some mammary cells prior to pregnancy , the activated form of STAT5 is detected in the great majority of epithelial mammary cells , and at increased levels , during pregnancy and lactation ( Liu et al . , 1995; Wagner and Rui , 2008 ) . This is because placental lactogen and prolactin ( PRL ) activate PRLR , leading to the recruitment of Jak2 and Jak2-mediated phosphorylation and activation of STAT5 ( Hennighausen and Robinson , 1998; Haricharan and Li , 2014 ) . At the onset of involution , STAT5 is rapidly deactivated to allow programmed death of excess cells accumulated during pregnancy and lactation ( Li et al . , 1997; Hennighausen and Robinson , 2001; Kreuzaler et al . , 2011 ) . Transgenic expression of constitutively active Stat5 mutants in mammary glands leads to increased mammary cell survival , hyperplasia , and even occasional tumors ( Iavnilovitch et al . , 2002; Dong et al . , 2010; Vafaizadeh et al . , 2012 ) , while Stat5 heterozygosity delays mammary tumorigenesis in mice carrying a transgenic oncogene ( Humphreys and Hennighausen , 1999; Ren et al . , 2002 ) . pSTAT5 is also detected frequently in human breast early lesions ( ductal carcinomas in situ ) and to a lesser degree in invasive breast cancers ( Cotarla et al . , 2004; Nevalainen et al . , 2004; Peck et al . , 2011 ) . Therefore , we asked whether preexisting precancerous cells in our parous cohort might evade apoptosis by exploiting this normally well-regulated cell survival pathway . The average percentage of pSTAT5+ cells per lesion was significantly higher in the parous mice at P7 . 5 , L12 , and I10 than in the corresponding age-matched virgin groups , while cells positive for total STAT5 were readily detectable with comparable frequency in virgin vs parous lesions as expected ( Kazansky et al . , 1995 ) ( Figure 3A , Figure 3—figure supplement 1 ) . Furthermore , >70% of the preneoplastic lesions in the parous cohort harbored >5% pSTAT5+ cells ( Figure 3B ) and were thus defined as pSTAT5+ lesions ( this cut-off was chosen because pSTAT5 positivity at this frequency correlated significantly with increased survival as shown below in Figure 3C ) , while <25% of the lesions in the virgin cohort were scored pSTAT5+ . These data suggest that premalignant mammary cells respond to pregnancy and lactation hormones by activating STAT5 , as their normal counterparts do ( Figure 3A inset ) ; however—unlike their normal counterparts—they continue to maintain high levels of pSTAT5 even during involution ( Figure 3—figure supplement 1B ) . 10 . 7554/eLife . 00996 . 010Figure 3 . Pregnancy causes preexisting early lesions to persistently activate STAT5 signaling . ( A and B ) Immunofluorescence for pSTAT5 in lesions ( A ) and accompanying quantification ( B ) . Insets show staining of normal ducts ( A ) . pSTAT5+ lesions are those that have >5% pSTAT5+ cells ( B ) . n ≥ 3 mice . Quantification of percentage of pSTAT5+ cells in lesions is presented in associated Figure 3—figure supplement 1 and analysis of pSTAT1 , pSTAT3 and pSTAT6 is presented in Figure 3—figure supplement 3 . ( C ) Generalized linear regression analysis for correlation between the square root of the number of pSTAT5+ cells in individual I10 lesions and percentage of CC3+ cells in their corresponding lesions . Dots represent individual lesions . Colors represent individual mice ( n = 4 ) . ( D ) Immunofluorescence for BclXL . Positive staining in cytoplasmic ( arrow ) and perinuclear ( arrowhead ) regions was observed . Bar graph shows the percentage of BclXL+ cells in early lesions . n = 4 . Supporting data in Figure 3—figure supplement 2 . ( E ) Immunohistochemical staining for GSK3-β phosphorylated at Ser9 . n = 4 mice . ( F ) Immunohistochemical staining for PRLR . Black arrows indicate cells with membrane PRLR . Bar graph shows the percentage of PRLR+ cells in early lesions . n = 4 mice . Supporting data in Figure 3—figure supplement 4 . Scale bars = 20 μm . All columns indicate the mean , and error bars represent SEM . p value for the P7 . 5 time point was derived by a Wilcoxon Rank Sum test . All other p values were generated by Student’s t test . Supporting Western blotting data are presented in associated figure supplements . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 01010 . 7554/eLife . 00996 . 011Figure 3—figure supplement 1 . caErbb2-induced early lesions activate STAT5 more robustly in parous mice than in virgin mice . ( A ) The percentage of pSTAT5+ and total STAT5a+ cells in 15 early lesions per mouse was quantified . Representative photomicrographs and additional quantification are shown in Figure 3A , B . n ≥ 3 mice . For pSTAT5+ cell frequency , Wilcoxon rank-sum test derived p values for P7 . 5 and L12 , and Student's t test measured p values for I10 . For total STAT5a+ cell frequency , Student's t test measured all p values . ( B ) Western blotting quantified pSTAT5 and total STAT5a and b isoforms in protein extracts from whole mammary glands of virgin and parous mice injected with RCAS-caErbb2 . Student's t test measured p value for pSTAT5 . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 01110 . 7554/eLife . 00996 . 012Figure 3—figure supplement 2 . caErbb2-induced early lesions activate downstream pSTAT5 targets more robustly in parous mice than in virgin mice . Western blotting quantified BclXL , Bcl2 , and cyclinD1 levels in protein extracts of early lesions-bearing mammary glands of virgin ( V ) and parous ( I10 ) mice . Bar graphs show the mean relative protein levels from three independent Western blots using ImageJ densitometer . Student's t test derived p values . Supporting histology is presented in Figure 3D . Columns indicate the mean , and error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 01210 . 7554/eLife . 00996 . 013Figure 3—figure supplement 3 . STAT1 , STAT3 and STAT6 activation is minimal and comparable between cErbb2-induced lesions of virgin and parous mice . Representative photomicrographs depicting immunofluorescence for nuclear pSTAT1 , pSTAT3 and pSTAT6 in lesions of virgin ( V ) and parous ( I10 ) mice . Inset shows pSTAT6 in mammary epithelial cells at early pregnancy as a positive control . Bar graphs show the mean percentage of positive cells in lesions from four mice in each group . Scale bar = 20 µm . White arrows indicate positive cells . Student's t test derived p values . Columns indicate the mean , and error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 01310 . 7554/eLife . 00996 . 014Figure 3—figure supplement 4 . caErbb2-induced early lesions activate PRLR persistently in parous mice . ( A ) Western blotting quantified PRLRL levels in protein extracts of early lesions-bearing mammary glands of virgin ( V ) and parous ( I10 ) mice . Bar graph shows the mean relative protein levels from three independent Western blots using ImageJ densitometer . ( B ) Bar graph indicating mean percentage of cells positive for immunohistochemical staining of PRLR specifically along the cell membrane . n = 4 mice . Student's t test derived p values . Supporting histology is presented in Figure 3 . Columns indicate the mean , and error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 014 Consistent with STAT5 activation , when early lesions in I10 mice were further analyzed , we detected increased levels of the STAT5 prosurvival transcriptional targets BclXL ( Socolovsky et al . , 2001 ) and Bcl2 ( Lord et al . , 2000; Hand et al . , 2010 ) as well as cyclinD1 ( Matsumura et al . , 1999; Brockman et al . , 2002 ) compared to early lesions in age-matched virgin mice ( Figure 3D , Figure 3—figure supplement 2 ) . Importantly , there was a strong inverse correlation between the percentage of pSTAT5+ cells and the level of apoptosis in these parous early lesions ( Figure 3C ) , suggesting a causal relationship between pSTAT5 levels and cell survival . Of note , <1% of cells in virgin and parous early lesions were pSTAT1+ or pSTAT6+ , and <2% were pSTAT3+ ( Figure 3—figure supplement 3 ) , indicating that among the STATs known to be activated in the mammary gland ( Chapman et al . , 1999; Watson , 2001; Hennighausen and Robinson , 2005; Khaled et al . , 2007 ) , STAT5 is the predominant family member activated in parous early lesions . We also tested whether increased cell survival and pSTAT5 occur in parous early lesions initiated by a different oncogene . In virgin mice infected by RCAS-Wnt1 , early lesions were not readily identified on an H&E-stained section within the first few months of infection , but by 6 months post infection , early lesions were easily detected . Compared to the lesions in virgin mice , the Wnt1-initiated lesions in age-matched parous mice exhibited improved survival ( Figure 2D ) and an increased percentage of pSTAT5+ cells ( Figure 2E , Figure 2—figure supplement 1B ) , while also showing reduced proliferation ( Figure 2—figure supplement 1A ) similar to their RCAS-caErbb2-injected counterparts . In addition , we asked whether parity leads to persistent STAT5 activation in classical transgenic models of breast cancer . pSTAT5 was readily detected in mammary epithelia of MMTV-Erbb2 transgenic mice at involution day 10 , but not in mammary glands of age-matched non-transgenic virgin mice ( Figure 2—figure supplement 2 ) . Together , these data suggest that parity-induced persistent activation of STAT5 in precancerous cells is not limited to a specific oncogenic mutation or a unique model of breast cancer . We next tested whether persistent STAT5 signaling in premalignant mammary cells is due to aberrant activation of upstream components that normally regulate STAT5 activity during pregnancy and lactation . As shown in Figure 3F , while being diminished in normal epithelia by I10 , PRLR was readily detected in caErbb2-induced early lesions , with the percentage of positive cells being threefold higher than that in early lesions of age-matched virgin mice ( Figure 3F , Figure 3—figure supplement 4B ) . This increase was confirmed by Western blotting comparing protein extracts of early lesions-bearing whole mammary glands of parous and virgin mice ( Figure 3—figure supplement 4A ) . These data suggest that oncogene-activated mammary cells , unlike their normal counterparts , fail to degrade PRLR at the onset of involution . PRLR is normally deactivated by GSK3β-initiated phosphorylation and proteosomal degradation as circulating prolactin diminishes at involution ( Plotnikov et al . , 2009 ) . However , transformed cells frequently phosphorylate and inactivate GSK3β and thus can aberrantly maintain PRLR ( Plotnikov et al . , 2008 ) . Therefore , we next tested whether oncogenic signaling inactivated GSK3β in early lesions , thereby allowing the PRLR molecules activated specifically in the parous mice to remain at the membrane longer than otherwise expected . Indeed , the inactivated form of GSK3β , pS9-GSK3β , was readily detected by immunohistochemistry in caErbb2-initiated early lesions in parous mice while it was not found in normal mammary cells ( Figure 3E ) . In early lesions of virgin mice , an elevated and similar level of pS9-GSK3β was also detected due to ErbB2-mediated oncogenic signaling . Together , these data suggest that during pregnancy and lactation preexisting premalignant mammary cells activate the PRLR-Jak2-STAT5 signaling cascade as normal mammary cells do , but after the onset of involution , these oncogene-activated cells aberrantly maintain an activated state of this pathway likely via oncoprotein-mediated inactivation of GSK3β preventing timely PRLR degradation . However , precancerous cells in virgin mice do not activate PRLR and therefore cannot exploit ErbB2-mediated GSK3β inactivation for aberrant activation of PRLR-STAT5 . Of note , by 6 months post involution , as few pSTAT5+ cells ( 5 . 7 ± 2 . 2% ) could be detected in early lesions of parous mice , as in virgin mice ( p=0 . 3; n = 4 mice; data not shown ) , suggesting that with time , parity-induced elevated levels of pSTAT5 in parous mice subside . To directly test whether STAT5 activation is sufficient to mimic pregnancy in promoting the survival and tumorigenesis of oncogene-activated mammary cells , we used three complementary in vivo approaches . First , we asked whether forced activation of STAT5 in virgin mice lowers apoptotic levels in caErbB2-expressing mammary epithelial cells . We used RCAS-caStat5a ( Dong et al . , 2010 ) and RCAS-caErbb2 viruses to co-infect MMTV-tva virgin mice ( the co-infection efficiency is 8 ± 4%; data not shown ) . At 2 weeks post infection , the resulting early lesions displayed either caErbB2 alone ( 69% ) or both caErbB2 and caSTAT5a ( 31% ) , and none had caSTAT5a alone ( data not shown ) , confirming our previous observation that the activation of STAT5 alone is not strongly tumorigenic ( Dong et al . , 2010 ) . Enrichment of early lesions positive for both caErbB2 and caSTAT5a relative to the low frequency of co-infection suggests that caSTAT5a promotes lesion initiation by caErbB2 . Importantly , the lesions that were positive for both caErbB2 and caSTAT5a were less apoptotic than the lesions induced in the contralateral glands by caErbB2 alone ( Figure 4—figure supplement 1A ) . Proliferation rates in these two sets of lesions were similar ( Figure 4—figure supplement 1B ) . As expected , the mice that were infected only with RCAS-caStat5a did not form tumors , while the co-infected mice developed tumors much more rapidly than the mice infected with RCAS-caErbb2 alone ( p=0 . 0002; Figure 4A ) , consistent with previous reports of a pro-tumorigenic role for STAT5 in the breast ( Furth et al . , 2011 ) . Significantly , the Kaplan–Meier tumor-free survival curve of the co-infected virgin mice became superimposable on that of the parous cohort of mice from Figure 1A that was infected with RCAS-caErbb2 alone and subsequently impregnated ( Figure 4A , red line ) . These results suggest that forced STAT5 activation in virgin mice is sufficient to mimic breast cancer promotion by pregnancy . 10 . 7554/eLife . 00996 . 015Figure 4 . STAT5 activation in virgin mice mimics pregnancy’s promotion of caErbb2-induced mammary tumorigenesis . ( A ) Kaplan–Meier tumor-free survival curves of virgin mice injected with RCAS-caErbb2 alone ( black ) or with both RCAS-caErbb2 and RCAS-caStat5a ( blue ) . The parous group from Figure 1A is shown for comparison ( red ) . Comparison of lesion apoptosis and proliferation between the two virgin groups is presented in associated Figure 4—figure supplement 1 . ( B ) Kaplan–Meier tumor-free survival curves of the virgin cohort from Figure 1A that was stratified into pSTAT5hi ( blue ) and pSTAT5lo ( black ) groups based on baseline pSTAT5 levels in the normal mammary glands of each of these mice ( Figure 4—figure supplement 2 ) . The parous group from Figure 1A is shown for comparison ( red ) . ( C and D ) Immunofluorescence detected pSTAT5+ cells ( C ) and apoptotic cells ( D ) in lesions induced by RCAS-caErbb2 in MMTV-tva and WAP-tva mice . n = 6 ( C ) and 4 ( D ) mice . Scale bar=20 μm . Columns represent the mean , and error bars indicate SEM . Student’s t test determined p values . Proliferation in lesions is shown in Figure 4—figure supplement 3 , and relevant baseline characteristics of these two mouse lines are delineated in Figure 4—source data 1 . ( E ) Kaplan–Meier tumor-free survival curves comparing MMTV-tva and WAP-tva virgin mice injected with RCAS-caErbb2 . Tumor multiplicity is presented in Figure 4—figure supplement 4 . ( F ) Kaplan-Meier tumor-free survival curves comparing WAP-tva mice injected with RCAS-caErbB2 and then either kept as virgin or impregnated . The p value for all tumor-free survival comparisons was generated using generalized Gehan-Wilcoxon test . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 01510 . 7554/eLife . 00996 . 016Figure 4—source data 1 . Comparison of uninfected and infected cells in MMTV-tva vs WAP-tva mice . Table indicates percentages of infected mammary epithelial cells ( identified by the HA tag in RCAS-β-actin ) that are positive for pSTAT5 in MMTV-tva and WAP-tva virgin mice that were injected with RCAS-β-actin ( HA ) ( n = 4 mice ) 4 days earlier , percentages of pSTAT5+ cells in uninfected control ducts ( n = 5 mice ) , and percentages of infected cells in mammary glands of MMTV-tva and WAP-tva virgin mice that were injected with RCAS-GFP and analyzed by FACS 4 days later ( n = 10 mice ) . Student's t test derived the p values . Associated histological and graphical data are presented in Figure 4C–F and associated figure supplements . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 01610 . 7554/eLife . 00996 . 017Figure 4—figure supplement 1 . Exogenous STAT5 activation in virgin mice recapitulates pregnancy’s promotion of breast cancer . Graphs showing apoptosis ( A ) and proliferation ( B ) in early lesions of virgin mice that were induced by either RCAS-caErbb2 alone or by both RCAS-caErbb2 and RCAS-caStat5a . Immunohistochemical staining identified the lesions that were induced by both caErbB2 and caSTAT5a in co-infected mammary glands . Contralateral glands were injected with RCAS-caErbb2 alone . Each line in panel a represents a mouse . Paired t test ( A ) and Student's t test ( B ) determined significance . Associated tumor latency data are presented in Figure 4A . Columns represent mean and error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 01710 . 7554/eLife . 00996 . 018Figure 4—figure supplement 2 . pSTAT5 in normal ducts is correlated with pSTAT5 in early lesions; CC3 in early lesions is inversely correlated with pSTAT5 . Linear correlation between the percentage of pSTAT5+ cells in early lesions vs percentage of pSTAT5 in normal ducts ( A ) of virgin mice , and between the percentage of pSTAT5+ cells in early lesions and percentage of apoptotic cells ( as identified by CC3 ) in early lesions ( B ) . Each dot represents a mouse . Statistical significance and R2 were derived using a generalized linear regression model . Associated tumor latency data are presented in Figure 4B . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 01810 . 7554/eLife . 00996 . 019Figure 4—figure supplement 3 . caErbB2 leads to early lesions with similar levels of proliferation in WAP-tva virgin mice and MMTV-tva virgin mice . ( E ) Bar graph showing the percentage of BrdU+ cells in early lesions induced by RCAS-caErbb2 ( n = 4 MMTV-tva and 5 WAP-tva mice ) . Statistical significance was determined by Student's t test . Columns represent mean and error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 01910 . 7554/eLife . 00996 . 020Figure 4—figure supplement 4 . caErbB2 leads to a higher tumor multiplicity in WAP-tva virgin mice than in MMTV-tva virgin mice . Tumor multiplicity in virgin MMTV-tva and WAP-tva mice . Fisher's Exact test determined p value . Associated tumor latency data are presented in Figure 4E . Columns represent mean and error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 020 We then asked whether physiological levels of STAT5 activity are sufficient to mimic pregnancy’s impact on mammary tumorigenesis initiated by caErbB2 . Among the experimental MMTV-tva virgin mice that were euthanized at 8 . 5 weeks post RCAS-caErbb2 infection ( described in Figure 1C–E ) , a small subset ( 20% ) exhibited high baseline levels of pSTAT5 in both the normal mammary epithelium and early lesions , while the rest had barely detectable pSTAT5 in either ( Figure 4—figure supplement 2A ) . We termed these mice the pSTAT5hi and pSTAT5lo groups , respectively . Apoptosis levels were inversely correlated with levels of pSTAT5 in early lesions from these two groups of mice ( Figure 4—figure supplement 2B ) . Moreover , when we reanalyzed the data among the experimental virgin mice from Figure 1A , the pSTAT5hi sub-cohort exhibited a shorter tumor-free survival than the pSTAT5lo mice and , remarkably , became indistinct from that of the parous cohort from Figure 1A ( Figure 4B ) . These observations suggest that high baseline pSTAT5 levels in virgin mice mimic stimulation of breast carcinogenesis by pregnancy . Not surprisingly , in the parous group from Figure 1A , mice with pSTAT5lo and pSTAT5hi baseline levels ( same criteria as in virgin mice ) had similar tumor latency ( data not shown ) , probably because profound STAT5 activation during pregnancy and lactation masks the effect of baseline pSTAT5 variations among individuals . To directly demonstrate that in virgin mice high levels of endogenous pSTAT5 can mimic pregnancy in increasing breast cancer risk , we took advantage of our newly characterized transgenic WAP-tva mouse line ( Haricharan et al . , 2013 ) , which expresses tva from the promoter of the Wap ( whey acidic protein ) gene—a classical transcriptional target of STAT5 ( Hennighausen and Robinson , 2008 ) —and used it for selective delivery of caErbb2 into the pSTAT5+ subset of mammary epithelial cells in virgin mice . As predicted , the infected cells in this line harbored significantly more pSTAT5+ cells ( threefold ) than did the infected cells in MMTV-tva mice , while the pSTAT5+ proportion among the uninfected cells remained comparable between the two lines of mice ( Figure 4—source data 1 ) . We further confirmed that RCAS infected equivalent numbers of mammary epithelial cells in WAP-tva and MMTV-tva mice ( Figure 4—source data 1 ) . As predicted , there were many more pSTAT5+ cells in the resulting early lesions in WAP-tva mice than in MMTV-tva mice ( Figure 4C ) . The apoptosis rate was lower in the early lesions in WAP-tva mice ( Figure 4D ) , while proliferation remained comparable ( Figure 4—figure supplement 3 ) . Mammary tumors also arose more rapidly ( Figure 4E ) and with greater multiplicity ( Figure 4—figure supplement 4 ) in WAP-tva mice than in MMTV-tva mice , consistent with previous evidence of alveolar or alveolar progenitor cells as the preferred cell of origin for ErbB2-initiated mammary tumors ( Li et al . , 2003; Andrechek et al . , 2004; Henry et al . , 2004; Jeselsohn et al . , 2010 ) . These data demonstrate that the STAT5-activated subset of mammary epithelial cells has a compromised apoptotic response to oncogenic insult and is highly vulnerable to oncogene-induced carcinogenesis even without a pregnancy . More importantly , in WAP-tva mice infected by RCAS-caErbb2 , pregnancy could no longer accelerate the already hastened tumorigenesis ( Figure 4F ) , directly implicating STAT5 in mediating pregnancy’s stimulation of breast carcinogenesis . Together , we conclude that STAT5 activation during pregnancy is sufficient to increase the survival of precancerous cells and to promote their progression to cancer . To further test the role of STAT5 in mediating pregnancy’s stimulation of carcinogenesis from preexisting precancerous cells , we asked whether the gene encoding STAT5a , the predominant form of STAT5 in the mammary gland and in the lesions generated in our model system ( Figure 3—figure supplement 1 ) , is required for the observed impact of pregnancy on early lesions . The Stat5a knockout mice on the FVB background showed normal mammary development during pregnancy , lactation , and involution ( Figure 5—figure supplement 1 ) unlike those on the 129 background ( Liu et al . , 1997 ) . When these Stat5a−/− mice were bred to MMTV-tva mice , infected with RCAS-caErbb2 , impregnated 4 days post infection , and analyzed at I10 , the resultant early lesions exhibited significantly lower levels of pSTAT5 than did the lesions in Stat5a+/+ mice ( Figure 5A ) confirming the predominance of STAT5a in the mammary gland . These lesions showed a level of apoptosis both comparable to that of early lesions in the age-matched virgin control mice and significantly higher than that of early lesions of Stat5a+/+ mice at I10 ( Figure 5B ) . The number and size of early lesions were also similar in parous vs virgin Stat5a−/− mice ( Figure 5C , D ) . Importantly , tumor incidence and tumor latency were similar in parous vs virgin Stat5a−/− mice ( Figure 5E , F ) . Together , these data demonstrate that Stat5a is necessary for pregnancy-mediated promotion of preneoplastic cell survival , lesion progression , and tumorigenesis . Of note , we did not observe a difference in frequency of pSTAT1+ , pSTAT3+ , or pSTAT6+ cells in lesions of parous Stat5a−/− animals when compared to lesions of wild-type animals ( p=0 . 4 , 0 . 5 , and 0 . 3 , respectively; Figure 5—figure supplement 2 ) . 10 . 7554/eLife . 00996 . 021Figure 5 . Stat5a genetic ablation negates pregnancy’s promotion of caErbB2-initiated mammary tumorigenesis . ( A ) Immunofluorescent staining for pSTAT5 in early lesions of mice with the indicated genotype . n = 5 . Scale bar = 20 µm . ( B ) Levels of apoptosis in early lesions were quantified via immunostaining for CC3 . n = 5 mice . ( C and D ) Average lesion number ( C ) and area ( D ) were quantified using images of immunostaining for RCAS-caErbb2-HA . n = 5 mice . ( E ) Tumor incidence at day 66 post injection with RCAS-caErbb2 . Black region indicates percentage of mice with palpable tumors . n = 18 parous mice and 19 virgin mice . ( F ) Kaplan–Meier tumor-free survival curves comparing Stat5a−/− parous and virgin mice . p values were determined by Student’s t test ( B–D ) , Fisher’s exact test ( E ) , and generalized Gehan–Wilcoxon test ( F ) . Columns represent means and error bars SEM except for ( E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 02110 . 7554/eLife . 00996 . 022Figure 5—figure supplement 1 . Stat5a ablation ( FVB background ) does not affect mammary gland development during pregnancy , lactation , and involution . ( A ) Representative images of whole mounts of lactating ( L12 ) and involuting ( I10 ) mammary glands from MMTV-tva mice with the indicated STAT5a genotype . n = 4 . Scale bars = 0 . 1 mm . ( B ) Representative images of hematoxylin and eosin staining in lactating MMTV-tva mice with the indicated STAT5a genotype . n = 4 . Scale bars = 20 µm . ( C ) Quantification of average litter size ( n = 6 Stat5a+/+ and 12 Stat5a−/− litters ) and average litter weight ( n = 11 Stat5a+/+ and 4 Stat5a−/− ) at the time of weaning ( after 3 weeks of lactation ) represented as mean ± SEM . Student's t test determined p values . Related quantification of early lesion attributes and tumor formation are presented in Figure 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 02210 . 7554/eLife . 00996 . 023Figure 5—figure supplement 2 . Stat5a ablation ( FVB background ) does not affect pSTAT1 , pSTAT3 , or pSTAT6 positivity in early lesions of parous mice at involution day 10 . Bar graphs showing the percentage of pSTAT1 , pSTAT3 , and pSTAT6+ cells in the early lesions of parous mice of indicated genotype at involution day 10 based on immunofluorescent analysis . Columns represent the mean and error bars , the SEM . Student's t test identified p values . n = 4 STAT5a+/+ and 3 STAT5a−/− mice . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 023 To build preclinical evidence for clinical trials to test inhibitors of STAT5 signaling for lowering breast cancer risk in women who have undergone a late-age pregnancy or have abnormally high baseline levels of pSTAT5 signaling , we tested the efficacy of pharmacological inhibitors in preventing mammary tumorigenesis in our mouse model . We first used two inhibitors to block the upstream activators of STAT5 and assessed the effect of these inhibitors on lesion cell apoptosis and lesion progression . AG490 is a small molecule inhibitor that blocks several tyrosine kinases including Jak2 ( Gazit et al . , 1991 ) ; Ruxolitinib is a small molecule inhibitor that specifically suppresses Jak kinases ( Quintas-Cardama et al . , 2010 ) . AG490 ( 75 , or 200 mg/kg body weight ) ( Levitzki and Gazit , 1995 ) or diluent ( DMSO ) was administered every other day for 5 days starting from I2 to early lesion-bearing mice that had already completed lactation . AG490 decreased pSTAT5 , BclXL , and cell survival in parous early lesions , and induced lesion regression ( Figure 6A–C , Figure 6—figure supplement 1A ) . In virgin mice , AG490 did not have any significant impact on precancerous cell pSTAT5 positivity , apoptosis or early lesion regression ( Figure 6A–C ) , in accordance with the low levels of pSTAT5 in these lesions . This finding also indicates that AG490 , which has been reported to inhibit EGFR and ErbB2 , did not inhibit oncogenic signaling from caErbB2 in our model , perhaps due to the mutations in this ErbB2 variant ( Bargmann and Weinberg , 1988 ) . Of note , AG490 did not affect pSTAT1 , pSTAT3 or pSTAT6 positivity in either parous or virgin early lesions ( Figure 6—figure supplement 1B ) . Using a similar experimental approach , we found that in parous mice ruxolitinib also decreased pSTAT5 , blocked cell survival , and induced regression of early lesions ( Figure 6E–F ) . Together , these findings suggest that Jak2 signaling is required for early lesion survival and progression and that blocking the upstream tyrosine kinase activity of STAT5 can prevent the progression of early lesions in parous mice . 10 . 7554/eLife . 00996 . 024Figure 6 . Inhibition of STAT5 signaling prevents pregnancy’s promotion of breast cancer risk . ( A–F ) Bar graphs showing the percentage of pSTAT5+ RCAS-caErbb2-induced lesions ( A and D ) , apoptotic cells in lesions ( B and E ) , and average lesion area ( C and F ) of I10 or virgin mice that were treated with either AG490 ( A–C ) at the indicated doses or Ruxolitinib ( D–F ) . n ≥ 3 mice . Statistical significance was determined by Student’s t test ( A and B , D–F ) and ANOVA ( C ) . The impact of AG490 on BclXL in these lesions is presented in Figure 6—figure supplement 1 . ( G ) Bar graphs showing percentage of CC3+ cells in lesions of mice treated with DMSO or C188-9 . n = 4 mice . Student’s t test derived p value . The potency of C188-9 on STAT5 is shown in Figure 6—figure supplement 2 . The impact of C188-9 on BclXL and cell proliferation is shown in Figure 6—figure supplement 3 . ( H ) Early lesions in parous mice treated either with DMSO or C188-9 were identified by immunohistochemical staining for the HA tag of RCAS-caErbb2 , and their areas were then quantified using ImageJ . n = 4 mice . Scale bar = 50 μm . Student’s t test derived p value . ( I and J ) Kaplan–Meier tumor-free survival curves ( I ) and tumor incidence plot ( J ) of parous and age-matched virgin controls treated with either C188-9 or DMSO . Generalized Gehan–Wilcoxon test determined p-value for ( I ) , and Fisher’s Exact test determined p-value for ( J ) . Tumor latency and incidence for virgin mice treated with C188-9 or DMSO are presented in Figure 6—figure supplement 4 . For all bar graphs , columns represent the mean , and error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 02410 . 7554/eLife . 00996 . 025Figure 6—figure supplement 1 . Impact of short-term treatment with AG490 on STAT proteins , BclXL , and apoptosis in MMTV–tva mice and WAP–tva mice . ( A ) Bar graphs showing the percentage of BclXL+ ( A; n = 5 mice ) cells in lesions of parous mice treated with AG490 at the indicated doses . ( B ) Bar graphs showing the percentage of pSTAT1+ , pSTAT3+ , and pSTAT6+ cells in lesions of virgin and parous mice treated with AG490 at indicated doses ( B; n = 3 mice ) . Effects of AG490 on pSTAT5 , cell survival , and mean lesion area are presented in Figure 6A–C . ( C and D ) Bar graphs showing the percentage of pSTAT5+ lesions ( C; defined as early lesions with >5% pSTAT5+ cells ) and apoptotic cells ( D ) assayed based on TUNEL immunofluorescence in WAP-tva virgin mice treated with AG490 at the indicated dose . n = 4 mice . For all graphs , columns represent the mean , and error bars the SEM . Student's t test determined all p values . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 02510 . 7554/eLife . 00996 . 026Figure 6—figure supplement 2 . C188-9 , a small molecule inhibitor of STAT signaling , inhibits STAT5 phosphorylation . ( A ) Flow histograms for pSTAT5 and pSTAT3 in Kasumi-1 cells pre-treated with C188-9 at the concentrations indicated or with DMSO . Where indicated , G-CSF ( 30 ng/ml ) or media was added , and the cells were incubated for 15 min . Data were analyzed using Cytobank . ( B ) The percentage of positive cells was normalized to the G-CSF-stimulated and C188-9-untreated control , and then plotted . The IC50 for G-CSF-induced pSTAT5 ( 2 . 8 μM ) and for pSTAT3 ( 7 . 4 μM ) was determined using GraphPad Prism . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 02610 . 7554/eLife . 00996 . 027Figure 6—figure supplement 3 . Impact of short-term C188-9 on biomarkers and lesion multiplicity in parous mice . ( A ) Immunofluorescence detected a decrease in pSTAT5+ and BclXL+ cells in the lesions of parous mice treated with C188-9 . n = 4 mice . Scale bars = 20 μm . Student's t test determined p value . ( B–D ) Bar graphs showing insignificant effects of C188-9 on pSTAT1+ , pSTAT3+ , and pSTAT6+ cells in early lesions ( B ) , cell proliferation in early lesions ( C ) , and on the number of early lesions per section per mouse ( D ) . n ≥ 4 . Student's t test derived p values . For all bar graphs , columns represent the mean , and error bars the SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 02710 . 7554/eLife . 00996 . 028Figure 6—figure supplement 4 . Short-term C188-9 treatment does not affect tumorigenesis in virgin mice . Tumor incidence at day 100 after RCAS-caErbb2 infection ( A ) and tumor-free survival curves of virgin mice treated with or without C188-9 ( B ) . Black columns in ( A ) represent mice with detectable tumors . Fisher's Exact test ( A ) and generalized Gehan–Wilcoxon test ( B ) determined statistical significance . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 028 To directly test the impact of blocking STAT5 activity on early lesion progression in these post-lactational animals , we evaluated the effect of our recently developed direct , small molecule STAT inhibitor ( C188-9 ) ( Xu et al . , 2009; Redell et al . , 2011 ) . This inhibitor targets the phosphotyrosyl peptide-binding pocket within the SH2 domain of STAT5 , STAT3 , and possibly other STAT proteins , thereby blocking two steps in their activation—recruitment to ligand-activated receptor complexes and dimerization ( Figure 6—figure supplement 2 ) . C188-9 ( 100 mg/kg BW ) or diluent ( DMSO ) was injected intraperitoneally into mice bearing RCAS-caErbb2-initiated early lesions only twice ( I2 and I9 ) . 1 day after the second injection ( I10 ) , C188-9 caused a significant decrease in pSTAT5+ and BclXL+ cells ( Figure 6—figure supplement 3A ) but did not significantly reduce the low frequency of cells positive for pSTAT1 , pSTAT3 , or pSTAT6 ( p=0 . 2 , 0 . 6 , and 0 . 8 , respectively; Figure 6—figure supplement 3B ) . Further , C188-9 caused a significant rise in apoptosis in the lesions of parous mice , ( Figure 6G ) while having no detectable effect on proliferation ( Figure 6—figure supplement 3C ) . Importantly , C188-9 caused a dramatic regression of parous premalignant lesions ( Figure 6H , Figure 6—figure supplement 3D ) . This effect is unlikely to be caused by off-target impact: in virgin mice , even four doses of C188-9 did not result in a significant decrease of pSTAT5+ cells , increase of apoptosis , or decrease of size in early lesions ( p=0 . 6 , 0 . 8 , and 0 . 7 , respectively; n = 3 mice; data not shown ) . Therefore , C188-9 , like AG490 , can block mammary early lesion survival and progression specifically in parous mice . To demonstrate that these damaged early lesions were impaired in their ability to progress to cancer , we generated another cohort of these early lesion-bearing parous mice , and gave them four weekly injections of C188-9 or diluent ( DMSO ) starting at I2 . This short-term treatment led to a significant reduction in tumor incidence: by 3 months post ErbB2 activation , only three of the eight C188-9-treated mice had developed palpable tumors , while all 10 diluent-treated parous mice had done so ( Figure 6J ) . For comparison , eight of the 15 C188-9-treated virgin control mice had also developed palpable tumors at this time point . This regimen also led to a significant improvement in tumor-free survival in the parous group ( Figure 6I ) . In fact , the tumor-free survival curve became superimposable on that of the age-matched virgin group that was similarly treated . It is important to note that C188-9 did not significantly affect either tumor incidence or latency in virgin mice ( p=0 . 6 and 0 . 5 , respectively; Figure 6—figure supplement 4 ) . Therefore , a short-term treatment with C188-9 hindered the progression of preexisting early lesions and removed the stimulatory effect of pregnancy on breast cancer risk . Together , these preclinical data suggest that a short-term prophylactic treatment with inhibitors blocking Jak2-STAT5 signaling may lower breast cancer risk for women who have had a late-age first pregnancy and have already completed lactation . The extensive mouse model data presented here suggest that in women a pregnancy may also cause preexisting early lesions to aberrantly and persistently activate STAT5 , and chemoprevention targeting STAT5 activity may lower breast cancer risk in women who have had a late-age pregnancy as well as in those who have abnormally high levels of pSTAT5 . Indeed , when we compared 14 cases of DCIS ( preinvasive lesions ) from women who had a pregnancy 6–25 years prior to diagnosis with 13 cases from age-matched nulliparous women in a tissue bank from the University of Colorado ( Figure 7—source data 1A ) , we detected significantly more pSTAT5+ lesions in the parous cases than in the nulliparous cases ( p=0 . 01; Figure 7A , B ) . Of note , only the cases with >5 years between last pregnancy and diagnosis were included for comparison in order to focus on pregnancy’s long-term impact on cancer risk rather than on the relatively rare pregnancy-associated breast cancer subset . Interestingly , patients with higher percentages of pSTAT5+ cells within their lesions also had higher percentages of pSTAT5+ cells in the lesion-adjacent , histopathologically benign breast epithelium ( which may be abnormal at genetic , epigenetic and gene expression levels ) ( Figure 7C , D ) . This observation suggests that pSTAT5 levels in histopathologically ‘normal’ breast epithelia may predict breast cancer risk in women . Therefore , we also quantified pSTAT5+ cell frequencies in tumor-adjacent breast epithelia ( ≥5 . 0 cm away from an invasive cancer ) from a cohort of 24 breast cancer patients diagnosed at the M . D . Anderson Cancer Center ( Figure 7E ) . Higher pSTAT5 in these breast epithelia retrospectively predicted accelerated development of breast cancer following a full-term pregnancy ( p=0 . 007; Figure 7F ) with all other characteristics being comparable ( Figure 7—source data 1B ) . Although the sample size of these two cohorts is relatively small , they nevertheless suggest an involvement of STAT5 in pregnancy’s promotion of long-term risk of sporadic human breast . Of note , BRCA mutation tests were negative in all 10 patients screened in the University of Colorado cohort ( 27 cases ) , and in 22 out of 24 patients screened in the MD Anderson Cancer Center cohort . 10 . 7554/eLife . 00996 . 029Figure 7 . STAT5 activation in normal breast and early lesions of women is associated with parity , and pSTAT5 retrospectively predicts decreased intervals between pregnancy and cancer diagnosis . ( A–C ) Immunohistochemical staining for pSTAT5 ( A ) and accompanying quantification represented by index plots for DCIS ( B ) and benign breast epithelia ( C ) . Scale bar = 20 μm . Pearson’s chi-square test determined p values . ( D ) Regression analysis showing a linear correlation between the percentage of pSTAT5 in DCIS and the adjacent benign epithelia in parous ( red ) and nulliparous ( black ) women . R2 and p value were derived using a generalized linear regression model . ( E and F ) Immunohistochemical staining for pSTAT5 ( E ) in tumor-adjacent breast epithelia stratified patients into pSTAT5low and pSTAT5high . The graph shows the time between the most recent pregnancy and the diagnosis of breast cancer in patients . Age at first pregnancy was included as a confounding factor in analysis of survival curves of women . Generalized Gehan–Wilcoxon test determined p value . Epidemiological characteristics of these patients are presented in Figure 7—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 02910 . 7554/eLife . 00996 . 030Figure 7—source data 1 . Descriptive characteristics of patients . ( A ) Characteristics of nulliparous and parous women ( Figure 7A–D ) whose DCIS and adjacent benign epithelia were assessed for pSTAT5 by immunohistochemistry . ( B ) Characteristics of parous women ( Figure 7E , F ) whose tumor-adjacent breast tissue was analyzed for pSTAT5 by immunohistochemistry . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 030 Our mouse model data are not in conflict with the epidemiological observation that a first pregnancy before age 22 greatly reduces breast cancer risk ( MacMahon et al . , 1970 ) , because at this young age , the chance of having already accumulated precancerous cells is small . To ascertain that the normal mammary epithelium in fully involuted mammary glands responds differently to a future oncogenic insult compared to preexistent oncogene-activated cells responding to a pregnancy , we injected RCAS-caErbb2 into fully involuted mammary glands of MMTV-tva mice ( age = 20 weeks ) . The normal mammary cells in these involuted older mice had the same baseline levels of total STAT5a protein and pSTAT5 as the normal mammary cells in the 6-week-old mice used for oncogene introduction prior to a pregnancy ( Figure 8—figure supplement 1 ) . However , when challenged by an oncogene , the normal mammary epithelial cells in these involuted glands did not aberrantly activate STAT5 ( Figure 8A ) when compared to mammary cells that gained caErbB2 first and were later exposed to pregnancy hormones ( Figure 3A ) . Consequently , their pSTAT5 level was comparable to the low level in early lesions arising in age-matched virgin mice ( Figure 8A ) . The apoptotic index was also high and comparable to that of the lesions in the age-matched virgin control ( Figure 8B ) . Therefore , normal breast epithelial cells in the fully involuted mammary gland—in the event of suffering an oncogenic mutation—do not aberrantly activate STAT5 signaling and can successfully initiate an apoptosis response to evade carcinogenesis . 10 . 7554/eLife . 00996 . 031Figure 8 . Pregnancy reprograms normal mammary cells to resist transformation by a future oncogene . Mice that had completed a pregnancy , 3 weeks of lactation , and 2 months of involution were injected with RCAS-caErbb2 . 3 weeks later , the resulting lesions were compared with those in age-matched virgin controls for pSTAT5 ( A ) , apoptosis ( via TUNEL ) ( B ) , and Ki67 ( C and E ) . Levels of apoptosis ( B ) and proliferation in normal ducts ( D ) of uninfected mice are shown for comparison . Columns represent the mean , and error bars the SEM . Student’s t test measured p values . n = 5 mice . ( F ) Schematic Model . Breast cells with oncogenic activation ( red ) progress to cancer slowly due to the apoptosis anticancer barrier . However , with a pregnancy , these preexisting precancerous cells activate PRLR-Jak2-STAT5 signaling ( becoming pink ) , and maintain the activated state of this pathway even at involution likely through oncoprotein-initiated phosphorylation and inactivation of GSK3β . pSTAT5 overcomes both the apoptosis anticancer barrier and the apoptotic force unleashed by involution , consequently accelerating progression to malignancy . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 03110 . 7554/eLife . 00996 . 032Figure 8—figure supplement 1 . STAT5 is activated to similar levels by pregnancy and lactation in mammary glands of both young and older mice . Western blotting ( A ) and densitometric quantification ( B ) of both pSTAT5 and total STAT5a protein levels in mammary glands at lactation day 12 from mice at16 weeks ( O ) or 6 weeks ( Y ) of age . Columns represent mean and errors bars the SEM . Student's t test determined p values . DOI: http://dx . doi . org/10 . 7554/eLife . 00996 . 032 Furthermore , we found that the mammary cells in these involuted glands responded to caErbB2 with a lower proliferation rate than the cells in virgin mice ( Figure 8C–E ) , demonstrating that the slowly proliferating mammary cells of involuted mammary glands are also indolent to aberrant proliferation initiated by a future oncogenic mutation , as previously suggested ( Guzman et al . , 1999; Medina et al . , 2001 ) .
The results of this study demonstrate that with a pregnancy , preexisting precancerous cells in the breast subvert a normally tightly-controlled STAT5 signaling pathway to benefit their survival and consequently progress more rapidly to cancer ( Figure 8F ) . In contrast , if mammary cells activate oncogenic signaling following the completion of a pregnancy and lactation , they acquire no survival advantage over cells that activate oncogenic signaling in nulliparous mice ( Figure 8B ) . Moreover , these cells become indolent to oncogene-induced aberrant proliferation ( Figure 8E ) . These results , when combined with our data on pregnancy’s instigation of preexisting precancerous cells , suggest the following overall hypothesis: pregnancy causes mammary cells that have not mutated to become resistant to transformation by a future oncogenic insult , but instigates whichever mammary cells have already mutated to progress to cancer . Therefore , the age-dependent effect of pregnancy on breast cancer risk in women may be due to age-associated gradual accumulation of mutated cells and early lesions which are stimulated by a delayed pregnancy to progress to cancer . Envisioning an application in preventing breast cancer in women having a late age pregnancy who may have already accumulated precancerous cells , we prophylactically treated fully involuted , early lesion-bearing mice with several small molecule inhibitors targeting Jak2 or STAT5 activity . This brief treatment induced apoptosis in early lesions and caused them to regress , and importantly led to significant delay in tumor appearance . Although all mice eventually developed tumors , this significant delay in age of cancer diagnosis is what matters clinically in cancer prevention . Furthermore , our data predict that intermittent treatment with anti-Jak2/STAT5 may inhibit newly formed early lesions and further delay tumor appearance or may even completely prevent cancer in some of them . pSTAT5 is detected in human breast early lesions , especially in women who have had a pregnancy; therefore , prophylactic therapy targeting Jak2-STAT5 signaling may lower breast cancer risk in high risk women . This potential clinical implication of our study takes on additional significance because there has been little further success in developing breast cancer chemoprevention since the introduction of inhibitors of estrogen signaling ( Hutchinson , 2011; Lyons et al . , 2011 ) . Even for these inhibitors to be effective , women have to take them continuously for several years , sometimes with prolonged side-effects ( Hutchinson , 2011 ) . The alternative prevention modality indicated by our results may require only a short-term treatment after pregnancy and lactation; thus , it may be more acceptable to women who are as yet breast cancer-free . Additionally , unlike current prevention strategies ( Bode and Dong , 2009 ) , this alternative regimen may prevent ER-positive cancer as well as other subtypes: ER is negative in 71% of caErbB2-initiated tumors and in 63% of Wnt1-induced tumors in our parous mice although ER is present in the great majority of early lesions induced by either caErbB2 or Wnt1 ( data not shown ) . Therefore , this new chemoprevention strategy may potentially have a significant impact on breast cancer prevention . pSTAT5 levels are low in early lesions from virgin MMTV-tva mice infected by RCAS-caErbb2 , and not surprisingly , brief treatment with either Jak2 or STAT5 inhibitors had little effect in this very small cohort of mice ( Figure 6 ) . However , this result should not be interpreted to conclude that STAT5 inhibition has no value in preventing breast cancer in nulliparous mice or humans . Prophylactic treatment with Jak2/STAT5 inhibitors for a longer time , in a large cohort , or in a subgroup selected for high baseline pSTAT5 may very well yield significant preventive benefit . As shown in Figure 4A , forced STAT5 activation indeed accelerated tumor formation in virgin mice that were infected by RCAS-caErbb2 . Some virgin mice had higher baseline levels of pSTAT5 in the normal glands , and their early lesions harbored more pSTAT5+ cells and fewer apoptotic cells and consequently evolved into cancer faster than the early lesions in the virgin mice with lower baseline levels of pSTAT5 ( Figure 4B ) . In addition , in any virgin mouse , some of the early lesions had relatively high percentage of pSTAT5 cells , and they correspondingly showed reduced levels of apoptosis ( Figure 4—figure supplement 2 ) . These observations suggest that even in nulliparous mice or women , increased pSTAT5 in precancerous cells promotes tumorigenesis , and blocking STAT5 activity may also reduce their risk of breast cancer . In support of this notion , we have found that a 2-week-treatment with AG490 caused RCAS-caErbb2 early lesions in virgin WAP-tva mice to deactivate STAT5 ( Figure 6—figure supplement 1C ) and to induce apoptosis ( Figure 6—figure supplement 1D ) . Therefore , pSTAT5 promotes early lesion progression to cancer in both nulliparous and parous mice , and chemoprevention targeting Jak2/pSTAT5 may lower the risk of breast cancer in both nulliparous and parous women . The majority of human premalignant lesions ( such as ADH ) probably do not progress and do not pose a significant risk , and only 20% of patients with these premalignant lesions in their breasts are eventually diagnosed with breast cancer ( Degnim et al . , 2007 ) . Our data suggest that pSTAT5 in premalignant breast lesions in women may predict a higher chance of progression to clinical cancer , and therefore may help differentiate high risk vs low risk patients and may affect treatment and prevention options . Moreover , increased pSTAT5 even in histopathologically normal breast epithelia may predict a high risk of breast cancer: as discussed in the preceding paragraph , increased baseline levels of pSTAT5 in normal mammary epithelia in nulliparous mice were correlated with higher pSTAT5 levels and lower apoptosis in early lesions of nulliparous mice and predicted an accelerated evolution to overt cancer; pSTAT5 levels were also higher in histologically normal mammary epithelium in MMTV-Erbb2 mice; and importantly , pSTAT5 was elevated in histologically normal breast epithelia adjacent to DCIS with increased pSTAT5 , and high pSTAT5 in tumor-adjacent normal appearing breast epithelia predicted shorter time from the last pregnancy to cancer diagnosis . It is also possible that pSTAT5 levels in normal breast epithelia may serve as a biomarker for estimating the efficacy of anti-Jak2/STAT5 and other chemopreventive strategies . Observation of a larger drop in baseline pSTAT5 levels in breast biopsies following prophylactic chemotherapeutic intervention might indicate a more effective elimination of the threat of potential malignancy . In conclusion , the results presented in this study identify the stimulatory effect of pregnancy-associated STAT5 activation on cancer initiation in the parous breast , delineating novel potential preventative strategies to combat the increased risk of breast cancer faced by women who have a late-age pregnancy . Further , these data suggest a potential reason for pregnancy’s age-dependent effect on breast cancer risk , based on the time at which an oncogenic insult occurs relative to the time at which pregnancy commences , thereby shedding some light on the dual-role played by pregnancy in human breast cancer .
MMTV-tva ( MA ) and WAP-tva ( WA ) mice in a FVB genetic background used in these studies have been previously reported ( Du et al . , 2006; Haricharan et al . , 2013 ) . All animals were euthanized according to the NIH guidelines . The animal protocol was approved by the IACUC of Baylor College of Medicine , Houston , TX . RCAS virus was prepared as described earlier ( Du et al . , 2006 ) . Briefly , the retrovirus was transfected into DF1 chicken fibroblast cells using Superfect ( Qiagen , Gaithersburg , MD ) and viral particles were harvested over 1 week . The viral particles were concentrated by ultracentrifugation and frozen for titration and intra-ductal injection . Titration was carried out through limited dilution transduction of DF1 cells . The lentiviral vector ( FUCGW ) carrying either GFP alone or both GFP and caErbb2 was prepared as described earlier ( Bu et al . , 2009 ) . 4 to 7 days after intraductal injection of RCAS , the experimental group consisting of approximately half the mice was mated . Pups were weaned at lactation day 21 . All mice were palpated thrice weekly for tumor incidence . Tumor-free mice were euthanized 7 months ( caErbB2/caSTAT5a ) , 12 months ( caErbB2 ) , or 24 months ( Wnt-1 ) post injection . Mammary glands analyzed for the incidence of early lesions were removed from the experimental and control animals either 1 . 5 weeks ( P7 . 5 ) , 2 weeks ( caSTAT5a/caErbB2 ) , 5 . 5 weeks ( L12 ) , 8 . 5 weeks ( I10 ) , or 6 months ( Wnt-1 ) post injection . A small subset of parous animals ( <20% ) had palpable tumors at the time of weaning . To ensure that the presence of a tumor did not affect pregnancy , lactation , or involution in these mice , we monitored average litter size at the time of weaning and ensured they were comparable between mice that bore tumors at involution day 2 ( n = 9 ) and mice that did not ( n = 49 ) . AG490 was administered intraperitoneally based on previous in vivo studies ( Burdelya et al . , 2002 ) . C188-9 was administered at the maximum tolerated daily dose ( 100 mg/kg BW ) based on drug toxicity studies carried out in mice by Dr David J Tweardy . C188-9 was delivered intraperitoneally using Becton–Dickinson LoDose ½ cc U100 insulin syringes . Both drugs were freshly diluted in DMSO before each administration . Ruxolitinib was administered intraperitoneally once a day for 10 days at 100 mg/kg based on a modified protocol from previous in vivo studies ( Quintas-Cardama et al . , 2010 ) . Tumors and mammary tissue were fixed in 4% paraformaldehyde overnight at 4°C , paraffin-embedded , and sliced into 3-μm sections . The sections were deparaffinized in xylene , rehydrated in graded alcohol , and used for histology and immunostaining . The tumor tissue for Western blotting analysis was snap-frozen in liquid nitrogen and stored at −80°C . #4 mammary glands were whole-mounted and stained with neutral red as previously described ( Moraes et al . , 2007 ) . Whole-mounted tissue was analyzed for lesions and gland morphology under xylene and then paraffin-embedded and sectioned for immunostaining . Immunohistochemistry ( IHC ) and immunofluorescence ( IF ) were performed as described earlier ( Du et al . , 2006 ) . Antigen retrieval was carried out by heating sections in 10 mM sodium citrate , pH6 . 0 . MOM and VectaStain Elite ABC Rabbit kits ( cat . no . PK-2200 & PK-6101; Vector Labs ) were used according to manufacturer’s protocols . Primary antibodies used included mouse monoclonal antibody against HA ( 1:500; Covance ) , BclXL ( 1:40; Santa Cruz ) , FLAG ( 1:500; Sigma ) , and vWF ( 1:200; DAKO ) ; rabbit IgG specific for pSTAT5 ( which recognizes both pSTAT5a and pSTAT5b ) ( 1:200; Cell Signaling ) , cleaved caspase 3 ( 1:200; Cell Signaling ) , PRLR ( 1:500; Santa Cruz ) , pSTAT3 ( 1:200; Cell Signaling ) , STAT5a ( 1:200; Santa Cruz ) , pSTAT1 ( 1:200; Cell Signaling ) , pHistone3 ( 1:200; Millipore ) , and Ki67 ( 1:200; Novocastra ) ; and goat IgG specific for pSTAT6 ( 1:50; Santa Cruz ) . Incubation with the primary antibody for IF staining was overnight at 4°C , while incubation with primary antibody for IHC was 1 hr at room temperature . Nuclei were counterstained with 4′-6-diamidino-2-phenylindole ( DAPI ) -containing mounting medium and hematoxylin , respectively , for IF and IHC . IHC for pSTAT5 in all human samples was controlled for quality of fixation and paraffin embedding of tissues by scoring the quality of pHistone3 staining of all tissues that were negative for pSTAT5 . Only those samples that passed this quality control test were included in the final analysis . Bright field images were captured using a Leica DMLB microscope , and images were processed with Magnavision and Adobe Photoshop software . Fluorescent images were captured with a Zeiss Axiskop2 plus microscope . Images were processed with Axiovision and Adobe Photoshop software . Paraffin-embedded gland and tumor sections were treated in proteinase K and subjected to the terminal deoxynucleotidyl transferase dUTP nick-end labeling ( TUNEL ) assay using the ApopTag Red in situ TUNEL detection kit ( Chemicon , S7165 ) . Nuclei were counterstained with DAPI-containing mounting medium . For tumor samples , five random fields were viewed and a total of at least 5000 cells were counted per sample . For quantification of apoptotic , proliferative , pSTAT5+ , PRLR+ , and p53+ cells in early lesions , at least 2000 cells were counted per sample . For quantification of apoptotic and proliferative cells in normal ducts , at least 10 ducts were counted per section , and a total of at least 1000 cells were counted per sample . When co-staining was impractical , consecutive sections were cut from formalin-fix and paraffin-embedded ( FFPE ) tissue and used for staining . ImageJ software was used for counting , and either DAPI or hematoxylin nuclear staining was used to identify the total number of cells . ImageJ software was also used to quantify the total percentage of HA+ area in the entire mammary gland from sequential bright-field pictures taken at 2 . 5 × resolution . Fixed thresholds were set to analyze both experimental and control mammary glands . Protein was extracted from frozen tissue , and lysates analyzed as previously described ( Li et al . , 2001 ) . Primary antibodies used were mouse monoclonal HA ( 1:1000; Covance ) , BclXL ( 1:500; Santa Cruz ) , cyclinD1 ( Cell Signaling , 1:500 ) , and STAT5b ( Santa Cruz; 1:500 ) ; and rabbit polyclonal Bcl2 ( 1:500; Santa Cruz ) , PRLR ( 1:500; Santa Cruz ) , pSTAT5 ( 1:500; Cell Signaling ) , STAT5a ( 1:500; Santa Cruz ) , and GAPDH ( 1:2000; Santa Cruz ) . Secondary antibodies used were HRP-conjugated goat anti-mouse and anti-rabbit ( 1:5000; Pierce ) . Super-signal Femto-Chemiluminescence substrate ( Thermo Scientific ) was used to visualize bands on Western blots . Only 10/27 women in cohort 1 opted for BRCA testing and all 10 were BRCA negative . The remaining 17 have undetermined BRCA status . For cohort 2 , however , all women were tested for BRCA status , and only 2/24 were found positive . All numbers in the text were represented as mean ± standard error of the mean . Statistical analysis of quantification of stained sections was done using ANOVA or Student’s t test for independent samples with Holm’s correction for multiple comparisons when distribution of data was judged to be normal . For hypothesis confirmation in Figure 6—figure supplement 1B Student’s one-sided t test was used . Where distribution was not normal ( assessed using Q–Q plots with the Wilk-Shapiro test of normality ) , either Kruskal-Wallis or Wilcoxon’s Rank Sum test was used . In all cases , at least 10 lesions were quantified per mouse . Holm’s correction was also used where required when using non-parametric tests . For categorical data with <15 data points in each group , the Fisher’s Exact test was used . For categorical data with ≥15 data points in each group , the Pearson’s chi-square test was used . Tumor-free survival analysis was done using the Generalized Gehan-Wilcoxon test with rho = 1 , and Kaplan-Meier survival curves were generated in R . The regression equation for correlation between pSTAT5 and apoptosis was configured to include lesion size ( total number of cells in the lesion ) along with pSTAT5% as potential predictive factors of apoptosis , and only pertains to lesions from a single mouse ( red ) . All graphs and regression analyses were generated either in MS Excel or R . | Pregnancy changes the probability that a woman will later develop breast cancer . If a woman’s first pregnancy occurs before her 22nd birthday , the chances of developing breast cancer are reduced . However , if the first pregnancy occurs after her 35th birthday , there is an increased risk of breast cancer . It is not clear why this age-related difference exists , but as more women wait until their 30s to start a family , there is greater urgency to understand this difference . Breasts undergo extensive changes during pregnancy . This remodeling makes their cells less likely to multiply , and also less likely to develop tumors , which could explain the protective effect of pregnancy for younger women . But why would older women not reap the same benefits ? One hypothesis is that older first-time mothers are more likely than younger first-time mothers to already have breast tissue with cells carrying cancer-causing mutations , or to have clusters of abnormal precancerous cells . Now , Haricharan et al . have tested this hypothesis by inserting two cancer-causing genes into female mice . Half of the mice were then made pregnant and allowed to nurse their young , whilst the other half were never mated . Although , both groups of mice later developed tumors , the mice that had been pregnant developed more tumors and did so faster . The increased cancer levels in the mice that had been pregnant were not due to them having more precancerous cells at the early stages of pregnancy than the unmated mice of the same age . Further , the precancerous cells in the impregnated mice did not proliferate faster than those in the mice that were never pregnant . Instead , pregnancy weakened the protective process that culls pre-existing precancerous cells . These cells evaded destruction by activating a signaling pathway called the STAT5 pathway in response to pregnancy hormones . Haricharan et al . also examined tissue samples from women with a very early form of breast cancer and found elevated levels of STAT5 in tumors from women who had been pregnant compared to those who had not been pregnant . The good news is that precancerous cells do not always become cancerous . However , for those women with a high risk of developing breast cancer , Haricharan et al . suggest that temporarily reducing STAT5 activity after pregnancy with medication might reduce this risk . Treating mice with anti-STAT5 drugs for a few weeks after they finished nursing their young lessened the elevated cancer risk , and so the next challenge is to see if this approach will also be effective in human clinical trials . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2013 | Mechanism and preclinical prevention of increased breast cancer risk caused by pregnancy |
Many driver mutations in cancer are specific in that they occur at significantly higher rates than – presumably – functionally alternative mutations . For example , V600E in the BRAF hydrophobic activation segment ( AS ) pocket accounts for >95% of all kinase mutations . While many hypotheses tried to explain such significant mutation patterns , conclusive explanations are lacking . Here , we use experimental and in silico structure-energy statistical analyses , to elucidate why the V600E mutation , but no other mutation at this , or any other positions in BRAF’s hydrophobic pocket , is predominant . We find that BRAF mutation frequencies depend on the equilibrium between the destabilization of the hydrophobic pocket , the overall folding energy , the activation of the kinase and the number of bases required to change the corresponding amino acid . Using a random forest classifier , we quantitatively dissected the parameters contributing to BRAF AS cancer frequencies . These findings can be applied to genome-wide association studies and prediction models .
In a time of personalized medicine and tumor sequencing , determining which missense mutations affect disease phenotype and exploring the role of cellular and environmental context are crucial . In many oncogenes , mutations are enriched at specific amino acid positions ( ‘mutation hotspots’ ) , and it is not usually obvious if rare substitutions are passengers or disease-causing mutations . A striking example is the oncogenic serine/threonine kinase BRAF , for which the V600E mutation in its kinase activation segment ( AS ) accounts for >95% of all BRAF cancer mutations . BRAF is a serine/threonine protein kinase that is an upstream regulator of cellular responses such as cell division and differentiation and is mediated by the MEK/ERK signaling pathway ( Garnett and Marais , 2004; Wellbrock et al . , 2004 ) . BRAF kinase is found mutated in both germline diseases ( e . g . cardiofaciocutaneous and Noonan syndromes; ( Rauen , 2013 ) and somatic cancers of the thyroid , skin , colon , and lung ( Holderfield et al . , 2014; Ascierto et al . , 2012 ) . BRAF contains an N-terminal region with a Ras-binding domain , which is followed by a cysteine-rich motif and a C-terminal kinase domain . BRAF is autoinhibited in a closed conformation by the interaction of the N-terminal conserved region 2 ( following the Ras-binding and cysteine-rich domains ) with the kinase domain , mediated by the interaction of two phosphorylated residues , Ser365 and Ser729 , with a 14-3-3 dimer ( Figure 1—figure supplement 1 ) ( Brummer et al . , 2006 ) . Upon dephosphorylation of the N-terminal phosphorylated Ser365 by phosphatase PPII , the Ras-binding domain is free to interact with Ras at the plasma membrane . This releases autoinhibition and enables either homodimerization or heterodimerization with CRAF , ARAF , or KSR17; subsequent phosphorylation in the AS at Thr599 and Ser602 results in kinase activation ( Taylor and Kornev , 2011; Hmitou et al . , 2007; Zhang and Guan , 2000; for a recent review on the topic see Lavoie and Therrien , 2015 ) . Similar to other kinases , the BRAF kinase domain has two subdomains comprising a small N-terminal lobe and a large C-terminal lobe ( Figure 1A ) ( Scheeff and Bourne , 2005; Roskoski , 2010 ) . The N-terminal lobe contains the nucleotide-binding pocket and the phosphate-binding loop , while the C-terminal lobe binds the protein substrates and contains the catalytic loop . The two lobes , which are spatially connected through the AS , can move relative to each other in order to open or close the cleft . AS residues undergo hydrophobic interactions with the phosphate-binding loop and the ‘αC helix’ of the N-terminal lobe ( making the ‘hydrophobic pocket’ ) , locking the kinase in its inactive state . In addition , the misalignment of spatially conserved hydrophobic residues in the N- and C terminal lobes ( ‘hydrophobic spines’ ) prevents catalytic activation ( Lavoie and Therrien , 2015; Hu et al . , 2015 ) . Phosphorylation within the AS causes structural rearrangements of the AS , the αC helix and the phosphate-binding loop , reorienting the catalytic Asp of the DFG motif in a catalysis-competent orientation , thereby causing BRAF to become active . 10 . 7554/eLife . 12814 . 003Figure 1 . Overall structure of the kinase domain of BRAF , zoom into the hydrophobic pocket of BRAF , and active- and inactive-like BRAF kinase domain 3D structures used for structure-energy calculation . ( A ) Structure of the BRAF kinase , with functional regions indicated . The BRAF kinase domain has two subdomains , a small N-terminal lobe and a large C-terminal lobe . The small lobe contains the nucleotide-binding pocket and the phosphate-binding loop , while the large lobe binds the proteins substrates and contains the catalytic loop . The two lobes are spatially connected through the activation segment ( AS ) of the large lobe . Sequentially , the N- and C-terminal lobes are connected by the hinge , and the AS is part of the C-lobe that interacts with the N-lobe . Movement of the two lobes relative to each other opens and closes the cleft . ( B ) The hydrophobic pocket around amino acid Val600 represented using the backbone and side chain view . Backbone residues are colored according to their location in the protein ( see Figure 1A ) . Specifically , Leu597 , Ala598 , Val600 , and Trp604 of the AS together with , Phe468 , Leu525 , Leu485 , Val487 , Phe498 , and Ala497 of the N-terminal subdomain build the hydrophobic pocket . All BRAF structural representations were done with SwissPdbViewer , using PDB entry 4EHE ( chain B of the crystallographic unit ) . ( C ) Superimposition of active-like BRAF kinase structures . The structural representations were made using SwissPdbViewer ( PDB entries 4MNE , 3OG7 and 4MNF ) . ( D ) Superimposition of inactive-like BRAF kinase structures . Structural representations were made using SwissPdbViewer ( PDB entries 4EHE and 3TV6 ) . ( E ) Pairwise correlation of FoldX energies for mutations in the hydrophobic pocket derived from active and inactive structures . Similar correlation results were obtained from FoldX energies using a recently published 3D structure of inactive monomeric BRAF ( Thevakumaran et al . ( 2015 ) ; PDB entry 4WO5 , which is missing four residues in the AS/ data not shown ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 00310 . 7554/eLife . 12814 . 004Figure 1—figure supplement 1 . BRAF activation cycle . The 14-3-3 dimer binds to Ser365 at the N-terminus and to Ser729 at the C-terminus , maintaining the kinase in a closed , inactive conformation . Dephosphorylation of the N-terminal phospho-Ser365 by phosphatase PPII sets free the Ras-binding domain to interact with Ras at the plasma membrane , which releases autoinhibition and enables either homodimerization or heterodimerization with CRAF , ARAF , or KSR1 . Heterodimerization is favored by 14-3-3 dimer binding to the C-terminal , phosphorylated Ser729 of BRAF . The wild-type activation cycle is depicted with the N-terminal kinase lobe in grey and the C-terminal one in blue . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 00410 . 7554/eLife . 12814 . 005Figure 1—figure supplement 2 . Cancer mutation frequencies in the hydrophobic pocket of BRAF . Each position in the hydrophobic pocket region is shown ( rows ) and mutation frequencies for the respective mutations ( columns ) . The mutation frequencies are colored according to the absolute number ( yellow: >0 to <10; orange: 10 to <100; red: 100 to <1000; pink: V600E , > 20 . 000 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 00510 . 7554/eLife . 12814 . 006Figure 1—figure supplement 3 . Basic principles of the FoldX force field , FoldX-based modeling , and the application of structure-energy calculations on mutations in BRAF’s hydrophobic pocket . ( A ) Basic scheme of a folded and unfolded kinase and the associated folding energy ( △G ) . ( B ) List of intramolecular forces contributing and opposing folding , which is integrated into the FoldX force field ( see energy function in the 'Materials and methods' section ) . ( C ) Example of amino acid side chain mutations performed using FoldX . Different rotamers are shown for the mutation of Val600 to Phe , as well as for the movement of the neighboring residue , Trp604 . ( D ) Example of change in folding energy when comparing WT to Mutant [△△G ( BRAF Mutant-WT ) ] . ( E ) Pipeline of FoldX-based modeling of BRAF WT and mutants in the hydrophobic pocket . ( F ) Interpretation of FoldX energies and additional analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 006 There are two main hotspot regions for cancer-causing mutations in BRAF . Mutations in the phosphate-binding loop ( residues 464 to 472 ) correspond to <1% of all BRAF mutations in cancer . The more important hotspot is found in the AS , with V600E being the most frequent BRAF somatic cancer mutation ( 98% in the COSMIC database ) ( Supplementary file 1; Figure 1—figure supplement 2 ) ( Cantwell-Dorris et al . , 2011; Sarkozy et al . , 2009; Holderfield et al . , 2014; Lavoie and Therrien , 2015 ) . Less frequently found mutations at position Val600 are mutations to Asp , Lys , and Arg , which all require two nucleotide substitutions ( Davies et al . , 2002; Lavoie and Therrien , 2015 ) . In the inactive conformation , Val600 is buried in a hydrophobic pocket made by residues from the N-terminal subdomain ( Ala497 , Phe498 , Leu525 , Leu485 , Phe468 , and Val487 ) and the AS ( Leu597 , Ala598 , and Trp604 ) ( Figure 1B ) . Substitution of this residue by charged amino acids ( e . g . Glu ) disrupts these interactions and results in constitutive kinase activation ( Wan et al . , 2004 ) . BRAF V600E does not require RAF dimerization or interaction with Ras to be active ( Poulikakos et al . , 2011 ) yet has an increased propensity to form dimers ( Freeman et al . , 2013; Roring et al . , 2012; Thevakumaran et al . , 2015 ) . Whereas extensive research on BRAF in past years has provided enormous insight and understanding about the regulation of BRAF kinase and the abnormal activity of V600E ( Lavoie and Therrien , 2015 ) no studies exist explaining why other amino acid substitutions in the hydrophobic pocket are not found with a high frequency in cancer . In principle , other mutations at the AS ( such as Leu597 mutated into Glu ) , or in other parts of the hydrophobic pocket ( e . g V487 into Glu or Leu525 into Glu ) , should also release the AS and cause constitutive kinase activation . Thus , to answer this question , we performed combined structure-energy , experimental and statistical analyses of mutations in the hydrophobic pocket . We show that V600E is the only single nucleotide substitution ( Asp , Lys , and Arg , require two bases substitutions ) that opens the AS through destabilization of autoinhibitory interactions , without significantly impairing the folding of the inactive or active kinase domain . We show that other mutations requiring three base substitutions ( i . e . V600H ) have kinase activities similar to V600E . We provide a quantitative measure for all parameters that contribute to BRAF cancer mutation frequencies by evaluating their importance using a random forest classifier . We anticipate that our results can be translated to other kinases and disease-causing proteins , provided that high-resolution X-ray structures are available .
Previous work on BRAF has shown that the V600E mutation is frequently found in cancer because it causes a disruption to the surrounding hydrophobic environment ( Wan et al . , 2004 ) . To recapitulate what is already known in the literature and to have a quantitative measure for the destabilization of the hydrophobic pocket introduced by the V600E mutation , we used structure-based energy calculations . The protein design algorithm FoldX provides a quantitative estimation of the intermolecular forces and interactions contributing to the stability of proteins ( △G = folding energy ) based on high-resolution X-ray structures ( Figure 1—figure supplement 3A–B ) ( Guerois et al . , 2002; Schymkowitz et al . , 2005; Van Durme et al . , 2011 ) . FoldX also enables amino acid replacements through side-chain rotamer modeling , allowing one to evaluate the energetic impact of a disease mutation on protein and/or complex stability ( Figure 1—figure supplement 3C–D ) ( Alibes et al . , 2010; Pey et al . , 2007; Rakoczy et al . , 2011; Kiel and Serrano , 2014 ) . We performed FoldX-based molecular modeling of amino acid substitutions in the hydrophobic pocket of BRAF using active-like ( 4MNE ( Haling et al . , 2014 ) , 3OG7 ( Bollag et al . , 2010 ) and the V600E mutant 4MNF ( Haling et al . , 2014 ) ) and inactive-like ( 4EHE ( Mathieu et al . , 2012 ) and 3TV6 ( Wenglowsky et al . , 2011 ) ) BRAF ‘template’ structures ( Figure 1C–D; Figure 1—figure supplement 3E; Supplementary file 1 ) . Using FoldX , we mutated every amino acid residue in the hydrophobic pocket of the five selected active and inactive structures to all amino acids , including itself ( Figure 1—figure supplement 3E ) . This resulted in a total of 5 x 280 = 1400 structural models , and the change in folding energy ( △△G BRAF Mutant-WT ) was determined ( Supplementary file 1 ) . Pairwise correlations of energies derived from active structures or inactive structures , respectively , show a good overall correlation ( Figure 1E ) . In contrast , poor correlations were found when comparing energies from active and inactive structures , supporting the classification of the template structures . All structural models with a change in FoldX energy ( △△G BRAF Mutant – WT ) > 0 . 8 kcal were considered , as destabilizing mutants as this energy corresponds to a value twice the standard deviation of the energies calculated using the FoldX force field . To interpret the changes in FoldX energies , we needed to take into account several considerations ( Figure 1—figure supplement 3F ) . First , mutations that destabilize the inactive conformation ( △△G BRAF_inactive ) will drive the protein into a complex with chaperones ( i . e . HSP90; Grbovic et al . , 2006 ) and/or aggregation/degradation , thereby decreasing its overall effective concentration . Second , mutations that destabilize the active conformation ( △△G BRAF_active ) will also result in the protein having a decreased effective concentration ( unless they favor heterodimer formation and cause paradoxical activation ( Heidorn et al . , 2010; Poulikakos et al . , 2010 ) ) . Third , unfavorable energy changes in the AS loop of the inactive structures will favor its release and therefore kinase activation ( △△G BRAF_inactive_loop ) . Structural inspection of 28 BRAF structures with different inhibitors showed that the AS loop between Leu597 and Gly615 is moderately to highly flexible ( high B-factors ) and consequently is unsolved in many structures ( Figure 2; Supplementary file 1 ) . Position Val600 is moderately flexible ( 70% solved in X-ray structures ) . This confirms previous predictions that the AS loop belongs to a region within the kinase domain ( intra domain region ) that has a large tendency to be disordered ( Lu et al . , 2015 ) . Also , previous enhanced-sampling structure-based computational simulations proposed that the AS exhibited a significant tendency to switch from the ordered to unstructured conformation ( Marino et al . , 2015 ) . Mutations in regions of high flexibility will have less impact on the unfolding of BRAF compared to those in conformational restricted regions . Thus , for the inactive state , we corrected the folding energies of the mutations in the AS loop ( △△G BRAF_inactive ) by the frequency for which the corresponding position is solved in the 28 crystal structures . This correction was not applied to the active-like structures because for these three active structures residues were solved only until position 600 . The only available structure for which the loop had been solved ( 4MNE ) , had a high B-factor from position 601 onwards , but as there was no significant destabilization seen by FoldX , no correction was applied . After applying all these factors , we found several mutations that release the AS ( FoldX energies above the threshold of 0 . 8 kcal/mol ) and therefore could activate the kinase ( Supplementary file 1 ) . 10 . 7554/eLife . 12814 . 007Figure 2 . AS loop residues in 28 BRAF kinase structures and comparison with B-factors . ( A ) Percentage of the 28 BRAF X-ray structures that have a given AS residue solved . ( B ) Percentage of presence of AS loop amino acids in the X-ray structures , mapped onto a BRAF ribbon diagram ( see legend for the color code ) . The structural representation was made using SwissPdbViewer ( PDB entry 4EHE ) . ( C ) Normalized B-factor averages for loop residues from inactive structures ( PDB entries 4EHE and 3TV6 ) plotted against the percentage of presence in the 28 BRAF X-ray structures . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 007 The overall energy changes ( △△G BRAF_inactive ) as well as the energy changes in the AS loop alone ( △△G BRAF_inactive_loop ) that result from the introduction of mutations in the inactive structure , have very poor correlations with the occurrence of the corresponding mutation in tumors ( Figure 3A; Figure 3—figure supplement 1; Figure 3—figure supplement 2 ) . We suggest the following reasons for this:10 . 7554/eLife . 12814 . 008Figure 3 . Structure-energy predictions and experimental analysis of mutations in the hydrophobic pocket of BRAF . ( A ) Comparison of the number of cancer mutations ( >0 ) with destabilization of the hydrophobic pocket as predicted by FoldX ( average energy values of 1EHE and 3TV6 , ‘FoldX △△G BRAF_inactive_loop’ ) . ( B ) Representative Western blot ( upper panel ) for selected Val600 mutations expressed 24 hr in normal medium and quantified using ImageJ ( lower panel ) . Two out of at least six biological replicates are shown . Bar graph shows the results of six biological replicates for the abundance of MEK-P normalized to total BRAF . ( C ) Representative Western blot ( upper panel ) analysis for selected single and triple nucleotide substitution BRAF mutations expressed 24 hr in normal medium and quantified using ImageJ ( lower panel ) . Two out of at four biological replicates are shown . Bar graphs show the results of two biological and two technical replicates for the abundance of MEK-P normalized to total BRAF . ( D ) Correlation of FoldX energies with MEK phosphorylation normalized by the total BRAF levels . FoldX energies were calculated from the inactive loop energy [BRAF_inactive_loop] minus the FoldX energies derided from active structures [BRAF_active] plus the hydrophobic solvation energy as a factor in the FoldX force field [BRAF hydr_solv_energy] . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 00810 . 7554/eLife . 12814 . 009Figure 3—figure supplement 1 . Mutations causing destabilization of the inactive loop and comparison with cancer frequencies . ( A ) Mutations having destabilization of the inactive loop above the threshold ( ‘loop energy’ ) , sorted by decreasing energy value . Colors indicate the number of cancer mutations . ( B ) Comparison of number of cancer mutations ( >0 ) with destabilization of the hydrophobic pocket as predicted by FoldX ( average energy values of 1EHE and 3TV6 , normalized by its presence in the structure/B-factor ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 00910 . 7554/eLife . 12814 . 010Figure 3—figure supplement 2 . Mutations causing destabilization of the inactive structure above the threshold . Colors indicate the number of cancer mutations . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 01010 . 7554/eLife . 12814 . 011Figure 3—figure supplement 3 . Additional Western blots supporting Figure 3B . Western blots of biological replicates of BRAF WT , V600E , V600D , V600K , V600M , V600A , V600G , and V600W used for the quantifications shown in the bar diagram of Figure 3B . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 01110 . 7554/eLife . 12814 . 012Figure 3—figure supplement 4 . Additional Western blots supporting Figure 3C . ( A ) Western blots of technical replicates of BRAF WT , V600E , V600H , and L597Y used for the quantifications shown in the bar diagram of Figure 3C . ( B ) Additional biological replicates for all BRAF mutants studied in this work . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 012 To see if these factors are responsible for the poor correlation between FoldX predicted energy changes and cancer frequency , we did a series of experiments and analyses described below . We transiently expressed wild-type or mutant BRAF in HEK293 cells in normal growth medium and analyzed BRAF expression and the phosphorylation state of BRAF and MEK ( Figure 3B; Figure 3—figure supplement 3 ) . After correcting for differences in BRAF expression levels , we found that BRAF V600E phosphorylated MEK at higher levels than wild-type BRAF , and as predicted by FoldX , at similar levels to the double-nucleotide substitutions of V600D and V600K . In contrast , BRAF V600M and V600A yielded wild-type levels of MEK phosphorylation , suggesting that these are in fact passenger mutations ( Figure 3B ) . The remaining mutations gave intermediate MEK phosphorylation levels . Indeed , the V600G mutation , which is also found in the germline and causes CFC syndrome ( Champion et al . , 2011 ) is an intermediate MEK activity mutant . The fact that we found V600K and V600D mutants to be as active as the V600E mutant supports the hypothesis previously published ( Davies et al . , 2002 ) that the lower frequency of these mutants in cancer must be is due to the fact that two base substitutions are needed for changing Val600 into Lys or Asp , whereas only one is needed for V600E . We confirmed this further by identifying mutations that were not found in cancer ( at positions 597 and 600 in the AS ) , that required three base changes , and that were predicted to be as activating as the most frequent cancer mutation found at these positions . Expression of these mutants ( L597Y and V600H ) in cells resulted in medium and high kinase activity as predicted ( Figure 3C; Figure 3—figure supplement 4 ) . Replacement of V600 by bulkier hydrophobic residues ( e . g . Met , Leu , Trp ) resulted in weak ( V600W ) or no kinase activation . V600W , despite having a very high destabilizing AS loop FoldX energy in the inactive orientation , had a similar activity to that of V600G ( Figure 3B ) . This supports our hypothesis that structural movements in the flexible AS could partially accommodate bulkier hydrophobic residues in the inactive orientation . Thus , we included the chemical nature/hydrophobicity as another factor . Considering the energies and parameters discussed above , we observed an excellent correlation between the FoldX predictions and MEK phosphorylation normalized by total BRAF ( Figure 3D ) . Based on the data above , we can explain why V600E is the most frequent cancer mutation at position 600 . We next wanted to analyze why no other mutation in the hydrophobic pocket - in a different position to Val600 - is found frequently mutated in cancer . Based on FoldX structure-energy calculations , we predicted that mutations in the hydrophobic pocket that destabilize the pocket and may thereby release the AS , would also affect the folding of the inactive and/or active kinase , thereby reducing the effective concentration and thus resulting in lower MEK phosphorylation ( Supplementary file 1 ) . We experimentally tested three mutations ( V487E , L525E and F498S ) that required one , or two ( L525E ) base changes ( Figure 4A–B ) . F498S is predicted to be the most destabilizing , followed by L525E and V487E . By analyzing the soluble and insoluble fractions from transiently transfected HEK293 cells , we determined that the ratios between soluble and insoluble BRAF were similar for wild-type and the V600E and V600W mutants ( Figure 4C , E; Figure 4—figure supplement 1A ) , while the V487E and L252E mutants resulted in significantly more insoluble protein . It is important to note that due to the very low levels of the BRAF F498S protein , it could only be detected when loaded in a 10-fold excess of lysate compared to wild-type ( with no separation into soluble and insoluble fractions; Figure 4D; Figure 4—figure supplement 1B ) . Comparing the ratios of BRAF expressed in the soluble and insoluble fractions ( Figure 4E ) shows an inverse correlation with the folding energies as predicted by FoldX ( Figure 4B; R2∼0 . 67 , assuming for F498S a ratio of BRAF soluble/insoluble >0 and <0 . 4 , ∼0 . 2; Figure 4—figure supplement 2 ) . However , despite the low levels of soluble F498S protein , it phosphorylated MEK at approximately the same level as wild-type BRAF ( Figure 4D ) , while the V487E and L525E mutants , after normalizing by the total soluble protein gave higher MEK phosphorylation levels than wild-type ( Figure 4—figure supplement 3 ) . 10 . 7554/eLife . 12814 . 013Figure 4 . Structure-energy predictions and experimental analysis of mutations affecting the folding of BRAF and analysis of phosphorylation of Thr599 and Ser602 to keep the AS in a fixed active state . ( A ) Structural representations of the localization of Val487 , Leu525 , and Phe498 in BRAF ( PDB entry 4EHE ) . ( B ) Destabilization of inactive and active states for V487E , L525E , and F498S BRAF ( folding mutants ) as predicted by FoldX . ( C ) Western blot analysis for BRAF mutations affecting folding . ( D ) Western blot analysis for BRAF F498S folding mutations . ( E ) Plot of BRAF soluble to insoluble ratios for the WT and mutations shown in the Western blots from pane ( C ) and ( D ) , sorted in a similar order as in Figure 3B . Bar graphs show the results from two biological replicates . The soluble/insoluble value for BRAF F498S was estimated ( see main text and represented with a star ) . ( F ) Illustration of the salt bridges that are proposed to stabilize the active conformation . The structural representation was done with the SwissPdbViewer , using PDB entry 4MNE . ( G ) Western blot analysis for the selected V600E and V600K mutations in combination with the T599A/S602A mutations expressed 24h in normal medium . ( H ) Quantifications of MEK phosphorylation levels normalized by total BRAF from ( G ) using ImageJ . Bars represent at least four biological replicates for the abundance of MEK-P normalized to total BRAF . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 01310 . 7554/eLife . 12814 . 014Figure 4—figure supplement 1 . Original western blots of spliced out lanes shown in Figure 4C and D . ( A ) Original Western blot and experimental procedure supporting Figure 4C . Supernatant ( ‘SUP’ ) or pellet ( ‘PELLET’ ) fractions of BRAF WT and mutants V600E , V487E , V600W , and L525E were each run on one SDS gel ( five gels in total ) . Each gel was spliced into parts , one with the expected size of BRAF and one with the expected size of actin . All gel pieces for BRAF were placed on the iBlot ( Invitrogen ) Western blot membrane and transferred simultaneously . We proceeded similarly for the five actin gel pieces . As such , BRAF WT and mutants were treated similarly during gel transfer and Western blot antibody incubation , washing , and subsequent ECL development . ( B ) Original Western blot supporting Figure 4D . A 10-fold higher amount of lysate was loaded for BRAF mutant F498S compared to WT . WT , wild type . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 01410 . 7554/eLife . 12814 . 015Figure 4—figure supplement 2 . Comparing experimental protein solubility with FoldX predicted folding energies . Comparing the ratios of BRAF expressed in the soluble and insoluble fractions with the FoldX folding energies . The correlation coefficient is 0 . 67 . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 01510 . 7554/eLife . 12814 . 016Figure 4—figure supplement 3 . MEK phosphorylation of wild-type and V600E , V487E , and L525E mutant BRAF in the supernatant . ( A ) Western blot analysis of MEK-phosphorylation after expression for 24 hr in normal medium and ImageJ quantification ( using two biological replicates ) . ( B ) MEK-phosphorylation levels normalized by the soluble fraction of BRAF ( as shown in Figure 4C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 01610 . 7554/eLife . 12814 . 017Figure 4—figure supplement 4 . Conformations of Lys601 found in all structures having position 601 solved , and an overlay of ten active-like BRAF structures . ( A ) Close-up of the 4MNE structure indicating the salt bridge between Arg575 and Glu611 that is conserved in all Raf kinases . Lys601 points in the direction of this salt bridge . ( B ) Superimposition of all BRAF kinases that are solved at K601 . For 4MNE , the ribbon representation is shown in grey , and the residues are as in panel ( A ) . For the remaining structures , only the backbone and side chain of K601 is shown . The structural representations were made using SwissPdbViewer ( PDB entries 4MNE , 1UWJ , 3TV4 , 4E4X , 4EHE , 4G9R , 4PP7 , 4JVG , 4MBJ , 4H58 , 3Q4C , 4E26 , and 4G9C ) . ( C ) Overlay of 10 active-like BRAF structures . Residues 597–600 ( red ) are very similar between the different structures , while resides 601–616 ( blue ) are more flexible . The structural representation was made using SwissPdbViewer ( PDB entries 2FB8 , 3D4Q , 3OG7 , 3PPJ , 3PPK , 3PRI , 3PSB , 4H58 , 4MNE , and 4MNF ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 01710 . 7554/eLife . 12814 . 018Figure 4—figure supplement 5 . Biological replicates in minimal ( serum-free ) growth medium . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 01810 . 7554/eLife . 12814 . 019Figure 4—figure supplement 6 . Analysis of the interactions in the BRAF RD motif , and expression levels of BRAF wild-type and the single V600E , E611A , and double V600E/E6111A mutants . ( A ) Illustration of salt bridges that are proposed to stabilize the active conformation . The structural representation was done with the SwissPdbViewer , using PDB entry 4MNE . The bottom panel shows the salt bridge between Glu611 and Arg575 in the active conformation and the proposed participation of Lys601 when mutated to Glu . ( B ) Western blot analysis for wild-type and mutant BRAF expressed for 24 hr in normal medium . ( C ) Western blot analysis for wild-type and mutant BRAF expressed for 24 hr in minimal medium ( no serum ) with or without 5 min of EGF stimulation ( at 50 ng/ml ) before harvesting and lysis . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 019 Following the above analysis , we suggest that mutations that slightly destabilize both the folded conformation and the AS may cause small changes in ERK phosphorylation , which do not lead to cancer but may cause developmental defects . Indeed , three conservative RASopathy mutations are found in this region ( L485F , L485S , and V487G; Rauen , 2013 ) ( Supplementary file 1 ) . Thus , other mutations in the pocket could indeed activate the kinase , but as a consequence of the resultant destabilization of the protein , they end up causing aggregation . Mutations that mimic phosphorylation can activate the kinase by interacting with Arg575 , as shown for positions Thr599 and Ser602 ( Roskoski , 2010 ) . This is the mode of interaction for all so-called ‘RD’ kinases that become activated through phosphorylation within the activation segment ( Johnson and Lewis , 2001 ) . Structural inspection after superimposing all kinases suggests that mutating Lys601 to Glu could also lead to interaction with Arg575 , with a small conformational change ( Figure 4F; Figure 4—figure supplement 4 ) . Thus , we added favorable energies to the Asp and Glu mutations made at those positions ( the added energy value was determined by mutating phospho-Ser to Ser in the cAMP-dependent protein kinase structure ( PDB entry 1ATP; Zheng et al . , 1993; Figure 4–figure supplement 5A ) . This did not apply to position 600 , however , which always points away from Arg575 in the active conformation , irrespective of if it is a Val or Glu , similar to the equivalent position in many other active kinases ( Figure 4—figure supplement 4 ) , and whose contribution to the active conformation energy is null . Finally , although in the V600E structure ( PDB entry 4MNF ) Glu600 forms a salt bridge with Lys507 in the αC helix ( Haling et al . , 2014 ) , both residues are solvent exposed , and the Lys side chain is not structurally constrained . Therefore , the possible salt-bridge energy contribution is negligible ( FoldX energy calculations suggest no energetic contribution between Glu600 and Lys507 ( △△G ( E600A ) = 0 . 04+/- 0 . 1 kcal/mol ) . This explains why mutations to Lys or Arg are as activating as Glu and Asp . As V600E or V600K does not stabilize the active conformation , we tested whether it still requires phosphorylation at Thr599 and Ser602 to keep the AS in a fixed active state ( by interacting with Arg575 ) by mutating these residues to alanines ( to disable phosphorylation ) . In normal growth medium , we observed either no change ( V600E ) or a slight reduction ( V600K ) in MEK phosphorylation ( Figure 4G–H; Figure 4—figure supplement 5 ) . This suggests that by opening the AS and preventing its closure , the kinase becomes active , independent of phosphorylation . These results were additionally supported by mutating Glu611 to Ala in the context of V600E . As Glu611 forms a salt bridge with Arg575 , this interaction may partially stabilize the open conformation . However , as MEK phosphorylation did not change ( Figure 4G–H; Figure 4—figure supplement 6B–C ) , it is more likely that this salt bridge contributes little or nothing to stabilization . We used random forest predictions to analyze the quantitative contribution of individual factors to the prediction of cancer frequencies . In addition to the six parameters described above , we also included as a parameter the change in codon usage frequency due to a mutation ( Supplementary file 1 ) . If a frequent codon is mutated to a rare one , this could affect translation efficiency and protein levels ( Lampson et al . , 2013 ) . To see if a combination of the factors discussed above can be used to predict the observed mutation frequency in cancer , we constructed a random forest classifier ( Figure 5A–B ) . This ensemble learning technique identifies the contributions of individual ‘trees’ ( here , FoldX energies , nucleotide substitutions , and codon frequencies ) to an output ( here , cancer frequencies ) . As values not given in the training set cannot be extrapolated by the random forest method , we ran two sets of 100 predictions . For each prediction , we trained with a random subset of samples , using ∼70% of the data and balancing mutations with low and high cancer frequencies . The V600E mutation was included in only one set , and the importance values for all seven parameters for all sets were kept . Next , we ran the trained random forest on the remaining ∼30% of the data and calculated the root mean square deviation ( RMSD ) as well as the correlation between the real data and the predicted values . The ratio of this correlation to the RMSD was us as a performance indicator for each run . The importance values of the seven different parameters were similar between sets , suggesting that the presence or absence of V600E did not affect the training outcome ( Figure 5C; Figure 5—figure supplement 1 ) . The AS loop energy was the highest contributor to the random forest prediction of cancer frequencies ( parameter 3; ∼70% ) , while parameters 2 ( folding energy active conformation ) , 4 ( nucleotide substitution frequency ) , 5 ( change in codon usage frequency ) , and 7 ( hydrophobic solvation energy ) contributed almost equally , and 1 and 6 ( destabilization of inactive conformation and the salt bridge ) had very little contribution ( Figure 5C ) . Ensemble methods , such as random forests , have several advantages compared to non-ensemble machine-learning methods , such as better handling of small sample sizes and high dimensionality , increased robustness and limited overfitting . The contributions of the different features calculated in this study are quite robust , and in the two cases analyzed ( with and without V600E in the training set ) , they were found to be comparable and to follow the same order . 10 . 7554/eLife . 12814 . 020Figure 5 . Quantitative contribution of individual factors to the prediction of cancer frequencies . ( A ) Comparison of real and predicted cancer frequencies ( labelled ‘real value’ and ‘predicted value’ ) for one exemplary random forest prediction ( run 16 ) . Black dots represent mutations that were in the training set , blue dots the ones in the test set , and red are the mutations that were tested experimentally in this work ( some of them were included in the training set , some of them in the test set ) . ( B ) Plot of RMSD against correlation for all individual random forest runs with V600E in the training set . The correlation is the correlation between the predicted value by the random forest ( ’predicted value’ ) and the experimental value ( ‘real value’ ) , and the RMSD calculates the deviation of the predicted values from the real ones . ( C ) Results from random forest analyses with V600E in the training set . Abbreviation for parameters: 1 ) destabilization of inactive conformation and/or folding; 2 ) destabilization of active conformation ( folding ) ; 3 ) destabilization of inactive loop conformation; 4 ) nucleotide substitution frequency; 5 ) change in codon usage frequency; 6 ) salt bridge; and 7 ) change in hydrophobic solvation energy . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 02010 . 7554/eLife . 12814 . 021Figure 5—figure supplement 1 . Random forest analyses without V600E in the training set . ( A ) Results from the random forest analyses without V600E in the training set . Abbreviation for parameters: 1 , destabilization of inactive conformation/folding; 2 , destabilization of active conformation ( folding ) ; 3 , destabilization of inactive loop-conformation; 4 , nucleotide substitution frequency; 5 , change in codon usage frequency; 6 , salt bridge; 7 , change in hydrophobic solvation energy . ( B ) Plot of the RMSD against correlation for all random forest runs without V600E in the training set . The correlation is the correlation between the predicted value by the random forest ( ’predicted value’ ) and the experimental value ( ‘real value’ ) , and the RMSD calculates the deviation of the predicted values from the real ones . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 021
We provide a complete picture for the genotype-phenotype associations of the hydrophobic pocket of the BRAF kinase domain and emphasize the importance of a balance between increased activity and loss in stability and/or folding . By using structure-energy calculations and a number of nucleotide substitutions , we were able to reconcile why V600E is by far the most frequent cancer mutation . We show here , that the effect of a mutation on folding depends on the structural flexibility of the respective hydrophobic pocket where the mutated amino acid is located ( Figure 6 , left side ) . Position Val600 is in a region where the destabilization of the hydrophobic pocket causes activation of the kinase as structural flexibility still allows for proper folding . In contrast , those mutations in structurally rigid areas of the hydrophobic pocket only slightly increase the catalytic activity while increasing destabilization , and thus rapidly cause unfolding of BRAF ( Figure 6 , right side ) . 10 . 7554/eLife . 12814 . 022Figure 6 . Schematic diagram depicting the relationship between structural flexibility , destabilization of the hydrophobic pocket , and cancer frequencies . The effect of a mutation on folding depends on the structural flexibility of the respective hydrophobic pocket where the mutated amino acid is located . In a region with higher structural flexibility destabilization in the hydrophobic pocket will cause activation of the kinase and still allows proper folding . Thus , the disease-causing propensity ( cancer mutation frequency ) will increase with increasing destabilization of the hydrophobic pocket . In contrast , mutations in structurally rigid areas of the hydrophobic pocket will only slightly increase the catalytic activity with increasing destabilization , and will then cause unfolding . DOI: http://dx . doi . org/10 . 7554/eLife . 12814 . 022 Altogether , we propose that whether or not BRAF-activating mutations are found in disease depends on the properties of the AS , the associated possibility of disturbing these properties in a single nucleotide substitution and the impact on the stability of the active and inactive conformations . Our results underscore the importance of considering the number of base substitutions required for a given mutation in genome-wide association studies . Rare mutations can be passengers or drivers , depending on the number of base substitutions needed . Additionally , individuals with silent mutations at critical hotspot positions may acquire rare disease mutations infrequently found in cancer . Finally , mutations that both activate and slightly destabilize a protein may be rescued by random fluctuations in the chaperone levels present in a population ( Lehner , 2013 ) . Consideration of these factors in combination with protein design algorithms may also offer mechanistic explanations of why certain mutations are found at higher frequencies in other oncogenic proteins .
Somatic BRAF mutations were downloaded from COSMIC ( http://cancer . sanger . ac . uk/cancergenome/projects/cosmic/ ) . Germline mutations for BRAF were extracted from OMIM ( http://www . omim . org/ ) and Uniprot ( http://www . uniprot . org/ ) . Protein structures were retrieved from the Protein Data Bank ( http://www . rcsb . org/pdb/home/home . do ) . FoldX ( http://foldx . crg . es/ ) is a computer algorithm that allows interaction energies contributing to the stability of proteins and protein complexes to be calculated ( Guerois et al . , 2002; Schymkowitz et al . , 2005 ) . For details concerning the force field , please see the description in the online version and in related publications ( Kiel and Serrano , 2009; Kiel and Serrano , 2007; Rakoczy et al . , 2011 ) . The FoldX algorithm enables predictions of mutational affects for any of the 20 natural amino acids , but not for any backbone changes . Prior to any mutagenesis , we optimized the total energy of the protein using the RepairPDB option of FoldX , which identifies and repairs those residues with bad torsion angles and van der Waals clashes . Mutagenesis was performed using the BuildModel option of FoldX , with five repetitions per mutation . Stabilities were calculated using the Stability command of FoldX , and ΔΔG values were computed by subtracting the energy of the wild-type from that of the mutant . The FoldX energy function includes terms that have been found to be important for protein stability . The free energy of unfolding ( ΔG ) of a target protein is calculated using the equation: ΔG = Wvdw * ΔGvdw + WsolvH * ΔGsolvH + WsolvP * ΔGsolvP + ΔGwb + ΔGhbond + ΔGel + ΔGKon + Wmc * T * ΔSmc + Wsc * T * ΔSsc with: If interaction energies between complexes are calculated , two additional terms are needed: Random forest ( Breimann , 2001 ) construction and predictions were performed using the package ‘randomForest’ for R ( R Development Core Team , 2008 ) . Two sets of 100 random forests each were constructed . In one set , the V600E mutant was always included in the training set of the classifiers , while in the other it was not . Random forests used ∼70% of the samples for the training , with the remaining ∼30% was used for performance testing . All random forests were trained with the same parameters . The number of trees was set to 40 , as a further increase did not improve the performance of the predictor . The number of variables randomly sampled as candidates at each split of the trees was set to four . To assess the significance of each of the features used in the random forest and how they contribute to the prediction outcome , we determined the importance of each of them . This value is computed by calculating the total decrease in node impurities when splitting on a certain variable . This means that every time a specific variable is used for a split in any of the trees in the forest , the decrease in the impurity of the child nodes , respect to the parent node , is calculated . In regression random forests , this is done by calculating the residual sum of squares , comparing the predicted value of the forest with the real value , for each of the samples in the training . It is expected that the residual sum of squares decreases at each split , thus improving the tree . The larger the decrease , the better the split , and thus the variable used is considered more important . For each variable , the decrease in the node impurity is calculated every time it is used for a split in any of the trees , and the values are added to determine the importance of this variable . The features that contribute most to the random forest prediction will have larger importance values . BRAF complementary DNA was cloned into pDEST/N-SF-TAP v1 with N-terminal Strep and Flag tags ( provided by Dr . Gloeckner and Dr . Ueffing , HelmHoltz Zentrum Muenchen; ( Gloeckner et al . [2007] ) and fully sequenced . Single amino acid mutations were introduced with the QuikChange site-directed mutagenesis kit ( Stratagene ) using pDEST/N-SF-TAP BRAF as a template . HEK293 cells were cultured in Dulbecco’s modified Eagle’s medium ( Gibco ) supplemented with L-glutamine and 10% ( v/v ) heat-inactivated fetal calf serum ( here , normal growth medium ) . For each seeding-transfection- ( stimulation ) -lysis experiment , HEK293 cells were seeded on 35-mm dishes and transfected after 24 hr ( at 80% confluence ) with 1 μg of BRAF plasmid , using Lipofectamine 2000 ( Invitrogen , Thermo Fisher Scientific , Waltham , Massachusetts , USA ) according to the manufacturer’s instructions . After 24 hr , cells were washed twice with PBS and resuspended in 200 μl of lysis buffer ( 0 . 1% SDS , 25 mM Tris [pH 7 . 8] , 1:1000 protease inhibitor cocktail 1 and 2 [Sigma] ) . For EGF stimulation experiments , cells were transfected ( in serum-free medium ) and then , after 1 day , stimulated with 50 ng of EGF or HRG , in 3 ml , for the indicated times , washed with PBS and lysed as above . To fractionate cells into soluble and insoluble fractions , cells were first lysed in hypertonic lysis buffer ( 20 mM Tris pH 7 . 5 , 5 mM MgCl2 , 5 mM CaCl2 , 1 mM DTT , 1 mM EDTA , 1:1000 protease inhibitor cocktails 1 and 2 [Sigma] ) , sonicated for 5 min , and centrifuged for 5 min at 3000 rpm , after which the supernatant was removed ( ‘soluble fraction’ ) . The pellet was resuspended in SDS lysis buffer ( ‘insoluble fraction’ ) . Cell lysates were loaded for Western blot analysis . Blots were incubated with an enhanced chemiluminescence reagent ( SuperSignal West Femto , Thermo 34096 ) and visualized with a LAS-3000 imager ( Fujifilm Co . ) . Two to three biological sample replicates were generated in each seeding-transfection-lysis experiment and analyzed on the identical Western blot ( ‘biological replicates performed at the same day’ ) . Up to eight different seeding-transfection-lysis experiments were performed ( ‘biological replicates performed at different days’ ) . The intensity of protein bands for MEK-p and flag ( for total BRAF levels ) was quantified with ImageJ . MEK-p levels were normalized by total BRAF levels ( using the flag antibody ) . To compare biological replicates performed at different days MEK-p/BRAF total intensities were referenced to WT ( =100% ) . While the relative intensity changes between WT and mutants always followed the same trend in all biological replicates performed at different days , the quantitative intensity spread could vary ( e . g . for V600E between 180% to 400% compared to WT ) . To compare intensities from Western blots from different days , we averaged experiments that had a similar intensity spread . The following antibodies were used for Western blotting: Flag ( Sigma , F1804 ) , phospho-BRAF Ser445 ( Cell Signaling , #2330 ) , phospho-MEK Ser217 and Ser221 ( Cell Signaling , #9121 ) , β-actin ( Thermo , MA5-15739 ) , and total BRAF ( SIGMA , HPA001328 ) . | Mutations in the gene that encodes a protein called BRAF are commonly found in certain cancers , such as melanomas . The same BRAF mutation is found in nearly all of these cancers . This mutation causes the 600th amino acid in the BRAF protein – an amino acid called a valine – to be replaced with another amino acid , a glutamate . BRAF is a type of enzyme called a kinase , and it transmits signals inside cells to promote cell growth . Kinases work by adding a phosphate group to other proteins to alter their activity . The structure of the BRAF kinase contains a pocket-like shape , and the valine at position 600 sits buried inside this pocket when the enzyme is inactive . The “valine-to-glutamate” mutation ( often called V600E for short ) disrupts the interactions that create this pocket . This in turn results in a permanently active form of BRAF and uncontrolled cell growth . However , it remains unclear why the valine-to-glutamate mutation is so much more common in cancer cells than any other mutation that could affect the pocket in BRAF . To address this question , Kiel et al . used a computational tool to generate three-dimensional models for all the different amino acid substitutions that could occur in BRAF’s pocket . Each mutation was then assessed to see how it might destabilize the structure of BRAF . Only the mutations that affected the 600th amino acid were predicted to be able to open the pocket without destabilizing the part of the enzyme that adds phosphate groups to other proteins . Kiel et al . validated their computational predictions by introducing normal or mutant versions of the BRAF-encoding gene into human cells grown in the laboratory . These experiments showed that a mutation that introduced an amino acid called histidine into position 600 could activate BRAF as much the valine-to-glutamate mutation . Kiel et al . suggest that this “valine-to-histidine” substitution is not found in cancers because it requires three changes to the DNA sequence of the BRAF gene , whereas the valine-to-glutamate substitution only requires one . The results underscore the importance of considering changes at both the DNA and protein level when attempting to understand why certain cancer-causing mutations are more common than others . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Conclusions",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics",
"computational",
"and",
"systems",
"biology"
] | 2016 | The yin–yang of kinase activation and unfolding explains the peculiarity of Val600 in the activation segment of BRAF |
KCNQ channels are critical determinants of neuronal excitability , thus emerging as a novel target of anti-epileptic drugs . To date , the mechanisms of KCNQ channel modulation have been mostly characterized to be inhibitory via Gq-coupled receptors , Ca2+/CaM , and protein kinase C . Here we demonstrate that methylation of KCNQ by protein arginine methyltransferase 1 ( Prmt1 ) positively regulates KCNQ channel activity , thereby preventing neuronal hyperexcitability . Prmt1+/- mice exhibit epileptic seizures . Methylation of KCNQ2 channels at 4 arginine residues by Prmt1 enhances PIP2 binding , and Prmt1 depletion lowers PIP2 affinity of KCNQ2 channels and thereby the channel activities . Consistently , exogenous PIP2 addition to Prmt1+/- neurons restores KCNQ currents and neuronal excitability to the WT level . Collectively , we propose that Prmt1-dependent facilitation of KCNQ-PIP2 interaction underlies the positive regulation of KCNQ activity by arginine methylation , which may serve as a key target for prevention of neuronal hyperexcitability and seizures .
Epilepsy imposes a major burden at both global and individual levels . Worldwide , about 1% of the population suffers from epilepsy , and nearly 4% of the population will experience epilepsy at some point during their lifetime ( Malkki , 2014 ) . In almost 30% of patients with epilepsy , anti-seizure medications do not provide sufficient seizure control ( Malkki , 2014 ) . Thus , understanding the etiology of epilepsy is essential both for clinical management of patients and for conducting neurobiological research that will direct future therapies . The etiology of epilepsy was formerly regarded as unknown in about three-quarters of patients; however , massively parallel gene-sequencing studies showed the importance of gene mutations in the etiology of epilepsy ( Thomas and Berkovic , 2014 ) . Among those epileptic conditions linked to channelopathies , mutations in potassium channel subunits represent the largest category ( Brenner and Wilcox , 2012; Cooper , 2012; Noebels , 2003 ) . Together with KCNA1 , KCNQ2 and KCNQ3 are the earliest identified K+ channels associated with idiopathic epilepsy ( Rogawski , 2000; Schroeder et al . , 1998; Browne et al . , 1994 ) . More than 30 mutations have been detected in KCNQ channels , most of which were found in KCNQ2: only four mutations were found in KCNQ3 ( Maljevic et al . , 2010 ) . These mutations caused a reduction in channel activity by different molecular mechanisms , by which all can lead to membrane depolarization and increased neuronal firing . In the central nervous system ( CNS ) , heteromeric KCNQ2/3 potassium channels form the M-current , a subthreshold potassium current ( Delmas and Brown , 2005; Jentsch , 2000 ) . Due to the slow gating kinetics , M-currents effectively oppose sustained membrane depolarization and repetitive action potential ( AP ) firing ( Brown and Passmore , 2009 ) . Thus , breakdown of M-channels by loss of function mutations or pharmacological inhibitors leads to neuronal hyperexcitability ( Schroeder et al . , 1998; Delmas and Brown , 2005 ) . KCNQ potassium channels are standouts for epileptologists , in that they are both mutated in human epilepsy and principal targets of an approved anti-epileptic drug ( ezogabine/retigabine ) ( Gunthorpe et al . , 2012; Soldovieri et al . , 2011 ) . Therefore , regulation of KCNQ channel activity has been studied by many groups ( Delmas and Brown , 2005 ) . Notably , M-channels are inhibited by muscarinic acetylcholine receptor agonists , leading to a profound increase in cellular excitability that can be reversed by the withdrawal of receptor agonist . Some subtypes of receptors for dopamine , serotonin , glutamate , and several peptide neurotransmitters , including luteinizing hormone-releasing hormone and bradykinin , are also found to be capable of inhibiting the M-channels . An important property of M-channels is that they require membrane phosphatidylinositol-4 , 5-bisphosphate ( PIP2 ) to open ( Delmas and Brown , 2005; Suh and Hille , 2008 ) , and the certain receptors mentioned above suppress M-current by depletion of PIP2 ( Delmas and Brown , 2005 ) . In addition , M-channels are also inhibited by CaM in a Ca2+-dependent manner , providing another mode of intracellular modulation of M-channel activities ( Gamper and Shapiro , 2003 ) . Recent work indicates that protein kinase C ( PKC ) phosphorylation of the KCNQ2 subunit reduces the affinity for its PIP2 binding and activities ( Kosenko et al . , 2012; Lee et al . , 2010 ) . Likewise , sumoylation of KCNQ channels results in reduced channel activity ( Qi et al . , 2014 ) . Thus , all regulatory pathways discovered so far act on the KCNQ channels to reduce the channel activity and increase neuronal excitability . Given its important role in stabilization of neuronal excitability , there has been much speculation for regulatory mechanisms to increase the channel-PIP2 interaction and channel activities , however , to date there is no evidence supporting such regulatory mechanism . Protein methylation , along with phosphorylation , controls a variety of cellular functions ( Nicholson et al . , 2009 ) . Protein arginine methyltransferases ( Prmts ) are enzymes that catalyze the transfer of a methyl group to arginine residues of histone or non-histone substrates ( Boisvert et al . , 2005 ) . In mammals , nine Prmts have been characterized . Among these , Prmt1 , originally identified as a histone H4 methyltransferase , methylates many non-histone proteins and implicated in diverse cellular processes including RNA processing , transcriptional regulation , oncogenesis , cell survival , insulin signaling , and metabolism ( Boisvert et al . , 2005; Bedford and Clarke , 2009; Krause et al . , 2007 ) . Although Prmt1 is a predominant Prmt in mammalian cells and is highly expressed in the CNS ( Nicholson et al . , 2009; Bedford and Clarke , 2009 ) , its functional significance in the CNS has not yet been identified . The positively charged ( basic ) arginines or lysines are candidates for mediating electrostatic interaction with PIP2 in channels such as KCNQ ( Hernandez et al . , 2008 ) , Kir2 ( Hansen et al . , 2011; Huang et al . , 1998; Lopes et al . , 2002 ) , and GIRK ( Whorton and MacKinnon , 2011 ) . Considering that each additional methyl group to an arginine residue can readily modulate their physical properties ( Bedford and Clarke , 2009 ) , methylation of arginine residues in PIP2 binding domain may alter KCNQ channels' affinity for PIP2 . However it is not known whether such methylation really occurs and regulates the channel activity and whether it is implicated in common disease phenotypes . In the present study , the role of arginine methylation of KCNQ in regulation of channel activities and neuronal excitability was investigated . Prmt1-heterozygous ( +/- ) mice exhibit epileptic seizures . We also found that Prmt1 depletion causes a decreased interaction between PIP2 and KCNQ channels , consequently causing a reduction in KCNQ channel activity . Prmt1 interacts and methylates at 4 arginine residues of KCNQ channels . Hippocampal neurons from the heterozygote mice lack KCNQ currents , and the current can be restored by exogenous PIP2 addition , accompanied by concomitant rescue of normal excitability . Furthermore a pharmacological inhibition of methylation or methylation-deficient mutants of KCNQ2 reduce PIP2 binding and activities of KCNQ channels . These data demonstrate that protein arginine methylation facilitates KCNQ channel-PIP2 interaction , leading to seizure suppression . We propose that Prmt1-dependent regulation of KCNQ channels represents an important mechanism of neuronal protection against over-excitability .
To assess the physiological importance of Prmt1 in the CNS , we utilized mutant mice for the Prmt1 gene in a C57BL/6J background ( Choi et al . , 2012 ) . As Prmt1 homozygous knockout mice are embryonic lethal ( Choi et al . , 2012 ) , we used heterozygous mice ( Prmt1+/- ) for our study . Immunoblot analysis demonstrated that Prmt1 proteins were highly expressed in wild-type ( WT ) brain but significantly reduced in the brain of Prmt1+/- mice ( Figure 1a ) . We used long-term video- electroencephalographic ( EEG ) recording ( 24 hr per day for 6 days ) to monitor behavior and spontaneous EEG activity in freely moving Prmt1+/- ( n = 8 ) and WT mice ( n = 6 ) , which led to observation of spontaneous seizure activity from Prmt1+/- mice ( n = 8 ) ( Figure 1b and c ) . Epileptiform spikes with a delta frequency range ( 1–3 Hz ) appeared in Prmt1 +/- mice ( Figure 1d , Figure 1—source data 1 ) , and lasted for 96 . 7 ± 12 . 5 s ( range , 30–400 s , Figure 1e , Figure 1—source data 1 ) . These results are in parallel with human seizures that usually last less than 3 min ( Bromfield EB and Sirven , 2006 ) . The number of seizure occurrences was 4 . 1 ± 1 . 4 per day in Prmt1+/- mice ( range , 0 . 3–12 per day , Figure 1f , Figure 1—source data 1 ) . Seizure activity was not observed in WT mice ( Figure 1d–f , Figure 1—source data 1 ) . During these seizure activities of Prmt1+/- mice , we typically observed partial seizure behaviors ( stages 1–3 seizures on the Racine scale ) or no changes in behavior ( Figure 1g , Figure 1—source data 1 ) . Convulsive seizures ( stages 4 and 5 ) were rarely observed in Prmt1+/- mice ( Figure 1g ) . 10 . 7554/eLife . 17159 . 003Figure 1 . Spontaneous seizures and increased locomotor activities in Prmt1+/- mice . ( a ) reduced expression of Prmt1 in Prmt1+/- mice , compared to WT control mice . ( b ) Representative traces of seizure activities: short ( upper ) or long ( middle and lower ) duration of epileptiform activity . ( c ) A colored power spectrum of the trace shown in b . ( d–g ) The mean bar graphs of seizure spike frequency ( d ) duration ( e ) , number of seizures per day ( f ) , and seizure scores ( g ) in WT ( n = 6 ) and Prmt1+/- mice ( n = 8 ) . ( h ) The WT and the Prmt1+/- mice spent similar amounts of times in the center of the open-field box . ( i ) The Prmt1+/- mice moved a longer distance than did the WT mice . ( j ) Total distance moved for 30 min . *p<0 . 05 by Student’s t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 00310 . 7554/eLife . 17159 . 004Figure 1—source data 1 . Source data for Figure 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 004 Behavioral hyperactivity , such as an increase in open-field locomotion , frequently accompanies seizures ( Kim et al . , 2011; Peñagarikano et al . , 2011; Peters et al . , 2005 ) . In an open-field test , male mutants showed a slight increase in locomotor activity ( Figure 1h–j , Figure 1—source data 1 ) . The Prmt1+/- mice ( n = 10 ) spent a similar amount of time in the center of the open-field box compared with the control mice ( n = 9 ) ( Figure 1h ) . However , the Prmt1+/- mice moved a longer distance than WT mice ( Figure 1i ) . The total travel distance was increased significantly in Prmt1+/- mice , compared to that in WT mice ( p<0 . 05; Figure 1j ) . Thus , Prmt1+/- mice display signs of a persistent neuronal hyperexcitability , including spontaneous seizures and a slightly increased locomotor activity . Hyperexcitability of hippocampal neurons is a characteristic feature of most epilepsies ( Noebels , 2003; McCormick and Contreras , 2001 ) . To identify neural mechanisms underlying spontaneous seizure activities in Prmt1+/- mice , we performed electrophysiological recordings from the dentate gyrus granule cells ( GCs ) of hippocampal slices from WT and mutant mice . GCs from WT mice typically displayed tonic firing patterns in response to a 1-s square current pulse injection: AP frequency elevated as the magnitude of the square pulse increased ( Figure 2a and c , Figure 2—source data 1 ) . GCs from Prmt1+/- mice showed significantly higher AP frequency than those from WT mice ( Figure 2b and c , Figure 2—source data 1 ) . The averaged AP frequency in response to a 200 pA depolarizing current in WT GCs was 8 . 1 ± 1 . 4 Hz ( n = 16 ) , while it increased significantly to 31 . 2 ± 2 . 2 Hz ( n = 18 , p<0 . 01 ) in Prmt1+/- GCs ( Figure 2c ) . These data indicate that the heterozygous deletion of the Prmt1 gene enhanced excitability of hippocampal neurons . To determine whether the increased firing resulted from a change in the threshold current , we compared the magnitude of the current injection to reach a threshold for AP firing ( AP threshold current ) in WT and Prmt1+/- GCs . Representative traces in Figure 2d , e clearly show that the current for spike initiation was reduced in Prmt1+/- neurons . Analysis of pooled results revealed a significant decrease in the mean threshold current in Prmt1+/- GCs compared to that in WT GCs ( Figure 2f , Figure 2—source data 1 ) . As illustrated in Figure 2g , the change in threshold current was correlated with an increase in the input resistance . The amplitude and duration of APs obtained in WT were not altered by the reduction of the Prmt1 gene dose ( Figure 2h and i , Figure 2—source data 1 ) . Therefore , the change in the threshold current in Prmt1+/- neurons reflects the effect on input resistance , which facilitates their AP initiation and , hence , increase in spike number . 10 . 7554/eLife . 17159 . 005Figure 2 . Comparison of neuronal excitability in WT and Prmt1+/- mice . ( a–b ) a and b panels show representative trace in the whole-cell current-clamp recording from WT ( a ) and Prmt1+/- ( b ) mature dentate GCs in response to 1-s depolarizing current injection ( 200 pA ) , respectively . ( c ) the mean number of action potentials ( AP No . ) plotted against the eliciting currents ( from 100 pA to 400 pA , + 50 pA increment , during 1-s ) . At all amplitudes , the Mean ± S . E . M . AP No . is significantly higher in Prmt1+/- ( ■; n = 18 , seven mice ) than WT dentate GCs ( □; n = 16 , seven mice; p<0 . 01 ) . ( d–e ) the threshold current for single AP elicited by a short depolarizing ( 100 ms ) step pulse of various amplitude in WT ( d ) and Prmt1+/- ( e ) dentate GCs . ( f–i ) the mean value of threshold current for AP generation ( 100 ms duration; f ) , input resistance ( g ) , AP height ( h ) , and AP half-width ( i ) from WT and Prmt1+/- mature dentate GCs . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 00510 . 7554/eLife . 17159 . 006Figure 2—source data 1 . Source data for Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 006 Next , we explored the ionic mechanisms underlying the increased excitability . An increase in neuronal excitability without a change in AP height and width indicated that the kinetics and amplitude of voltage-gated currents associated with the spike itself were not affected . Increased cation current or leak current was unlikely to be involved in the phenotype , either , as this would be expected to result in a decrease in input resistance . Thus , the most probable target was the KCNQ/M channel . KCNQs are voltage-dependent K+ channels that partially open at the resting membrane potential . Inhibiting the KCNQ channel increases both input resistance and neuronal excitability ( Jentsch , 2000; Brown and Passmore , 2009 ) . Thus , we examined whether M-channel deficiency contributed to increased neuronal excitability of Prmt1+/- mice . The M-current amplitude was measured using a standard deactivation voltage protocol ( Peters et al . , 2005; Lawrence et al . , 2006; Adams et al . , 1982 ) . In GCs from WT mice , a step to -60 mV forced deactivation of a slowly relaxing outward current that decayed , consistent with the presence of M-current . The M-channel antagonist XE991 ( 10 μM ) caused a complete loss of detectable M-current ( Figure 3a ) , accompanied by a reduction in the holding current . However , Prmt1+/- GCs showed little measurable M-currents ( p<0 . 001 ) ( Figure 3b–c , Figure 3—source data 1 ) . 10 . 7554/eLife . 17159 . 007Figure 3 . KCNQ current deficiency contributed to the persistent hyperexcitability in Prmt1+/- mice . ( a–b ) Representative current traces of voltage clamp recordings from WT ( a; n = 22 , seven mice ) and Prmt1+/- ( b; n = 12 , four mice ) mature dentate GCs in response to the voltage protocol depicted below . The upper panel shows the overlay of currents in the absence ( black ) or presence ( blue ) of 10 μM XE991 . To illustrate the kinetic components of M-current relaxation and activation , the corresponding difference current is shown in an expanded scale below each current trace . ( c ) Summary statistics from experiments shown in a–b reveal loss of M-current in Prmt1+/- mice . Error bars , S . E . M . ***p<0 . 001 by Student’s t-test . ( d–e ) Changes in input resistance ( d ) and threshold current ( e ) in WT ( +/+ ) and Prmt1+/- GCs ( +/- ) in response to 10 μM XE991 . Each connected line represents an individual neuron . Right panel displays the summary of changes in input resistance or threshold current upon application of 10 μM XE991 . XE991 was less effective in mutants than in WT . ( f–g ) Average numbers of APs generated during incremental 1-s depolarizing current steps before and after application of XE991 in WT ( f ) and Prmt1+/- GCs ( g ) . XE991 produced a substantial change in firing rate of WT GCs ( f; n = 7~11 , seven mice ) , whereas little change in firing rate occurred with XE991 in Prmt1+/- GCs ( g; n = 9 , seven mice ) . *p<0 . 05; **p<0 . 01 by paired Student’s t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 00710 . 7554/eLife . 17159 . 008Figure 3—source data 1 . Source data for Figure 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 00810 . 7554/eLife . 17159 . 009Figure 3—figure supplement 1 . Changes in membrane potential in WT ( +/+ ) and Prmt1+/- GCs ( +/- ) in response to 10 μM XE991 . Each connected line represents an individual neuron . Right panel gives the averaged changes . XE991 was less effective in mutants than in WT . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 00910 . 7554/eLife . 17159 . 010Figure 3—figure supplement 1—source data 1 . Source data for Figure 3—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 010 Consistent with the major contribution of the KCNQ channel ( Delmas and Brown , 2005 ) , XE991 application led to a depolarization ( ΔVm , 4 . 5 ± 1 . 6 mV ) in WT neurons . In contrast , such an effect of XE991 was abolished in the mutant ( ΔVm , −0 . 9 ± 0 . 7 mV ) ( Figure 3—figure supplement 1 , Figure 3—figure supplement 1—source data 1 ) . Similarly , application of XE991 increased input resistance in the WT neurons ( Δinput resistance , 51 . 7 ± 4 . 6 MΩ ) but had less of an effect in the mutant ( Δinput resistance , 29 . 2 ± 6 . 3 MΩ ) ( Figure 3d , Figure 3—source data 1 ) . The Prmt1+/- neurons’ threshold currents were also insensitive to XE991 treatment ( Figure 3e , Figure 3—source data 1 ) . These results suggest that the increase in the input resistance of Prmt1+/- neurons might be a result of the reduced level of KCNQ channel activity . We further examined the effect of XE991 on firing frequency . The firing frequency in WT neurons increased significantly after applying 10 μM XE991 ( Delmas and Brown , 2005 ) ( Figure 3f , Figure 3—source data 1 ) . For example , the AP frequency in response to a 200 pA-depolarizing current was 8 . 4 ± 1 . 9 Hz in the control and increased to 12 . 6 ± 1 . 9 Hz following the application of 10 μM XE991 . This XE991-induced spiking was largely absent in the mutant neurons ( Figure 3g , Figure 3—source data 1 ) . Before and after the XE991 treatment , AP frequency in response to a 200 pA depolarizing current was 34 ± 3 . 2 Hz and 35 . 2 ± 5 . 6 Hz , respectively . Thus , these results showed that Prmt1+/- GCs displayed a high firing rate at baseline , and their firings did not further increase during XE-991 application , suggesting that defective M-current contributes to the neuronal hyperexcitability observed in the Prmt1+/- mice . However , we cannot exclude the potential involvement of other Prmt1 target ( s ) in the neuronal hyperexcitability observed in the Prmt1+/- mice . To assess the effect of protein methylation on M-currents , WT hippocampal slices were treated with a pan-methyltransferase inhibitor 5-deoxy-5- ( methylthio ) adenosine ( MTA ) or a more specific blocker of Prmt1 , furamidine dihydrochloride . The treatment of WT hippocampal slices with MTA ( 100 μM ) or furamidine dihydrochloride ( 20 μM ) for 1 hr completely abolished M-currents ( Figure 4a–c; p<0 . 001 vs WT control , Figure 4—source data 1 ) . We then studied their effects on neuronal excitability . The application of MTA ( 100 μM , 1 hr ) enhanced neuronal excitability of WT GCs ( Figure 4d–e ) . After the MTA treatment , AP frequency in response to a 200 pA depolarizing current was 34 . 7 ± 4 . 6 Hz . Consistent with ablation of M-currents by MTA , subsequent application of XE991 had no further effect ( AP frequency was 31 . 8 ± 3 . 9 Hz ) ( Figure 4d–e , Figure 4—source data 1 ) . The MTA-induced increase in excitability was accompanied by the declined AP threshold currents and increased input resistance ( threshold current , 110 . 5 ± 5 . 5 pA , n = 4; input resistance , 234 . 0 ± 5 . 8 MΩ , n = 4 ) , which were relatively insensitive to a subsequential treatment with XE991 ( p>0 . 05; Figure 4j–k , Figure 4—source data 1 ) . Furthermore , the treatment of WT hippocampal slices with furamidine dihydrochloride exhibited a similar effect on neuronal excitability , input resistance , and threshold current , confirming the results obtained with MTA treatment ( Figure 4f–g and j–k , Figure 4—source data 1 ) . However , methylation suppression with MTA had no effect on AP firing ( Figure 4h–i , Figure 4—source data 1 ) , threshold current ( Figure 4j , Figure 4—source data 1 ) , or input resistance ( Figure 4k , Figure 4—source data 1 ) in Prmt1+/- GCs . Thus , these data suggest that the reduction of Prmt1 activity in the hippocampal neurons might be responsible for the neuronal hyperexcitability in Prmt1+/- neurons . 10 . 7554/eLife . 17159 . 011Figure 4 . Methylation suppression with MTA or furamidine , increases neuronal excitability via KCNQ channel . ( a–b ) Representative current traces recorded from MTA- ( a ) or furamidine- ( b ) pretreated WT GCs using the voltage protocol depicted below in the absence ( black ) or presence ( blue ) of 10 μM XE991 . ( c ) Summary of M-currents from experiments shown in a–b . Data for M-currents of WT control shown in Figure 3c is also shown as a reference . ***p<0 . 001 by Student’s t-test . ( d–i ) APs were evoked by applying 1-s depolarizing current pulses of different intensities ( 100–400 pA ) in WT ( d–g ) or Prmt1+/- ( h–i ) neurons . Panel d , f , and h illustrate representative traces after incubation for 1 hr with 100 μM MTA ( d and h ) or 20 μM furamidine ( f ) . The effect of applying XE991 once the maximal effect of MTA or furamidine is achieved was shown in right . Scale bar indicate 40 mV . Summarized data compare the number of APs before and after application of XE991 in MTA-pretreated WT ( e ) , furamidine-pretreated WT ( g ) , and MTA-pretreated mutants ( i ) . The gray lines shows the number of APs observed in untreated WT and Prmt1+/- neurons . Genotypes are given on each line . Firing rate in MTA-treated WT cells ( □; n = 4 , two mice ) or furamidine-treated WT cells ( □; n = 4 , three mice ) increased to the level of Prmt1+/- cells ( gray line ) . Further , XE991 had no effect in MTA-treated WT cells ( ○; n = 4 , paired Student’s t tests ) or furamidine-treated WT cells ( △; n = 4 , paired Student’s t tests ) . MTA had no further effects on Prmt1+/- neurons . NS , not significantly different . ( j–k ) the mean value of threshold current for AP generation ( j ) and input resistance ( k ) . ANOVA Tukey test . **p<0 . 02; ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 01110 . 7554/eLife . 17159 . 012Figure 4—source data 1 . Source data for Figure 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 012 Importantly , KCNQ2 and KCNQ3 protein levels were unchanged in hippocampi from Prmt1+/- mice , suggesting the involvement of post-translational mechanisms in the regulation of KCNQ currents ( Figure 5—figure supplement 1 ) . Previous studies have suggested that methylation of some ion channel proteins could affect their activities/functions ( Beltran-Alvarez et al . , 2013; Sariban-Sohraby et al . , 1984 ) . To investigate a potential interaction between KCNQ2 and Prmt1 , co-immunoprecipitation experiments were conducted with HEK293T cells . KCNQ2 and Prmt1 proteins were coprecipitated when co-expressed in HEK293T cells ( Figure 5a ) . The intracellular C-terminal region of KCNQ2 encompassing amino acids 320–840 , designated as KCNQ2-C , was sufficient to interact with Prmt1 ( Figure 5b and c ) . To further confirm the interaction between KCNQ2 and Prmt1 proteins in the native neuronal environment , co-immunoprecipitation was performed with mouse hippocampal lysates . We found that KCNQ2 was co-immunoprecipitated endogenously with Prmt1 in hippocampus ( Figure 5d ) . 10 . 7554/eLife . 17159 . 013Figure 5 . Prmt1 binds to and methylates KCNQ2 at R333 , R345 , R353 , and R435 . ( a–c ) Immunoblotting analysis showing the physical association of KCNQ2 and Prmt1 . HEK293T cells were transfected with expression vectors , as indicated . Whole-cell lysates were immunoprecipitated and immunoblotted by either anti-flag or anti-HA antibody . Representative data from at least three independent experiments are shown . ( d ) Western blotting analysis showing endogenous interaction of KCNQ2 and Prmt1 . Cell lysates from mouse hippocampus were immunoprecipitated with anti-Prmt1 antibody and were immunoblotted with anti-KCNQ2 antibody or anti-Prmt1 antibody . Representative data from at least three independent experiments are shown . ( e ) In vitro methylation assays with GST-KCNQ2 ( amino acids 320–449 ) and a series of GST-Prmts ( 1–6 ) in the presence of [3H]S-adenosylmethionine ( SAM ) . Total amounts of GST-KCNQ2 ( arrowhead ) and GST-Prmts are shown by Coomassie brilliant blue staining . ( f ) In vitro methylation assays with GST-Prmt1 together with GST-KCNQ2 ( amino acids 320–449 ) WT , R333K/R353K , R345K/R353K R345K/R435K . R333K/R345K/R353K ( RRRR-KKKR ) , or R333K/R345K/R353K/R435K ( 4RK ) in the presence of [3H]SAM . ( g ) Immunoblotting analysis showing the decreased asymmetric dimethylation of KCNQ2 in Prmt1 knockdown cells . HEK293T cells were transfected with control or HA-KCNQ2 in combination with control or Prmt1 shRNA expression vectors , followed by immunoprecipitation with KCNQ2 antibodies and immunoblotting with antibodies to ADMA , KCNQ2 , Prmt1 and GAPDH . ( h ) Immunoblotting analysis showing the enhanced asymmetric dimethylation of KCNQ2 in Prmt1 overexpressing cells . HEK293T cells were transfected with control or HA-KCNQ2 in combination with control or Prmt1 expression vectors , followed by immunoprecipitation with KCNQ2 antibodies . ( i ) Immunoblotting analysis showing decreased asymmetric dimethylation of KCNQ2 in Prmt1+/- brain , compared to the WT control . Brain lysates from Prmt1+/+ and Prmt1+/- mice were immunoprecipitated with anti-KCNQ2 antibodies or control IgG . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 01310 . 7554/eLife . 17159 . 014Figure 5—figure supplement 1 . Expression of KCNQ2 and KCNQ3 in the hippocampus of WT and Prmt1+/- mice . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 01410 . 7554/eLife . 17159 . 015Figure 5—figure supplement 2 . Identification of in vivo arginine methylation sites of KCNQ2 MS/MS spectra of the methylated peptides . ( a ) 332RRNPAAGLIQSAWR345 , ( b ) 334NPAAGLIQSAWR345 , ( c ) 346FYATNLSR353 and ( d ) 435RSPSADQSLEDSPSK449 . R* represents monomethylated arginine and R** represents dimethylated arginine . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 01510 . 7554/eLife . 17159 . 016Figure 5—figure supplement 3 . The sequence alignment of 5 difference human KCNQ isoforms . The arginine residues critical for methylation in KCNQ2 are marked in red . KCNQ3 and KCNQ4 have 3 conserved arginine residues while KCNQ5 and KCNQ1 have two or one conserved arginine residues , respectively . Note that KCNQ2 and KCNQ3 form M currents in brain . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 01610 . 7554/eLife . 17159 . 017Figure 5—figure supplement 4 . In vitro methylation assays with GST-Prmt1 or myc-Prmt5 together with GST-KCNQ2 ( 320-449 aa ) . In vitro methylation assays with myc-Prmt5 purified from HEK293T cells and GST-Prmt1 by using GST-KCNQ2 ( aa 320–449 ) as a substrate and histone as a positive control . Coomassie blue stainings show the loading controls for each protein . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 01710 . 7554/eLife . 17159 . 018Figure 5—figure supplement 5 . Expression of Prmt8 in Prmt1+/+ and Prmt1+/- brain lysates . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 018 To assess whether KCNQ2 is methylated in vivo and to identify the in vivo methylation sites , we performed a liquid-chromatography-coupled tandem mass spectrometric analysis . Four arginine residues such as Arg333 ( R333 ) , Arg345 ( R345 ) , Arg353 ( R353 ) , and Arg435 ( R435 ) , were identified as methylated in vivo ( Table 1 and Figure 5—figure supplement 2 ) . To determine whether Prmt1 methylates KCNQ2 , we performed in vitro methylation assays using various bacterially purified GST-Prmts ( Prmt1-6 ) or myc-tagged Prmt5 purified from HEK293T cells and GST-KCNQ2 ( amino acids 320–449 ) that contain all of four arginine residues identified by mass spectrometry . GST-KCNQ2 was methylated by Prmt1 , but not by other Prmts ( Figure 5e and Figure 5—figure supplement 4 ) . We next examined whether the newly identified methylation sites were direct targets for methylation by Prmt1 . KCNQ2 mutants in each of which either R333 , R345 , R353 , or R435 were substituted with lysine ( further referred as RK mutants ) were generated and analyzed by in vitro methylation assay with Prmt1 ( Figure 5f ) . While the single arginine-to-lysine substitutions did not result in observable decrease of methylation ( data not shown ) , the reduction in KCNQ2 methylation was readily detected with the double RK mutants , R333K/R353K , R345K/R353K , or R345K/R435K . This was further declined in the triple mutant of R333K/R345K/R353K , compared to double mutants . Furthermore , substitution of the four Arg’s in KCNQ2 ( 4RK , R333K/R345K/R353K/R435K ) completely abolished the methylation by Prmt1 , suggesting that these four arginine residues are critical targets for Prmt1-dependent methylation . 10 . 7554/eLife . 17159 . 019Table 1 . Mono- and dimethylated peptides identified from KCNQ2 . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 019Peptide*ModificationCharge state332RRNPAAGLIQSAWR345Mono-methylated R333+3334NPAAGLIQSAWR345Mono-methylated R345+2346FYATNLSR353Di-methylated R353+2435RSPSADQSLEDSPSK449Di-methylated R435+3*Methylated amino acid residues are highlighted in bold . To test whether Prmt1 methylates KCNQ2 in vivo , HEK293T cells expressing HA-KCNQ2 were transfected with control or Prmt1 shRNA , followed by immunoprecipitation with anti-KCNQ2 antibodies and immunoblotting with antibodies to asymmetric dimethylarginine ( ADMA ) and KCNQ2 . Prmt1 knockdown specifically decreased the ADMA-positive KCNQ2 levels , suggesting that KCNQ2 methylation is sensitive to Prmt1 levels ( Figure 5g ) . Conversely , Prmt1 overexpression enhanced KCNQ2 methylation ( Figure 5h ) . We next examined the arginine methylation status of KCNQ2 channels in WT or Prmt1+/- brains ( Figure 5i ) . The immunoprecipitation analysis showed that ADMA-positive KCNQ2 proteins were readily detected in the WT brain , which were significantly reduced in that of Prmt1+/- . These results strongly advocate that Prmt1 interacts with and methylates KCNQ2 . Prmt1 and Prmt8 often share substrates in vitro , and PRMT8 is neuron-specific ( Kousaka et al . , 2009 ) . We analyzed whether Prmt8 contributes to hypo-methylation of KCNQ2 proteins in Prmt1+/- brains . The protein expression of Prmt8 in the Prmt1+/- brain was unaltered from WT brain ( Figure 5—figure supplement 5 ) . Thus , Prmt8 might not be involved in the decreased KCNQ2 methylation in Prmt1+/- brain . Taken together , these data indicate that reduced Prmt1 levels cause hypo-methylation of KCNQ2 in the Prmt1+/- brain . To examine the functional role of Prmt1-mediated methylation of KCNQ2 , KCNQ2 channel activities were assessed in Prmt1-knockdown HEK293T cells ( Figure 6a–b , Figure 6—source data 1 ) . Using the conventional whole-cell patch clamp technique , the channel function was measured by applying 'step' pulses from –70 to +40 mV in 10-mV increments at a holding potential of –60 mV for 1 s , followed by a tail pulse to –60 mV . Figure 6a demonstrates representative whole-cell current traces recorded from KCNQ2-transfected HEK293T cells . Consistent with previous studies ( Lee et al . , 2010 ) , KCNQ2-transfected cells displayed slowly activating outward currents and tail currents . In non-transfected cells , we observed small endogenous outward currents at positive voltage step pulses , but no tail currents ( data not shown ) . The shRNA-mediated depletion of Prmt1 decreased KCNQ current density from 57 . 6 ± 7 . 6 pA/pF of control transfected cells to 26 . 3 ± 7 . 6 pA/pF . Rectification of the I-V curves was not changed by Prmt1 knockdown . To verify the role of Prmt1-mediated methylation in KCNQ2 channel function , the effect of Prmt1 inhibitors on KCNQ2 current was examined . In agreement with Prmt1 knockdown data , the treatment of MTA or furamidine dramatically reduced the KCNQ2 currents ( Figure 6c–d , Figure 6—source data 1 ) . The KCNQ2 currents were confirmed by XE991 treatment . Assuming that the XE991-sensitive portion entirely represents KCNQ2 currents , 100 μM MTA or 20 μM furamidine inhibited KCNQ2 currents by 79 . 9 ± 13 . 6% and 72 . 1 ± 2 . 1% , respectively . 10 . 7554/eLife . 17159 . 020Figure 6 . Methylation of KCNQ2 regulates its channel activity . ( a ) Representative current recordings from HEK293T cells expressing KCNQ2 with control shRNA vector ( left ) or a Prmt1 shRNA ( Prmt1 sh ) ( right ) . Currents were elicited by voltage steps from -70 mV to +40 mV with a subsequent step to -60 mV . ( inset ) Control immunoblotting for Prmt1 expression . ( b ) I-V relationships in control cells ( ● ) or Prmt1 knockdown cells ( ○ ) or nontransfected cells ( △ ) are shown . ( c–d ) Treatment of MTA ( c ) or furamidine ( d ) induces a reduction of KCNQ2 currents ( left ) . Both of them suppress KCNQ2 currents without altering shape of I-V curves ( right ) . XE991 , KCNQ channel blocker . ( e ) Representative current recordings from cells expressing WT and methylation defective mutants . ( inset ) The control western blot shows the expression of various KCNQ vectors . ( f ) The current density of KCNQ2 at +40 mV for each mutant shown in e . ANOVA Tukey test , ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 02010 . 7554/eLife . 17159 . 021Figure 6—source data 1 . Source data for Figure 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 021 We further examined the correlation between the Prmt1-induced methylation and KCNQ2 channel function by recording whole-cell currents in HEK293T cells expressing WT or KCNQ2 mutants . Compared to WT , the KCNQ2 mutants with reduced methylation exhibited remarkably decreased KCNQ2 currents ( Figure 6e ) . The current density at +40 mV was 304 . 4 ± 11 . 0 pA/pF ( n = 5 ) for WT KCNQ2 channels . We observed significantly decreased KCNQ2 currents in cells expressing RK mutants ( R333K , R345K , R353K , R435K , or 4RK ) : in these cells the current density ( pA/pF ) of steady-state currents at +40 mV were 18 . 2 ± 0 . 3 ( n = 3 , p<0 . 001 ) , 13 . 7 ± 0 . 5 ( n = 3 , p<0 . 001 ) , 14 . 8 ± 3 . 9 ( n = 3 , p<0 . 001 ) , 18 . 0 ± 2 . 2 ( n = 4 , p<0 . 001 ) , and 12 . 5 ± 1 . 6 ( n = 4 , p<0 . 001 ) , respectively ( Figure 6f , Figure 6—source data 1 ) . These data suggest that the arginine methylation of KCNQ2 is required for its channel activity . Importantly , KCNQ2 mutants showed no alteration in sensitivity to retigabine , KCNQ channel activator ( Figure 9d–g , summary data in Figure 9i , Figure 9—source data 1 ) , indicating that methylation regulates channel function through a distinct molecular mechanism not overall potential of channel activation . To assess whether methylation of KCNQ2 regulates its cell surface localization , we labeled total plasma membrane proteins of intact cells with membrane-impermeable biotinylation reagents , and isolated membrane proteins with streptavidin beads . The ratio of channels on the membrane to those in the total lysates was not significantly different between control cells and Prmt1 knockdown cells ( Figure 7—figure supplement 1 , left panel ) . KCNQ2-4RK also showed a similar amount of biotinylation when expressed in HEK293T cells ( Figure 7—figure supplement 1 , right panel ) . The absence of biotinylation on the cytosolic marker protein heat shock protein-90 confirmed specificity of the surface labeling . As the surface-resident channels remained unchanged , the KCNQ conductance should be determined by the open probability . KCNQ/M channels require a certain level of PIP2 in the cell membrane to maintain their activity ( Delmas and Brown , 2005; Suh and Hille , 2002; Winks et al . , 2005; Zhang et al . , 2003 ) . PIP2 acts to stabilize the open state of KCNQ2 channels resulting in increased open probability at all voltage levels . Thus , the depletion of PIP2 ( Suh and Hille , 2002; Winks et al . , 2005; Zhang et al . , 2003 ) or the reduction of PIP2 affinity of channel ( Kosenko et al . , 2012; Lee et al . , 2010; Hernandez et al . , 2008 ) can suppress KCNQ currents . Interestingly , the methylated arginine residues , R333 , R345 , R353 , and R435 , reside in the PIP2-binding domain of KCNQ2 ( Suh and Hille , 2008; Hernandez et al . , 2008 ) . This led us to postulate that methylation might affect the PIP2 binding affinity of KCNQ2 thereby regulating channel activities . To appraise the affinity of PIP2 for KCNQ2 , we utilized neomycin , which is widely used to sequester PIP2 ( Liscovitch et al . , 1994; Haider et al . , 2007; Suh and Hille , 2007 ) . Since neomycin is a polycation that binds specifically to PIP2 , it has been used to determine the PIP2 content in biological membranes ( Arbuzova et al . , 2000 ) . The neomycin sensitivity of ion channels such as Kir and KCNQ is well regarded as a measure of its PIP2 affinity ( Haider et al . , 2007; Suh and Hille , 2007; Kosenko et al . , 2012; Schulze et al . , 2003 ) . Accordingly , ion channels with high PIP2 affinities are expected to be less sensitive to neomycin than those with low PIP2 affinity . Inclusion of 10 μM neomycin in the patch pipette solution led to a moderate inhibition of the WT KCNQ2 current in HEK293T cells ( Figure 7a , Figure 7—source data 1 ) . In contrast , mutant channels with reduced methylation showed greatly increased inhibition by 10 mM neomycin . So , the remaining current at 10 min time points after rupturing the membrane is significantly reduced from 81 . 2 ± 1 . 9% ( WT , n = 5 ) to 48 . 7 ± 4 . 2% ( R333K , n = 5 , p<0 . 01; Figure 7a ) . We measured dose-responses to neomycin . In comparison with WT , methylation-deficient mutant KCNQ2 ( R333K , R345K , R353K , R435K ) exhibited a higher sensitivity to neomycin ( Figure 7b , Figure 7—source data 1 ) , suggesting lower PIP2 affinity . To directly examine whether these channels have different PIP2 sensitivity , we included 20 μM or 200 μM diC8-PIP2 in the patch pipette solution and measured current augmentation after rupturing the plasma membrane ( Figure 7c and d , Figure 7—source data 1 ) . WT KCNQ2 did not show increase in current by diC8-PIP2 even at 200 μM , suggesting that endogenous PIP2 might be at a saturating concentration . In contrast , methylation-deficient mutant KCNQ2 ( R333K , R345K , R353K , R435K , 4RK ) showed a dose-dependent increase in current but they did not reach to WT level at corresponding concentrations of PIP2 . Only R333K reached to the WT level when introduced to 200 μM PIP2 . On the other hand , 4RK appeared to be most severely affected in PIP2 binding . Altered PIP2 affinity in RK mutants was also assessed by a voltage-sensitive phosphatase from Danio rerio ( Dr-VSP ) , which hydrolyzes PIP2 at highly depolarized voltages ( e . g . , +100 mV ) and transiently reduces the PIP2 level ( Falkenburger et al . , 2010; Rjasanow et al . , 2015 ) . Dr-VSP was coexpressed with WT KCNQ2 or RK mutants and its activity was elicited by membrane depolarization . Consistent with a previous report ( Falkenburger et al . , 2010 ) , activation of Dr-VSP reduced WT KCNQ2 currents and currents were recovered quickly after PIP2 resynthesis on repolarization ( Figure 7—figure supplement 2 , Figure 7—figure supplement 2—source data 1 ) . When subjecting RK mutants to the same VSP activation protocol , the overall behavior was similar to the WT KCNQ2 channel; however , the recovery of current was slowed 2–4 folds ( Figure 7—figure supplement 2 , Figure 7—figure supplement 2—source data 1 ) . The mean time to the 70% maximum current ( T70 ) for R333K , R345K , R353K and R435Kwas 22 . 8 ± 3 . 7 , 24 . 1 ± 1 . 9 , 32 . 1 ± 2 . 6 , and 17 . 9 ± 1 . 5 s , respectively ( p<0 . 05 vs WT:9 . 5 ± 3 . 1 s ) . The slowed recovery after VSP reflects the reduced PIP2 affinity ( Rjasanow et al . , 2015 ) , further supporting for the reduced PIP2 affinity of RK mutants . Taken together , these data suggest that methylation of KCNQ2 channel regulates its interaction with PIP2 . 10 . 7554/eLife . 17159 . 022Figure 7 . Methylation of KCNQ2 determines its PIP2 affinity . ( a ) Pooled data show 10 μM neomycin-induced rundown of WT ( ● ) and R333K ( ○ ) KCNQ2 currents . KCNQ2 currents were normalized to KCNQ2 current at t = 0 . ( b ) Dose-response curves for neomycin measured at 10 min after rupturing the plasma membrane . Indicated concentration of neomycin was included in the patch pipette . Solid lines are Hill fits to the mean data for WT , R333K , R345K , R353K , and R435K giving IC50 values of 59 . 1 ± 8 . 3 , 9 . 2 ± 4 . 2 , 5 . 2 ± 0 . 2 , 5 . 4 ± 3 and 6 . 2 ± 8 μM and slopes of 0 . 8 ± 0 . 05 , 0 . 8 ± 0 . 3 , 0 . 6 ± 0 . 1 , 0 . 6 ± 0 . 2 , and 0 . 5 ± 0 . 2 , respectively . ( c ) Representative traces showing a significant increase in R333K KCNQ2 mutant by 200 μM diC8-PIP2 addition to patch pipette ( red line ) compared to controls ( black line ) . KCNQ2 ( WT ) did not show apparent augmentations . Currents were elicited by voltage steps from -60 mV to +40 mV . ( d ) from data such as shown in c current densities for 0 , 20 , and 200 μM diC8-PIP2 in the recording pipette were determined and plotted as bars with S . E . M . Currents were measured at >20 min after rupturing the plasma membrane . The numbers in parentheses indicate the number of cells . ANOVA Tukey test . **p<0 . 01 versus corresponding concentration to WT . #p<0 . 05; ##p<0 . 01; ###p<0 . 001 . ( e–g ) Effects of Prmt1 ( e-f ) or 4RK mutation ( g ) on the binding of KCNQ2 and PIP2 . Different amounts ( 10–500 pmol ) of PIP2 were spotted onto nitrocellulose membranes and analyzed by a protein-lipid overlay procedure using cell lysate prepared from HEK293T cells transfected with indicated expression vectors or the control pcDNA vector ( left panels ) . The control western blots are shown in right panels . ( h ) Left , representative traces show augmentation of M-current by 200 μM diC8-PIP2 in the patch pipette . Right , summary of M-current density from WT neurons and Prmt1+/- neurons with 0 , 20 , and 200 μM diC8-PIP2 in the recording pipette . Mean ± S . E . M . ANOVA Tukey test , ***p<0 . 001 . ( i–j ) the mean value of input resistance ( i ) , and threshold current for AP generation ( 100 ms duration; j ) from WT ( +/+ ) neurons , mutants ( +/- ) , and mutants loaded with 20 μM diC8-PIP2 ( +/-; PIP2 ) . ( k ) the mean number of APs in response to 1-s depolarizing current injection ( 200 pA ) from a WT ( +/+ ) neuron , a mutant ( +/- ) , and a mutant loaded with 20 μM diC8-PIP2 ( +/-; PIP2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 02210 . 7554/eLife . 17159 . 023Figure 7—source data 1 . Source data for Figure 7 . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 02310 . 7554/eLife . 17159 . 024Figure 7—figure supplement 1 . Surface expression of KCNQ channels was not affected by Prmt1 knockdown or 4RK mutation . Membrane surface biotinylation assays show that the ratio of channels on the membrane to those in the total lysates was not significantly different in control cells and Prmt1 knockdown cells ( left ) . Also , 4RK ( R333K/R345K/R353K/R435K ) substitution in KCNQ2 does not affect the correct trafficking of channel to the plasma membrane ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 02410 . 7554/eLife . 17159 . 025Figure 7—figure supplement 2 . Quantitative determination of the sensitivity of KCNQ2 channels to activation of Dr-VSP in HEK293T cells . Tail current amplitudes were used to measure current inhibition by Dr-VSP activation and its recovery . ( a–e ) Time course of tail current amplitude in a cell transfected with Dr-VSP and WT KCNQ2 ( a ) , R333K ( b ) , R345K ( c ) , R353K ( d ) , or R435K ( e ) . Membrane was held at -60 mV and depolarized to -20 mV for 300 ms every 1 s , except for shaded area where membrane was held at +100 mV for 2 s . Tail currents were measured during slow channel deactivation at -60 mV . Right , superimposed currents at time points before VSP activation ( a ) , after VSP activation ( b ) , and during recovery ( c ) . ( f ) Summary of tail current density at baseline ( initial current ) and right after VSP activation ( VSP ) . ( g ) time to the 70% maximum current amplitude ( T70 ) after VSP activation was measured for indicated WT KCNQ2 and RK mutants for recovery kinetics comparison ( n = 5–9 ) . Mean ± SEM shown . ANOVA Holm-Sidak test . *p<0 . 05 versus WT . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 02510 . 7554/eLife . 17159 . 026Figure 7—figure supplement 2—source data 1 . Source data for Figure 7—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 02610 . 7554/eLife . 17159 . 027Figure 7—figure supplement 3 . Comparison of XE991 sensitivity in GCs . Changes in firing rate ( a ) and threshold current ( b ) in WT ( +/+ ) neurons , mutants ( +/- ) , and mutants loaded with 20μM diC8-PIP2 ( +/-; PIP2 ) in response to 10 μM XE991 . Each connected line represents an individual neuron . Closed circles give the averaged changes . No change in input resistance and firing rate occurred with 10 μM XE991 in mutants , whereas addition of 10 μM XE991 to mutants after PIP2 dialysis produced a substantial decrease in threshold current and increase in firing rate , as in WT GCs . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 02710 . 7554/eLife . 17159 . 028Figure 7—figure supplement 3—source data 1 . Source data for Figure 7—figure supplement 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 02810 . 7554/eLife . 17159 . 029Figure 7—figure supplement 4 . Application of exogenous PIP2 had little effect on the excitability in WT neurons . ( a ) representative trace in the whole-cell current-clamp recording from WT GCs with no PIP2 ( left ) or 20 μM diC8-PIP2 in the patch pipette ( right ) in response to 1-s depolarizing current injection ( 200 pA ) . ( b ) the mean number of action potentials ( AP No . ) plotted against the eliciting currents ( from 100 pA to 400 pA , + 50 pA increment , during 1-s ) . ( c–f ) summary of input resistance ( c ) , threshold current for AP generation ( 100 ms duration; d ) , AP half-width ( e ) , and AP height ( f ) from WT mature dentate GCs when loaded with no PIP2 or 20 μM diC8-PIP2 . Mean ± S . E . M . The numbers in parentheses indicate the number of cell tested . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 02910 . 7554/eLife . 17159 . 030Figure 7—figure supplement 4—source data 1 . Source data for Figure 7—figure supplement 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 03010 . 7554/eLife . 17159 . 031Figure 7—figure supplement 5 . The effects of SK channel block and BK channel block on neuronal excitability of control or furamidine-pretreated GCs . Spike trains were evoked by injecting 1-s depolarizing current pulses of different intensities ( 100–250 pA ) into the cell before and after application of apamin ( a–f ) or paxilline ( g–l ) in control or furamidine ( 20 μM , 1 hr ) -pretreated GCs . Examples of APs during 200-pA current injection without ( left ) and with the SK channel blocker apamin ( 100 nM ) or the BK channel blocker paxilline ( 5 μM ) ( a , d , g , j ) . Number of APs evoked after injection of different currents ( b , e , h , k ) . Apamin increases the firing frequency in both control and furamidine-pretreated GCs . Paxilline has little effect on firing frequency in both control and furamidine-pretreated GCs . The mean value of input resistance ( left ) and threshold current for AP generation ( 100 ms duration; right ) ( c , f , i , l ) . Error bars , S . E . M . scale bar; 40 mV , 300 msec . *p<0 . 05; **p<0 . 01 by paired Student’s t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 03110 . 7554/eLife . 17159 . 032Figure 7—figure supplement 5—source data 1 . Source data for Figure 7—figure supplement 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 032 To further confirm that arginine methylation modulates the PIP2 binding affinity of KCNQ2 , we used a protein-lipid overlay ( PLO ) assay with an HA-KCNQ2-C ( Figure 7e–g ) . PIP2 strips were incubated with lysates from control or Prmt1 knockdown cells expressing HA-KCNQ2-C . We found that the PIP2 affinity of KCNQ2 decreased in the Prmt1 knockdown cells compared to that in the control cells ( Figure 7e ) . In contrast , overexpression of Prmt1 with HA-KCNQ2 enhanced the affinity of KCNQ2 for PIP2 ( Figure 7f ) . Compared to that of WT KCNQ2 , methylation-deficient mutant channels also had very little affinity for PIP2 ( Figure 7g ) , consistent with electrophysiological measurements . We then tested whether PIP2 loading could also rescue M-currents in neurons of Prmt1 +/- mice . As shown in Figure 7h , M-current density in Prmt1+/- neurons was restored to the WT level by PIP2 loading . Introduction of 20 and 200 μM diC8-PIP2 to Prmt1+/- GCs increased M-currents to 27 . 3 ± 2 . 9 ( n = 3 ) and 29 . 9 ± 5 . 9 pA ( n = 5 ) , respectively , which are insignificantly different from that of WT ( 31 . 65 ± 2 . 03 pA , n = 22 , p>0 . 05; Figure 7h , Figure 7—source data 1 ) , while exogenous PIP2 had little effects on M-currents in WT GCs ( p>0 . 05 , Figure 7h , Figure 7—source data 1 ) . As expected from opening of the M-channels by PIP2 , the input resistance decreased significantly from 214 . 8 ± 7 . 7 MΩ to 163 . 8 ± 16 . 4 MΩ ( Figure 7i , Figure 7—source data 1 ) , and the AP threshold current increased from 115 . 6 ± 3 . 3 pA to 175 ± 12 . 1 pA ( Figure 7j , Figure 7—source data 1 ) . These results indicate that declined methylation caused by Prmt1 depletion impairs KCNQ channel activity via reduction in channel-PIP2 interaction . Since PIP2 restored the diminished KCNQ currents , we asked whether PIP2 could also reduce the observed neuronal hyperexcitability in Prmt1+/- neurons . We found that PIP2 loading decreased the firing rate toward normal values in Prmt1+/- GCs . For example , the AP frequency of Prmt1+/- GCs in response to a 200 pA depolarizing current was 31 . 2 ± 2 . 2 Hz , and decreased to 15 . 5 ± 2 . 9 Hz following PIP2 loading ( Figure 7k , Figure 7—source data 1 ) . In addition , Prmt1+/- neurons also regained the sensitivity to XE991 ( Figure 7—figure supplement 3 , Figure 7—figure supplement 3—source data 1 ) . In contrast , PIP2 application had little effects on the excitabilities of WT neurons ( Figure 7—figure supplement 4 , Figure 7—figure supplement 4—source data 1 ) . These results suggest that the impaired KCNQ channel function in the hippocampus of Prmt1+/- mice can be restored by PIP2 addition . Although the recovery of excitability with PIP2 application was significant it was incomplete , suggesting the possible involvement of other Prmt1 target ( s ) in neuronal hyperexcitability . As immediate candidates , we tested other K+ channels such as SK and BK channels in Prmt1 effects which are known to regulate membrane excitability of dentate gyrus GCs ( Brenner et al . , 2005 ) . The data obtained by using apamin and paxilline , specific SK and BK channel blockers , respectively , ruled out that the contribution of these channels in hyperexcitability of Prmt1-deficient GCs ( Figure 7—figure supplement 5 , Figure 7—figure supplement 5—source data 1 ) . Further study will be required to identify additional targets of Prmt1 . Taken together , genetic deletion or pharmacological inhibition of Prmt1 reduces the affinity between the KCNQ channel and PIP2 , leading to reduction of KCNQ currents accompanied by neuronal hyperexcitability . Prmt1 seems to play a key role in physiological functions of KCNQ channels by controlling its interaction with PIP2 at the basal state . KCNQ channels are also known to contribute to cellular protections in pathologic condition . In particular , the enhancement of KCNQ currents in response to oxidative stress led to a dramatic reduction of the AP firing frequency that may prevent neuronal death ( Patel , 2004 ) . We then examined whether Prmt1-mediated methylation is functionally associated with the mechanisms of cytoprotective enhancement of KCNQ currents under oxidative stress . To investigate the role of Prmt1 in reactive oxygen species ( ROS ) -induced augmentation of KCNQ channel and consequential neuronal silencing , neuronal excitabilities in WT and Prmt1+/- GCs are evaluated in response to H2O2 . Consistent with previous studies ( Gamper et al . , 2006 ) , H2O2 treatment strongly reduced neuronal excitabilities in WT GCs ( Figure 8a–b ) . The AP firing rate during injection of a 200 pA depolarizing current was 9 . 3 ± 2 . 1 Hz , and decreased to 1 . 4 ± 0 . 9 Hz ( Figure 8c , Figure 8—source data 1 ) after H2O2 treatment . Consistent with the KCNQ current augmentation , the input resistance was decreased from 163 . 8 ± 24 . 9 to 116 . 3 ± 21 . 8 MΩ ( Figure 8d , Figure 8—source data 1 ) . In contrast , H2O2 application did not lower firing rates and input resistance in GCs of Prmt1+/- mice , although their firing rates and input resistance were elevated compared to WT ( Figure 8e–h , Figure 8—source data 1 ) . Consistently , the pharmacological inhibition of arginine methylation with MTA ( Figure 8i–l , Figure 8—source data 1 ) or furamidine ( Figure 8m–p , Figure 8—source data 1 ) also blocked neuronal silencing and input resistance reduction in WT GCs in response to the oxidative stress inflicted by H2O2 . These data indicate that Prmt1-mediated protein arginine methylation is necessary for ROS-induced neuronal silencing via KCNQ channel activation . 10 . 7554/eLife . 17159 . 033Figure 8 . Prmt1 is involved in Oxidative stress-induced silencing of neurons . Spike trains were evoked by injecting 1-s depolarizing current pulses of different intensities ( 100–250 pA ) into the cell before ( black ) and after ( red ) application of H2O2 . Left panels show representative recordings in normal extracellular medium ( a , e , i , m ) and middle panels after application of 500 μM H2O2 for 10 min ( b , f , j , n ) . H2O2 had little effect on firing rates in Prmt1+/- GCs ( f; n = 5 , four mice ) , MTA ( j; n = 4 , four mice ) - or furamidine ( n; n = 7 , four mice ) -treated WT GCs than in untreated WT GCs ( b; n = 8 , four mice ) . Summary diagrams ( right column ) compare firing rates or input resistance before ( black ) and after ( red ) application of H2O2 in WT ( c , d ) , Prmt1+/- ( g , h ) , MTA ( k , l ) - or furamidine ( o , p ) -treated WT dentate GCs . **p<0 . 01; ***p<0 . 001 by Student’s t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 03310 . 7554/eLife . 17159 . 034Figure 8—source data 1 . Source data for Figure 8 . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 034 Consistent with a previous study ( Gamper et al . , 2006 ) , H2O2 induced a sharp augmentation of WT KCNQ2 channels heterologously expressed in HEK293T cells ( Figure 9a ) . The mean current augmentation induced by 500 μM H2O2 was 2 . 6 ± 0 . 2-fold ( Figure 9h ) . At the concentration of 500 μM , the H2O2 effect usually reached a plateau within 7~8 min of application . The H2O2 effect was completely reversed by the addition of a reducing agent , dithiothreitol ( DTT , 2 mM ) , suggesting that this effect is due to a reversible oxidative modification . Notably , DTT-reduced KCNQ2 currents in H2O2-treated cells were back to the resting level within several minutes , but did not inhibit them further ( Figure 9a ) . Also , DTT had no effect on KCNQ2 currents in non- H2O2-treated cells ( data not shown ) . When protein arginine methylation was blocked with the treatment of MTA ( Figure 9b ) or furamidine ( Figure 9c ) , KCNQ2 channels appeared to be insensitive to H2O2 with the currents augmented by H2O2 ( 500 μM ) by 0 . 9 ± 0 . 02-fold and 0 . 8 ± 0 . 03-fold , respectively ( Figure 9h , Figure 9—source data 1 ) . We then examined whether the sensitivity to H2O2 is abrogated by the KCNQ2 mutants with reduced methylation . As shown in Figure 9d–g , the mutant channels were insensitive to H2O2 , compared to WT KCNQ2 . Currents of the R333K , R345K , R353K and R435K mutant channels were augmented by 1 . 6 ± 0 . 3- , 0 . 9 ± 0 . 1- , 0 . 7 ± 0 . 1- and 1 . 1 ± 0 . 1-fold , respectively ( Figure 9h , Figure 9—source data 1 ) . 10 . 7554/eLife . 17159 . 035Figure 9 . Activation of KCNQ2 channels by H2O2 is sensitive to methylation . ( a–g ) Time course for the effect of 500 μM H2O2 on KCNQ2 currents . WT KCNQ2 ( a-c ) , R333K ( d ) , R345K ( e ) , R353K ( f ) or R435K ( g ) were expressed in HEK293T cells . For experiment ( b ) & ( c ) , cells were pretreated with MTA ( 100 μM ) or furamidine ( 20 μM ) , respectively . Whole-cell currents were monitored by 1-s hyperpolarizing steps to -60 mV from a holding potential of -20 mV at 5-s intervals . H2O2 ( 500 μM ) , DTT ( 2 mM ) , the M-channel blocker , XE991 ( 50 μM ) , and M-channel activator , retigabine ( RTG; 10 μM ) , were applied during the periods indicated by the bars . Plotted are normalized current amplitudes versus time during the experiment . Representative current traces from each experiment ( not normalized ) are depicted in the insets . ( h ) Summarized data for KCNQ2 activation induced by H2O2 . Letters indicate statistically distinct groups ( ANOVA Tukey , p<0 . 01 ) . ( i ) Summarized data for KCNQ2 activation induced by retigabine . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 03510 . 7554/eLife . 17159 . 036Figure 9—source data 1 . Source data for Figure 9 . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 03610 . 7554/eLife . 17159 . 037Figure 9—figure supplement 1 . Time course for the effect of 500 μM H2O2 on Triple-Cys mutant of KCNQ2 currents ( left ) . Summarized data for Triple-Cys mutant of KCNQ2 currents activation induced by H2O2 or retigabine ( right ) . ( b ) Treatment of MTA induces a reduction of KCNQ2 ( WT ) and Triple-Cys mutant of KCNQ2 currents ( left ) . Both of them suppress KCNQ2 currents without altering inhibition levels by MTA ( right ) . XE991 , KCNQ channel blocker . ( c ) Treatment of furamidine induces a reduction of KCNQ2 ( WT ) and Triple-Cys mutant of KCNQ2 currents ( left ) . Both of them suppress KCNQ2 currents without altering inhibition levels by furamidine ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 03710 . 7554/eLife . 17159 . 038Figure 9—figure supplement 1—source data 1 . Source data for Figure 9—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 17159 . 038 Recently , Gamper et al . ( Gamper et al . , 2006 ) showed that a triple cysteine pocket in the S2-S3 linker is critical for the H2O2-induced enhancement of KCNQ channel . Interestingly , MTA and furamidine reduced activities of the triple-Cys mutant of KCNQ2 by a similar extent as it did those of the WT KCNQ2 channel , while the mutant was insensitive to H2O2 as expected ( Figure 9—figure supplement 1 , Figure 9—figure supplement 1—source data 1 ) . This result indicates that the H2O2 activation pathway is unnecessary for the arginine methylation-mediated regulation of Prmt1 . Taken together , Prmt1 is prerequisite for cytoprotective enhancement of KCNQ currents in the milieu of oxidative stress .
In this study , we describe a novel regulatory mechanism whereby methylation of KCNQ2 affects KCNQ/M channel activities and overall excitabilities of the brain . The heterozygous reduction of the Prmt1 gene dose in mice caused spontaneous seizures . The elevated neuronal excitability in Prmt1+/- mice appears to be a direct result of reduced KCNQ/M channel activities in that KCNQ2 from the Prmt1+/- brain was less asymmetrically methylated and showed diminished affinity to PIP2 . Furthermore , the activation of KCNQ channel under oxidative stress as well as its basal activity at rest depends on protein arginine methylation . It is well known that ROS generated under oxidative stress induces augmentation of KCNQ currents in neurons ( Gamper et al . , 2006 ) . The resulting reduction in neuronal discharge lowers electrical activity of the cell and energy consumption , thereby providing an important mechanism of protecting neurons from the oxidative stress-induced death ( Gamper et al . , 2006; Won et al . , 2002 ) . Considering that oxidative stress is also involved in induction of seizures ( Patel , 2004 ) , the impairment of neuroprotective KCNQ activation in response to oxidative stress , may contribute to spontaneous seizures together with the reduced basal activity of KCNQ channels in Prmt1+/- mice . Protein arginine methylation is a unique post-translational modification that increases the KCNQ/M channel function unlike other known regulatory signaling pathways , and thus provides a platform for the design of novel therapeutic strategies for epilepsy and other neuronal hyperexcitability disorders . Our results find that 4 arginine residues ( R333 , R345 , R353 , and R435 ) are the major methylation sites in KCNQ2 . Interestingly , all of the methylation sites reside within the C-terminal PIP2 binding domain . This is known as a 'hot spot' , as it serves as binding sites for multiple signaling molecules as well as PIP2 ( Delmas and Brown , 2005; Hernandez et al . , 2008; Zhang et al . , 2003 ) . The observation that decreased methylation lowers PIP2 affinity proposes that Prmt1 methylation on the arginine residues would increase channel-PIP2 interaction . Indeed , previous computational analyses revealed that methylation renders the guanidinium of arginines to be more electron-rich and thus to exhibit higher pKa values ( Shearer , 2008 ) , which would stabilize the positive charge of the arginine residue in proteins in the physiological condition and thereby likely facilitate their electrostatic interaction with negatively charged molecules . Thus , it is conceivable that arginine methylation is required for the KCNQ channel activity because it promotes electrostatic binding of KCNQ proteins to PIP2 . Various KCNQ channel mutations associated with epilepsy have been shown to cause a reduction in M-channel activity , which all can lead to membrane depolarization and increased neuronal firing via diverse mechanisms . Mutations in the C-terminal region of KCNQ channels reduce channel activities by interfering with channel targeting to the surface membrane ( Schwake et al . , 2000 ) , protein stability ( Soldovieri et al . , 2006 ) , interaction with calmodulin ( Jentsch , 2000 ) , or modulation by cAMP ( Schroeder et al . , 1998 ) . Mutations affecting the pore region of KCNQ2 may reduce currents by affecting ion channel conductance , while mutations at the S4 domain affect channel gating and increase the threshold for channel activation ( Maljevic et al . , 2010 ) . It is interesting in this context that R353G , a mutant of one of the major methylation sites identified in this study , is linked to familial epilepsy , likely due to reduced calmodulin binding and a consequent defect in membrane trafficking ( Jentsch , 2000; Etxeberria et al . , 2008 ) . However , a recent study proposed an additional pathophysiological mechanism involving a diminished PIP2 affinity of the channel for this epileptic KCNQ2 mutation ( Kosenko et al . , 2012 ) . This is in agreement with our current study indicating that disruption of KCNQ2 channel-PIP2 interaction caused by decreased methylation is a key mechanism linked to neuronal hyperactivity . Like KCNQ2 , KCNQ3 channel contains conserved methylation sites within the PIP2 biding domain ( Figure 5—figure supplement 3 ) suggesting that arginine methylation might be also involved in regulation of channel-PIP2 interaction in KCNQ3 channels as well . Considering that the KCNQ2 and KCNQ3 complex produces M-currents , hypo-methylation of either KCNQs might significantly diminish M-currents and neuronal hyperexcitability . The positively charged arginines are also important for electrostatic interaction with PIP2 in other PIP2-sensitive channels and transporters such as Kir2 ( Hansen et al . , 2011; Huang et al . , 1998; Lopes et al . , 2002 ) , and GIRK ( Whorton and MacKinnon , 2011 ) . PIP2 is required for proper function of many plasma membrane ion channels and transporters ( Suh and Hille , 2008; Hilgemann et al . , 2001 ) . Thus , methylation can be a general mechanism for regulating ion flux physiology . Consistent with this , our results show that the effects of KCNQ channel block on firing rates are not the same as those of Prmt1 block ( Figure 3 ) , implying a possible involvement of other Prmt1 target ( s ) in neuronal hyperexcitability and seizures . PIP2 loading to Prmt1+/- GCs recovered the firing properties significantly but incompletely . It appears that SK and BK channels are not involved in this matter . Thus further studies are required to identify other Prmt1 targets in the control of neuronal excitability . Regardless , the results presented here provide the first direct evidence of functional role for protein arginine methylation in the control of neuronal excitability and also offers a platform to further investigate the physiological and pathophysiological roles of KCNQ/M channel activity in CNS .
Prmt1+/- mice were obtained from EUCOMM consortium ( C57BL/6N agouti background ) and backcrossed onto C57BL/6J background for at least 5 generations before being used for the experiment as previously described ( Choi et al . , 2012 ) . All animal experiments were approved by the Institutional Animal Care and Research Advisory Committee at Sungkyunkwan University School of Medicine Laboratory Animal Research Center ( Approval No . IACUC-11-39 ) . Mice were maintained in C57BL/6J background and bred through heterozygous matings . For this study , we have used only male mice to avoid the complication due to the menstural cycles with female mice . For EEG surgery and recording , long-term video-EEG monitoring was performed as described previously ( Jeon et al . , 2011 ) . Male Prmt1+/- mice and their littermate WT control mice were anesthetized by intraperitoneal injection of 1% ketamine ( 30 mg/kg ) and xylazine hydrochloride ( 4 mg/kg ) . Surgery was performed using a stereotaxic apparatus ( Kopf Instruments , USA , California ) . Recordings were obtained using skull screws ( stainless , 1 . 0 mm in diameter ) , which were positioned in AP −1 . 8 mm and L −2 . 1 mm from the bregma with grounding over the cerebellum . Electrical activities were recorded after being amplified ( ×1200 ) , bandpass-filtered at 0 . 1–70 Hz , and digitized with a 400-Hz sampling rate using a digital system ( Comet XL , Astro-Med , Warwick , RI , USA ) . Video-EEG signals were continuously recorded 24 hr per day for 7 days , and the waveforms and epileptiform activities were analyzed offline using Matlab , PSG Twin 4 . 3 ( Astro-Med , USA , West Warwick , RI ) , and pClampfit 10 . 2 ( Axon Instruments , USA , California ) . To obtain colored power spectra , EEG signals were filtered from 1 to 70 Hz . Colored power spectra were calculated and drawn with Fourier transformation of 2-s window sizes . The duration of seizures , the number of seizures , the frequency of seizure spikes , and seizure score were measured and analyzed . The seizure score was classified on the basis of Racine’s scale ( Racine , 1972 ) : stage 0 , no changes in behavior; stage 1 , arrest , immobility and rigid posture; stage 2 , jerking , tail rattling , staring with mouth clonus , and head nodding; stage 3 , forelimb clonus; stage 4 , rearing with forelimb clonus; stage 5 , body shaking , wild running and jumping; stage 6 , tonic-clonic seizures . Two Prmt1+/- mice only showed interictal-like activities without ictal activities and were removed from analysis . The open-field task was used to assess locomotor activity ( Jung et al . , 2013 ) . The open-field box was made of white plastic ( 40 × 40 × 40 cm ) and the open field was divided into a central field ( center , 20 × 20 cm ) and an outer field ( periphery ) . Male Prmt1+/- mice and their littermate WT control mice were used . Individual mice were placed in the periphery of the field and the paths of the animals were recorded with a video camera . The total distance traveled for 30 min and the time spent in the central area for 5 min were analyzed using the program EthoVision XT ( Noldus , USA , Virginia ) . Brain slices were prepared from male Prmt +/- mice and their littermate WT control mice aged 4–6 weeks old . Mice were killed by decapitation after being anesthetized with pentobarbital sodium , and the whole brain was immediately removed from the skull and chilled in artificial cerebrospinal fluid ( aCSF ) at 4°C . Transverse hippocampal slices ( 350 μm thick ) were prepared using a vibratome ( VT1200S , Leica , Germany , Nussloch ) . Slices were incubated at 35°C for 30 min and thereafter maintained at 32°C until in situ slice patch recordings and fluorescence microscopy . Hippocampal granule cells of the dentate gyrus were visualized using an upright microscope equipped with differential interference contrast optics ( BX51WI , Olympus , Japan , Tokyo ) . Whole-cell current clamp techniques were used for excitability of dentate granule cells ( GCs ) . The pipette solution contained ( in mM ) : 143 K-gluconate , 7 KCl , 15 HEPES , 4 MgATP , 0 . 3 NaGTP , 4 Na-ascorbate , and 0 . 1 EGTA/or 10 BAPTA with the pH adjusted to 7 . 3 with KOH . The bath solution ( or aCSF ) for the control experiments contained the following ( in mM ) : 125 NaCl , 25 NaHCO3 , 2 . 5 KCl , 1 . 25 NaH2PO4 , 2 CaCl2 , 1 MgCl2 , 20 glucose , 1 . 2 pyruvate , and 0 . 4 Na-ascorbate , pH 7 . 4 when saturated with carbogen ( 95% O2 and 5% CO2 ) . All bath solution included in 20 μM bicuculine and 10 μM CNQX to block the inhibitory synaptic signal . The perfusion rate of the bathing solution and the volume of the recording chamber for slices were 2 . 2 ml/min and 1 . 2 ml , respectively . Patch pipettes with a tip resistance of 3–4 MΩ were used . The series resistance ( Rs ) after establishing whole-cell configuration was between 10 and 15 MΩ . Electrophysiological recordings were made in somata with EPC-8 amplifier ( HEKA Instruments , Germany , Lambrecht/Pfalz ) . Experiments were performed at 32 ± 1°C . The following parameters were measured: ( 1 ) the resting membrane potential , ( 2 ) the input resistance ( Rin , membrane potential changes ( V ) for given hyperpolarizing current ( −35 pA , 600 ms ) input ) , ( 3 ) AP threshold current ( current threshold for single action potential generation , 100 ms duration ) , ( 4 ) AP height; defined as the peak relative to the most negative voltage reached during the afterhyperpolarization immediately after the spike , ( 5 ) AP half-width; measured as the width at half-maximal spike amplitude , ( 6 ) F-I curve ( firing frequencies ( F ) against the amplitude of injected currents ( I ) ; 100–400 pA ) . We excluded data for analysis when series resistance exceeded 20 MΩ or when resting membrane potential was more positive than −60 mV . The whole-cell voltage clamp technique was used to measure K+ currents . Whole-cell K+ currents , evoked in response to voltage step to potentials ranging from –70 mV to +30 mV ( in 10 mV increments , 1 s duration ) from a holding potential of –60 mV , were examined in WT and Prmt1+/- dentate GCs . Expression vectors encoding human KCNQ2 and Triple-Cys mutant of KCNQ2 channel were generously provided by Dr . Shapiro ( University of Texas Health Science Center ) ( Falkenburger et al . , 2010 ) . To construct the HA -tagged KCNQ2 and HA-tagged KCNQ2-C ( aa320-840 ) expression vectors , the entire coding region of KCNQ2 or the corresponding sequence of KCNQ2 aa320-840 were amplified by polymerase chain reaction ( PCR ) and the PCR products were cloned into pcDNA3 . 1-HA vector ( Clontech , USA , California ) . A series of KCNQ2 arginine-to-lysine mutants were generated by site-directed mutagenesis using a QuickChange kit ( Stratagene , La Jolla , USA , California ) . To generate GST fusion protein of KCNQ2 aa320-449 , the corresponding region was amplified and cloned into PGEX-4T-1 vector ( Clontech , USA , California ) . Flag-tagged Prmt1 and Dr-VSP-GFP vectors were described in the previous reports ( Malkki , 2014; Keum et al . , 2016 ) . Stable cell lines exhibiting knockdown of Prmt1 were generated using Mission shRNA vector ( Sigma , USA , Missouri ) . The PLK0 . 1 TR control vector ( random 18mer ) was used for generating the control cell lines . 293T cells were seeded at 3 × 10 ( Rogawski , 2000 ) cells per 10 cm plate and transfected with 10 μg of the shRNA vector together with 5 μg of VSVg and 5 μg of Δ8 . 2 helper plasmids using Effectene ( Qiagen , Germany , Hilden ) . The medium was changed 12 hrs after transfection . After 48 hrs , the lentivirus-containing culture media were harvested . 293T cells were infected by 0 . 45 μm-filtered supernatant from virus-producing cells in the presence of 8 μg/ml polybrene . After 2 days , the puromycin-resistant cells were maintained with 2 μg/ml puromycin . Medium containing puromycin was replaced every 2 days until puromycin-resistant stable cell lines were established . Human embryonic kidney cell line HEK293T cells ( ATCC , USA , Virginia ) were maintained in Dulbecco’s modification of Eagle’s medium ( Gibco-BRL , USA , California ) supplemented with 10% fetal bovine serum at 37°C in 5% CO2 . For transient transfections , cells were transfected using Lipofectamine 2000 reagents ( Invitrogen , USA , California ) and green fluorescent protein was used as a reporter to label the transfectants . The KCNQ currents from HEK293T cells were measured with the whole-cell patch clamp technique . Voltage clamp was performed using an EPC-10 amplifier ( HEKA Instruments , Germany , Lambrecht/Pfalz ) at a sampling rate of 10 kHz filteredat 1kHz . Data were acquired using an IBM-compatible computer running Patchmaster software ( HEKA Instruments , Germany , Lambrecht/Pfalz ) . The patch pipettes were pulled from borosilicate capillaries ( Hilgenberg-GmbH , Germany , Malsfeld ) using a Narishige puller ( PC-10 , Narishige , Japan , Tokyo ) . The patch pipettes had a resistance of 2–3 MW when filled with the pipette solution containing ( in mM ) 140 KMeSO4 , 20 KCl , 20 HEPES , 0 . 5 Na-GTP , 5 Mg-ATP , 4 vitamin C , and 10 1 , 2-bis ( 2-aminophenoxy ) ethane N , N , N_ , N_-tetraacetic acid ( BAPTA ) , pH 7 . 4 adjusted with KOH . The normal external solution was as follows ( in mM ) : 143 NaCl , 5 . 4 KCl , 5 HEPES , 0 . 5 NaH2PO4 , 11 . 1 glucose , 0 . 5 MgCl2 , and 1 . 8 CaCl2 , pH 7 . 4 adjusted with NaOH . Pipette capacitance was compensated after formation of a gigaohm seal . Access resistance was typically 2 . 8–3 . 2 MΩ . The perfusion system was a homemade 100-μl perfusion chamber through which solution flowed continuously at 5 ml/min . The currents from HEK293T cells were studied by holding the cell at −60 mV , and 1-s steps from −70 to 40 mV in 10-mV increments were applied , followed by 1-s pulses to −60 mV . All recordings were carried out at room temperature ( RT ) . Currents were analyzed and fitted using Patch master ( HEKA Instruments , Germany , Lambrecht/Pfalz ) and Origin ( ver . 6 . 0 , Microcal , USA , Massachusetts ) software . All values are given as mean ± standard error . I/V relationship were obtained by plotting the outward current at the end of a 1-s test pulse as a function of the test potential . Cells were lysed in NETN lysis buffer containing 1 mM phenylmethylsulfonyl fluoride , 1 mM sodium fluoride , 1 mM sodium orthovanadate , 2 μg/ml aprotinin , 2 μg/ml leupeptin , and 1 μg/ml pepstatin A for 1 hr at 4ºC . Following lysis , the cells were centrifuged at 13 , 000 rpm for 10 min at 4°C . Total cell extracts were incubated with appropriate antibodies for 4 hrs at 4°C and Protein G plus/Protein A-agarose ( Calbiochem , USA , California ) was added , and the incubation was continued for 2 hrs . After incubation , immunoprecipitates were washed three times with NETN lysis buffer and subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis ( SDS-PAGE ) and transferred to nitrocellulose membranes . The membranes were blocked with TBS containing 5% skim milk and 0 . 1% Tween 20 for 30 min at RT . The membranes were incubated with appropriate primary antibody for 2 hr at RT . After washing with TBS containing 0 . 1% Tween 20 three times for 10 min , the membranes were incubated with horseradish peroxidase-conjugated secondary antibody for 1 hr at RT . After washing with TBS containing 0 . 1% Tween 20 three times for 10 min , proteins were detected using the enhanced chemiluminescent ( ECL ) system ( GE Healthcare , USA , Chicago ) . Primary antibodies used in this study were as follows: Prmt1 ( 1:1000 , Millipore , USA , Massachusetts ) , ADMA ( 1:1000 , Cell Signaling , USA , Massachusetts ) , α-tubulin ( 1:2000 , Cell signaling , USA , Massachusetts ) , KCNQ2 ( 1:1000 Abcam , UK , Cambridge ) , Hsp90 and PRMT8 ( 1:1000 , Santa Cruz , USA , Texas ) , α-Flag , α-Myc , α-HA and α-GST ( 1:1000 , Abfrontier , South Korea , Seoul ) , and GAPDH ( 1:2000 , Abfrontier , South Korea , Seoul ) . For mass analysis of KCNQ2 proteins , HA-KCNQ2 expression vectors were tranfected into HEK293T cells . Cells were lysed in NETN lysis buffer ( 50 mM Tris–HCl , pH 7 . 5 , 150 mM NaCl , 0 . 5% Nonidet P-40 ) containing 1 mM phenylmethylsulfonyl fluoride , 1 mM sodium fluoride , 1 mM sodium orthovanadate and 2 μg/ml aprotinin , 2 μg/ml leupeptin and 1 μg/ml pepstatin A for 1 hr at 4°C . Following lysis , cells were centrifuged at 13 000 r . p . m . for 10 min at 4°C . Total cell extracts were incubated with anti-HA antibody for 2 hrs at 4°C and protein G plus/protein A-agarose ( Calbiochem , USA , California ) was added , and the incubation was continued for 2 hrs . After incubation , immunoprecipitates were washed three times with NETN lysis buffer and subjected to SDS-PAGE electrophoresis . In-gel digestion was carried out with 12 . 5 ng/µl sequencing grade modified trypsin ( Promega , USA , Wisconsin ) in 50 mM NH4HCO3 buffer ( pH 7 . 8 ) at 37°C for overnight . Produced tryptic peptides were extracted with 5% formic acid in 50% ACN solution at room temperature for 20 min . The supernatants were collected and dried with SpeedVac . Resuspended samples in 0 . 1% formic acid were purified and concentrated using C18 ZipTips ( Millipore , USA , Massachusetts ) before MS analysis . The tryptic peptides were loaded onto a fused silica microcapillary column ( 12 cm × 75 µm ) packed with C18 reversed phase resin ( 5 µm , 200 Å ) . Peptide separation was conducted with a series of step gradients composed of initial isobaric flow for 5 min with 3% solvent B ( 0 . 1% formic acid in acetonitrile ) , then linear gradient from 3% to 40% for 40 min . At the end of each running , 90% of solvent B was eluted for 10 min with the flow rate 250 nL/min . The% gradient of solvent B was against solvent A ( 0 . 1% formic acid in H2O ) . The column was directly connected to LTQ linear ion-trap mass spectrometer ( ThermoFinnigan , USA , New Jersey ) equipped with a nano-electrospray ion source . The electrospray voltage was set at 1 . 95 kV , and the threshold for switching from MS to MS/MS was 500 . The normalized collision energy for MS/MS was 35% of main radio frequency amplitude ( RF ) and the duration of activation was 30 ms . All spectra were acquired in data-dependent scan mode . Each full MS scan was followed by five MS/MS scan corresponding from the most intense to the fifth intense peaks of full MS scan . The acquired LC-ESI-MS/MS fragment spectra were searched against a modified NCBI protein reference database using Mascot program . The searching conditions were trypsin enzyme specificity , a permissible level for three missed cleavages , peptide tolerance; the amu , a mass error of amu on fragment ions , fixed modification of carbamidomethylation of cysteine , and variable modifications of oxidation of methionine , monomethylation of arginine and dimethylation of arginine . Purified recombinant GST-KCNQ2 ( amino acids 320–449 ) proteins were incubated for 3 hrs at 37°C with 1 μg of recombinant GST-Prmts in 30 μl methylation buffer ( 50 mM Tris/HCl , pH 7 . 5 ) supplemented with 0 . 5 μCi of S-adenosyl-L-[methyl- 3H]methionine ( PerkinElmer Life Sciences , USA , Massachusetts ) . Reactions were stopped by adding 2× SDS-PAGE sample buffer and heating . Samples were analyzed by SDS-PAGE and fluorography . GST-KCNQ2 and GST-tagged Prmts were bacterially expressed purified . Myc-tagged Prmt5 was expressed in HEK293T cells and purified by immunoprecipitation and elution . Protein lipid overlay assays were performed as described previously ( Dowler et al . , 2002 ) . Briefly , extracts were prepared from HEK293T control cells or Prmt1 knockdown cells transfected with HA-KCNQ2 expression vectors by resuspending in modified NETN buffer ( 20 mM Tris , pH 7 . 5 , 150 mM NaCl , 0 . 5% Nonidet P-40 , 1 mM EDTA , 1 mM sodium fluoride , 1 mM Na3VO4 , 1 mM phenylmethylsulfonyl fluoride , 1 µg/ml Leupeptin , 0 . 2 µg/ml Pepstatin A ) . PIP2 ( 10–500 pmol ) was spotted on Hybond-C extra membrane ( Amersham Biosciences , USA , New Jersey ) . Spotted membranes were blocked with 0 . 2% fatty acid-free BSA ( Sigma USA , Missouri ) in TBST for 1 hr at room temperature . The membranes were incubated with cell lysates for 2 hrs at room temperature , followed by washing and detection with anti-HA antibody . Data are presented as mean ± standard error of mean . Student’s t-test or ANOVA Tukey tests were performed with Sigmaplot 12 . 0 after normal distribution of data was examined Shapiro-Wilk test . To detect a difference in mean value between two independent groups with 80% power , a sample-size of 4 cells in each group is needed when Student’s t-test is used with parameter values that are typical for the present study ( alpha = 0 . 05 , the effect size d = 3 ) . When ANOVA Tukey tests is used with parameter values that are typical for the present study ( alpha = 0 . 05 , the effect size f = 30 ) , sample-size of 3 cells in each group is needed for a difference in mean value between two independent groups with 80% power . We used G*power program ( Faul et al . , 2009 ) ( http://www . gpower . hhu . de/ ) for this estimation . We obtained data from more than 4 cells for Student's t-test or more than 3 cells for ANOVA Tukey test when statistical significance was tested . | In the brain , cells called neurons transmit information along their length in the form of electrical signals . To generate electrical signals , ions move into and out of neurons through ion channel proteins – such as the KCNQ channel – in the surface of these cells , which open and close to control the electrical response of the neuron . Abnormally intense bursts of electrical activity from many neurons at once can cause seizures such as those experienced by people with epilepsy . A significant proportion of patients do not respond to current anti-seizure medications . Openers of KCNQ channels have emerged as a potential new class of anti-epileptic drugs . A better understanding of how KCNQ channels work , and how their opening by PIP2lipid signals is regulated , could help to develop more effective therapies for epilepsy . A process called methylation controls many biological tasks by changing the structure of key proteins inside cells . Although methylation occurs throughout the brain , its role in controlling how easily neurons are activated ( a property known as “excitability” ) remains unclear . Kim , Jeong , Kim , Jung et al . now show that a protein called Prmt1 methylates the KCNQ channels in mice , and that this methylation is essential for suppressing seizures . Mice born without the Prmt1 protein developed epileptic seizures and the KCNQ channels in their neurons featured a reduced level of methylation . However , increasing the amount of PIP2 in these neurons restored their excitability back to normal levels . The methylation of KCNQ channel proteins increases their affinity for PIP2 , which is critical to open KCNQ channels . Kim et al . propose that these “opening” controllers balance the action of known “closers” of KCNQ channels to maintain neurons in a healthy condition . In future , Kim et al . plan to investigate whether methylation affects the activity of other ion channels controlled by PIP2 . Such experiments will complement a more widespread investigation into other ways in which the Prtmt1 protein may control the activity of neurons . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | Protein arginine methylation facilitates KCNQ channel-PIP2 interaction leading to seizure suppression |
The centromere , responsible for chromosome segregation during mitosis , is epigenetically defined by CENP-A containing chromatin . The amount of centromeric CENP-A has direct implications for both the architecture and epigenetic inheritance of centromeres . Using complementary strategies , we determined that typical human centromeres contain ∼400 molecules of CENP-A , which is controlled by a mass-action mechanism . This number , despite representing only ∼4% of all centromeric nucleosomes , forms a ∼50-fold enrichment to the overall genome . In addition , although pre-assembled CENP-A is randomly segregated during cell division , this amount of CENP-A is sufficient to prevent stochastic loss of centromere function and identity . Finally , we produced a statistical map of CENP-A occupancy at a human neocentromere and identified nucleosome positions that feature CENP-A in a majority of cells . In summary , we present a quantitative view of the centromere that provides a mechanistic framework for both robust epigenetic inheritance of centromeres and the paucity of neocentromere formation .
Centromeres are essential for proper cell division . During mitosis , a transient structure called the kinetochore is assembled onto centromeric chromatin , which mediates the interaction between DNA and the mitotic spindle ( Allshire and Karpen , 2008; Cheeseman and Desai , 2008 ) . Intriguingly , although centromeres are directly embedded in chromatin , specific DNA sequences are neither necessary nor sufficient for centromere function . This is best exemplified by the rare occurrence , within the human population , of neocentromeres: functional centromeres that have repositioned to atypical loci on the chromosome ( Amor et al . , 2004; Marshall et al . , 2008; du Sart et al . , 1997; Voullaire et al . , 1993 ) . Rather than centromeric sequences , the primary candidate for epigenetic specification of centromeres is the histone variant CENP-A , which replaces canonical H3 in centromeric nucleosomes ( Palmer et al . , 1987 , 1991; Stoler et al . , 1995; Henikoff et al . , 2000; Yoda et al . , 2000 ) . CENP-A chromatin is sufficient for recruitment of the downstream centromere and kinetochore complexes ( Foltz et al . , 2006; Okada et al . , 2006; Carroll et al . , 2009 , 2010; Barnhart et al . , 2011; Guse et al . , 2011; Mendiburo et al . , 2011 ) . In addition , CENP-A is stably transmitted at centromeres during mitotic ( Jansen et al . , 2007; Bodor et al . , 2013 ) and meiotic ( Raychaudhuri et al . , 2012 ) divisions , and its assembly is tightly cell cycle controlled ( Jansen et al . , 2007; Schuh et al . , 2007; Silva et al . , 2012 ) . Importantly , targeting of this protein to an ectopic site of the genome is sufficient to initiate an epigenetic feedback loop , recruiting more CENP-A to this site ( Mendiburo et al . , 2011 ) . However , little is known about the quantity of CENP-A present at centromeres , despite this being an essential parameter for a functional understanding of both centromeric architecture and epigenetic inheritance . Here , we use multiple , independent approaches to determine the absolute copy number of CENP-A at centromeres . In addition , we provide novel insights in the mechanisms of centromere size control .
To determine absolute centromeric CENP-A levels in human cells , we set out to build cell lines in which the entire CENP-A pool is fluorescent . To accomplish this , we removed a significant and essential portion of the CENP-A gene to create a knock-out allele in stably diploid , human retinal pigment epithelium ( RPE ) cells ( Figure 1A , bottom ) . Subsequently , a fluorescent knock-in allele was created by placing GFP or YFP encoding sequences in frame with the sole remaining CENP-A gene ( Figure 1A , middle ) . Specifically , we have built the following endogenously targeted RPE cell lines: CA+/− , CAG/− , CAY/− , and CA+/F ( where + = wild-type; − = knock-out; G = GFP knock-in; Y = YFP knock-in; F = floxed [to control for potential gene-targeting artifacts]; Figure 1—figure supplement 1A ) . Western blot analysis confirms that CAG/− and CAY/− cells exclusively contain tagged CENP-A ( of ∼43 kDa ) , while CA+/+ ( wild-type ) , CA+/F , and CA+/− cells only express wild-type CENP-A ( ∼16 kDa ) protein ( Figure 1B ) . Importantly , heterozygous expression or tagging of endogenous loci did not interfere with cell viability . 10 . 7554/eLife . 02137 . 003Figure 1 . CENP-A levels are regulated by mass-action . ( A ) Schematic of gene-targeting strategy that allowed for the creation of CENP-A knockout and fluorescent knock-in alleles . The region encoding the essential CENP-A targeting domain ( CATD , Black et al . , 2007 ) is indicated . ( B ) Quantitative immunoblots of CENP-A , HJURP , and Mis18BP1 in differentially targeted RPE cell lines . α-tubulin is used as a loading control . ( C ) Immunofluorescence images of same cell lines as in B . CENP-A intensity is represented in a heat map as indicated on the right . The fold difference ± SEM ( n is biological replicates ) compared to wild-type RPE cells is indicated below . Scale bar: 10 μm . Note that in contrast to quantification of immunoblots , immunofluoresce detection of untagged and tagged CENP-A is directly comparable . ( D ) Quantification of centromeric CENP-A levels ( from C ) by immunofluorescence ( IF ) and total CENP-A levels ( n = 4–9 independent experiments as in B ) by western blot ( WB ) . All cell lines expressing untagged CENP-A are normalized to CA+/+ while those expressing tagged CENP-A are normalized to the centromeric CAY/− levels measured in C , as indicated by dashed lines . ( E ) Correlation of centromeric and total cellular CENP-A levels as measured in D . Dashed line represents a predicted directly proportional relationship with indicated correlation coefficients . Throughout , the average ± SEM is indicated . ( F ) Quantification of centromeric CENP-A levels in synchronized HeLa cells ( based on anti-CENP-A staining ) within a single cell cycle after transient transfection of indicated proteins . Asterisk indicates statistically significant increase compared to control or indicated transfections ( one-tailed t test; p<0 . 05; n = 3 ) ; NS indicates no significant increase . Average ± SEM of three independent experiments is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 02137 . 00310 . 7554/eLife . 02137 . 004Figure 1—figure supplement 1 . CENP-A expression is the rate limiting factor for centromeric CENP-A levels . ( A ) Pedigree of targeted RPE cell lines used in this study . Uninterrupted lines indicate single gene-targeting events , interrupted lines indicate multiple sequential gene-targeting events , and dashed lines indicate stable ectopic protein expression . ( B–C ) Correlation of centromeric CENP-A and total cellular HJURP ( B ) or Mis18BP1 levels ( C ) . Insets show quantification of total protein levels from Figure 1B; n = 3–5 independent experiments . Dashed lines represent hypothetical directly proportional relationships with indicated correlation coefficients . In the insets , the average ± SEM ( n = 3–5 ) is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 02137 . 004 While CENP-A is an essential and constitutive component of centromeres , how the size of the centromeric chromatin domain is controlled is not known . We analyzed the consequences of different CENP-A expression levels in our CENP-A heterozygous knock-out and knock-in lines , as well as in a cell line that ectopically overexpressed CENP-A-YFP ( CAY/−+OE; Figure 1B; Figure 1—figure supplement 1A ) . First , we measured the total protein pool of CENP-A in our cell lines by quantitative immunoblotting . While we found the detection output for CENP-A to be linear over at least a 32-fold range ( Figure 2E ) , due to differences in protein transfer efficiencies this method does not allow for a comparison between proteins of different sizes , for example ( GFP- or YFP- ) tagged and untagged ( wild-type ) CENP-A ( Figure 2—figure supplement 3 ) . Nevertheless , we could directly compare CAG/− , CAY/− , and CAY/−+OE cell lines ( Figure 2—figure supplement 3 ) and found that cellular CENP-A content spans a sixfold range ( Figure 1B , D ) . 10 . 7554/eLife . 02137 . 005Figure 2 . Human centromeres contain 400 molecules of CENP-A . ( A ) Schematic outline of strategy allowing for the quantification of the centromeric fraction of CENP-A compared to the total cellular pool . Scale bars: 5 μm . ( B ) Quantification of the centromeric fraction of CENP-A in CAY/− cells . Each circle represents one centromere; circles on the same column are individual centromeres from the same cell . Dashed line indicates average of all centromeres . ( C ) Quantification of the centromeric fraction of CENP-A in indicated cell lines . Each square represents the average centromeric signal from one cell; squares on the same column are individual cells from the same experiment ( Exp ) . Figure 2—figure supplement 2 shows quantification of individual centromeres in CAG/− and CAY/−+OE cells . ( D ) Representative quantitative immunoblot of purified recombinant CENP-A and endogenous CENP-A from whole cell extracts ( WCE ) . ( E ) Quantification of D . Solid line represents the best fit linear regression . Dashed line represents the amount of CENP-A from 150 , 000 cells . ( F ) Quantification of the total cellular CENP-A copy number . Each diamond represents one replicate experiment; measurement from E is indicated as a gray diamond . ( G ) Calculation of average CENP-A copy number per centromere ( CEN ) in wild-type RPE cells . Throughout , the average ± SEM is indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 02137 . 00510 . 7554/eLife . 02137 . 006Figure 2—figure supplement 1 . Representative fluorescence lifetime imaging ( FLIM ) micrograph of a CENP-A-YFP expressing cell ( left ) and quantification of indicated cellular regions ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02137 . 00610 . 7554/eLife . 02137 . 007Figure 2—figure supplement 2 . Measurements of individual centromeres and CENP-A levels for different cell lines . ( A and B ) Graphs as in Figure 2B for CAG/− ( A ) and CAY/−+OE ( B ) cells . ( C ) Graph showing the absolute amount of centromeric CENP-A for indicated cell lines . DOI: http://dx . doi . org/10 . 7554/eLife . 02137 . 00710 . 7554/eLife . 02137 . 008Figure 2—figure supplement 3 . Transfer efficiency of recombinant and cellular CENP-A . Immunoblots of recombinant and cellular CENP-A from CA+/+ , CAG/− , and CAY/− cells , after protein transfer onto a stack of three membranes . The fraction of CENP-A retained on the first membrane ( compared to the total signal from all three membranes ) is quantified below . While YFP- or GFP-tagged CENP-A barely passes through the membrane at all , untagged CENP-A from cell extracts or recombinant protein preps is retained equally well on the first membrane . DOI: http://dx . doi . org/10 . 7554/eLife . 02137 . 008 Given its essential role in centromere function , we predicted a tight control of centromeric CENP-A levels . However , instead of maintaining a fixed amount of CENP-A at centromeres , the levels varied extensively ( Figure 1C ) . Both CA+/− and CAG/− cells , which contain a single intact allele , have decreased centromeric CENP-A levels , while the parental CA+/F cells maintain wild-type levels . Surprisingly , despite expressing CENP-A from a single allele , CAY/− cells have increased CENP-A levels , which may be due to adaptations that arose during the creation of this cell line . As expected , CENP-A levels are further elevated in CAY/−+OE cells ( Figure 1C ) . Remarkably , we found a very high correlation ( r2 = 84% ) for a hypothetical directly proportional relationship between centromeric and total cellular CENP-A-GFP or CENP-A-YFP levels ( Figure 1D , E ) . Similarly , despite an only approximately twofold range of expression , we still observe a high correlation with direct proportionality ( r2 = 71% ) for cells expressing untagged CENP-A ( Figure 1D , E ) . Thus , our observations indicate that centromeric levels are determined by a mass-action mechanism , where the amount of centromeric CENP-A varies in direct proportion with the cellular content . An alternative hypothesis is that stable cell lines have undergone long-term adaptation to altered CENP-A expression , which has led to re-equilibrated centromeric levels . For example , proteins involved in depositing CENP-A at the centromere may have adapted to CENP-A expression levels . Indeed , we see a weak correlation between the levels of CENP-A and its histone chaperone HJURP ( Dunleavy et al . , 2009; Foltz et al . , 2009; Barnhart et al . , 2011 ) in our cell lines ( Figure 1B , Figure 1—figure supplement 1B ) . Conversely , no correlation was detected for Mis18BP1 ( Figure 1B , Figure 1—figure supplement 1C ) , another essential protein for CENP-A assembly ( Fujita et al . , 2007; Maddox et al . , 2007 ) , arguing that it is a non-stoichiometric component of the loading pathway . To test whether centromeric CENP-A levels require long-term adaptation , we analyzed the effect of CENP-A and/or HJURP overexpression in a single round of CENP-A assembly . Therefore , we transiently expressed CENP-A and/or HJURP and measured the level of centromeric CENP-A after a single cell cycle in HeLa cells , which can be effectively synchronized in S phase using thymidine . While induction of CENP-A expression leads to a prompt increase in centromeric levels , no ( additional ) effect was observed by expression of HJURP ( Figure 1F ) . Together , our results strongly suggest that centromeric CENP-A levels are directly regulated by its protein expression levels . To understand how CENP-A chromatin is self-propagated and nucleating the kinetochore , it is critical to establish the absolute amount of CENP-A present . In vertebrates , previous estimates range from a few tens of molecules ( in chicken DT40 cells , Ribeiro et al . , 2010 ) to a potential maximum of tens of thousands ( in HeLa cells , Black et al . , 2007 ) . To directly determine the copy number of CENP-A on human centromeres , we developed a 3D imaging strategy ( Figure 2A ) , which we adapted from a method used previously to quantify cytokinesis proteins in fission yeast ( Wu and Pollard , 2005; Wu et al . , 2008 ) . In brief , we use a non-cell permeable dye ( Figure 2A , I ) to determine the 3D shape of cells ( Figure 2A , II ) and measure the total amount of fluorescence within the entire cell volume ( Figure 2A , III ) . Total cellular fluorescence of CAY/− cells ( Figure 2A , III ) was corrected for autofluorescence measured in wild-type RPE cells ( Figure 2A , IV ) , thus resulting in a measure of total CENP-A-derived fluorescence . Next , centromere-specific fluorescence was measured after correction for local background ( Figure 2A , V; Hoffman et al . , 2001 ) and axial oversampling . Importantly , fluorescence lifetime of CENP-A-YFP is similar between highly concentrated centromeric and diffuse cytoplasmic pools ( Figure 2—figure supplement 1 ) , arguing that clustering does not lead to changes in fluorescence efficiency . In effect , our 3D-integrated fluorescence strategy measures the centromeric fraction of CENP-A compared to the total cellular pool . We find that while CENP-A is enriched at centromeres , on average only 0 . 44% of cellular CENP-A is present per centromere in CAY/− cells ( Figure 2B ) . Very similar fractions were observed in CAG/− and CAY/−+OE cells ( 0 . 38% in both cases; Figure 2C , Figure 2—figure supplement 2A , B ) , which provides an additional line of evidence in support of a mass-action mechanism for CENP-A assembly . Furthermore , these findings show that a surprising minority , about one-fifth of the CENP-A protein content ( 0 . 44% × 46 ) is present on the functionally relevant subcellular location , i . e . at the centromeres . To convert centromeric fractions to absolute amounts , we determined the total number of CENP-A molecules in RPE cells . To this end , we prepared whole cell extracts of RPE cells and analyzed these alongside highly purified recombinant CENP-A/H4-complexes of known concentration by quantitative immunoblotting ( Figure 2D ) . Importantly , we ensured that recombinant and cellular CENP-A have the same transfer efficiency and can be directly compared to each other ( Figure 2—figure supplement 3 ) . Fitting the cellular amount of CENP-A onto a linear regression curve of purified protein ( Figure 2E ) shows that CA+/+ cells contain an average of ∼9 . 1 ± 1 . 1 × 104 ( n = 10 ) molecules of CENP-A per cell ( Figure 2F ) . Because the centromeric fraction of CENP-A is fixed , we can calculate the absolute amount of CENP-A per centromere in our cell lines ( Figure 2G , Figure 2—figure supplement 2C ) and show that wild-type RPE cells contain ∼400 molecules of CENP-A on an average centromere . Both the expression and centromeric loading of CENP-A are cell cycle regulated ( Figure 3A ) . In human cells , cellular protein levels of CENP-A peak in late G2 ( Shelby et al . , 2000 ) , while incorporation into centromeric chromatin occurs in early G1 phase ( Jansen et al . , 2007 ) . Thus , it is possible that part of the cell-to-cell variation of the centromeric CENP-A ratio observed in Figure 2C is due to differing cell cycle stages . We tested this possibility using a previously developed fluorescent ubiquitin-based cell cycle indicator ( FUCCI ) that can be used in live cells ( Sakaue-Sawano et al . , 2008 ) . In particular , we used hCdt1 ( 30/120 ) -RFP , which is expressed ubiquitously throughout the cell cycle , but is specifically degraded in S , G2 , and M phases ( Sakaue-Sawano et al . , 2008 ) . As a result , protein levels increase as cells enter and progress through G1 phase , peak at the G1/S boundary , and then drop until cells re-enter G1 ( Figure 3A ) . We expressed this protein in CAY/− cells and tracked the RFP fluorescence intensity over time ( Figure 3B , Figure 3—figure supplement 1A ) to identify cells that entered S phase ( see ‘Materials and methods’ for details ) . We compared the centromeric fraction of CENP-A of S Phase cells to that of randomly cycling cells and found that neither the mean nor the variance differs significantly between these two populations ( Figure 3C ) . Importantly , expression of the FUCCI marker itself has no effect on the measurements performed ( Figure 3—figure supplement 1B ) . While the centromeric fraction of CENP-A is likely low in G2 phase and high just after assembly in early G1 , we find that the variation observed in Figure 2C is not a consequence of such cell cycle-induced effects and may instead reflect inherent variation between cells . 10 . 7554/eLife . 02137 . 009Figure 3 . Centromeric CENP-A levels are equivalent between S phase and randomly cycling cells . ( A ) Cartoon depicting changes in cell morphology and nuclear levels of hCdt1 ( 30/120 ) -RFP ( in red ) throughout the cell cycle ( Sakaue-Sawano et al . , 2008 ) . Approximate timing of CENP-A expression ( Shelby et al . , 2000 ) and centromeric loading ( Jansen et al . , 2007 ) are indicated in orange and blue , respectively . The stage at which cells were analyzed to measure the centromeric fraction of CENP-A is indicated in green . ( B ) An example trace of a cell entering S phase ( indicated by a sudden decrease in RFP levels ) is shown . The centromeric fraction of CENP-A was measured at this point as outlined in Figure 2A . Peak expression is normalized to 100 and background fluorescence to 0 . Micrographs of hCdt-1 ( 30/120 ) -RFP at indicated timepoints are shown below . ( C ) As in Figure 2C . Orange squares represent cells that have passed the G1-S transition point , as indicated by decreasing levels of hCdt-1 ( 30/120 ) -RFP . Gray squares represent randomly cycling cells . No statistically significant differences ( NS ) were observed between randomly cycling cells and S phase cells . DOI: http://dx . doi . org/10 . 7554/eLife . 02137 . 00910 . 7554/eLife . 02137 . 010Figure 3—figure supplement 1 . hCdt-1 ( 30/120 ) -RFP expression allows for accurate determination of cell cycle stages and measurements of centromeric CENP-A ratios . ( A ) An example trace of a cell that had entered G1 phase at the beginning of the experiment ( as determined by cellular morphology using DIC ) is shown . Graph as in Figure 3B . ( B ) Baculoviral transduction of hCdt-1 ( 30/120 ) -RFP does not affect measurements of CENP-A-YFP . Centromeric CENP-A ratio measurements of non-transduced cells were compared to measurements of unstaged ( i . e . , randomly cycling ) cells expressing hCdt-1 ( 30/120 ) -RFP . Graph as in Figure 3C . DOI: http://dx . doi . org/10 . 7554/eLife . 02137 . 010 Although the method we employed to measure centromeric ratios is internally controlled , it relies on measurement of integrated fluorescence of whole cells , including highly dilute cytoplasmic CENP-A . To exclude potential errors in measurements of low protein concentration , we stably expressed H2B-RFP in CAY/− cells ( Figure 4A , inset ) and determined that 0 . 73% of nuclear CENP-A is present on each centromere ( Figure 4A ) . In addition , low salt fractionation experiments indicate that ∼74% of cellular CENP-A co-pellets with other chromatin components in CAY/−+H2B-RFP cells ( Figure 4B ) , indicating that this represents the stable nuclear pool . Combined , we find a similar number of CENP-A molecules per centromere when analyzing the nuclear pool ( 492 molecules; Figure 4C ) as when measuring total cellular CENP-A . This argues that the measurements performed above are not significantly influenced by a potential inaccuracy in determining the cytoplasmic pool . Interestingly , it has recently been shown that detectable levels of CENP-A are assembled into non-centromeric chromatin of HeLa cells ( Lacoste et al . , 2014 ) . We now find that , at least in RPE cells the proporation of chromatin bound CENP-A outside of the centromere is surprisingly high ( ∼66% in this cell line ) . 10 . 7554/eLife . 02137 . 011Figure 4 . Measurement of nuclear CENP-A confirms centromeric copy number . ( A ) As in Figure 2B , except that the centromeric fraction compared to total nuclear pool is indicated . Inset shows a representative image of a CAY/−+H2B-RFP cell ( scale bar: 2 . 5 μm ) . ( B ) Quantitative immunoblot showing the soluble fraction and a dilution series from the insoluble fraction of CENP-A-YFP in CAY/−+H2B-RFP cells ( left ) . Tubulin is used as a marker for the soluble fraction and H4K20me2 ( exclusively found in chromatin , Karachentsev et al . , 2007 ) for the insoluble fraction . Quantification of insoluble fraction of CENP-A is shown to the right . ( C ) Calculation of the average CENP-A copy number per centromere ( CEN ) in wild-type RPE cells , based on results from CAY/−+H2B-RFP cells . DOI: http://dx . doi . org/10 . 7554/eLife . 02137 . 011 To further validate that the strategy described above accurately measures centromeric CENP-A copy numbers , we used two additional independent quantification methods . First , we applied a method that employs the statistical properties of fluorescence redistribution ( Rosenfeld et al . , 2005 , 2006 ) . This method relies on the fact that random segregation of fluorescent molecules leads to each daughter receiving an ( unequal ) fraction , where the distribution of differences relates to the total number of molecules ( as outlined in Figure 5A ) . During mitosis , sister centromeres form individually resolved spots by light microscopy , allowing us to measure the fluorescence intensity of individual sisters ( Figure 5B ) . We find that rather than accurately segregating exactly half of pre-assembled CENP-A onto each daughter chromatid , the difference between sister centromeres follows a random distribution ( Figure 5B , C ) . Previously , Rosenfeld et al . have provided mathematical evidence that measurements of this deviation allow for the determination of the fluorescence intensity of a single heritable , segregating unit ( Figure 5A , Rosenfeld et al . , 2005 , 2006 ) . We measured an average of 75 . 4 segregating units of CENP-A-GFP per centromere in CAG/− cells ( Figure 5D ) . Because each segregating unit consists of one or more nucleosomes , containing two molecules of CENP-A each ( Sekulic et al . , 2010; Tachiwana et al . , 2011; Bassett et al . , 2012; Hasson et al . , 2013; Padeganeh et al . , 2013 ) , an average CAG/− centromere has a minimum of 150 . 8 molecules of CENP-A . Correcting the amount of CENP-A measured in CAG/− cells for wild-type levels ( Figure 1C ) results in ≥377 molecules of CENP-A per centromere ( Figure 5D , right y-axis ) . Importantly , these measurements differ significantly if random centromere pairs are chosen for which no statistical correlation exists ( Figure 5—figure supplement 1E ) . This confirms that fluorescence intensities at sister centromeres co-vary and renders this type of analysis suitable for centromere quantification . Stochastic fluctuation measurements in CAY/− and CAY/−+OE cells indicates that wild-type cells contain ≥188 and ≥149 CENP-A molecules per centromere , respectively ( Figure 5—figure supplement 1A–D ) . Importantly , the number of co-segregating CENP-A nucleosomes is unknown , which can be one or more . Therefore , despite the variation between the cell lines used here , all results obtained from this method provide a minimum estimate of the centromeric CENP-A copy number that is in agreement with the 400 centromeric molecules of CENP-A measured above ( Figure 2G ) . 10 . 7554/eLife . 02137 . 012Figure 5 . Independent quantification methods confirm centromeric CENP-A copy number . ( A ) Stochastic fluctuation method: cartoon depicting inheritance and random redistribution of parental CENP-A nucleosomes onto sister chromatids during DNA replication . A hypothetical distribution of the absolute difference between the two sister centromeres , as well as the formula for calculating the fluorescence intensity per segregating unit ( α ) are indicated on the right . ( B ) Representative image of mitotic CENP-A-YFP expressing cell . CENP-B staining allows for identification of sister centromeres . Blowup to the right represents a single slice of the boxed region showing that CENP-B is located in between the CENP-A spots of sister centromeres . ( C ) Frequency distribution of the difference between CENP-A-GFP intensity of sister centromeres in CAG/− cells . ( D ) Quantification of centromeric CENP-A-GFP based on the stochastic fluctuation method . Each circle represents one centromere; circles on the same column are individual centromeres from the same cell . Left y-axis indicates segregating CENP-A-GFP units in CAG/− cells; right y-axis shows the conversion to minimum number of centromeric CENP-A molecules in CA+/+ ( WT ) cells . ( E ) Fluorescent standard method: representative fluorescence images of 4kb-LacO , LacI-GFP S . cerevisiae and human CAG/− cells . ( F ) Quantification of fluorescence signals of LacI-GFP and CENP-A-GFP spots ( n = 2 biological replicates ) . The left y-axis indicates the fluorescence intensity normalized to LacI-GFP; the right y-axis shows the conversion to maximum number of centromeric CENP-A molecules in wild-type cells . ( G ) Comparison of independent measurements for the centromeric CENP-A copy number ( corrected for CA+/+ levels; Stoch . fluctuations = stochastic fluctuation method [Figure 5A]; Integr . fluorescence = integrated fluorescence method [Figure 2A] ) . Levels from all strategies are corrected for CA+/+ ( WT ) levels . Throughout , the average ± SEM and scale bars of 2 . 5 μm are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 02137 . 01210 . 7554/eLife . 02137 . 013Figure 5—figure supplement 1 . Stochastic fluctuations of CENP-A segregation allows for copy number measurements . ( A–D ) Results as in Figure 5C–D for CAY/− ( A–B ) and CAY/−+OE cells ( C–D ) . ( E ) Quantification of segregating units in CAG/− cells based on sister centromeres ( dark green ) or random centromere pairs ( light green; random pairs were assigned independently three times ) . Asterisks indicate a significant difference from sister centromere result ( t test; p<0 . 0001 in all cases ) . Each circle represents one centromere pair . Throughout , the average ± SEM is indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 02137 . 013 Next , we used a yeast strain that harbors a chromosomally integrated 4 kb LacO-array and expresses GFP-LacI as a calibrated fluorescent standard ( Lawrimore et al . , 2011 ) . While there is a potential for 204 molecules of GFP-LacI to be bound to this array ( Lawrimore et al . , 2011 ) , it is unlikely that the entire array is fully occupied at any moment . Because CAG/− cells express the same version of GFP as this yeast strain , direct comparison of fluorescent foci ( Figure 5E ) provides a maximum estimation of the centromeric CENP-A-GFP copy number . In this way , we determined that CAG/− centromeres contain at most 215 ± 32 CENP-A-GFP molecules , which translates to ≤538 CENP-A molecules in wild-type cells ( Figure 5F ) . Importantly , the copy number that we measure directly by our 3D integrated fluorescence approach is in close agreement with minimum and maximum estimates of the stochastic fluctuation and fluorescent standard approaches , respectively ( Figure 5G ) . This provides confidence that 400 molecules of CENP-A per centromere in wild-type RPE cells is an accurate measure . While cells are able to survive with a sixfold range of CENP-A levels ( Figure 1D ) , centromere function may be compromised when levels are not accurately maintained . Indeed , based on a conserved ratio of centromere and kinetochore proteins and kinetochore microtubules between multiple yeast species as well as chicken DT40 cells , it has been hypothesized that centromeres form modular structures by repeating individual structural subunits ( Joglekar et al . , 2008; Johnston et al . , 2010 ) , as originally proposed by Zinkowski et al . ( 1991 ) . Thus , the amount of CENP-A would directly reflect the number of downstream centromere and kinetochore proteins and microtubule attachment sites . Conversely , experiments in human cells indicate that the centromere is assembled by multiple independent subcomplexes ( Foltz et al . , 2006; Liu et al . , 2006 ) . Here , we analyzed whether altering the levels of CENP-A has an effect on the recruitment of other , downstream centromere or kinetochore proteins . Both CENP-C and CENP-T rely on CENP-A for their centromeric recruitment ( Régnier et al . , 2005; Liu et al . , 2006; Fachinetti et al . , 2013 ) and have recently been shown to be responsible for mitotic recruitment of the KMN network ( Gascoigne et al . , 2011 ) , including the key microtubule binding protein Hec1/NDC80 ( Cheeseman et al . , 2006; DeLuca et al . , 2006 ) . Interestingly , we found that none of these three proteins were significantly affected by altering the levels of CENP-A between 40% and 240% of wild-type levels ( Figure 6A , Figure 6—figure supplement 1 ) . In line with previous findings ( Liu et al . , 2006; Fachinetti et al . , 2013 ) , these results argue against a modular centromere architecture where CENP-A nucleosomes form individual binding sites for downstream components . Rather , a >2½-fold excess of CENP-A appears to be present for recruitment of centromere and kinetochore complexes of fixed pool size . 10 . 7554/eLife . 02137 . 014Figure 6 . Reduction of CENP-A leads to a CENP-C , CENP-T , and Hec1 independent increase in micronuclei . ( A ) Quantification of centromeric CENP-A ( from Figure 1 ) , CENP-C , CENP-T , and Hec1 levels for indicated cell lines; n = 4 independent experiments in each case . Note that cell lines carrying tagged CENP-A have a slight , yet non-significant impairment in recruiting CENP-C , CENP-T , and Hec1 . However , this does not correlate with the CENP-A levels themselves . Below , representative images of indicated antibody staining from CA+/+ cells are shown . Representative images from all cell lines can be found in Figure 6—figure supplement 1 . ( B ) Quantification of the fraction of cells containing micronuclei ( MN ) for indicated cell lines . Asterisk indicates statistically significant increase compared to wild-type ( paired t test; p<0 . 05; n = 3–4 independent experiments [500–3000 cells per experiment per cell line] ) ; NS indicates no significant difference . Throughout , the average ± SEM is indicated and dashed lines represent wild-type levels . Scale bars: 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02137 . 01410 . 7554/eLife . 02137 . 015Figure 6—figure supplement 1 . Representative images for quantifications in Figure 6B . Images of indicated cell lines are shown for immunofluorescence staining of ( A ) CENP-C , ( B ) CENP-T , and ( C ) Hec1 ( mitotic cells ) . Scale bars: 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02137 . 015 Intriguingly , despite no quantitative effect on centromeric proteins , we observed that decreasing CENP-A levels leads to an increase in the fraction of cells containing micronuclei ( MN; Figure 6B ) . MN often arise as a consequence of mitotic errors , such as lagging chromosomes during anaphase ( Ford et al . , 1988 ) , breakage of anaphase bridges ( Hoffelder et al . , 2004 ) , or multipolar mitoses ( Utani et al . , 2010 ) . The presence of MN can be scored by DAPI staining ( Figure 6B , bottom ) . A baseline fraction of 0 . 53% ± 0 . 07% ( n = 4 ) of wild-type CA+/+ cells contain MN ( Figure 6B ) . Both cell lines that have decreased CENP-A levels show a significantly increased fraction of cells with MN with 2 . 77% ± 0 . 48% ( n = 3 ) and 1 . 95% ± 0 . 50% ( n = 4 ) in CA+/− and CAG/− cells , respectively . Importantly , these two cell lines were derived independently from the parental CA+/F cell line ( Figure 1—figure supplement 1A ) , which has wild-type levels of CENP-A and no significant increase in MN ( Figure 6 ) . In addition , neither cell line with increased CENP-A levels has a larger fraction of MN than CA+/F cells . While the essential role for CENP-A in centromere function is well established ( Régnier et al . , 2005; Liu et al . , 2006; Black et al . , 2007 ) , our results indicate that a critical level of CENP-A is passed after reducing the levels to ∼50% . However , the molecular mechanism responsible for MN formation remains unclear , as downstream centromere and kinetochore components of CENP-A remain unaffected . Interestingly , we find that not all centromeres of the same cell have equal amounts of CENP-A ( e . g . , Figure 5D ) . This could either be due to in cis features driving differential regulation of CENP-A on individual centromeres , or by stochastic , yet unbiased , effects at centromeres . To distinguish between these possibilities , we measured the centromeric levels of endogenous CENP-A on specific chromosomes . First , we analyzed a monoclonal HCT-116 cell line that has an integrated Lac-array in a unique position in the genome ( Thompson and Compton , 2011 ) . While the site of integration is unknown , expressing LacI-GFP allows for the identification of the same chromosome in a population of cells ( Figure 7A ) . Both the average and variance of CENP-A at this centromere does not differ statistically from the bulk ( Figure 7B , Figure 7—figure supplement 1A ) , arguing against centromere specific features driving CENP-A levels on the Lac-marked chromosome . Conversely , we found that the Y-centromere , uniquely identified by the lack of CENP-B ( Figure 7C; Earnshaw et al . , 1987 ) , of two independent male cell lines had a slight yet significant reduction of CENP-A ( 19% in wild-type HCT-116 and 13% in DLD-1; Figure 7D , Figure 7—figure supplement 1B , C ) , consistent with an earlier report ( Irvine et al . , 2004 ) . Finally , we used a human patient-derived fibroblast cell line ( PDNC-4 ) where one centromere of chromosome 4 has repositioned to an atypical location ( Amor et al . , 2004 ) , which we designate as NeoCEN-4 ( Figure 7E ) . As has been observed in other cell lines derived from this patient ( Amor et al . , 2004 ) , we found that the NeoCEN-4 has a ∼25% decrease in centromeric CENP-A ( Figure 7F , Figure 7—figure supplement 1D ) . Taken together , these results show that while CENP-A expression drives centromeric levels , local sequence or chromatin features can also contribute to the average amount of CENP-A at specific centromeres . Nevertheless , even on these centromeres , the variance in CENP-A levels is maintained , indicating that other stochastic processes contribute to CENP-A levels . 10 . 7554/eLife . 02137 . 016Figure 7 . Centromere and cell specific distribution of CENP-A . ( A , C , E ) Representative micrograph of mitotic spreads for LacI-GFP::LacO expressing HCT-116 cells ( A ) ; wild-type HCT-116 cells ( C ) ; and PDNC-4 cells ( E ) . Blowups show the chromosome containing the integrated Lac-array ( A ) ; the Y-chromosome ( outline indicated; CENP-B negative ) as well as an autosome ( CENP-B positive ) ( C ) ; and the neocentric chromosome 4 , containing 2 pairs of ACA spots ( staining both CENP-A and CENP-B ) , but only 1 pair of CENP-A spots ( E ) . ( B , D , F ) Quantification of CENP-A levels on the centromere of the chromosome containing the Lac-array ( CEN-Lac; n = 29; B ) ; the Y-chromosome ( CEN-Y; n = 18; D ) ; and neocentric chromosome 4 ( NeoCEN-4; n = 39; F ) of indicated cell lines compared to all other centromeres within the same cell ( Other CENs; n = 1008 , 620 , and 1592 , respectively ) . Median ( line ) , interquartile distance ( box ) , 3 × interquartile distance or extremes ( whiskers ) , and outliers ( spots ) are indicated . Figure 7—figure supplement 1 shows results of individual centromeres . Asterisk indicates statistically significant difference ( t test; p<0 . 05 ) ; NS indicates no significant difference . ( G ) Representative images of CENP-A antibody staining in indicated cell types . Images of RPE cells are shown as independent reference . Primary fibrobl . indicates primary human foreskin fibroblasts . ( H ) Quantification of G . Mean ± SEM for n = 3–4 independent experiments is shown . Left y-axis represents centromeric CENP-A levels normalized to RPE cells; right y-axis shows number of CENP-A molecules per centromere ( CEN ) . ( I ) Combined results from A–H allow for the determination of CENP-A copy numbers on individual chromosomes as indicated . ( J ) Statistical map of the distribution of 216 CENP-A nucleosomes on the NeoCEN-4 at three different scales . The top 216 peaks are indicated in blue . Y-axis indicates the probability of CENP-A occupancy for each nucleosome . ( K ) Histogram of the CENP-A nucleosome occupancy . Inset shows the distribution of 216 neocentric CENP-A nucleosomes on the 10% highest occupancy peaks ( green ) and 90% lowest occupancy peaks ( red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02137 . 01610 . 7554/eLife . 02137 . 017Figure 7—figure supplement 1 . Measurements of individual centromeres for graphs in Figure 7A–F . CENP-A levels are normalized to the average of each individual cell for CEN-Lac in HCT-116 cells ( A ) , CEN-Y in wild-type HCT-116 cells ( B ) , CEN-Y in DLD-1 cells ( C ) , and NeoCEN-4 in PDNC-4 cells ( D ) . Each circle represents one centromere; circles on the same column are individual centromeres from the same cell . Colored circle represents uniquely identified chromosome . Averages ± SEM are indicated . Graph to the right in C as in Figure 7D for DLD-1 cells ( n = 26 and 927 for CEN-Y and Other CENs , respectively ) . Dashed line indicates average of all centromeres . DOI: http://dx . doi . org/10 . 7554/eLife . 02137 . 017 Next , to determine whether the CENP-A copy number of our model cell line is representative for functionally different cells , we performed comparative immunofluorescence against CENP-A ( Figure 7G ) . We analyzed four different cancer cell lines ( HeLa , U2OS , HCT-116 , and DLD-1 ) , as well as the PDNC-4 neocentromere cell line discussed above , and primary human foreskin fibroblasts that were cultured for a limited number of passages ( <15 ) since their isolation from a patient ( Figure 7G ) . Using these cell lines , we found a sixfold range of centromeric CENP-A levels ( Figure 7H ) , indicating that there is substantial variance between different cell lines . However , we find that the primary cells have a similar amount of CENP-A as RPEs ( Figure 7H ) , arguing that our measure of absolute CENP-A copy numbers made in RPE cells is relevant for healthy , human tissues as well . We combined these results with our measurements of individual centromeres and determined that , while an average centromere in PDNC-4 cells contains ∼579 molecules of CENP-A , the NeoCEN-4 only contains ∼432 . Average Y-centromeres contain ∼143 or ∼87 molecules in HCT-116 and DLD-1 cells , respectively ( Figure 7I ) . In conclusion , we find evidence that cis-elements can have an effect on CENP-A levels , at least on human Y- and neo-centromeres . The number of CENP-A nucleosomes we find at individual centromeres is much smaller ( ∼25-fold , see Figure 8A ) than the total number of nucleosome positions on human centromeric DNA . This indicates that either CENP-A is randomly distributed at a low level throughout the centromere domain or that it occupies specific ‘hotspots’ . Due to their repetitive nature , it is not possible to map individual CENP-A nucleosomes on canonical centromeres . However , a recent high-resolution ChIP-seq analysis of the ( non-repetitive ) NeoCEN-4 identified 1113 unique CENP-A nucleosome positions spanning a ∼300 kb locus ( Hasson et al . , 2013 ) . By combining the relative height of individual peaks with the total number of CENP-A nucleosomes at this neocentromere , we were able to determine the fraction of cells containing CENP-A at each nucleosome position ( Figure 7J ) . This statistical map of CENP-A occupancy shows that , while the median is ∼6% ( Figure 7K ) , individual positions feature CENP-A with a surprisingly high occupancy ( up to 80% of all cells; Figure 7J , arrow ) . Remarkably , more than one third of all CENP-A nucleosomes are located on the top 10% potential positions ( Figure 7K , inset ) . This strongly suggests that , at least on the NeoCEN-4 , a number of nucleosome positioning sequences exist that strongly favor CENP-A over other H3 variants . 10 . 7554/eLife . 02137 . 018Figure 8 . A quantitative view of human centromeric chromatin . ( A ) Distribution of CENP-A . Estimated ratio of CENP-A ( red ) to H3 ( gray ) at the centromere and on non-centromeric loci ( genome ) in interphase cells . Estimations are calculated assuming 2 CENP-A molecules per nucleosome ( Sekulic et al . , 2010; Tachiwana et al . , 2011; Bassett et al . , 2012; Hasson et al . , 2013; Padeganeh et al . , 2013 ) , an average nucleosome positioning distance of 200 base pairs , an average centromere size of 2 . 5 × 106 base pairs ( Sullivan et al . , 1996; Lee et al . , 1997 ) of which approximately 40% ( 1 Mbp ) contains CENP-A ( Sullivan et al . , 2011 ) , a diploid genome size of 6 × 109 base pairs , 200 CENP-A nucleosomes per centromere , and 2 . 5 × 104 CENP-A nucleosomes outside of centromeres ( 9 . 1 × 104 CENP-A molecules per cell [Figure 2F] , of which 74% is in chromatin [Figure 4B] and 0 . 44% in each centromere [Figure 2B] ) . The fraction of CENP-A on centromeres , non-centromeric chromatin , and unincorporated CENP-A are indicated in green , blue , and black , respectively . CENP-A nucleosomes are represented as though evenly spread throughout the centromeric domain . Alternatively , they could be distributed into one or more clusters within this domain . ( B ) Mitotic organization of centromeric chromatin . 200 nucleosomes are redistributed to 100 nucleosomes per centromere on replicated sister chromatids ( Jansen et al . , 2007; Bodor et al . , 2013 ) . The exact CENP-A copy number at the centromere depends on the available total pool ( mass-action ) . Excess CENP-A could either lead to an increased CENP-A domain or lead to a higher density of CENP-A within a domain of fixed size . This pool forms an excess to recruit downstream centromere and kinetochore complexes and ultimately provides microtubule binding sites for ∼17 kinetochore microtubules ( McEwen et al . , 2001 ) . To avoid mitotic errors , a critical amount of CENP-A is required ( dashed lines ) . ( C ) Graph representing the chance of at least one chromosome in a cell ( with 46 chromosomes ) reaching critically low levels of CENP-A by random segregation of pre-existing CENP-A nucleosomes . Calculations were performed for varying levels of critical nucleosome numbers at a fixed steady state of 200 ( left ) , or by varying the steady state number at a fixed critical level of 22 ( right ) . Red bars represent identical calculations . DOI: http://dx . doi . org/10 . 7554/eLife . 02137 . 018
It has been proposed that centromeres in budding yeast feature a single nucleosome of CENP-ACse4 ( Meluh et al . , 1998; Furuyama and Biggins , 2007 ) . For this reason , the yeast centromere cluster has been extensively used to calibrate fluorescence intensities of CENP-A and other proteins from a number of species ( Joglekar et al . , 2006 , 2008; Johnston et al . , 2010; Schittenhelm et al . , 2010 ) . However , the single nucleosome hypothesis has recently been challenged ( Coffman et al . , 2011; Lawrimore et al . , 2011; Haase et al . , 2013 ) . To avoid dependency on any single reference , we used three independent methods to measure the human centromeric CENP-A copy number . One strategy uses intrinsically controlled fluorescence ratios of cellular and centromeric CENP-A-YFP signals ( Figure 2A ) . The second method does not rely directly on fluorescence intensities , but rather on the stochastic redistribution of CENP-A ( Figure 5A ) . Finally , we compared CENP-A signals directly to a calibrated fluorescent standard ( Figure 5E ) . Importantly , despite the independent nature of these strategies , they all come to a very similar conclusion . Therefore , we demonstrate that typical centromeres in human RPE cells contain ∼400 molecules of CENP-A . While there is a continuing debate on the composition of CENP-A nucleosomes ( Black and Cleveland , 2011; Henikoff and Furuyama , 2012 ) , current evidence , at least in human cells , strongly favors an octameric arrangement harboring two copies of CENP-A ( Sekulic et al . , 2010; Tachiwana et al . , 2011; Bassett et al . , 2012; Hasson et al . , 2013; Padeganeh et al . , 2013 ) . Hence , our numbers , correspond to 200 CENP-A nucleosomes in interphase , which are split into 100 nucleosomes on mitotic chromosomes ( Figure 8B ) . Epigenetic centromere inheritance is achieved by quantitative inheritance of CENP-A across cell division cycles ( Jansen et al . , 2007; Bodor et al . , 2013 ) . We find that rather than accurately ensuring that each daughter receives exactly half , redistribution of CENP-A occurs in a random fashion ( Figure 5B , C ) . Because this type of regulation has the potential for individual centromeres to stochastically inherit critically low levels of CENP-A , the steady state must be sufficiently high to avoid chromosome loss . Although the critical amount of CENP-A is not known , we have previously shown that HeLa cell viability is lost if CENP-A levels are reduced to ∼33% ( Black et al . , 2007 ) , i . e . 44 nucleosomes ( see Figure 7H ) . Conversely , we show here that CAG/− cells are viable at 40% of RPE levels ( 80 nucleosomes ) . Consequently , we estimate that the critical number of nucleosomes that must be inherited , which is half of the steady state level and is replenished during G1 phase , lies between 22 and 40 . We used these values to calculate the chance that any one centromere per cell inherits critically low levels of CENP-A for different steady state and critical CENP-A nucleosome levels ( Figure 8C ) . We demonstrate that at a steady state of 200 CENP-A nucleosomes per centromere , less than one in 1016 cell divisions will give rise to a centromere containing 40 CENP-A nucleosomes or less ( Figure 8C , left ) . Thus , the chance of inheriting a critical amount of CENP-A at wild-type steady state levels is negligible . Conversely , with 100 CENP-A nucleosomes at steady state , the chance of a chromosome inheriting even the most stringent critical level of 22 nucleosomes is close to 10−6 ( Figure 8C , right ) , which may pose a significant problem , for example during the development of a human organism . Conversely , although critical levels would be reached even less frequently if centromeres contained a steady state of , for example 300 CENP-A nucleosomes , this degree of accuracy may be superfluous and not outweigh the cost of maintaining a large centromere size ( Figure 8C , right ) . Therefore , we argue that the number of CENP-A molecules found on human centromeres is optimized for robust epigenetic inheritance and centromeric function . Previously , it has been shown that CENP-A is interspersed with both H3 . 1 and H3 . 3 at the centromere ( Blower et al . , 2002; Sullivan and Karpen , 2004; Ribeiro et al . , 2010; Dunleavy et al . , 2011; Sullivan et al . , 2011 ) . Indeed , based on the average size of the centromeric chromatin domain , we estimate that 200 CENP-A nucleosomes represent only ∼4% of all centromeric nucleosomes ( see Figure 8A for calculation ) . Surprisingly , we find that the majority of chromatin bound CENP-A is located outside the centromere . Indeed , a recent study found that a proportion of CENP-A containing nucleosomes also exist in non-centromeric chromatin of HeLa cells , and is assembled by DAXX , a major chaperone of histone H3 . 3 ( Lacoste et al . , 2014 ) . In addition , detectable levels of non-centromeric CENP-A have been observed in budding yeast ( Camahort et al . , 2009 ) and chicken DT40 cells ( Shang et al . , 2013 ) . Here , we quantify this pool in human RPE cells and while there is more than twice as many non-centromeric CENP-A nucleosomes than there are centromeric ones , this only represents <0 . 1% of all nucleosomes in the genome and thus CENP-A is ∼50-fold enriched ( per unit length of DNA ) at centromeres ( Figure 8A ) . This result may explain how , despite being outnumbered 25:1 by other H3 variants at the centromere , CENP-A can still accurately specify the centromeric locus . This hypothesis may potentially be tested by creating artificial CENP-A binding sites ( e . g . , using the LacO/LacI system ) of different known sizes and determining the threshold at which centromeres can be formed . Interestingly , the study by Lacoste et al . showed that the extra-centromeric CENP-A is not randomly distributed , but enriched at sites of high histone turnover ( Lacoste et al . , 2014 ) . Our finding that CENP-T , CENP-C , and Hec1 do not quantitatively correlate with CENP-A levels ( Figure 6A ) argues that not each ( non-centromeric ) CENP-A nucleosome is able to recruit downstream centromere components . It would be interesting to determine to what extent other centromere and kinetochore proteins are present throughout the genome and whether they are also enriched at extra-centromeric CENP-A ‘hotspots’ . This question is particularly relevant since it has been observed that downstream centromere components may affect centromeric CENP-A levels ( Okada et al . , 2006; Carroll et al . , 2009 , 2010; Hori et al . , 2013 ) . A critical combination of components at such hotspots may trigger neocentreomere formation , the mechanisms of which are still unresolved . Previously , it has been observed that at very high levels of overexpression , CENP-A ceases to be centromere restricted ( Van Hooser et al . , 2001; Heun et al . , 2006; Gascoigne et al . , 2011 ) . Nevertheless , here we show that within a sixfold range of expression levels , the CENP-A loading machinery has a constant efficiency , which maintains a strict ratio between the centromeric and total pools of CENP-A . Thus , within a physiological range , centromeric CENP-A levels are regulated by a mass-action mechanism , where the loading efficiency is independent of the expression levels . This mechanism ensures that with fluctuating expression levels , CENP-A remains mainly centromere restricted , and may prevent potential neocentromere seeding . Remarkably , varying the amount of CENP-A at centromeres during perpetual growth in culture does not affect the levels of several other centromeric proteins . One possible explanation for this is that there is a fixed subset of ‘active’ CENP-A nucleosomes that is responsible for recruiting downstream components , even if the total amount of CENP-A is variable . Alternatively , an excess of CENP-A could form a chromatin domain that provides a ‘platform’ for recruitment of a centromere complex of fixed size . Surprisingly , however , we find that a critical amount of CENP-A for prevention of micronuclei is reached even before downstream centromere and kinetochore protein levels are affected ( Figures 6 and 8B ) . Our analysis indicates that the distribution of CENP-A among centromeres within one cell is generally uniform . However , in agreement with prior publications , we find that both the Y-centromere and a human neocentromere have lower CENP-A levels ( Amor et al . , 2004; Irvine et al . , 2004 ) . Interestingly , both these centromere types are atypical in that they are formed on relatively small genomic loci: ∼600 kb for the Y-centromere ( Abruzzo et al . , 1996 ) and ∼300 kb for the NeoCEN-4 ( Hasson et al . , 2013 ) , whereas autosomes and the X-chromosome have alpha-sattellite arrays of several magabases in size ( Wevrick and Willard , 1989; Mahtani and Willard , 1990; Lo et al . , 1999 ) . In addition , in contrast to canonical centromeres , neither the Y-centromere nor neocentromeres recruit the sequence-specific DNA binding protein CENP-B ( Earnshaw et al . , 1987; Amor et al . , 2004 ) , which has been hypothesized to alter the 3D structure of centromeric chromatin ( Pluta et al . , 1992 ) . The presence of CENP-B binding sites has recently been shown to have a role in phasing CENP-A nucleosomes ( Hasson et al . , 2013 ) , and to cooperate with CENP-A in kinetochore function ( Fachinetti et al . , 2013 ) , and may therefore be involved in regulation of centromeric CENP-A levels as well . Furthermore , high resolution analysis of a human neocentromere reveals a non-random distribution of CENP-A ( Hasson et al . , 2013 ) , where individual nucleosome positions are occupied in anywhere between 0 . 5% and 80% of cells ( Figure 7J , K ) . Thus , despite specific DNA sequences being neither sufficient nor required for centromere identity ( Earnshaw and Migeon , 1985; Voullaire et al . , 1993; Amor et al . , 2004; Marshall et al . , 2008 ) , the amount of CENP-A at centromeres likely results from a combination of a systematic cellular mechanism with a contribution of local sequence or chromatin features . In conclusion , several key mechanistic insights follow from our findings . First , while CENP-A nucleosomes are highly enriched at the centromere , most CENP-A is distributed at low levels throughout chromatin . This indicates that there is no exclusive pathway that restricts CENP-A assembly to centromeres . Nevertheless , we propose that the ample number of CENP-A nucleosomes facilitates a robust epigenetic signal that can absorb fluctuations in CENP-A inheritance and assembly in order to faithfully maintain centromere identity . Secondly , the requirement for a sizable number of CENP-A nucleosomes to perpetuate an active centromere reduces the likelihood for inadvertent detrimental neocentromere seeding without the need for a tightly restricted assembly mechanism . In addition , the fixed ratio between total and centromeric CENP-A levels may prevent excess CENP-A from accumulating at high density at non-centromeric loci , thus further reducing the probability of neocentromere formation . Finally , the number of centromeric CENP-A nucleosomes represents an ample pool of which only a subset is required to nucleate otherwise self-organized centromere and kinetochore complexes . In summary , from our analysis an integrated view of centromeric architecture , size , and regulation emerges ( Figure 8 ) that provides a basis to explain the self-propagating nature of the epigenetic centromere .
All human cell lines used were grown at 37°C , 5% CO2 . Cells were grown in DMEM/F-12 ( RPE ) , DMEM ( HeLa , U2OS , PDNC-4 ) , MEM ( primary fibroblasts; Coriell GM06170 ) , McCoy's 5A ( HCT-116 ) , or RPMI-1640 ( DLD-1 ) cell culture media . Media were supplemented with 10% fetal bovine serum ( FBS ) , 2 mM glutamine , 1 mM sodium pyruvate ( SP ) , 100 U/ml penicillin , and 100 μg/ml streptomycin , with the following exceptions: for RPE cells SP was substituted for 14 . 5 mM sodium bicarbonate; for HeLa newborn calf serum was used instead of FBS; for fibroblasts 15% FBS was used; for DLD-1 cells SP was omitted; and both SP and glutamine were omitted for HCT-116 cells . During live cell imaging , culture medium was replaced with Leibowitz's L-15 medium containing 10% FBS and 2 mM glutamine . LacI-GFP::LacO HCT-116 cells ( gift from Duane Compton , Thompson and Compton , 2011 ) were selected alternatingly with 2 μg/ml blasticidin and 300 μg/ml hygromycin; PDNC-4 cells were selected with 100 μg/ml neomycin . All media and supplements were purchased from Gibco ( Paisley , UK ) . All targeted cell lines are derived from wild-type hTERT RPE cells ( CA+/+ ) . Gene targeting was achieved by Adeno-associated virus ( AAV ) mediated delivery of targeting constructs essentially as described ( Berdougo et al . , 2009 ) , except in the case if CAG/−cells ( see below ) . The CA+/F cell line was created by inserting loxP sites surrounding CENP-A exons 2 and 4 as described previously ( Fachinetti et al . , 2013 ) . The CA+/− cell line was created by targeting the floxed CENP-A allele of CA+/F cells with a construct lacking 1373 bp of the CENP-A gene ( from 43 bp upstream of exon 2 to 134 bp downstream of exon 4 ) encompassing the essential CENP-A targeting domain ( Black et al . , 2007 ) . CAY/− cells were created by sequential targeting of a first CENP-A allele with the targeting construct inserting loxP sites flanking exon 3 and 4 as described above and the second allele by targeting EYFP ( carrying citrine and monomerization mutations: Q69M , A206K ) in frame with the CENP-A gene , immediately prior to the stop codon in exon 4 . The floxed allele was subsequently removed by retroviral delivery of HR-MMPCreGFP , a ‘Hit and Run’ Cre vector , as described ( Silver and Livingston , 2001 ) . CAG/− cells were created from an independent CA+/− clone where the remaining intact CENP-A allele was targeted with EGFP using a FACS-based strategy that we developed previously ( Mata et al . , 2012 ) . Targeting resulted in insertion of the EGFP ORF directly downstream the last coding sequence in exon 4 , just upstream of the endogenous stop codon , without insertion of any selectable marker gene . CAY/−+OE cells were created by stable transfection of and selection ( 5 μg/ml blasticidin ) for a CENP-A-YFP expression vector ( pBOS-Blast ) bearing a CENP-A-YFP fusion protein identical to that of the endogenous knockin locus in CAY/− cells . CAY/−+H2B-RFP and CA+/++H2B-RFP cell lines were created by stable transfection of and selection ( 5 μg/ml puromycin ) for a H2B-RFP expression vector ( Black et al . , 2007 ) in CAY/− and CA+/+ cells , respectively . All cell lines were monoclonally sorted by FACS . For the transient transfection experiment ( Figure 1F ) , wild-type HeLa cells were first synchronized in S phase by addition of 2 mM thymidine . After 17 hr , cells were released using 24 μM deoxycytidine and simultaneously transfected with untagged , wild-type CENP-A and/or HJURP expression vectors ( or an empty vector ) in combination with an EYFP-CENP-C expression vector ( Shah et al . , 2004 ) ( 2:2:1 proportion ) . 9 hr later , thymidine was re-added for an additional 15 hr , at which point cells were again released with deoxycytidine for 9 hr . A final thymidine arrest was performed and after 15 hr cells were fixed . Only cells expressing the positive transfection marker EYFP-CENP-C were analyzed . All stable and transient transfections were performed using Lipofectamine LTX ( Invitrogen; Carlsbad , CA ) according to the manufacturer's instructions . All samples were prepared in 1X Laemmli sample buffer , separated by SDS-PAGE , and transferred onto nitrocellulose membranes . Whole cell extracts were prepared by lysing cells directly in sample buffer , to ensure that the entire cellular protein pool remained present in the sample . Recombinant CENP-A/H4-complexes were purified as described previously ( Black et al . , 2004 ) , concentration was determined by A280 measurement and mixed with protein extracts from chicken DT40 cells to bring the overall protein concentration of the purified CENP-A protein preps to a level comparable to the RPE cell extracts . Absence of cross-recognition of human CENP-A antibody to chicken protein was confirmed by omission of recombinant human CENP-A protein in DT40 extracts ( Figure 2D , second lane ) . Alternatively , recombinant CENP-A/H4 was spiked into RPE cell extracts . Results obtained from the two methods are comparable ( 95 . 3 ± 14 . 0 ng [n = 8] and 75 . 4 ± 5 . 4 ng [n = 2] , respectively; p>0 . 5 ) . Cellular CENP-A quantity was determined by comparison of fluorescence derived from cellular and purified CENP-A . The following antibodies and dilutions were used: CENP-A ( #2186; Cell Signaling Technology , Danvers , MA or Ando et al . , 2002 ) at 1:1000 or tissue culture supernatant at 1:400 , respectively; α-tubulin ( DM1A; Sigma-Aldrich , St . Louis , MO ) at 1:5000; HJURP ( gift from Dan Foltz , Foltz et al . , 2009 ) at 1:10 , 000; Mis18BP1 ( A302-825A; Bethyl Laboratories , Inc . , Montgomery , TX ) at 1:2000; H4K20me2 ( ab9052; Abcam , Cambridge , UK ) at 1:1000 . IRDye800CW-coupled anti-mouse or anti-rabbit ( Licor Biosciences ) and DyLight680-coupled anti-mouse or anti-rabbit ( Rockland Immunochemicals , Gilbertsville , PA ) secondary antibodies were used prior to detection on an Odyssey near-infrared scanner ( Licor Biosciences , Lincoln , NE ) . Immunoblot signals were quantified using the Odyssey software , and a linear response was confirmed over a 32-fold range ( Figure 2E ) . Target protein signals were normalized to the α-tubulin loading control signal to correct for slight deviations in cell concentration between extracts of different RPE cell lines . Cell fractionation was performed for CAY/−+H2B-RFP cells after cell lysis in ice cold buffer consisting of 50 mM Tris-HCl ( pH 7 . 5 ) , 150 mM NaCl , 0 . 5 mM EDTA , 1% Triton-X 100 , 1 mM DTT , and a mix of protease inhibitors ( 1 mM PMSF , 1 μg/ml leupeptin , 1 μg/ml pepstatin , and aprotinin [A6279; Sigma , 1:1000 dilution] ) . Soluble proteins were separated from the insoluble fraction by centrifugation at 21 , 000×g at 4°C and resuspended in an equal volume of lysis buffer . Both supernatant and pellet fractions were incubated with 1 . 25 U/μl of benzonase nuclease ( Novagen , San Diego , CA ) on ice for 30 min prior to denaturation in Laemmli sample buffer . Imaging was performed on an Andor Revolution XD system , controlling an inverted microscope ( Eclipse-Ti; Nikon , Tokyo , Japan ) , an iXonEM+ EMCCD camera ( DU-897; Andor , Belfast , UK ) , a CSU-X1 spinning disk unit ( Yokogawa , Tokyo , Japan ) , a laser combiner/multi-port switch system ( Andor ) and a motorized stage ( Prior Scientific , Cambridge , UK ) , controlled by MicroManager software ( Edelstein et al . , 2010 ) . Images were collected using a Nikon 100X , 1 . 4 NA , Plan Apo oil immersion objective ( fixed cell imaging ) or a Nikon 60X , 1 . 2 NA , Plan Apo VC water immersion objective ( live cell imaging ) at 1× binning . For live cell imaging , the temperature of the chamber was maintained at 37°C . Cells grown on glass coverslips were fixed and mounted as described ( Bodor et al . , 2012 ) and imaged using a Zeiss LSM710 coupled to a motorized stage of an upright Zeiss Axio Examiner microscope equipped with a Zeiss 63X , 1 . 4 NA , Plan Apo oil immersion objective lens . A Coherent Chameleon Vision II multi-photon Ti-Sapphire laser was used to excite EYFP samples . All images were 512 × 512 pixels in size , with a pixel size of 0 . 09 μm . For all samples , an optimal setting of the laser power and PMT voltage was chosen to avoid pixel saturation and minimize photobleaching . The CLSM settings were kept constant so that valid comparisons could be made between measurements from different samples . Fluorescence lifetime imaging microscopy ( FLIM ) was performed by measuring the decay rate of EYFP using a Becker & Hickl time-correlated single photon counting hybrid detector coupled to the confocal LSM710 setup . The SPCImage ( Becker & Hickl , Berlin , Germany ) software was utilized to perform single exponential fitting for each pixel location . Cell fixation , immunofluorescence , and DAPI staining was performed as described previously ( Bodor et al . , 2012 ) . The following antibodies and dilutions were used: CENP-A ( gift from Tatsuo Fukagawa , Ando et al . , 2002 ) tissue culture supernatant at 1:100 , rabbit polyclonal CENP-B ( sc22788; Santa Cruz Biotechnology , Dallas , TX ) at 1:100 , tissue culture supernatant from mouse hybridomas expressing monoclonal CENP-B ( Earnshaw et al . , 1987 ) at 1:4 , CENP-C ( Foltz et al . , 2009 ) at 1:10 , 000 , CENP-T ( gift from Dan Foltz , Barnhart et al . , 2011 ) at 1:1000 , Hec1 ( 9G3 . 23; MA1-23308; Pierce , Rockford , IL ) at 1:100 , ACA ( anti-centromere antibodies; 83JD , gift from Kevin Sullivan ) at 1:100 . Fluorescent secondary antibodies were obtained from Jackson ImmunoResearch ( West Grove , PA ) or Rockland ImmunoChemicals and used at a dilution of 1:200 . Immunofluorescence signals of Figures 1C , 5E , 6B , 7G were automatically quantified using the CRaQ method as described previously ( Bodor et al . , 2012 ) using CENP-T or CENP-C as a centromere reference . Hec1 levels were measured exclusively in prometaphase or metaphase ( based on DAPI staining ) of unperturbed cells . Micronuclei were scored based on DAPI staining . Mitotic spreads were performed after mitotic shake-off of cells arrested overnight ( ∼16 hr ) in 250 ng/ml nocodazole . 25 , 000 cells/ml were swollen in 75 mM KCl and 5000 cells were cytospun onto coverslips using a Cytopro 7620 cytocentrifuge ( Wescor Inc . , Logan , UT ) for 4 min , at 1200 rpm , high acceleration . Cells were then fixed and processed for immunofluorescence as described above . Average centromere signals of both sisters were measured after background correction , by subtracting the minimum pixel value from the maximum of a box of 5 × 5 pixels around each sister centromere . Specific chromosomal markers were used as described in the text to detect centromeres of interest and signals were normalized to the average of all centromeres of the same cell spread . CA+/+ cells were mixed with CAY/− , CAG/− , or CAY/−+OE cells at a ∼1:4 ratio on 35 mm glass-bottom petri dishes ( MatTek Corporation , Ashland , MA ) . Non-cell permeable dextran-AlexaFluor647 ( 10 , 000 MW; Molecular Probes , Eugene , OR ) was added at 2–4 μg/ml to stain the medium outside of cells ( Figure 2A , I ) . To minimize oversampling , individual live cells were imaged at 500 nm axial resolution ( close to the resolution limit of the objective ) spanning the entire cell volume . Images were flatfield corrected for unequal illumination using the signal of a uniform fluorescent slide and the ‘Shading Corrector’ plugin for FIJI . For each axial section , the cell outline was determined based on absence of dextran-AlexaFluor647 staining , and the integrated fluorescence intensities inside the cell outline as well as those of 1–3 independent background regions per section were determined . Background corrected signals from all sections were summed to determine the total cellular fluorescence . Fluorescence measurements of CAY/− , CAG/− , or CAY/−+OE cells were corrected for autofluorescence by subtraction of average per pixel fluorescence intensity of non-fluorescent CA+/+ cells from the same dish . Alternatively , CA+/++H2B-RFP and CAY/−+H2B-RFP cells were mixed and no dextran was added to the medium . In this case , the H2B-RFP signal was used to determine the nuclear volume , and the total nuclear fluorescence was determined as described above for the total cellular volume . Automated centromere detection was performed by an analogous algorithm to a previous study ( Bodor et al . , 2012 , 2013 ) , where diffraction limited spots are detected based on their size , circularity , and feret's diameter . Centromere signals were measured by integrating the intensity of a 5 pixel diameter surrounding each centromere in the appropriate axial section . Local background fluorescence was derived by measuring the difference in intensity between concentric circles of 5 and 7 pixel diameter , and subtracted from centromeric signals ( Hoffman et al . , 2001 ) . In addition , centromeric signals were corrected for axial oversampling . For this , diffraction limited spots of yellow/green PS-Speck fluorescent beads ( Molecular Probes ) were measured in multiple plains . The sum intensity of individual beads from all these plains was compared to the signal in the plain with the maximum signal ( i . e . , the focal plane ) . The percentage of centromeric fluorescence was determined in relationship to the total fluorescence of each individual cell . To allow for cell cycle staging of CAY/− cells , transduction with hCdt1 ( 30/120 ) -RFP was performed using the BacMam 2 . 0 baculovirus system ( Invitrogen ) . Expression levels of transduced protein were allowed to stabilize for 3 days prior to analysis . Individual cells were followed by live cell microscopy using DIC and RFP signals . Nuclear RFP signals were tracked every ∼2 hr to monitor their cell cycle progression . Imaging of CENP-A-YFP and cellular volume were performed as described above . Analysis of the centromeric CENP-A ratio was performed as described above , but restricted to cells in which RFP levels were decreasing at the specific timepoint of analysis ( to exclude cells in G1 phase ) and which did not enter mitosis or showed an increase in RFP levels for at least the following 3–4 hr ( to exclude cells in G2 phase ) . Centromeric ratio was compared to non-transduced , randomly cycling cells ( Figure 3C ) or randomly cycling cells that were transduced , but not followed over time ( Figure 3—figure supplement 1 ) . For these experiments , wild-type cells used to measure cellular autofluorescence were seeded on a separate dish . CAY/− , CAG/− or CAY/−+OE cells were treated with nocodazole ( 250–500 ng/μl ) for ∼9 hr , after which cells were fixed and processed for immunofluorescence as described above . Sister centromere pairs were identified by CENP-B staining and GFP or YFP fluorescence intensity of each sister was measured and background corrected by subtracting the minimum pixel value of a 5 pixel diameter circle from the maximum value . The difference ( δ ) in fluorescence intensity and the sum ( Σ ) intensity of the two sisters were determined . The fluorescence intensity per segregating unit ( α ) was determined from the average δ2/Σ of all centromere pairs of the same experiment and cell line . The number of segregating units on each centromere was calculated as Σ/α , as described previously ( Rosenfeld et al . , 2005 , 2006 ) and in Figure 5A . In addition to sister centromeres , three independent rounds of random centromere pairing between all centromeres measured in a single experiment on CAG/− cells were performed and centromeric CENP-A-GFP units based on these pairings were quantified in Figure 5—figure supplement 1E . 4 kb-LacO , LacI-GFP Saccharomyces cerevisiae ( gift from Kerry Bloom , Lawrimore et al . , 2011 ) were grown in minimal synthetic media ( Yeast nitrogen base [Sigma] + complete synthetic defined single drop-out medium lacking uracil and histidine [MP Biomedicals , Solon , OH] ) , supplemented with 2% D ( + ) Glucose ( Merck , Darmstadt , Germany ) . Prior to imaging , log-phase cells ( OD600 of ∼0 . 7 ) were transferred onto a 2% low melting agarose pad and sealed under a coverslip with VALAP ( 1:1:1 vaseline:lanolin:paraffin ) . CAG/− cells were grown on 35-mm glass-bottom petri dishes and yeast and human cells were imaged using identical settings during the same microscopy session . Fluorescence intensity of centromeres and Lac-arrays were quantified after background correction ( maximum minus minimum of a 5 × 5 pixel box ) . CENP-A ChIP-Seq data from the PDNC-4 neocentromere cell line ( Accession #GSE44724 ) was processed as previously described ( Hasson et al . , 2013 ) . Briefly , paired-end ChIP-Seq reads were aligned to the human genome build hg19 with Bowtie2 version 2 . 0 . 0 using paired-end mode . Reads were aligned by using a seed length of 50 bp , and only the single best alignment per read with up to two mismatches was reported in the SAM file . The aligned mate pairs were joined in MATLAB by requiring ≥95% overlap identity . The joined reads were aligned to the PDNC-4 neocentromere and only reads which mapped with 100% identity were used in the subsequent analysis . Nucleosome positions at the neocentromere were determined using the ‘findpeaks’ function in MATLAB . The probability of CENP-A occupancy at a given position was determined according to the following formula: ( total reads overlying that position ) × ( 216 CENP-A nucleosomes [Figure 7I] ) / ( total reads mapping to the entire neocentromere ) . All calculations represented in Figure 8C were performed in R . For these calculations we assume that CENP-A is inherited following a binominal distribution , consistent with our findings ( Figure 5 , Figure 5—figure supplement 1A , C ) . To determine the chance ( X ) of any chromosome reaching critical levels of CENP-A , the ‘pbinom’ function was used to calculate the fraction of a binomial distribution ( where p=0 . 5 and n [steady state number of nucleosomes] = 200 or was varied as indicated ) that is either below a critical value ( c = 22 , or varied as indicated ) or above a critical value ( n−c ) . To determine the chance that any chromosome in a cell ( containing 46 chromosomes ) reaches critical levels , we calculated the chance that 46 independent centromeres do not reach critical levels and subtracted this chance from 1; [1 − ( 1 − X ) 46] . | The genetic information in a cell is packed into structures called chromosomes . These contain strands of DNA wrapped around proteins called histones , which helps the long DNA chains to fit inside the relatively small nucleus of the cell . When a cell divides , it is important that both of the new cells contain all of the genetic information found in the parent cell . Therefore , the chromosomes duplicate during cell division , with the two copies held together at a single region of the chromosome called the centromere . The centromere then recruits and coordinates the molecular machinery that separates the two copies into different cells . Centromeres are inherited in an epigenetic manner . This means that there is no specific DNA sequence that defines the location of this structure on the chromosomes . Rather , a special type of histone , called CENP-A , is involved in defining its location . Bodor et al . use multiple techniques to show that human centromeres normally contain around 400 molecules of CENP-A , and that this number is crucial for ensuring that centromeres form in the right place . Interestingly , only a minority of the CENP-A molecules are located at centromeres; yet this is more than at any other region of the chromosome . This explains why centromeres are only formed at a single position on each chromosome . When the chromosomes separate , the CENP-A molecules at the centromere are randomly divided between the two copies . In this way memory of the centromere location is maintained . If the number of copies of CENP-A inherited by one of the chromosomes drops below a threshold value , a centromere will not form . However , Bodor et al . found that the number of CENP-A molecules in a centromere is large enough , not only to support the formation of the centromere structure , but also to keep it above the threshold value in nearly all cases . This threshold is also high enough to make it unlikely that a centromere will form in the wrong place because of a random fluctuation in the number of CENP-A molecules . Therefore , the number of CENP-A molecules is crucial for controlling both the formation and the inheritance of the centromere . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2014 | The quantitative architecture of centromeric chromatin |
Centrioles are composed of long-lived microtubules arranged in nine triplets . However , the contribution of triplet microtubules to mammalian centriole formation and stability is unknown . Little is known of the mechanism of triplet microtubule formation , but experiments in unicellular eukaryotes indicate that delta-tubulin and epsilon-tubulin , two less-studied tubulin family members , are required . Here , we report that centrioles in delta-tubulin and epsilon-tubulin null mutant human cells lack triplet microtubules and fail to undergo centriole maturation . These aberrant centrioles are formed de novo each cell cycle , but are unstable and do not persist to the next cell cycle , leading to a futile cycle of centriole formation and disintegration . Disintegration can be suppressed by paclitaxel treatment . Delta-tubulin and epsilon-tubulin physically interact , indicating that these tubulins act together to maintain triplet microtubules and that these are necessary for inheritance of centrioles from one cell cycle to the next .
The major microtubule organizing center of mammalian cells , the centrosome , is composed of a pair of centrioles with associated appendages and pericentriolar material . The centrioles have a nine-fold symmetry and are formed , in part , of long-lived microtubules , which persist through multiple cell divisions ( Kochanski and Borisy , 1990; Balestra et al . , 2015 ) . In most organisms , including humans , the centriolar microtubules have a triplet structure , found only in centrioles . This structure consists of a complete A-tubule and associated partial B-tubule attached to the A-tubule wall , and a partial C-tubule attached to the B-tubule wall . The molecular mechanisms involved in making triplet microtubules are not well-understood , even in the well-characterized somatic centriole cycle of mammalian cells . In these cells centrioles duplicate once per cycle , such that daughter cells receive exactly one pair of centrioles . Centriole duplication is initiated at the G1-S transition when the kinase PLK4 localizes to a single focus on the mother centriole ( Sonnen et al . , 2012 ) . Subsequently , the cartwheel , formed by SASS6 oligomerization , assembles to template the 9-fold symmetry of the newly-formed procentriole ( Guichard et al . , 2017; Hilbert et al . , 2016 ) . Microtubules are added to the cartwheel underneath a cap of CP110 ( Kleylein-Sohn et al . , 2007 ) . By G2-M , the triplet microtubules are completely formed ( Vorobjev and Chentsov YuS , 1982 ) . Subsequently , the A- and B-tubules elongate to the full ~500 nm length of the centriole , forming a distal compartment with doublet microtubules and marked by POC5 ( Azimzadeh et al . , 2009 ) . In mitosis , the cartwheel is lost , and the newly-formed centriole becomes disengaged from its mother and acquires pericentriolar material ( Vorobjev and Chentsov , 1980; Vorobjev and Chentsov YuS , 1982; Khodjakov and Rieder , 1999; Tsou and Stearns , 2006; Tsou et al . , 2009 ) . In G2-M of the following cell cycle , the centriole acquires appendages , marking its maturation into a centriole that can nucleate a cilium ( Graser et al . , 2007; Guarguaglini et al . , 2005 ) . Members of the tubulin superfamily are critical for centriole formation and function . All eukaryotes have alpha- , beta- and gamma-tubulin , but the tubulin superfamily also includes three less-studied members , delta-tubulin , epsilon-tubulin , and zeta-tubulin . These tubulins are found in a subset of eukaryotes , and are evolutionarily co-conserved , making up the ZED tubulin module ( Turk et al . , 2015 ) . In the unicellular eukaryotes Chlamydomonas , Tetrahymena , Paramecium and Trypanosoma , mutations in delta-tubulin or epsilon-tubulin result in centrioles that lack triplet microtubules ( Dupuis-Williams et al . , 2002; Dutcher and Trabuco , 1998; Dutcher et al . , 2002; Gadelha et al . , 2006; Garreau de Loubresse et al . , 2001; Goodenough and StClair , 1975; Ross et al . , 2013 ) . Humans and other placental mammals have delta-tubulin and epsilon-tubulin , but lack zeta-tubulin ( Findeisen et al . , 2014; Turk et al . , 2015 ) . Here , we show that human cells lacking delta-tubulin or epsilon-tubulin also lack triplets , that this results in unstable centrioles and initiation of a futile cycle of centriole formation and disintegration , and identify an interaction between delta-tubulin and epsilon-tubulin .
To determine the roles of delta-tubulin and epsilon-tubulin in the mammalian centriole cycle , null mutations in TUBD1 and TUBE1 were made using CRISPR/Cas9 genome editing in hTERT RPE-1 human cells . Recent work has established that loss of centrioles in mammalian cells results in a p53-dependent cell-cycle arrest ( Bazzi and Anderson , 2014; Lambrus et al . , 2015; Wong et al . , 2015 ) . We found that homozygous null mutations of delta-tubulin or epsilon-tubulin could only be isolated in TP53−/− cells , thus all subsequent experiments use RPE-1 TP53−/− cells as the control . Three TUBD1−/− and two TUBE1−/− cell lines were generated ( Figure 1—figure supplement 1 ) . Sequencing of the alleles in these lines demonstrated that they were all consistent with independent cutting by Cas9 and processing by non-homologous end-joining of the two alleles in a diploid cell . The TUBD1−/− lines are all compound heterozygotes bearing small deletions of less than 20 base pairs proximal to the cut site on one chromosome and insertion of one base pair on the other , resulting in frameshift and premature stop mutations . The two TUBE1−/− lines are also compound heterozygotes bearing large deletions surrounding the cut site , that in each case remove an entire exon and surrounding DNA , including the ATG start site . In all cases , the next ATG is not in-frame . We conclude that these alleles are likely to be null , or strong loss-of-function mutations . We next assessed the phenotype of TUBD1−/− and TUBE1−/− cells stably expressing GFP-centrin as a marker of centrioles . Many cells in an asynchronous population had multiple , unpaired centrin foci ( Figure 1A ) . These foci also labeled with the centriolar proteins CP110 and SASS6 ( see Figures 2 and 3 ) . To determine whether these foci are centrioles , and to assess their ultrastructure , we analyzed them using correlative light-electron microscopy . In serial sections of interphase TUBE1−/− ( Figure 1A ) and TUBD1−/− ( Figure 1B ) cells , some of the centrin-positive foci corresponded to structures that resemble centrioles , but were narrower than typical centrioles and lack appendages . Strikingly , only singlet microtubules were identified in the two centriole cross-sections observed , both from TUBD1−/− cells ( Figure 1C ) . The measured diameters of other centriole sections from both TUBD1−/− and TUBE1−/− mutant cells were also consistent with singlet microtubule structure ( Figure 1D , E ) . Centrioles in TUBD1−/− and TUBE1−/− cells were of similar outer diameter: 172 . 5 nm ±13 nm in TUBD1−/− cells ( n = 17 ) , 174 . 6 nm ±8 nm in TUBE1−/− cells ( n = 13 ) . In contrast , centrioles in control TP53−/− cells had a larger diameter: 222 . 9 ± 9 nm ( n = 24 ) for mother centrioles , and 212 . 1 ± 10 nm ( n = 10 ) for procentrioles , similar to previous measurements of mammalian cell centrioles ( Loncarek et al . , 2008; Wang et al . , 2015 ) . The reduced outer diameter of these aberrant centrioles is consistent with the presence of only singlet microtubules ( Vorobjev and Chentsov YuS , 1982 ) . We also noted that there is a slightly reduced central lumen diameter in mutant centrioles ( Figure 1E ) . It is less clear why the mutant centrioles would have a reduced lumenal diameter , but we note that this result is consistent with the observation that normal procentrioles with singlet microtubules , prior to the elaboration of triplets , also have a reduced lumenal diameter ( Vorobjev and Chentsov YuS , 1982 ) . These results demonstrate that cells lacking either delta-tubulin or epsilon-tubulin form defective centrioles that lack normal triplet microtubules . This is similar to the defects reported for delta-tubulin and epsilon-tubulin mutants in unicellular eukaryotes ( Dupuis-Williams et al . , 2002; Dutcher and Trabuco , 1998; Dutcher et al . , 2002; Gadelha et al . , 2006; Garreau de Loubresse et al . , 2001; Goodenough and StClair , 1975; Ross et al . , 2013 ) . The length profiles of centrioles in both tubulin mutants revealed important aspects of the defect associated with lack of normal microtubule triplet structure . In interphase , the length of centrioles from TUBD1−/− cells ( 222 nm ±37 nm; n = 18 ) and TUBE1−/− cells ( 339 nm ±131 nm; n = 15 ) was similar to that of control procentrioles ( 207 nm ±20 nm; n = 9 ) ( Figure 2A ) . All were shorter than control mother centrioles ( 485 . 6 nm ±43 nm; n = 14 ) . We next analyzed the ultrastructure of centrioles in a TUBE1−/− prometaphase cell using correlative light-electron microscopy ( Figure 2B ) . These centrioles ( n = 3 ) exhibited a remarkable morphological phenotype , consisting of two electron-dense segments , one of ~50 nm and the other of ~200 nm , connected by singlet microtubules spanning a gap of ~250 nm . The combined length ( ~500 nm ) of these structures approximates that of typical mature mammalian centrioles ( Figure 2A ) . We hypothesized that the aberrant centrioles formed in TUBD1−/− and TUBE1−/− cells elongate in G2-M , but that only the A-tubule is present . The shorter density might correspond to the CP110 cap , and the longer density to the centriole end containing the cartwheel . Procentrioles in control cells reached a full length of 403 nm ±22 nm in G2-M ( Figure 2A ) , an increase of approximately 200 nm from their interphase state . We tested whether TUBD1−/− and TUBE1−/− mutant centrioles exhibited this same 200 nm increase in the separation between CP110 and SASS6 foci by mitotic entry . We found that in TUBD1−/− and TUBE1−/− and control interphase cells , the centroids of CP110 and SASS6 foci were separated by a mean distance of 0 . 3 μm , whereas in mitotic cells the foci were separated by a mean distance of 0 . 5 μm ( Figure 2C , D ) . Thus , despite their structural defects , centrioles in TUBD1−/− and TUBE1−/− cells undergo the normal cell cycle-dependent elongation . The elongation of centrioles in G2/M creates a distal compartment that is a feature of centrioles in some , but not all , organisms . In mammalian cells this compartment is defined by the centrin-binding protein POC5 ( Azimzadeh et al . , 2009 ) . The lack of electron-dense structure between the two centriole segments joined by singlet microtubules in mitotic TUBE1−/− mutant cells might be due to a failure to recruit distal compartment components . Consistent with this , we found that POC5 was present in centrioles from mitotic control cells and absent from those in TUBD1−/− and TUBE1−/− cells ( Figure 2E ) . This suggests that the doublet microtubules of the extended centriole distal end are required for defining this compartment . Together , these results indicate that the primary centriolar defect in cells lacking delta-tubulin or epsilon-tubulin is the absence of triplet microtubules . To determine the consequences of this defect on the centriole cycle , we determined the distribution of centrioles in asynchronously dividing cells , as determined by staining for centrin and CP110 . Control cells had a centriole number distribution typical of TP53−/− cells , with approximately 50% of cells having two centrioles , corresponding to cells in G1 phase , 40% having three to four centrioles , corresponding to cells in S through M phases , and 10% having more than four centrioles ( Figure 3A , B ) . In contrast , in TUBD1−/− and TUBE1−/− cells , approximately 50% of cells had five or more centriole foci , whereas 50% of cells had no detectable centriole foci ( Figure 3A , B ) . Similar centriole distributions were found in several independently derived TUBD1−/− and TUBE1−/− cell lines , and this phenotype could be rescued by expression of delta-tubulin and epsilon-tubulin , respectively ( Figure 3—figure supplement 1A , B ) . We reasoned that a possible explanation for the centriole distribution in TUBD1−/− and TUBE1−/− cells is that the centriole structures we observed by EM are produced de novo in each cell cycle , and that these aberrant centrioles are unstable and do not persist into the next cell cycle . This hypothesis predicts that the aberrant centrioles in TUBD1−/− and TUBE1−/− cells would ( 1 ) not be paired , since de novo centrioles only form in the absence of an existing centriole , ( 2 ) lack markers of maturation such as distal appendages , since they would not persist to the point of acquiring such proteins , ( 3 ) fail to recruit substantial pericentriolar material , since the centriole-centrosome conversion occurs at entry to the next cell cycle , and ( 4 ) would be formed in S phase , and be lost at some point prior to the subsequent S phase . In agreement with this hypothesis , the centrioles in mutant cells , as visualized by centrin and CP110 were never observed to be closely apposed , as is typical of wild-type cells ( Figure 3A ) . Rather , in interphase they appeared to be distributed within the central region of the cell ( Figure 3A ) . The centrioles in asynchronous TUBD1−/− and TUBE1−/− cells all lacked Cep164 , a component of the centriolar distal appendage and marker of mature centrioles that have progressed through at least one cell cycle ( Figure 3C ) , whereas approximately 40% of all centrioles were positive for Cep164 in asynchronous control cells , consistent with the cycle of distal appendage acquisition ( Nigg and Stearns , 2011 ) . Lastly , most of the centrioles in TUBD1−/− and TUBE1−/− cells lacked detectable gamma-tubulin ( Figure 3C ) , and those that stained positive had less than centrioles in control cells ( Figure 3—figure supplement 1C ) . In addition , we noted that SASS6 , the cartwheel protein that is present in nascent and recently-formed centrioles , but is lost from centrioles at the mitosis-interphase transition in human cells , was present in most of the centrioles in TUBD1−/− and TUBE1−/− cells , consistent with these centrioles originating in the observed cell cycle , but not having successfully persisted into the subsequent cell cycle . To investigate the fate of newly-formed centrioles in TUBD1−/− and TUBE1−/− cells , we next tested the cell cycle-dependence of the formation and loss of aberrant centrioles in mutant cells ( Figure 3D ) . As in previous experiments , about 50% of TUBD1−/− and TUBE1−/− cells in an asynchronous population had centrin and CP110-positive centriole foci . Cell cycle stages were analyzed as follows: G0/G1 , synchronized by serum withdrawal; S phase , identified from asynchronous culture by PCNA labeling; G2 , synchronized by the CDK1 inhibitor RO-3306; and M , identified from asynchronous culture by presence of condensed chromatin ( Figure 3D ) . TUBD1−/− and TUBE1−/− cells in G0/G1 mostly lacked centriole structures , whereas cells in S-phase , G2 and mitosis had them . These results indicate that in TUBD1−/− and TUBE1−/− cells , aberrant centrioles are formed in S-phase , persist into mitosis , and are absent in G1 . We note that this loss of centriole structure is likely due to a specific event that occurs at the mitosis-interphase transition , rather than simply time since formation , since cells were arrested in G2 for 24 hr , which is substantially longer than the normal progression through mitosis to G1 , nevertheless the centriole structures persisted ( Figure 3D ) . The timing of centriole loss in the mitosis-interphase transition was more finely determined in both fixed time-point and live imaging experiments . Control or TUBE1−/− cells were synchronized by mitotic shakeoff , and the presence of centriole foci was assessed over time as cells entered G1 ( Figure 3E ) . In control cells , the number of centrioles followed the pattern expected from the centriole duplication cycle . In TUBE1−/− cells , the majority of mitotic cells had centrioles . By 1 hr after shakeoff , the fraction of interphase cells without centrioles had increased to 50% , and this fraction continued to increase at 2 hr and 3 hr after shakeoff . By 12 hr after shakeoff , 56 ± 12% of cells had entered S-phase , and centriole structures began to appear , consistent with de novo centriole formation . We also imaged control and mutant cells expressing GFP-centrin to visualize centrioles in live cells ( Figure 3F , Videos 1 and 2 ) . Centrioles in control cells segregated normally in mitosis , and the mitotic interval was 46 min ±6 min ( n = 11 ) . In contrast , centrioles in TUBE1−/− cells did not persist into the next interphase , and the mitotic interval was longer , at 106 min ±43 min ( n = 10 ) . The prolonged time in mitosis is similar to that observed for acentriolar human cells ( Lambrus et al . , 2015 ) . Thus , delta-tubulin and epsilon-tubulin are not required to initiate centriole formation in human cells , but the aberrant centrioles that form in their absence are unstable and disintegrate during progression from M phase to the subsequent G1 phase . We note that this phenotype is specific to loss of TUBD1 and TUBE1 , rather than a property of de novo centrioles in general . De novo centrioles formed after washout of the centriole duplication inhibitor centrinone persisted through mitosis and the subsequent G1 ( Figure 3—figure supplement 1D ) , consistent with previous reports ( La Terra et al . , 2005 ) . We hypothesized that centriole disintegration in the absence of TUBD1 and TUBE1 may instead result from instability of the elongated singlet centriolar microtubules that we observed in mitotic cells . It follows that if these microtubules could be stabilized , the centrioles might persist into the next cell cycle , despite their structural defects . To test this , G2-M stage TUBE1−/− cells were treated with the microtubule-stabilizing drug paclitaxel and the presence of centrioles assessed after forcing progression into interphase . Paclitaxel treatment did not prevent centriole elongation , as measured by the separation between CP110 and SASS6 foci , as in Figure 2 ( 0 . 49 µm ± 0 . 2 µm; n = 105; not significantly different from TUBE1−/− mitotic cells in Figure 2 by unpaired two-tailed t-test ) . After 3 hr of paclitaxel treatment , cells were treated with the CDK inhibitor RO-3306 , which resulted in exit from mitosis as evidenced by flattening of cells and formation of micronuclei . The effect of paclitaxel was evident as bundling of microtubules compared to control cells ( Figure 4A ) . Centrioles in these cells were still present 3 hr after RO-3306 treatment , whereas control RO-3306-treated cells that were not treated with paclitaxel lacked centrioles ( Figure 4A , B ) . It has been suggested recruitment of pericentriolar material is important to stabilize centrioles , and that centrioles that fail to recruit PCM can be stabilized by induced retention of the SASS6-containing cartwheel ( Izquierdo et al . , 2014 ) . However , the TUBE1−/− centrioles stabilized by paclitaxel treatment still failed to recruit high levels of gamma-tubulin ( Figure 4C ) , and lost their SASS6 cartwheel ( 97% ± 2% cells completely lack SASS6 foci , three independent experiments with 100 cells each; and Figure 4D ) as expected for centrioles that have transited mitosis . We propose that preventing depolymerization of the centriolar microtubules in TUBE1−/− cells stabilizes the structure of these aberrant centrioles such that they survive into the next cell cycle . One important observation of this work is that the phenotypes of delta-tubulin and epsilon-tubulin null mutants are similar . This suggests that the proteins work together to accomplish their function . To test this hypothesis , we assessed the ability of delta-tubulin and epsilon-tubulin to interact by co-immunoprecipitation from human HEK293T cells co-expressing tagged versions of the proteins . Epsilon-tubulin could be immunoprecipitated with delta-tubulin , and not with GFP from control cells ( Figure 4E ) . These results indicate that epsilon-tubulin and delta-tubulin can interact , and we speculate that they may dimerize to form higher-order structures , as do alpha-tubulin and beta-tubulin . Interestingly , comparisons of the predicted surfaces of delta-tubulin and epsilon-tubulin that correspond to the interaction surfaces of alpha-tubulin and beta-tubulin revealed both similarities and differences that might influence their potential for interaction with themselves or other tubulins ( Inclán and Nogales , 2001 ) . Triplet microtubules are absent in delta-tubulin or epsilon-tubulin mutant cells in all organisms that have been examined , and our results suggest that delta-tubulin and epsilon-tubulin are required either to form the triplet microtubules , or to stabilize them against depolymerization . The former seems unlikely , since the presence of triplet centriolar microtubules is not strictly correlated with the presence of delta-tubulin and epsilon-tubulin in evolution ( Figure 4—figure supplement 1 ) . Among the organisms that lack delta-tubulin and epsilon-tubulin , C . elegans lacks triplet microtubules , but both Drosophila and the plant Ginkgo biloba have triplet microtubules in their sperm cells . Since loss of these tubulins must have occurred independently in the dipteran insect and plant lineages , the most parsimonious interpretation is that triplet microtubule formation itself does not require delta-tubulin or epsilon-tubulin , rather than that these two lineages independently evolved mechanisms of triplet formation in their absence . Thus , we consider it more likely that delta-tubulin and epsilon-tubulin are required for stabilization of the centriolar triplets in most organisms . We do not yet know the molecular basis of this differential requirement for delta-tubulin or epsilon-tubulins with respect to microtubule triplet stability . However , we note that those few centriole-bearing organisms that lack delta-tubulin and epsilon-tubulin have simpler centriole structures that lack typical distal appendages , and also often lack a distal compartment that is typical of more complex centrioles . Why do centrioles disintegrate in delta-tubulin and epsilon-tubulin mutant cells ? We have shown that in these cells , aberrant centrioles with elongated singlet microtubules become unstable as cells progress through mitosis , and that disintegration can be suppressed by treatment with the microtubule stabilizing drug paclitaxel . We do not yet know the basis for the cell-cycle dependence of this effect , but here consider three possible , non-exclusive , explanations for why centrioles from the mutants might be more sensitive to disintegration . First , it is possible that doublet and triplet microtubules are inherently more stable than singlet microtubules , and that this inherent stability is responsible for the fact that centriolar microtubules are non-dynamic . To our knowledge , this has not been directly tested , in the absence of other proteins that might also affect dynamics . We note that ciliary axonemes , made of doublet microtubules , can be dynamic , although in Chlamydomonas , where this has been best characterized , disassembly of the axoneme is slow , and is under complex regulatory control ( Lefebvre et al . , 1978; Marshall et al . , 2005; Hu et al . , 2015 ) . Second , it is possible that doublet and triplet microtubules provide unique interaction surfaces to recruit stabilizing proteins . Consistent with this possibility , non-tubulin densities have been identified in cryoEM structures of centrioles and ciliary axonemes ( Li et al . , 2012; Ichikawa et al . , 2017 ) . We found that POC5 , a component of the distal end compartment , is not recruited to singlet microtubule centrioles; perhaps POC5 binds to and stabilizes the doublet microtubules in the distal compartment of centrioles . A final possibility is that centrioles lacking the normal triplet structure would likely also lack the A-C linker , which bridges the A- and C-tubules of adjoining triplets . Perhaps the A-C linker stabilizes centriolar microtubules by direct interaction with them , in addition to providing higher-order organization to the structure . No components of the A-C linker have been identified , but the poc1 mutant in Tetrahymena causes partial loss of this linker and defects in triplet microtubule organization ( Meehl et al . , 2016 ) . Each of these models has in common that the triplet microtubules of the centrioles are more stable , either intrinsically and/or by recruitment of stabilizing proteins , than typical singlet microtubules; further work will be required to determine the nature of this stability , and why it is particularly critical at the mitosis-interphase transition . Here we have shown that delta-tubulin and epsilon-tubulin likely work together in a critical aspect of centriole structure and function , and that cells lacking either tubulin undergo a futile cycle of de novo centriole formation and disintegration . Our results show that in human cells , delta-tubulin and epsilon-tubulin act to stabilize centriole structures necessary for inheritance of centrioles from one cell cycle to the next , perhaps by stabilizing the main structural feature of centrioles , the triplet microtubules .
hTERT RPE-1 TP53−/− cells were a gift from Meng-Fu Bryan Tsou ( Memorial Sloan Kettering Cancer Center ) and were cultured in DMEM/F-12 ( Corning ) supplemented with 10% Cosmic Calf Serum ( CCS; HyClone ) . HEK293T/17 cells ( RRID:CVCL_1926 ) for lentivirus production ( see below ) were obtained from the ATCC and cultured in DMEM ( Corning ) supplemented with 10% CCS . hTERT RPE-1 and HEK293T/17 cells were authenticated using STR profiling using CODIS loci . All other cell lines used were derived from hTERT RPE-1 TP53−/− cells . Stable TP53−/−; TUBE1−/− and TP53−/−; TUBD1−/− knockout cell lines were made in the hTERT RPE-1 TP53−/− cells by CRISPR/Cas9 ( see below ) . For rescue experiments , clonal knockout cell lines were rescued using lentiviral transduction ( see below ) . All cells were cultured at 37°C under 5% CO2 , and are mycoplasma-free ( Uphoff and Drexler , 2004 ) . Recombinant lentiviruses were made by cotransfection of HEK293T cells with the respective transfer vectors , second-generation lentiviral cassettes ( packaging vector psPAX2 , pTS3312 and envelope vector pMD2 . G , pTS3313 ) using 1 µg/µL polyethylenimine ( PEI; Polysciences ) . The medium was changed 6–8 hr after transfection , and viral supernatant was harvested after an additional 48 hr . hTERT RPE-1 TP53−/− GFP-centrin cells were made by transduction with mEGFP-centrin2 ( pTS4354 ) lentivirus and 8 µg/mL Sequabrene carrier ( Sigma-Aldrich ) . Cells were cloned by limiting dilution into 96-well plates . TUBD1−/− cell lines were generated using lentiCRISPRv2 ( Addgene plasmid #52961 ( Sanjana et al . , 2014; Shalem et al . , 2014 ) with the sgRNA sequence CTGCTCTATGAGAGAGAATG ( pTS4617 ) . hTERT RPE-1 TP53−/− GFP-centrin cells were transduced with lentivirus and 8 µg/mL Sequabrene for 72 hr , then passaged into medium containing 6 µg/mL puromycin . Puromycin-containing culture medium was replaced daily for 5 days until all cells in uninfected control had died . Puromycin-resistant cells were cloned by limiting dilution into 96-well plates , followed by genotyping and phenotypic analysis . TUBE1−/− cell line 1 was generated using pX330 ( Addgene plasmid #42230 Cong et al . , 2013 ) with the sgRNA sequence GGGTAGAGACCTGGTCGCCG ( pX330-TUBE1 , pTS3752 ) . hTERT RPE-1 TP53−/− cells were transiently co-transfected with pX330-TUBE1 and EGFP-expressing vector pEGFP-N1 ( Clontech , pTS3627 ) at 9:1 ratio using Continuum Transfection Reagent ( Gemini Bio-Products ) . GFP-positive cells were clonally sorted into single wells of 96-well plates by FACS , followed by genotyping and phenotypic analysis . Cells were subsequently transduced with GFP-centrin2 lentivirus for CLEM . TUBE1−/− cell line 2 was generated using lentiCRISPRv2 with the sgRNA sequence GCGCACCACCATGACCCAGT ( pTS4615 ) . Transduction and selection were carried out as for TUBD1−/− cell lines . Both rescue construct transfer vectors contained opposite orientation promoters: EF-1alpha promoter driving monomeric Kusabira Orange kappa ( mKOk ) with rabbit beta-globin 3’UTR , as well as mouse PGK promoter driving the rescue construct with WPRE . For the delta-tubulin rescue construct , silent mutations were made in the PAM and surrounding sequence such that it was no longer complementary to the lentiCRISPR sgRNA ( C117G and A120T ) using QuikChange Lightning Site-Directed Mutagenesis Kit ( Agilent ) ( pTS4665 ) . For the epsilon-tubulin rescue construct , full-length TUBE1 cDNA was used ( pTS4666 ) . Using these transfer vectors , lentivirus was produced and TUBD1−/− and TUBE1−/− cells , respectively , were transduced . For rescue experiments , cells expressing mKOk were counted . Correlative light and electron microscopy ( CLEM ) was performed as described previously ( Kong and Loncarek , 2015 ) , using hTERT RPE-1 TP53−/− TUBD1−/− and TP53−/− TUBE1−/− GFP-centrin cells . Cells in Rose chambers were enclosed in an environmental chamber at 37°C and imaged on an inverted microscope ( Eclipse Ti; Nikon , Tokyo , Japan ) equipped with a spinning-disk confocal head ( CSUX Spinning Disk; Yokogawa Electric Corporation , Tokyo , Japan ) . After analysis by live imaging , Rose chambers were perfused with freshly prepared 2 . 5% glutaraldehyde , and 200 nm thick Z-sections spanning the entire cell were recorded to register the position of centrioles . Cell positions on coverslips were then marked by diamond scribe . Rose chambers were disassembled , and cells were washed in PBS , followed by staining with 2% osmium tetroxide and 1% uranyl acetate . Samples were dehydrated and embedded in Embed 812 resin . The same cells identified by light microscopy were then serially sectioned . The 80 nm-thick serial sections were transferred onto copper slot grids , stained with uranyl acetate and lead citrate , and imaged using a transmission electron microscope ( H-7650; Hitachi , Tokyo , Japan ) . Cells were grown on poly-L-lysine-coated #1 . 5 glass coverslips ( Electron Microscopy Sciences ) . Cells were washed with PBS , then fixed with −20°C methanol for 15 min . Coverslips were then washed with PBS and blocked with PBS-BT ( 3% BSA , 0 . 1% Triton X-100 , 0 . 02% sodium azide in PBS ) for 30 min . Coverslips were incubated with primary antibodies diluted in PBS-BT for 1 hr , washed with PBS-BT , incubated with secondary antibodies and DAPI diluted in PBS-BT for 1 hr , then washed again . Samples were mounted using Mowiol ( Polysciences ) in glycerol containing 1 , 4 , -diazobicycli-[2 . 2 . 2]octane ( DABCO , Sigma-Aldrich ) antifade . Primary antibodies used for immunofluorescence: mouse IgG2b anti-centrin3 , clone 3e6 ( 1:1000 , Novus Biological , RRID:AB_537701 ) , mouse IgG2a anti-centrin , clone 20H5 ( 1:200 , EMD Millipore , RRID:AB_10563501 ) , rabbit anti-CP110 ( 1:200 , Proteintech ) , mouse IgG2b anti-SASS6 ( 1:200 , Santa Cruz ) , mouse IgG1 anti-gamma-tubulin , clone GTU-88 ( 1:1000 , Sigma-Aldrich , RRID:AB_477584 ) , rabbit anti-POC5 ( 1:500 , Bethyl Laboratories , RRID:AB_10949152 ) , rabbit anti-CEP164 ( 1:500 , described previously ( Lee et al . , 2014 ) , mouse IgG2a anti-PCNA ( 1:500 , BioLegend , RRID:AB_314692 ) , mouse IgG1 anti-alpha-tubulin , clone DM1A ( 1:1000 , Sigma-Aldrich , RRID:AB_477583 ) . Primary antibodies used for Western blotting: goat anti-GFP ( 1:500 , Rockland , RRID:AB_218182 ) , mouse IgG1 anti-myc , clone 9e10 ( 1:100 , Developmental Studies Hybridoma Bank , RRID:AB_2266850 ) . For immunofluorescence , AlexaFluor conjugated secondary antibodies ( Thermo-Fisher ) were diluted 1:1000 . For Western blotting , IRDye conjugated donkey secondary antibodies ( LiCOR ) were diluted 1:20 , 000 . For cell cycle analyses , TUBD1−/− or TUBE1−/− cells were seeded onto coverslips , then synchronized in G0/G1 by serum withdrawal for 24 hr , or in G2 with 10 µM RO-3306 ( Adipogen ) for 24 hr . Cells were fixed for immunofluorescence and analyzed for centrin/CP110 presence . Mitotic shakeoff was performed on asynchronously growing cells . One pre-shake was performed to improve synchronization . Cells were fixed at indicated times and analyzed for centrin/CP110 presence . Centrinone ( Wong et al . , 2015 ) was a gift from Andrew Shiau and Karen Oegema ( Ludwig Institute for Cancer Research and UC San Diego ) . hTERT RPE-1 TP53−/− cells were treated with 125 nM centrinone for ≥2 weeks , and centrinone-containing medium was replaced on top of cells daily . For centrinone washout , cells were washed twice with PBS , then mitotic shakeoff was performed with centrinone-free medium . A subset of cells were fixed for immunofluorescence 12 hr after shakeoff , when cells had entered S-phase . 19 hr after shakeoff , a second shakeoff was performed to harvest cells that entered mitosis . Cells were fixed 3 hr post-second shakeoff for immunofluorescence , and analyzed for centrin/CP110 presence . For paclitaxel experiments , mitotic cells were removed by shakeoff from an asynchronous population , then 15 μM paclitaxel ( Tocris ) or DMSO was added to the cells remaining on the dish . For both populations , G2-phase cells were allowed to enter mitosis , and then harvested in mitosis by shakeoff 3 hr later . Cells were plated on coverslips and forced to exit mitosis by treatment with 10 µM RO-3306 , then fixed for immunofluorescence 3 hr later . Cells with micronuclei were analyzed for centrin/CP110 presence in both conditions . Cells were seeded onto glass-bottom dishes ( World Precision Instruments ) 1 day prior to imaging . 30 min prior to imaging , the medium was changed to phenol-free DMEM-F12 ( Life Technologies ) supplemented with 10% CCS . Images were acquired as 0 . 5 μm Z-stacks collected every 10 min using a Zeiss Axio Observer microscope with a confocal spinning-disk head ( Yokogawa ) , PlanApoChromat 63x/1 . 4 NA objective , and a Cascade II:512 EM-CCD camera ( Photometrics ) , run with MicroManager software ( Edelstein et al . , 2014 ) . During image acquisition , cells were incubated at 37°C under 5% CO2 . HEK293T cells were co-transfected with GFP-delta-tubulin ( pTS3753 ) and myc-epsilon-tubulin ( pTS4111 ) , or GFP ( pTS3517 ) and myc-epsilon-tubulin ( pTS4111 ) using PEI . 48 hr after transfection , cells were harvested and lysed in lysis buffer ( 50 mM Hepes pH7 . 4 , 150 mM NaCl , 1 mM DTT , 1 mM EGTA , 1 mM MgCl2 , 0 . 25 mM GTP , 0 . 5% Triton X-100 , 1 μg/ml each leupeptin , pepstatin , and chymostatin , and 1 mM phenylmethylsulfonyl fluoride ) . Insoluble material was pelleted , and soluble material was incubated at 4°C with GFP-binding protein ( Rothbauer et al . , 2008 ) coupled to NHS-activated Sepharose 4 Fast Flow resin ( GE Healthcare ) for 2 hr . Beads were pelleted at 500 g for 1 min , washed three times with lysis buffer , then eluted in sample buffer and the eluate was run on SDS-PAGE gels . Western blots were scanned on a LiCOR imager and analyzed using ImageJ . | Most structures inside a cell have a short lifespan and are continually replaced . Centrioles – specialized structures that help cells divide , and send and receive signals – are among the few exceptions and can persist through many cell generations . Centrioles are cylindrical structures that are made up of protein tubes called microtubules . Specifically , nine groups of three microtubules , known as triplet microtubules , are linked together to make the walls of the cylinder . The triplets of microtubules are only found in centrioles , and until now it was not known what role this specific formation plays . Now , Wang et al . studied two lesser known members of the protein family that build the microtubules , called delta-tubulin and epsilon-tubulin . When either of these proteins was removed from human cells grown in the laboratory , the centrioles only had single microtubules rather than the usual triplets . The centrioles still formed at the correct time , but disappeared soon after the cell had divided . When the cells were then treated with a drug that stabilizes the microtubules , the centrioles no longer disappeared once the cell had divided . This suggests that the triplet microtubule formation is needed to stabilize and maintain the centrioles through the cell divisions . Moreover , the results were similar for delta- and epsilon-tubulin , and it appears that the proteins work together to help stabilize the triplet microtubules . Defects in centrioles are associated with many diseases , including some types of cancer and many genetic conditions that can lead to heart or kidney disease , obesity , diabetes and many others . Deeper knowledge of centriole structure and its role may help us to better understand these diseases . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2017 | Centriole triplet microtubules are required for stable centriole formation and inheritance in human cells |
Stress is a potent modulator of the mammalian brain . The highly conserved stress hormone response influences many brain regions , particularly the hippocampus , a region important for memory function . The effect of acute stress on the unique population of adult neural stem/progenitor cells ( NPCs ) that resides in the adult hippocampus is unclear . We found that acute stress increased hippocampal cell proliferation and astrocytic fibroblast growth factor 2 ( FGF2 ) expression . The effect of acute stress occurred independent of basolateral amygdala neural input and was mimicked by treating isolated NPCs with conditioned media from corticosterone-treated primary astrocytes . Neutralization of FGF2 revealed that astrocyte-secreted FGF2 mediated stress-hormone-induced NPC proliferation . 2 weeks , but not 2 days , after acute stress , rats also showed enhanced fear extinction memory coincident with enhanced activation of newborn neurons . Our findings suggest a beneficial role for brief stress on the hippocampus and improve understanding of the adaptive capacity of the brain .
Stress is a powerful and essential mediator of mammalian behavior . Proper response to a perceived stressor facilitates survival at the individual level and species propagation at the population level . Despite this necessity , stress responses can become maladaptive . Chronic stress , for example , leads to a host of adverse health consequences , including cardiovascular disease , obesity , depression , and exacerbation of neurodegeneration ( McEwen , 2004 ) . Acute stress , defined as a single exposure on the scale of minutes to hours without cycles of recovery and re-exposure , has proven more enigmatic . One model of stress effects on the brain , an inverted U function , explains the variable consequences of acute stress for brain health ( Lupien and McEwen , 1997 ) . In this model , while severe or prolonged stressors are detrimental , brief or moderate stressors actually enhance neural function . Behavioral studies focusing on the memory functions of the hippocampus have demonstrated such a relationship in rodents , where moderate stress enhances memory performance yet more severe stress impairs it ( Conrad et al . , 1999 ) . The hippocampus is exquisitely sensitive to stress and the primary stress hormone class , glucocorticoids ( GCs ) . Within the dentate gyrus ( DG ) sub-region , in particular , there exists a high density of GC receptors that respond to elevated circulating GCs ( De Kloet et al . , 1998 ) . In addition , the DG is strongly connected via the entorhinal cortex and medial septum to the basolateral amygdala ( BLA ) , a brain region involved in emotional processing and an important mediator of many stress effects on the hippocampus ( McGaugh , 2004 ) . Both of these mediators of stress ( GCs and BLA input ) have been shown to regulate the unique population of neural progenitor cells ( NPCs ) that reside in the adult DG ( Kirby and Kaufer , 2009; Kirby et al . , 2012b ) . Dentate NPCs proliferate and give rise to new neurons throughout the lifespan in several mammalian species , including rats , mice and primates ( Abrous et al . , 2005; Kirby and Kaufer , 2009 ) . They become electrophysiologically active , integrate into local circuitry , play important modulatory roles in hippocampal memory function ( Abrous et al . , 2005 ) and can respond to stress and stress hormones at multiple phases of development ( Kirby and Kaufer , 2009 ) . Moreover , recent work indicates that newborn , immature neurons integrate multiple signals more readily than mature neurons and that they have enhanced excitability , possibly contributing to a disproportionately large role in new memory formation ( Marín-Burgin et al . , 2012 ) . Numerous studies show that chronic stress , in addition to impairing memory function , suppresses proliferation , survival and differentiation of new neurons in the adult DG ( Wong and Herbert , 2004; Mirescu and Gould , 2006; Kirby and Kaufer , 2009 ) . The effect of acute stress on neurogenesis , however , is unclear . While early work indicated suppression of proliferation following acute stress or GC injection ( Cameron and Gould , 1994; Gould et al . , 1997 ) , subsequent studies have yielded mixed results , often reporting no change in proliferation following a variety of acute stressors ( Thomas et al . , 2006; Thomas et al . , 2007; Dagyte et al . , 2009; Hanson et al . , 2011 ) . In contrast , investigations of hippocampal growth factor secretion have shown that acute stress enhances expression of mitogenic growth factors such as basic fibroblast growth factor ( FGF2 ) and nerve growth factor ( NGF ) ( Mocchetti et al . , 1996; Molteni et al . , 2001 ) , implying a potential for increased neurogenesis following acute stress . Indeed , a number of interventions that stimulate GC release such as acute exercise ( Kronenberg et al . , 2006 ) and acute sexual experience ( Leuner et al . , 2010 ) actually increase neurogenesis in the adult hippocampus . Combined , these studies suggest that adult hippocampal neurogenesis may follow an inverted U function similar to hippocampal memory—decreasing following chronic stress yet increasing in response to acute stressors . We examined the effect of several forms of acute stress on adult hippocampal neurogenesis , seeking to resolve the contradictory evidence for enhanced vs impaired hippocampal plasticity . We found that acute stress or corticosterone ( CORT , the primary rat GC ) administration increased dorsal but not ventral hippocampus cell proliferation in adult rats . This increase was not dependent on input from the BLA and was accompanied by an increase in FGF2 expression in dorsal hippocampal astrocytes . Furthermore , we show that astrocyte-secreted FGF2 is necessary for CORT-induced enhancement in NPC proliferation in vitro . 2 weeks after acute stress , when newborn neurons are first becoming functional , we also find enhancement in hippocampus-dependent memory accompanied by enhanced activation of newborn neurons . These findings have important implications for understanding regulation of hippocampal plasticity in the face of environmental challenge and in distinguishing adaptive vs pathological stress responses . Moreover , they suggest that stress effects on adult neurogenesis may follow an inverted U function similar to that already demonstrated for hippocampal memory function ( Lupien and McEwen , 1997 ) .
To investigate the effect of acute stress on adult neurogenesis in the DG ( Figure 1 ) , we chose three common models of acute stress in rodents: 30 min novel environment , 30 min footshock or 3 hr immobilization . Rats were handled for 5 days prior to stress exposure then perfused 3 hr after the beginning of the stressor ( Figure 2A ) . Immobilization stress significantly increased the number of cells immunopositive for the proliferation marker Ki67 in the dorsal DG 3 . 23-fold above control ( 27 . 51 ± 2 . 74 control vs 88 . 99 ± 12 . 49 immob ) while novel environment or footshock did not significantly alter Ki67+ count compared to control ( Figure 2B , C ) . Plasma CORT was significantly elevated above control levels both 30 min and 3 hr after stressor initiation in immobilized rats but not in novel environment- or footshock-exposed rats ( Figure 2D ) . This finding suggested that an increase in CORT might underlie the immobilization-induced increase in DG proliferation . To test that hypothesis , we next habituated handled rats to daily oil injections for 3 days , injected them with exogenous CORT ( 0 , 5 or 40 mg/kg body weight ) and assessed cell proliferation 3 hr later ( Figure 2A ) . 40 mg/kg CORT significantly increased the number of Ki67+ proliferative cells in the dorsal DG 1 . 92-fold above oil-injected controls ( 26 . 19 ± 2 . 83 0 mg/kg CORT vs 50 . 29 ± 7 . 74 40 mg/kg CORT; Figure 2C , E ) . Plasma CORT levels were also consistently elevated 30 min and 3 hr after injection ( Figure 2F ) , similar to the levels seen in immobilized rats . Notably , injection of 5 mg/kg CORT yielded similar plasma CORT levels to footshock ( approximately 119 and 121 ng/ml , respectively ) and also did not produce an increase in the numbers of Ki67+ cells . If rats were not habituated to injection , no difference in cell proliferation was found ( Figure 2J , K ) . We next assessed short-term survival of newly-born cells following acute stress . 3 hr after the start of immobilization or CORT injection , rats were injected with the proliferative marker 5-bromodeoxyuridine ( BrdU ) then perfused 24 hr later . Both immobilization ( Figure 2G , I ) and 40 mg/kg CORT injection ( Figure 2H , I ) significantly increased the number of cells immunopositive for BrdU surviving 24 hr after termination of the stressor in the dorsal DG by 1 . 92 and 1 . 48-fold , respectively ( 34 . 88 ± 8 . 70 control vs 68 . 96 ± 10 . 23 immob; 39 . 80 ± 6 . 89 0 mg/kg CORT vs 58 . 95 mg/kg CORT ) . 10 . 7554/eLife . 00362 . 003Figure 1 . Dorsal versus ventral DG . Areas of dorsal ( A ) and ventral ( B ) dentate gyrus used for cell proliferation quantification are highlighted in red . Images are adapted from brainatlas . org . DOI: http://dx . doi . org/10 . 7554/eLife . 00362 . 00310 . 7554/eLife . 00362 . 004Figure 2 . Acute stress increases adult cell proliferation in dorsal hippocampus . ( A ) Experimental timeline . ( B ) Acute immobilization increased Ki67+ cell count in the adult dorsal DG while exposure to novel environment or footshock did not significantly change Ki67+ cell count . One-way ANOVA , p<0 . 0001; ***q = 5 . 975 , p<0 . 0001 . ( C ) Representative images of Ki67+ cells ( black arrows ) in the dorsal DG ( dashed outline ) of control , immobilized , 0 mg/kg and 40 mg/kg CORT-injected rats . ( D ) Acute immobilization increased plasma CORT levels 30 min and 3 hr after the stressor began . CORT elevations caused by novel environment and footshock were not significant . One-way ANOVA , p<0 . 0001; ***q = 5 . 56 , p<0 . 0001; **q = 4 . 02 , p<0 . 001 . ( E ) Acute injection of 40 mg/kg CORT increased Ki67+ cell count in the adult dorsal DG compared to 0 mg/kg oil control while 5 mg/kg CORT did not significantly alter Ki67+ cell count . One-way ANOVA , p=0 . 007 . *q = 3 . 15 , p<0 . 05 . ( F ) 40 mg/kg CORT injection led to a sustained increase in plasma CORT 30 min and 3 hr after injection . The change in plasma CORT following 5 mg/kg CORT injection was not significantly different from oil injection . Two-way ANOVA , effect of CORT dose p<0 . 0001 . **q = 3 . 62 and 3 . 61 , p<0 . 001 , 40 mg/kg 30 min and 3 hr , respectively . ( G ) The number of BrdU-labeled newborn cells surviving 24 hr after the end of immobilization was greater in immobilized rats than controls . *p=0 . 03 ( H ) the number of BrdU-labeled newborn cells surviving 24 hr after CORT/oil injection was greater in rats given 40 mg/kg CORT compared to 0 mg/kg CORT . *p=0 . 04 . ( I ) Representative images of BrdU+ cells ( black arrows ) in the dorsal DG ( dashed outline ) of control , immobilized , 0 mg/kg and 40 mg/kg CORT-injected rats . ( J ) Experimental timeline . Rats were handled for 3 days , injected with CORT or oil vehicle then perfused 3 hr later . ( K ) No difference in Ki67+ cell number was found in the adult dorsal DG with increasing CORT dose . All values are average ± SEM . Scale bar is 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00362 . 004 Recent work suggests different functional roles for the dorsal and ventral hippocampus in mediating spatial memory vs emotion regulation , respectively ( Fanselow and Dong , 2010 ) . Adult neurogenesis may also be differentially regulated in dorsal vs ventral hippocampus; while environmental enrichment enhances neurogenesis in both dorsal and ventral DG , chronic mild stress preferentially suppresses proliferation in the ventral subregions ( Tanti et al . , 2012 ) . In our acute stress models , we found no effect of novel environment , footshock , immobilization or CORT injection ( Figure 3A–C ) on proliferative Ki67+ cell number in the ventral hippocampus ( Figure 1B ) . 10 . 7554/eLife . 00362 . 005Figure 3 . Acute stress does not increase adult cell proliferation in ventral hippocampus . ( A ) None of the stressors affected Ki67+ cell count in the ventral DG . ( B ) CORT did not affect Ki67+ cell count in the ventral DG . ( C ) Representative images of Ki67+ cells ( black arrows ) in the ventral DG ( dashed outline ) of control , immobilized , 0 mg/kg and 40 mg/KG CORT-injected rats . All values are average ± SEM . Scale bar is 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00362 . 005 Work by McGaugh and colleagues shows that acute stress-induced alterations in hippocampal memory and long-term potentiation depend on input from the BLA ( Quirarte et al . , 1997; Roozendaal et al . , 1999; Roozendaal et al . , 2009b ) , a fear and emotion processing center . In addition , we have previously shown that BLA input regulates adult DG neurogenesis under basal conditions and is required for the activation of newborn neurons in a fear conditioning paradigm ( Kirby et al . , 2012b ) . To test whether BLA input is necessary for the acute stress-induced increase in proliferation , we performed unilateral excitotoxic lesions of BLA in adult rats ( as described in [Kirby et al . , 2012a , 2012b] ) . Animals were then exposed to immobilization stress or no stress control ( Figure 4A ) . At the end of the stressor , each rat received a BrdU injection and was perfused 2 hr later . Plasma CORT response to immobilization was similar in lesioned and sham-operated rats ( Figure 4B ) , suggesting intact hormonal stress response in BLA-lesioned , immobilized rats . Consistent with our findings in intact rats , sham-operated immobilized rats had a significant 2 . 6-fold increase in BrdU+ proliferative cells over no stress controls ( 6 . 37 ± 1 . 14 control vs 16 . 69 ± 2 . 11 immob; Figure 4C ) . BLA lesion suppressed proliferation ipsilateral to the lesion by approximately 1 . 5-fold ( 1 . 66-fold control , 1 . 44-fold immob ) , as we have previously reported , but did not block the stress-induced increase in proliferation ( Figure 4D ) . These findings suggest that input from the BLA is not necessary for acute stress effects on adult hippocampal cell proliferation . 10 . 7554/eLife . 00362 . 006Figure 4 . Acute stress increases cell proliferation independent of BLA input . ( A ) Experimental timeline . ( B ) Plasma CORT elevation after immobilization was similar between sham-operated and unilaterally BLA-lesioned rats . Two-way ANOVA effect of time , *p=0 . 04 . ( C ) In sham-operated rats , acute immobilization increased the number of BrdU+ cells in the adult DG . **p=0 . 001 . ( D ) Unilateral excitotoxic lesion of the BLA decreased the number of BrdU+ cells in the ipsilateral DG , but did not interact with stress . Two-way ANOVA , effect of lesion , **p=0 . 002 . All values are average ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 00362 . 006 Given that the stress-induced increase in proliferation was not dependent on BLA input , we next tested whether stress-induced proliferation is a cell autonomous phenomenon . This was accomplished by quantifying the response of isolated adult rat hippocampal NPCs grown in vitro ( as described in Gage , 2000 ) to acute CORT exposure . NPCs were FGF2 deprived for 24 hr then treated with 1 μM CORT ( equivalent to approximately 350 ng/ml , the level measured in plasma of immobilized rats ) or EtOH vehicle in either high ( 20 ng/ml ) or low ( 0 ng/ml ) FGF2 . After 3 hr , BrdU was added to the cells and they were fixed 2 hr later ( Figure 5A ) . Cultured rat hippocampal NPCs depend on FGF2 signaling for proliferation ( Chipperfield et al . , 2002 ) . We found that while high FGF2 increased the percentage of proliferating BrdU+ NPCs over low FGF2 ( 1 . 21-fold in EtOH , 1 . 48-fold in 1 μM CORT ) , exposure to CORT did not affect the percent of proliferative cells in either high or low FGF2 media as compared to vehicle ( Figure 5B ) . These data indicate that NPCs do not respond to acute CORT in isolation , but likely rely on input from another cell type in the neurogenic niche . A growing body of work indicates that astrocytes can strongly regulate NPC dynamics through secreted factors ( Song et al . , 2002 ) , so we next tested whether astrocytes might participate in the regulation of NPCs in response to CORT . We treated primary hippocampal astrocytes cultured as described in ( McCarthy and de Vellis , 1980 ) with 1 μM CORT or EtOH vehicle for 3 hr then extracted astrocyte conditioned media ( ACM ) . When administered to isolated NPCs , CORT-treated ACM increased the percent of BrdU+ NPCs significantly over EtOH-treated coculture media ( CoC ) control by 1 . 52-fold ( 29 . 56 ± 2 . 81 CoC-EtOH vs 44 . 93 ± 2 . 82 ACM-1 μM CORT; Figure 5C , D ) . These findings suggest that astrocytes could mediate the in vivo increase in NPC proliferation following acute stress through secreted factors . 10 . 7554/eLife . 00362 . 007Figure 5 . ACM from CORT-treated astrocytes increases NPC proliferation . ( A ) Experimental timeline . ( B ) Treatment of isolated hippocampal NPCs with 1 µM CORT for 3 hr did not alter the percent of proliferating BrdU+ cells compared to EtOH vehicle . 3 hr of treatment with 20 ng/ml human recombinant FGF2 increased the percent of proliferative BrdU+ cells . Two-way ANOVA , effect of FGF2 , **p=0 . 005 . ( C ) ACM was extracted from astrocytes treated with 1 µM CORT or EtOH vehicle . Treatment of NPCs with ACM from CORT-treated astrocytes increased the percent of proliferative BrdU+ cells compared to EtOH , CoC-treated control NPCs . Two-way ANOVA , effect of CORT , p=0 . 02; effect of media , p=0 . 0024 . **q = 4 . 23 , p<0 . 001 vs EtOH-CoC . ( D ) Representative images of NPCs treated with CoC or ACM , EtOH or CORT . BrdU+ cells are orange and DAPI is blue . All values are average ± SEM . Scale bar is 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00362 . 007 Given that astrocytes secrete a variety of growth factors that support cell proliferation , we next investigated hippocampal growth factor expression in stressed rats . We quantified levels of the following growth factors previously reported to be mitogenic and/or elevated by acute stress ( Mocchetti et al . , 1996; Molteni et al . , 2001; Ma et al . , 2009 ) in the dorsal hippocampus of stressed and control rats: FGF2 , brain derived neurotrophic factor ( BDNF ) , nerve growth factor ( NGF ) , FGF receptor 1 ( FGFR1 ) , FGFR2 , FGFR3 , FGFR4 , vascular endothelial growth factor ( VEGF ) , growth arrest and DNA damage–inducible 45β ( GADD45β ) as well as the early immediate gene , CFOS . Acute stress caused a significant increase in FGF2 mRNA and protein levels in the dorsal hippocampus ( Figure 6A–D , I ) . Immobilization increased fgf2 mRNA 1 . 77 ( ±0 . 20 ) fold over control while 40 mg/kg CORT increased fgf2 mRNA by 1 . 70 ( ±0 . 23 ) fold over 0 mg/kg CORT ( Figure 6A , B ) . Immobilization and 40 mg/kg CORT also significantly increased FGF2 protein levels in the dorsal hippocampus by 1 . 59 ( ±0 . 24 ) fold and 2 . 54 ( ±0 . 34 ) fold , respectively ( Figure 6C , D , I ) . In contrast , bdnf exon IV mRNA levels were significantly decreased by immobilization and CORT ( Figure 6E , F ) . BDNF protein level , however , was unchanged by either manipulation ( Figure 6G–I ) . Acute stress did not consistently alter mRNA levels of ngf , fgfr1 , fgfr2 , fgfr3 , fgfr4 , vegf , bdnf exon IX , or gadd45β ( Figure 7A–P ) . mRNA of the immediate early gene cfos was generally increased at 30 min after any manipulation , including oil injection ( Figure 7Q , R ) . 10 . 7554/eLife . 00362 . 008Figure 6 . Acute stress increases FGF2 expression in dorsal hippocampus . ( A ) 3 hr of immobilization increased fgf2 mRNA expression over control in dorsal hippocampus . Other groups did not significantly differ from control . One-way ANOVA , p=0 . 05 . **q = 3 . 54 , p<0 . 01 . ( B ) 40 mg/kg CORT increased fgf2 mRNA expression in the dorsal hippocampus 3 hr after CORT injection compared to 30 min after 0 mg/kg CORT injection . Other groups did not significantly differ from oil-injected controls . Two-way ANOVA , effect of CORT , p=0 . 03; effect of time , p<0 . 0001; interaction , p=0 . 0021 . ***q = 4 . 34 , p<0 . 001 . ( C ) FGF2 protein levels in dorsal hippocampus increased with 3 hr of immobilization over control . Other groups did not significantly differ from control . One-way ANOVA , p>0 . 05 . *q = 2 . 79 , p<0 . 05 . ( D ) FGF2 protein levels in dorsal hippocampus increased 3 hr after 40 mg/kg CORT injection compared to 30 min after 0 mg/kg vehicle injection . Two-way ANOVA , effect of CORT , p=0 . 01; effect of time , p=0 . 03 . *q = 3 . 18 , p<0 . 05 . ( E ) 3 hr of immobilization decreased bdnf exon IV expression over control in dorsal hippocampus . Other groups did not significantly differ from controls . One-way ANOVA , p=0 . 0007 . *q = 3 . 05 , p<0 . 05 . ( F ) There was an overall significant decrease in bdnf exon IV mRNA expression with increasing CORT dose in dorsal hippocampus . Two-way ANOVA , effect of CORT , *p=0 . 02 . ( G ) BDNF protein levels in dorsal hippocampus did not change with immobilization , novel environment or shock compared to control . ( H ) BDNF protein levels did not change compared to 0 mg/kg vehicle with increasing CORT dose . ( I ) Representative western bands of FGF2 , BDNF and ACTIN from the 3 hr time point in dorsal hippocampus . ( J ) 3 hr of immobilization increased fgf2 mRNA expression over control in ventral hippocampus . Other groups did not significantly differ from controls . One-way ANOVA , p=0 . 03 . *q = 2 . 87 , p<0 . 05 . ( K ) 40 mg/kg CORT increased fgf2 mRNA expression in the ventral hippocampus 3 hr after CORT injection compared to 30 min after 0 mg/kg CORT injection . Other groups did not significantly differ from oil-injected controls . Two-way ANOVA , effect of time , p=0 . 002 . **q = 3 . 34 , p<0 . 01 . ( L ) Immobilization , novel environment and shock did not alter FGF2 protein levels in ventral hippocampus . ( M ) CORT did not alter FGF2 protein levels in ventral hippocampus 3 hr after injection . ( N ) Representative western bands of FGF2 and ACTIN from the 3 hr time point in ventral hippocampus . All values are average ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 00362 . 00810 . 7554/eLife . 00362 . 009Figure 7 . mRNA expression levels in dorsal hippocampus following acute stressors . ( A ) There was no change in ngf mRNA with novel environment , shock or immobilization . ( B ) 5 mg/kg CORT significantly increased ngf mRNA at 3 hr over 0 mg/kg CORT , 30 min *q = 2 . 79 , p<0 . 05 . There was no change in fgfr1 ( C , D ) , fgfr2 ( E , F ) , fgfr3 ( G , H ) , fgfr4 ( I , J ) , vegf ( K , L ) , bdnf exon IX ( M , N ) Or gadd45β ( O , P ) mRNA expression in dorsal hippocampus . ( Q ) Exposure to novel environment or footshock increased cfos expression in the dorsal hippocampus . One-way ANOVA , p<0 . 0001 . ***q = 5 . 92 and q = 8 . 54 , novel and shock , respectively , p<0 . 001 . All values are average ± SEM . ( R ) All injection conditions showed a decrease in cfos mRNA over time . Two-way ANOVA , effect of time , p<0 . 0001 . ***q = 4 . 42 , 4 . 68 , 4 . 91 , 0 mg/kg 3 hr , 5 mg/kg 3 hr , 40 mg/kg 3 hr , respectively , p<0 . 001 . All values are average ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 00362 . 009 We next quantified FGF2 expression in ventral DG where we previously saw no effect of stress on cell proliferation . Immobilization and 40 mg/kg CORT injections significantly increased fgf2 mRNA levels in the ventral hippocampus , similar to the dorsal hippocampus ( Figure 6J , K ) . However , protein levels of FGF2 were not increased by any of the acute stressors in the ventral portion of the hippocampus ( Figure 6L–N ) . These results provide intriguing correlative evidence that increased FGF2 protein levels following acute immobilization or CORT exposure could underlie the observed increase in dorsal hippocampal proliferation . To determine whether astrocytes were the source of increased dorsal hippocampal FGF2 in vivo , we used confocal microscopy to quantify double immunohistochemical labeling for FGF2 and the astrocyte marker , glial fibrillary acidic protein ( GFAP ) in adult male rats treated as in Figure 2A . Consistent with previous reports ( Bhatnagar et al . , 1997 ) , FGF2-immunoreactive cells were found throughout the DG and the hilus , with staining primarily in the cell body and nucleus . The DG is primarily composed of granule neurons and while almost every cell expressed FGF2 , we found very few GFAP+ cells within the DG as expected . Mean optical density ( with background correction ) throughout the Z-stack of FGF2 expression in the DG revealed no effect of stress or CORT injection on DG FGF2 expression ( Figure 8A , D ) . In the hilus , a mixture of GFAP+ and GFAP- cells was found , with almost all GFAP+ cells being FGF2+ . By quantifying integrated optical density of individual FGF2+ cells that were either GFAP+ or GFAP− , we found that both immobilization and 40 mg/kg CORT injection increased FGF2 signal in GFAP+ cells significantly over their respective control groups ( Figure 8B , D ) . No effect of immobilization or CORT was observed in GFAP− cells in the hilus ( Figure 8C , D ) . These findings suggest that the stress-induced increase in FGF2 levels in the DG most likely comes from neighboring astrocytes in the hilus . 10 . 7554/eLife . 00362 . 010Figure 8 . Acute stress increases FGF2 expression in GFAP+ astrocytes in the dorsal hilus . ( A ) There was no change in mean optical density of FGF2-ir in the DG following 3 hr immobilization or CORT injection . ( B ) Both immobilization and 40 mg/kg CORT injection significantly increased integrated optical density of FGF2-ir in GFAP+ cells in the hilus . One-way ANOVA , p=0 . 0024 . *p=0 . 04 and 0 . 05 , con v immob and 0 V 40 mg/kg CORT , respectively . ( C ) FGF2-ir integrated optical density in GFAP- cells of the hilus did not change . ( D ) Representative images of FGF2+ cells ( green ) that are GFAP+ ( red; white arrows ) or GFAP- in DG and hilus of a control and an immobilized rat . DAPI is blue . Scale bar = 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00362 . 010 To determine whether FGF2 was the astrocyte-derived factor driving NPC proliferation , we next examined the role of FGF2 in the effect of CORT-treated ACM on isolated NPCs . Consistent with previous work ( Forget et al . , 2006 ) , media from untreated astrocytes showed no detectable FGF2 ( 0 . 12 ± 0 . 46 pg/ml ) . In contrast , astrocytes treated with 1 μM CORT for 3 hr secreted 3 . 5 ± 0 . 68 pg/ml FGF2 protein ( Figure 9A ) . We next tested whether levels of FGF2 as low as 4 pg/ml were sufficient to stimulate NPC proliferation . Treating with 4 pg/ml rat recombinant FGF2 caused a significant 1 . 39-fold increase in BrdU labeling compared to vehicle ( 0 pg/ml FGF2 ) , suggesting that the levels of FGF2 present in CORT-ACM were sufficient to stimulate NPC proliferation ( 11 . 95 ± 1 . 22% BrdU+ , 0 pg/ml; 16 . 65 ± 1 . 72% BrdU+ , 4 pg/ml , p=0 . 04 ) . We next investigated whether blocking FGF2 function could prevent acute stress-induced enhancement of NPC proliferation using a neutralizing antibody against FGF2 ( nAb ) . An ELISA for rat FGF2 revealed that the nAb decreased available FGF2 over a wide range of concentrations ( Figure 9B ) . The nAb did not , however , affect the availability of FGF1 ( Figure 9C ) , a closely related member of the FGF family . When tested in vitro , we found that FGF2 neutralization blocked FGF2-induced stimulation of NPC proliferation without affecting proliferation in FGF2-free conditions ( Figure 9D ) . These findings suggest that the nAb blocks FGF2 signaling specifically and does not have nonspecific toxic effects on NPC growth . We then pretreated ACM from EtOH- or CORT-treated astrocytes with FGF2 nAb before exposing NPCs to the treated media . Pretreatment with the FGF2 nAb blocked the CORT-ACM induced increase in NPC proliferation ( 30 . 67 ± 3 . 44% BrdU positive in EtOH , no nAb vs 30 . 33 ± 2 . 59% BrdU positive in 1 μM CORT with nAb; Figure 9E ) , suggesting that increased astrocytic FGF2 is the driving signal for CORT-ACM-induced proliferation of adult hippocampal NPCs . 10 . 7554/eLife . 00362 . 011Figure 9 . Blocking FGF2 prevents CORT-ACM induced increase in NPC proliferation . ( A ) ACM from EtOH-treated primary astrocytes had no FGF2 protein ( relative to blank ) while CORT-treated ACM contained 3 . 5 pg/ml FGF2 . *p=0 . 01 . ( B ) Availability of rat FGF2 was dramatically reduced by pretreating FGF2 protein with an FGF2 neutralizing antibody . Two-way ANOVA , p<0 . 0001 main effects of nAb and FGF2 concentration . Post-hoc Dunnett's multiple comparison tests with 0 μg/ml as control shown for 10 μg/ml ( upper row *s ) and 20 μg/ml ( lower row *s ) : *p<0 . 05 , ** p<0 . 01 , ***p<0 . 001 , ****p<0 . 0001 . ( C ) Availability of FGF1 was not affected by pretreating FGF1 protein with the same FGF2 neutralizing antibody . ( D ) Isolated NPCs were treated with the FGF2 nAb in either high ( 20 ng/ml ) or low ( 0 ng/ml ) FGF2 conditions . 20 μg/ml nAb effectively blocked the FGF2-induced increased in percent BrdU+ proliferating cells . Two-way ANOVA , effect of FGF2 , p=0 . 04 . **q = 3 . 37 , p<0 . 01 . The nAb did not affect proliferation in low FGF2 conditions . ( E ) Pretreatment with FGF2 nAb prevented the increase in percent BrdU+ proliferating NPCs caused by ACM from CORT-treated astrocytes . Two-way ANOVA , interaction , p=0 . 02 . *q = 3 . 06 , p<0 . 05 . All values are average ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 00362 . 011 Several studies suggest that newly-born neurons play an important functional role in hippocampal memory during an immature , highly plastic phase of their development ( Kee et al . , 2007; Deng et al . , 2009; Kitamura et al . , 2009; Stone et al . , 2010 ) . However , newborn neurons in the adult rat require two or more weeks to mature and become physiologically active ( Snyder et al . , 2009 ) , implying that the enhancement in neurogenesis we observed following acute stress should require several weeks to influence behavior . We therefore assessed hippocampal memory using a fear conditioning task either 2 days or 2 weeks after acute stress . Immobilized and control rats were given 10 unsignaled , 1 s , 1 mA shocks in a fear conditioning chamber . The next day , they received five 10 min extinction trials where they were exposed to the fear conditioning chamber without shock . On the third day of testing , they received a single 10 min extinction probe trial without shock . When rats were tested 2 days after immobilization , control and immobilized rats showed similar freezing behavior during training , extinction and the 24 hr extinction probe ( Figure 10A–C ) . When rats were tested 2 weeks after immobilization , immobilized rats did not differ from controls in percent time freezing during training or during the five extinction trials ( Figure 10D , E ) . During the extinction probe , however , immobilized rats froze significantly less than control rats , indicating better retention of the fear extinction ( Figure 10F ) . These results suggest that an incubation time is necessary for acute immobilization to improve fear memory and demonstrate enhanced hippocampal memory at a time when newborn neurons are highly plastic and sensitive to environmental input . 10 . 7554/eLife . 00362 . 012Figure 10 . Acute stress causes delayed enhancement of contextual fear extinction retention . Acute immobilization 2 days prior to contextual fear conditioning did not change percent time freezing during training ( A ) , extinction ( B ) or 24-hr extinction probe ( C ) compared to control . Two-way ANOVA for extinction , effect of trial , p<0 . 0001 . Acute immobilization 2 weeks prior to contextual fear conditioning did not change percent time freezing during training ( D ) or extinction ( E ) compared to control . Two-way ANOVA for extinction , effect of trial , p<0 . 0001 . ( F ) Immobilized rats froze significantly less than controls in the 24-hr extinction probe . *p=0 . 04 . ( G ) The number of surviving BrdU+ cells in the dorsal DG of immobilized rats did not differ from controls . ( H ) Immobilization does not alter the percent of BrdU+ cells co-expressing doublecortin ( DCX ) or glial fibrillary acidic protein ( GFAP ) . ( I ) The percent of DCX+ immature neurons expressing cfos was greater in immobilized rats compared to controls . *p=0 . 02 . All values are average ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 00362 . 012 To determine if the neurons born 2 weeks before fear conditioning might play a role in the enhancement of fear memory retention , we quantified cell fate and activation in a subset of rats perfused 1 hr after the fear extinction probe . The number of surviving BrdU+ cells ( Figure 10G ) and the percent of BrdU+ cells expressing the immature neuronal marker doublecortin ( DCX+ ) or the astrocytic marker GFAP+ ( Figure 10H ) were similar in immobilized and control rats . However , immobilized rats had significantly more immature ( DCX+ ) cells expressing the immediate early gene cfos than control rats ( 1 . 57 ± 0 . 52% cfos+ vs 5 . 39 ± 1 . 40% cfos+; Figure 10I , J ) , suggesting greater activation of immature neurons in immobilized rats . These results indicate enhanced utilization of the highly plastic pool of new neurons born around the time of an acute stressor and suggest that stress-stimulated proliferation may support later memory benefits .
The present study demonstrates that acute stress or exposure to the stress hormone CORT induces an increase in proliferation of hippocampal NPCs via increased secretion of astrocytic FGF2 . This increase in proliferation is correlated with selective activation of the hyper-plastic newborn neurons and enhanced retention of fear extinction 2 weeks after the stressor . Taken together , these findings suggest a beneficial role for acute stress on hippocampal plasticity . Consistent with our findings , previous studies using similar stress paradigms show stress-induced enhancements in memory consolidation and growth factor expression ( Roozendaal et al . , 1999; Molteni et al . , 2001; Roozendaal et al . , 2009b ) . Acute stress may also support hippocampal LTP , another factor implicated in stimulating adult hippocampal neurogenesis ( Korz and Frey , 2005 ) . Notably , other manipulations that increase stress hormone secretion , such as exercise , sexual experience and mild immune challenge , similarly stimulate adult neurogenesis ( Kronenberg et al . , 2006; Wolf et al . , 2009; Leuner et al . , 2010; Buwalda et al . , 2012 ) . Recent work in adult squirrel monkeys also shows that coping with intermittent social stress through multiple pair separations and new pair formations both stimulates adult hippocampal neurogenesis and improves hippocampal memory ( Lyons et al . , 2010 ) . These studies , along with our findings , evoke the interesting possibility that acute stress may be beneficial for brain health in general and hippocampal plasticity in particular . Previous studies of adult NPC response to acute stress have yielded mixed results . While some initial work suggested that acute stress causes a decrease in NPC proliferation ( Cameron and Gould , 1994; Gould et al . , 1997 ) , more recent publications have not replicated these findings ( Thomas et al . , 2006; Thomas et al . , 2007; Dagyte et al . , 2009; Hanson et al . , 2011 ) . Notably , while we found that acute CORT enhanced proliferation in well-handled , habituated rats , we did not find any effect of acute CORT on neurogenesis in rats that were not habituated to injection . Our findings therefore demonstrate that factors present before the stressor begins may alter the effect of acute stress , perhaps by changing the susceptibility of controls or stressed animals to manipulation . Not surprisingly , a rich literature exists regarding the role of handling on altering the rodent stress response ( Korosi and Baram , 2010 ) . Such experimental differences could explain discrepancies in the literature . Previous research concerning the long-term effects of acute stress in rodents focuses primarily on models of traumatic stress and the resultant PTSD-like symptoms . Most prominently , the single prolonged stress model of PTSD uses three acute stressors in series ( restraint , forced swim and ether ) and results in delayed deficits in fear extinction , as well as enhanced anxiety ( Knox et al . , 2012a , 2012b ) . However , if any one of the three components of this stressor protocol are eliminated , deficits are no longer evident ( Knox et al . , 2012a ) , suggesting that stressor effects depend on the severity and length of the stressor , perhaps following the inverted U function previously described ( Lupien and McEwen , 1997 ) . Previous work in primates has provided additional evidence for an inverted U relationship between stress and neural function . In multiple non-human primate species , for example , mild early life stress can lead to resilience against stress later in life , a phenomenon referred to as stress inoculation ( Lyons and Parker , 2007; Katz et al . , 2009; Parker and Maestripiero , 2011 ) . In contrast , more severe early life stress can have the opposite outcome , increasing stress vulnerability later in adulthood . When combined with our current findings , these studies fit well with the proposed inverted U function for stress effects on brain health ( Lupien and McEwen , 1997; Salehi et al . , 2010; Luksys and Sandi , 2011 ) . At high , traumatic levels , acute stress may result in maladaptive pathology ( e . g . , PTSD-like symptoms in the single prolonged stress model ) . At more moderate levels , though , such as immobilization in well-handled rats , acute stress may actually enhance function . Future research will be needed to more precisely define the limits of stimulating vs detrimental acute stress . Many acute-stress-induced changes in hippocampal plasticity rely on functional input from the BLA . For example , while stress enhances hippocampal memory consolidation and LTP , lesion of the BLA blocks these enhancements ( Quirarte et al . , 1997; Roozendaal et al . , 2009a , 2009b ) . We have recently reported that hippocampal cell proliferation and the activation of the newborn neurons in a fear-conditioning paradigm both depend on BLA neural input ( Kirby et al . , 2012b ) . In the current study , we found that while BLA lesion suppressed hippocampal cell proliferation as we have shown before , it did not affect the acute stress-induced increase in neurogenesis . These findings imply that acute stress regulation of neurogenesis may not rely on the same systems-level in vivo circuitry that mediates acute stress regulation of memory consolidation and LTP . In support of this hypothesis , we were able to model CORT-induced NPC proliferation in isolated NPCs . We found that while NPCs did not respond to acute CORT treatment independently in vitro , they did proliferate more in response to conditioned media from CORT-treated primary astrocytes , suggesting a role for secreted factors from local astrocytes in mediating stress effects on NPCs . The dynamic role of astrocytes in facilitating neuronal function through secreted factors has gained much recognition ( Eroglu and Barres , 2010 ) . In addition to aiding in synaptic glutamate recycling , astrocytes secrete several factors such as thrombospondins ( Eroglu and Barres , 2010 ) , Hevin , SPARC ( Kucukdereli et al . , 2011 ) and glypicans ( Allen et al . , 2012 ) that regulate synaptic formation and function in mature neurons ( Eroglu and Barres , 2010 ) . Astrocytes express the glucocorticoid receptor , GR , and when they are exposed to high levels of GCs , GC-bound GR translocates to the nucleus and enhances FGF2 gene transcription ( Molteni et al . , 2001; Gubba et al . , 2004; Unemura et al . , 2012 ) . FGF2 is a potent and necessary proliferative factor in adult NPCs ( Chipperfield et al . , 2002 ) . We found that acute stress stimulated FGF2 expression in the dorsal hippocampus and in primary hippocampal astrocytes . We further showed that in the DG and hilus , the enhancement in FGF2 levels following stress is largely restricted to GFAP+ astrocytes . Notably , changes in FGF2 protein levels following acute stress closely paralleled the neurogenic response; that is , FGF2 levels were increased only in stress conditions that also stimulated neurogenesis . Neutralizing the astrocyte-secreted FGF2 prevented enhanced proliferation in cultured NPCs . These findings suggest a novel role for astrocytes in supporting hippocampal plasticity in response to an environmental stressor through secreted FGF2 . Further research will be required to fully dissect the molecular mechanisms by which stress induces FGF2 secretion from astrocytes . Newly-born neurons are implicated in numerous hippocampal memory functions ( Aimone et al . , 2010 ) , particularly contextual fear conditioning . Contextual fear conditioning activates immature neurons ( Kirby et al . , 2012b ) and when new neurons are selectively knocked down , fear extinction is impaired ( Deng et al . , 2009; Stone et al . , 2010 ) . Sahay et al . have further shown that prolonging survival of newly-born neurons through targeted knockdown of apoptotic pathways enhances discrimination between similar contexts in a fear conditioning task ( Sahay et al . , 2011 ) . We found that 2 weeks , but not 2 days , after acute stress , immobilized rats showed enhanced contextual fear extinction retention compared to controls . We also found that immobilized rats had greater activation of newly born neurons in response to fear extinction recall . This time window coincides with the period in which the newborn cells are hyper-plastic and more likely to be recruited to active circuitry ( Kee et al . , 2007; Deng et al . , 2009; Kirby et al . , 2012b ) . Given that immobilized rats had a similar number of newborn neurons as controls , these data suggest that immobilized rats better utilize the pool of immature neurons , possibly contributing to their enhanced memory . However , this connection does not preclude the potential contribution of other aspects of hippocampal plasticity to the enhanced memory . Future research will be necessary to fully determine the role of immature neurons vs existing circuitry in enhancement of hippocampal memory . A recent hypothesis posits that the dorsal and ventral hippocampus support different behavioral functions , with the dorsal region being important for spatial and declarative memory while the ventral region facilitates affective regulation ( Fanselow and Dong , 2010 ) . This division may also apply to neural stem cell regulation within the DG of each region where chronic stress preferentially effects the ventral over the dorsal hippocampus , perhaps reflecting the negative affective consequences of chronic stress ( Tanti et al . , 2012 ) . In the present study , we found that acute stress or stress hormone exposure increased neurogenesis and FGF2 expression in the dorsal , but not in the ventral , hippocampus . Moreover , the memory benefits observed 2 weeks after stress were in contextual fear extinction , a task that relies on the kind of spatial memory hypothesized to depend on the dorsal DG ( Fanselow and Dong , 2010 ) . This selective involvement of the dorsal hippocampus implies that our acute stressor model functions as a cognitive stimulant for the declarative domains of hippocampal function rather than as a modulator of emotional responsivity . The brain's response to acute stress can define the line between life-saving adaptation and long-term pathology . The current study suggests that moderate , acute stress may stimulate heightened brain plasticity via increased neurogenesis . These findings have important implications for understanding adaptive vs pathological responses to stress .
Adult male Sprague-Dawley rats ( Charles River ) were pair-housed on a 12 hr light dark cycle with lights on at 07:00 hr . Rats were allowed to acclimate to the animal facility for 1 week before handling began . All animal procedures were approved by the UC Berkeley Animal Care and Use Committees . Novel environment , footshock , immobilization and control rats were all handled for 5 days . On the sixth day , they were exposed to a stressor or left undisturbed in the case of controls . The novel environment and footshock exposure both lasted 30 min and occurred in fear conditioning chambers described in the ‘Fear conditioning' . For footshock , rats were exposed to 1 mA , 1 s duration unsignaled shock 30 times . Rats were returned to their home cage after the 30 min exposure until the time of sacrifice . Immobilized rats were confined for 3 hr in decapicone bags ( Braintree Scientific , Braintree , MA ) . For most experiments , rats were handled for 2 days , given a subcutaneous ( SC ) needle stick on day 3 and then injected with sesame oil ( SC ) for 2 days ( days 4 and 5 ) . On the sixth day , rats received 0 , 5 or 40 mg/kg corticosterone ( SC , Sigma , St . Louis , MO ) suspended in sesame oil . CORT was either at 5 mg/ml or 40 mg/ml such that rats receiving 5 or 40 mg/kg corticosterone received equal volumes of oil relative to body weight . In the case of the rats not pre-injected , rats were handled for 3 days and then injected with 0 , 5 or 40 mg/kg corticosterone as above on the fourth day . 5-Bromo-2′-deoxyuridine ( BrdU , Sigma ) was dissolved in physiological saline . Rats were injected with BrdU ( intraperitoneally , 200 mg/kg ) 3 hr after the beginning of the stressor or injection unless otherwise noted . Excitotoxic lesions of the BLA were performed using unilateral stereotaxic infusion of N-methyl-d-aspartate ( Sigma ) as per ( Kirby et al . , 2012a , 2012b ) . Coordinates for BLA infusion were: −2 . 8 mm anterior/posterior ( A/P ) , ±5 . 1 mm medial/lateral ( M/L ) relative to bregma; −6 . 8 mm ( 2 min ) and −6 . 5 mm ( 1 min ) relative to dura . Following 3 weeks of recovery , during which rats were handled regularly , rats were immobilized for 3 hr ( n = 7 each , sham and unilateral lesion ) or left undisturbed in their home cage ( n = 7 each , sham and unilateral lesion , respectively ) . Tail vein blood samples were taken at the beginning and end of immobilization for plasma corticosterone quantification . At the end of immobilization , all rats received one injection of 100 mg/kg BrdU and were perfused 2 hr later . Rats were anesthetized with Euthasol euthanasia solution and transcardially perfused with ice cold 0 . 1 M phosphate buffered saline ( PBS ) followed by 4% paraformaldehyde in 0 . 1 M PBS . Brains were post-fixed for 24 hr at 4°C , equilibrated in 30% sucrose in 0 . 1 M PBS and then stored at −20°C . Immunostaining was performed on a 1 in 6 series of free-floating 30 µm cryostat sections . Ki67 staining was done for sections from control ( n = 10 ) , novel environment ( n = 6 ) , footshock ( n = 4 ) , immobilized ( n = 6 ) , 0 mg/kg ( n = 6 ) , 5 mg/kg ( n = 6 ) and 40 mg/kg ( n = 6 ) CORT-injected rats as per ( Kirby et al . , 2012b ) with a few additions . Sections were antigen-retrieved using 10 mM citrate buffer , pH 8 . 0 at 95°C for 20 min prior to peroxidase blocking and the primary antibody used was rabbit anti-Ki67 ( 1:500; Novus , St . Louis , MO ) . BrdU staining was done as per ( Kirby et al . , 2012b ) for sections from control ( n = 6 ) , immobilized ( n = 5 ) , 0 mg/kg ( n = 6 ) , and 40 mg/kg ( n= 5 ) rats who were perfused 24 hr after the BrdU injection , which occurred 3 hr after the start of the stressor . Sections from control ( n = 10 ) and immobilized ( n = 8 ) rats who were perfused 2 weeks after immobilization/BrdU injection , 1 hr after the final fear extinction probe ( see ‘Fear conditioning' ) , were also stained for BrdU using the same procedure . All stained BrdU and Ki67 sections were mounted on gelatin-coated slides , dehydrated in alcohols and coversliped with permount . Double immunohistochemical labeling for FGF2 quantification in GFAP+ cells was done on sections from control ( n = 2 ) , immobilized ( n = 6 ) , 0 mg/kg CORT injected ( n = 6 ) and 40 mg/kg CORT injected ( n = 6 ) rats as per ( Kirby et al . , 2012b ) with the following deviations . Primary antibodies were rabbit anti-FGF2 ( 1:500; Abcam , Cambridge , UK ) , and mouse anti-GFAP ( 1:1000; Cell Signalling , Danvers , MA ) . Secondary antibodies were AlexaFluor 488 donkey anti-rabbit ( 1:200; Invitrogen , Carlsbad , CA ) and Cy3 donkey anti-mouse ( 1:200; Jackson ImmunoResearch , West Grove , PA ) . Following secondary incubation and rinsing , sections were mounted on gelatin-coated slides and coverslipped with Vectashield mounting medium with DAPI ( Vector , Burlingame , CA ) . Triple immunohistochemical labeling for cell fate analysis was done on sections from control ( n = 6 ) and immobilized ( n = 6 ) rats 2 weeks after immobilization/BrdU injection , after the final fear extinction probe ( see ‘Fear conditioning' ) as per ( Kirby et al . , 2012b ) with a few exceptions . Primary antibodies were goat anti-DCX ( 1:200; Santa Cruz Biotechnology , Dallas , TX ) , mouse anti-GFAP ( 1:100; Cell Signalling ) and rat anti-BrdU ( 1:500; Abcam ) . Secondary antibodies were AlexaFluor 594 anti-goat , AlexaFluor 647 anti-mouse and biotin anti-rat ( 1:500; Jackson ImmunoResearch ) . Tertiary antibody was Streptavidin Alexa Fluor 488 ( 1:1000; Jackson ImmunoResearch ) . Double immunohistochemical staining for DCX and cfos was performed similarly , ( n = 5 control , n = 7 immobilized ) with goat anti-DCX as above and mouse anti-cfos ( 1:50; Santa Cruz Biotechnology ) . Secondary antibodies were as above for anti-goat and AlexaFluor 647 anti-mouse ( 1:500; Jackson ImmunoResearch ) . Sections were mounted on gelatin-coated slides and coversliped with DABCO antifading medium . Ki67- and BrdU-positive cells were counted in the dorsal and ventral dentate gyrus and subgranular zone using a 40× air objective ( Zeiss , Oberkochen , Germany ) . The area sampled was calculated using StereoInvestigator software ( MicroBrightfield , Williston , VT ) and used to calculate the number of positive cells per micrometer ( Lupien and McEwen , 1997 ) . For quantification of FGF2 immunoreactivity , 18 μm Z-stacks of 1 μm slices in the dorsal DG and hilus were acquired using a 20× air objective . Mean DG optical density was measured in ImageJ software using the summed Z-stack of FGF2 immunoreactivity . Mean optical density of FGF2- areas of tissue were subtracted from the DG intensity value to correct for background . Integrated optical density of GFAP+ and GFAP- negative cells in the hilus were determined by confirming GFAP expression in the Z-stack and then acquiring integrated optical density from the summed Z-stack of each individual cell . 51 to 97 GFAP+ cells and 16 to 57 GFAP- cells were sampled per rat . To quantify cFos expression in new neurons , BrdU positive cells were located in the dorsal dentate gyrus for each animal using a 40× oil objective and assessed in z-series of <1 . 0 μm slices to determine if other markers ( DCX , GFAP ) were co-expressed . Confocal images were captured on a Zeiss 510 META/NLO confocal microscope with a 40× oil objective . Rats were lightly anesthetized with isoflurane and rapidly decapitated 30 min or 3 hr after the beginning of their respective stressors ( n = 6/grp ) . Bilateral hippocampi were dissected and flash frozen in liquid nitrogen . Trunk blood was collected for plasma corticosterone quantification . One hippocampus per rat was used for mRNA expression quantification . The other was used for western blot analysis , right and left side being counterbalanced among groups . Real-time reverse transcriptase PCR was run on Trizol-extracted RNA with primers detailed in Table 1 . 10 . 7554/eLife . 00362 . 013Table 1 . List of primersDOI: http://dx . doi . org/10 . 7554/eLife . 00362 . 013GeneDirectionSequencebdnfexonIV+GGAGTGGAAAGGGTGAAACA−GGATTCAGTGGGACTCCAGAbdnfexonIX+GAGAAGAGTGATGACCATCCT−TCACGTGCTCAAAAGTGTCAGcfos+GGCAAAGTAGAGCAGCTATCTCCT−TCAGCTCCCTCCTCCGATTCfgf2+CGGTACCTGGCTATGAAGGA−CTCCAGGCGTTCAAAGAAGAfgfr1+ACCTGAGGCATTGTTTGACC−GTGAGCCACCCAGAGTGAATfgfr2+GGCCTCTCTGAATGCTAACG−ACGAGACAATCCTCCTGTGGfgfr3+TCTGGTCCTTTGGTGTCCTC−TGAGGATGCGGTCTAAATCCfgfr4+GTGGCTGTGAAGATGCTGAA−GAGGAATTCCCGAAGGTTTCgadd45β+GTCACCTCCGTCTTCTTGGA−GAGGCGGTGGGACTTACTTTngf+GGACGCAGCTTTCTATCCTGG−CCCTCTGGGACATTGCTATCTGrplp+CCAAAGGTTTGGGAGAACAA−GGGTCATGGCATAGAGCAATvegf+GAGGAAAGGGAAAGGGTCAAA−CACAGTGAACGCTCCAGGATT Primer sequences were designed using Primer1 software and checked for specificity using BLAST . Extracted RNA was treated with DNase ( DNA-free , Ambion , Carlsbad , CA ) , and two-step PCR was used , following manufacturer instructions for iScript cDNA synthesis kit ( BioRad , Hercules , CA ) and then iQ SYBR Green Supermix . Samples were run in a BioRad IQ5 real-time PCR machine . After the PCR was complete , specificity of each primer pair was confirmed using melt curve analysis . Ct values were determined using PCR miner ( Zhao and Fernald , 2005 ) and normalized to the reference ribosomal RNA , RPLP . Fold change in mRNA expression is relative to no stress control rats . Protein was extracted by homogenizing in RIPA buffer with protease inhibitor ( 1:100; Calbiochem , Billerica , MA ) and phosphatase inhibitor ( 1:10; Roche , Basel , Switzerland ) . Following 30 min incubation on ice , samples were centrifuged at 12 , 000×g for 30 min at 4°C . The extracted protein was stored at −80°C . Total protein content was assessed using a BCA kit ( Pierce , Waltham , MA ) . Samples were diluted 1:1 in laemmli buffer ( Biorad ) + 5% β-mercaptoethanol ( Fisher , Waltham , MA ) and run on 4–20% Mini-PROTEAN TGX gels ( Biorad ) at 100 V for 1 . 5 hr in 1× Tris-glycine-SDS buffer . They were then transferred to nitrocellulose membrane ( Biorad ) at 100 V for 1 hr in 1× Tris-glycine-SDS buffer with 20% methanol . Membranes were blocked for 1 hr with 5% milk in 0 . 1 M Tris buffered saline with 1% Tween-20 ( Fisher ) ( TBS-t ) . Membranes were incubated in primary ( rabbit anti-FGF2 , 1:100; Abcam; mouse anti-actin , 1:10 , 000; Roche; rabbit anti-bdnf , 1:500; Abcam ) in blocking solution overnight at 4°C . The next day , membranes were rinsed three times with TBS-t then incubated in secondary ( LiCor [Lincoln , NE] IRDye 680LT Donkey anti-mouse or LiCor IRDye 800CW Donkey anti-rabbit , 1:20 , 000 ) for 1 hr . After three final rinses , membranes were visualized using a LiCor Odyssey scanner . The correct band size was found relative to a LiCor IRDye ( 680/800 ) protein marker ladder . All bands were quantified using LiCor Odyssey software , corrected for background and expressed relative to their corresponding actin band . Fold change in protein expression was then calculated relative to no stress control . All blood samples were centrifuged at 2000×g for 15 min and plasma was extracted and stored at −20°C until assayed . Corticosterone was measured using a Corticosterone EIA kit ( Enzo Life Sciences , Farmingdale , NY ) . Isolation of neural stem/progenitor cells from adult rodents are described in detail in ( Gage , 2000 ) . Progenitors used in these experiments were purchased from Millipore ( Billerica , MA; SCR022 ) . Cells were cultured under standard conditions ( 37°C , 5% CO2 ) on poly-ornithine ( Sigma ) and laminin ( Invitrogen ) coated plates in N2-supplemented ( Invitrogen ) Dulbecco's modified Eagle medium ( DMEM ) /F-12 ( 1:1 ) ( Invitrogen ) with 20 ng/ml recombinant human FGF-2 ( PeproTech , Rocky Hill , NJ ) . Primary astrocyte cultures were prepared from P1–2 day old Sprague Dawley rat pup hippocampi using the method described by McCarthy and Vellis ( McCarthy and de Vellis , 1980 ) . Briefly , hippocampi were dissected in ice-cold media , chopped and digested using papain from papaya latex extract ( Sigma ) in HBSS ( Invitrogen ) for 20 min at 37°C . Papain was inactivated using 10% horse serum , cells were centrifuged for 1 min at 350×g and resuspended in HBSS and triturated by passing through serological and flame-polished pipettes of progressively smaller bores . Cells were then plated in DMEM ( Invitrogen ) supplemented with 10% fetal bovine serum ( Axenia BioLogix , Dixon , CA ) and 1% Penicillin/Streptomycin ( Invitrogen ) at a density of 3 × 106 in T75 flasks . After reaching confluency , flasks were shaken on an orbital shaker at 225 rpm for 2 hr at 37°C . Cells were then washed 5× with warm PBS to remove suspended microglia . Astrocytes were then trypsinized and re-plated in 100 mm dishes . 24 hr after plating , astrocytes were treated with 1 µM CORT or equivalent volume of EtOH vehicle for 3 hr . ACM was then collected , filtered with a 0 . 2 µm sterile filter and stored at −20°C . In all studies , NPCs were FGF2 deprived for 24 hr then treated for 3 hr with the appropriate media . They were then pulsed with 30 µM BrdU and fixed 2 hr later with 4% paraformaldehyde for 10 min . Cell treatments were: 0 ng/ml FGF2 ( +EtOH , n = 6; +CORT , n = 5 ) , 20 ng/ml FGF2 ( +EtOH , n = 5; +CORT , n = 5 ) , CoC media ( +EtOH , n = 6; +CORT , n = 6 ) , ACM ( +EtOH , n = 6; +CORT , n = 6 ) . Treatments for the rat recombinant FGF2 experiment were 0 pg/ml FGF2 ( n = 8 ) and 4 pg/ml FGF2 ( n = 7 ) . Fixed cells were rinsed with 0 . 1 M PBS , denatured in 1 N HCl at 37°C , rinsed and blocked in 5% normal donkey serum , 0 . 3% triton-100 in PBS . Cells were then incubated overnight at 4°C in mouse anti-brdu ( 1:500; BD Biosciences , Franklin Lakes , NJ ) in 2% normal donkey serum in PBS . Cells were then rinsed and incubated in Cy3 anti-mouse ( 1:500; Jackson Immunoresearch ) in 2% normal donkey serum in PBS , rinsed , counterstained with DAPI ( 1:20 , 000 in PBS ) and coverslipped with DABCO anti-fading medium . BrdU+ and DAPI+ cells were counted in randomly sampled sites within each well using StereoInvestigator software ( Microbrightfield ) and a 20× air objective ( Zeiss ) . CoC media or ACM from primary astrocytes treated with EtOH or CORT was incubated with neutralizing FGF2 antibody ( Millipore ) for 1 hr at 37°C prior to use on NPCs . For nAb dose testing , NPCs were treated with 0 ng/ml FGF2 ( 0 µg/ml nAb , n = 8; 10 µg/ml nAb , n = 5 , 20 µg/ml nAb , n = 3 ) or 20 ng/ml FGF2 ( 0 µg/ml nAb , n = 8; 10 µg/ml nAb , n = 7 , 20 µg/ml nAb , n = 8 ) . For ACM treatment with the FGF2 nAb , all n = 6 . For nAb specificity tests , the recommended standard curves of FGF1 and FGF2 from their respective rat-specific ELISAs ( Antibodies Online ) were preincubated in 0 , 10 or 20 μg/ml nAb at 37°C for 1 hr before being quantified in the ELISA according to manufacturer's instructions . ACM from CORT ( n = 3 ) and EtOH ( n = 3 ) treated astrocytes was analyzed using the same rat FGF2 ELISA ( Antibodies Online , Atlanta , GA ) . Rats were immobilized for 3 hr or left undisturbed , then injected with 200 mg/kg BrdU . Either 2 days or 2 weeks after immobilization , rats were trained in contextual fear conditioning as per ( Kirby et al . , 2012b; 2 days: n = 10 con , n = 10 immob; 2 weeks: n = 20 con , n = 18 immob ) . The next day , rats were exposed to the fear conditioning context five times for 10 min each time with no shock ( extinction ) . 24 hr later , they were re-exposed to the conditioning chamber for 10 min with no shock ( extinction probe ) . A subset of rats ( 2 days: n = 10 con , n = 10 immob; 2 weeks: n = 10 con , n = 8 immob ) was tested on an elevated plus maze 1 day prior to contextual fear conditioning ( data not shown ) . In most studies , data were analyzed using a one-way ANOVA followed by Dunnett's posthoc tests with the appropriate control group as reference . For in vivo CORT injection studies with 30 min and 3 hr time points , data were analyzed using two-way ANOVA followed by Dunnett's posthoc test ( 0 mg/kg-30 min used as reference ) . For cell culture studies , data were analyzed using two-way ANOVA followed by Dunnett's posthoc tests . For BLA lesion studies , a repeated measures two-way ANOVA was used with hemisphere being the paired variable . When only two groups were being compared , unpaired t-tests were used in all cases . p≤0 . 05 was considered significant . | A little stress can be good for you . Just over 100 years ago , psychologists Robert Yerkes and John Dodson suggested that cognitive performance improves as stress increases , although it falls off again if stress levels continue to rise . The hippocampus is a key brain region for both memory and the regulation of emotion , and is highly sensitive to the main class of stress hormones , glucocorticoids . One particular subregion of the hippocampus , the dentate gyrus , contains a high density of glucocorticoid receptors , and is also notable for being one of only two regions in the adult mammalian brain that can give rise to new neurons via a process called neurogenesis . Chronic stress is known to impair memory and to reduce neurogenesis . However , the effects of acute stress are less clear-cut: early studies suggested that it suppressed the generation of new neurons , whereas several recent studies have observed no effect . Other work has shown that acute stress increases the expression of growth factors—substances that stimulate cellular growth and proliferation—which would seem to suggest that stress could enhance neurogenesis . Now Kirby et al . have obtained further insights into the effects of acute stress on the proliferation of cells in the dentate gyrus . Exposing rats to a moderate acute stressor , namely being restrained for a few hours , led to increased neurogenesis in the dorsal , but not ventral , hippocampus . Injecting rats with the stress hormone corticosterone had the same effect . In both cases , the enhanced neurogenesis was accompanied by increased expression of a growth factor called FGF2 , which is produced by glial cells called astrocytes . Intriguingly , Kirby and co-workers found that the stressed rats performed better than control animals in a memory test . Moreover , the beneficial effects were seen if the rats performed the task 2 weeks after their stressful experience , but not if they performed the task 2 days after being stressed . This is pertinent because new neurons in the dentate gyrus become functional 2 weeks after being generated , which suggests that the stress-induced increase in neurogenesis could account for the rats' improved memory . The work of Kirby and co-workers has thus identified a mechanism by which moderate acute stress could have beneficial effects on cognition . Given that acute stress can be harmful in other instances—leading , for example , to post-traumatic stress disorder—further work is required to identify the factors that determine whether a response to stress is adaptive or pathological . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2013 | Acute stress enhances adult rat hippocampal neurogenesis and activation of newborn neurons via secreted astrocytic FGF2 |
Enhancers and silencers often depend on the same transcription factors ( TFs ) and are conflated in genomic assays of TF binding or chromatin state . To identify sequence features that distinguish enhancers and silencers , we assayed massively parallel reporter libraries of genomic sequences targeted by the photoreceptor TF cone-rod homeobox ( CRX ) in mouse retinas . Both enhancers and silencers contain more TF motifs than inactive sequences , but relative to silencers , enhancers contain motifs from a more diverse collection of TFs . We developed a measure of information content that describes the number and diversity of motifs in a sequence and found that , while both enhancers and silencers depend on CRX motifs , enhancers have higher information content . The ability of information content to distinguish enhancers and silencers targeted by the same TF illustrates how motif context determines the activity of cis-regulatory sequences .
Active cis-regulatory sequences in the genome are characterized by accessible chromatin and specific histone modifications , which reflect the action of DNA-binding transcription factors ( TFs ) that recognize specific sequence motifs and recruit chromatin-modifying enzymes ( Klemm et al . , 2019 ) . These epigenetic hallmarks of active chromatin are routinely used to train machine learning models that predict cis-regulatory sequences , based on the assumption that such epigenetic marks are reliable predictors of genuine cis-regulatory sequences ( Ernst and Kellis , 2012; Ghandi et al . , 2014; Hoffman et al . , 2012; Kelley et al . , 2016; Lee et al . , 2011; Sethi et al . , 2020; Zhou and Troyanskaya , 2015 ) . However , results from functional assays show that many predicted cis-regulatory sequences exhibit little or no cis-regulatory activity . Typically , 50% or more of predicted cis-regulatory sequences fail to drive expression in massively parallel reporter assays ( MPRAs ) ( Moore et al . , 2020; Kwasnieski et al . , 2014 ) , indicating that an active chromatin state is not sufficient to reliably identify cis-regulatory sequences . Another challenge is that enhancers and silencers are difficult to distinguish by chromatin accessibility or epigenetic state ( Doni Jayavelu et al . , 2020; Gisselbrecht et al . , 2020; Pang and Snyder , 2020; Petrykowska et al . , 2008; Segert et al . , 2021 ) , and thus computational predictions of cis-regulatory sequences often do not differentiate between enhancers and silencers . Silencers are often enhancers in other cell types ( Brand et al . , 1987; Doni Jayavelu et al . , 2020; Gisselbrecht et al . , 2020; Huang et al . , 2021; Jiang et al . , 1993; Ngan et al . , 2020; Pang and Snyder , 2020 ) , reside in open chromatin ( Doni Jayavelu et al . , 2020; Huang et al . , 2019; Huang et al . , 2021; Pang and Snyder , 2020 ) , sometimes bear epigenetic marks of active enhancers ( Fan et al . , 2016; Huang et al . , 2021 ) , and can be bound by TFs that also act on enhancers in the same cell type ( Alexandre and Vincent , 2003; Grass et al . , 2003; Huang et al . , 2021; Iype et al . , 2004; Jiang et al . , 1993; Liu et al . , 2014; Martínez-Montañés et al . , 2013; Peng et al . , 2005; Rachmin et al . , 2015; Rister et al . , 2015; Stampfel et al . , 2015; White et al . , 2013 ) . As a result , enhancers and silencers share similar sequence features , and understanding how they are distinguished in a particular cell type remains an important challenge ( Segert et al . , 2021 ) . The TF cone-rod homeobox ( CRX ) controls selective gene expression in a number of different photoreceptor and bipolar cell types in the retina ( Chen et al . , 1997; Freund et al . , 1997; Furukawa et al . , 1997; Murphy et al . , 2019 ) . These cell types derive from the same progenitor cell population ( Koike et al . , 2007; Wang et al . , 2014 ) , but they exhibit divergent , CRX-directed transcriptional programs ( Corbo et al . , 2010; Hennig et al . , 2008; Hughes et al . , 2017; Murphy et al . , 2019 ) . CRX cooperates with cell type-specific co-factors to selectively activate and repress different genes in different cell types and is required for differentiation of rod and cone photoreceptors ( Chen et al . , 2005; Hao et al . , 2012; Hennig et al . , 2008; Hsiau et al . , 2007; Irie et al . , 2015; Kimura et al . , 2000; Lerner et al . , 2005; Mears et al . , 2001; Mitton et al . , 2000; Murphy et al . , 2019; Peng et al . , 2005; Sanuki et al . , 2010; Srinivas et al . , 2006 ) . However , the sequence features that define CRX-targeted enhancers vs . silencers in the retina are largely unknown . We previously found that a significant minority of CRX-bound sequences act as silencers in an MPRA conducted in live mouse retinas ( White et al . , 2013 ) , and that silencer activity requires CRX ( White et al . , 2016 ) . Here , we extend our analysis by testing thousands of additional candidate cis-regulatory sequences . We show that while regions of accessible chromatin and CRX binding exhibit a range of cis-regulatory activity , enhancers and silencers contain more TF motifs than inactive sequences , and that enhancers are distinguished from silencers by a higher diversity of TF motifs . We capture the differences between these sequence classes with a new metric , motif information content ( Boltzmann entropy ) , that considers only the number and diversity of TF motifs in a candidate cis-regulatory sequence . Our results suggest that CRX-targeted enhancers are defined by a flexible regulatory grammar and demonstrate how differences in motif information content encode functional differences between genomic loci with similar chromatin states .
The cis-regulatory activities of CRX-targeted sequences vary widely ( Figure 1a ) . We defined enhancers and silencers as those sequences that have statistically significant activity that is at least twofold above or below the activity of the basal Rho promoter ( Welch’s t-test , Benjamini-Hochberg false discovery rate ( FDR ) q < 0 . 05 , Supplementary file 3 ) . We defined inactive sequences as those whose activity is both within a twofold change of basal activity and not significantly different from the basal Rho promoter . We further stratified enhancers into strong and weak enhancers based on whether or not they fell above the 95th percentile of scrambled sequences . Using these criteria , 22% of CRX-targeted sequences are strong enhancers , 28% are weak enhancers , 19% are inactive , and 17% are silencers; the remaining 13% were considered ambiguous and removed from further analysis . To test whether these sequences function as CRX-dependent enhancers and silencers in the genome , we examined genes differentially expressed in Crx-/- retina ( Roger et al . , 2014 ) . Genes that are de-repressed are more likely to be near silencers ( Fisher’s exact test p = 0 . 001 , odds ratio = 2 . 1 , n = 206 ) and genes that are down-regulated are more likely to be near enhancers ( Fisher’s exact test p = 0 . 02 , odds ratio = 1 . 5 , n = 344 , Materials and methods ) , suggesting that our reporter assay identified sequences that act as enhancers and silencers in the genome . We sought to identify features that would accurately classify these different classes of sequences . Neither CRX ChIP-seq-binding status nor DNA accessibility as measured by ATAC-seq strongly differentiates between these four classes ( Figure 1b ) . Compared to CRX ChIP-seq peaks , ATAC-seq peaks that lack CRX binding in the adult retina are slightly enriched for inactive sequences ( Fisher’s exact test p = 2 × 10–7 , odds ratio = 1 . 5 ) and slightly depleted for strong enhancers ( Fisher’s exact test p = 1 × 10–21 , odds ratio = 2 . 2 ) . However , sequences with ChIP-seq or ATAC-seq peaks span all four activity categories , consistent with prior reports that DNA accessibility and TF binding data are not sufficient to identify functional enhancers and silencers ( Doni Jayavelu et al . , 2020; Huang et al . , 2019; Huang et al . , 2021; Pang and Snyder , 2020; White et al . , 2013 ) . We examined whether the number and affinity of CRX motifs differentiate enhancers , silencers , and inactive sequences by computing the predicted CRX occupancy ( i . e . expected number of bound molecules ) for each sequence ( White et al . , 2013 ) . Consistent with our previous work ( White et al . , 2016 ) , both strong enhancers and silencers have higher predicted CRX occupancy than inactive sequences ( Mann-Whitney U test , p = 6 × 10–10 and 6 × 10–17 , respectively , Figure 1c ) , suggesting that total CRX motif content helps distinguish silencers and strong enhancers from inactive sequences . However , predicted CRX occupancy does not distinguish strong enhancers from silencers: a logistic regression classifier trained with fivefold cross-validation only achieves an area under the receiver operating characteristic ( AUROC ) curve of 0 . 548 ± 0 . 023 and an area under the precision recall ( AUPR ) curve of 0 . 571 ± 0 . 020 ( Figure 2a and Figure 2—figure supplement 1 ) . We thus sought to identify sequence features that distinguish strong enhancers from silencers . We performed a de novo motif enrichment analysis to identify motifs that distinguish strong enhancers from silencers and found several differentially enriched motifs matching known TFs . For motifs that matched multiple TFs , we selected one representative TF for downstream analysis , since TFs from the same family have PWMs that are too similar to meaningfully distinguish between motifs for these TFs ( Figure 2—figure supplement 2 , Materials and methods ) . Strong enhancers are enriched for several motif families that include TFs that interact with CRX or are important for photoreceptor development: NeuroD1/NDF1 ( E-box-binding bHLH ) ( Morrow et al . , 1999 ) , RORB ( nuclear receptor ) ( Jia et al . , 2009; Srinivas et al . , 2006 ) , MAZ or Sp4 ( C2H2 zinc finger ) ( Lerner et al . , 2005 ) , and NRL ( bZIP ) ( Mears et al . , 2001; Mitton et al . , 2000 ) . Sp4 physically interacts with CRX in the retina ( Lerner et al . , 2005 ) , but we chose to represent the zinc finger motif with MAZ because it has a higher quality score in the HOCOMOCO database ( Kulakovskiy et al . , 2018 ) . Silencers were enriched for a motif that resembles a partial K50 homeodomain motif but instead matches the zinc finger TF GFI1 , a member of the Snail repressor family ( Chiang and Ayyanathan , 2013 ) expressed in developing retinal ganglion cells ( Yang et al . , 2003 ) . Therefore , while strong enhancers and silencers are not distinguished by their CRX motif content , strong enhancers are uniquely enriched for several lineage-defining TFs . To quantify how well these TF motifs differentiate strong enhancers from silencers , we trained two different classification models with fivefold cross-validation . First , we trained a 6-mer support vector machine ( SVM ) ( Ghandi et al . , 2014 ) and achieved an AUROC of 0 . 781 ± 0 . 013 and AUPR of 0 . 812 ± 0 . 020 ( Figure 2a and Figure 2—figure supplement 1 ) . The SVM considers all 2080 non-redundant 6-mers and provides an upper bound to the predictive power of models that do not consider the exact arrangement or spacing of sequence features . We next trained a logistic regression model on the predicted occupancy for eight lineage-defining TFs ( Supplementary file 4 ) and compared it to the upper bound established by the SVM . In this model , we considered CRX , the five TFs identified in our motif enrichment analysis , and two additional TFs enriched in photoreceptor ATAC-seq peaks ( Hughes et al . , 2017 ) : RAX , a Q50 homeodomain TF that contrasts with CRX , a K50 homeodomain TF ( Irie et al . , 2015 ) and MEF2D , a MADS box TF which co-binds with CRX ( Andzelm et al . , 2015 ) . The logistic regression model performs nearly as well as the SVM ( AUROC 0 . 698 ± 0 . 036 , AUPR 0 . 745 ± 0 . 032 , Figure 2a and Figure 2—figure supplement 1 ) despite a 260-fold reduction from 2080 to 8 features . To determine whether the logistic regression model depends specifically on the eight lineage-defining TFs , we established a null distribution by fitting 100 logistic regression models with randomly selected TFs ( Materials and methods ) . Our logistic regression model outperforms the null distribution ( one-tailed Z-test for AUROC and AUPR , p < 0 . 0008 , Figure 2—figure supplement 3 ) , indicating that the performance of the model specifically requires the eight lineage-defining TFs . To determine whether the SVM identified any additional motifs that could be added to the logistic regression model , we generated de novo motifs using the SVM k-mer scores and found no additional motifs predictive of strong enhancers . Finally , we found that our two models perform similarly on an independent test set of CRX-targeted sequences ( White et al . , 2013; Figure 2—figure supplement 3 ) . Since the logistic regression model performs near the upper bound established by the SVM and depends specifically on the eight selected motifs , we conclude that these motifs comprise nearly all of the sequence features captured by the SVM that distinguish strong enhancers from silencers among CRX-targeted sequences . To understand how these eight TF motifs differentiate strong enhancers from silencers , we first calculated the total predicted occupancy of each sequence by all eight lineage-defining TFs and compared the different activity classes . Strong enhancers and silencers both have higher total predicted occupancies than inactive sequences , but total predicted occupancies do not distinguish strong enhancers and silencers from each other ( Figure 2b , Supplementary file 5 ) . Since strong enhancers are enriched for several motifs relative to silencers , this suggests that strong enhancers are distinguished from silencers by the diversity of their motifs , rather than the total number . We considered two hypotheses for how the more diverse collection of motifs function in strong enhancers: either strong enhancers depend on specific combinations of TF motifs ( ‘TF identity hypothesis’ ) or they instead must be co-occupied by multiple lineage-defining TFs , regardless of TF identity ( ‘TF diversity hypothesis’ ) . To distinguish between these hypotheses , we examined which specific motifs contribute to the total motif content of strong enhancers and silencers . We considered motifs for a TF present in a sequence if the TF predicted occupancy was above 0 . 5 molecules ( Supplementary file 4 ) , which generally corresponds to at least one motif with a relative KD above 3% . This threshold captures the effect of low affinity motifs that are often biologically relevant ( Crocker et al . , 2015; Farley et al . , 2015; Farley et al . , 2016; Parker et al . , 2011 ) . As expected , 97% of strong enhancers and silencers contain CRX motifs since the sequences were selected based on CRX binding or significant matches to the CRX PWM within open chromatin ( Figure 2c ) . Compared to silencers , strong enhancers contain a broader diversity of motifs for the eight lineage-defining TFs ( Figure 2c ) . However , while strong enhancers contain a broader range of motifs , no single motif occurs in a majority of strong enhancers: NRL motifs are present in 23% of strong enhancers , NeuroD1 and RORB in 18% each , and MAZ in 16% . Additionally , none of the motifs tend to co-occur as pairs in strong enhancers: no specific pair occurred in more than 5% of sequences ( Figure 2d ) . We also did not observe a bias in the linear arrangement of motifs in strong enhancers ( Materials and methods ) . Similarly , no single motif occurs in more than 15% of silencers ( Figure 2c ) . These results suggest that strong enhancers are defined by the diversity of their motifs , and not by specific motif combinations or their linear arrangement . The results above predict that strong enhancers are more likely to be bound by a diverse but degenerate collection of TFs , compared with silencers or inactive sequences . We tested this prediction by examining in vivo TF binding using published ChIP-seq data for NRL ( Hao et al . , 2012 ) and MEF2D ( Andzelm et al . , 2015 ) . Consistent with the prediction , sequences bound by CRX and either NRL or MEF2D are approximately twice as likely to be strong enhancers compared to sequences only bound by CRX ( Figure 2e ) . Sequences bound by all three TFs are the most likely to be strong or weak enhancers rather than silencers or inactive sequences . However , most strong enhancers are not bound by either NRL or MEF2D ( Figure 2f ) , indicating that binding of these TFs is not required for strong enhancers . Our results support the TF diversity hypothesis: CRX-targeted enhancers are co-occupied by multiple TFs , without a requirement for specific combinations of lineage-defining TFs . Our results indicate that both strong enhancers and silencers have a higher total motif content than inactive sequences , while strong enhancers contain a more diverse collection of motifs than silencers . To quantify these differences in the number and diversity of motifs , we computed the information content of CRX-targeted sequences using Boltzmann entropy . The Boltzmann entropy of a system is related to the number of ways the system’s molecules can be arranged , which increases with either the number or diversity of molecules ( Phillips et al . , 2012 , Chapter 5 ) . In our case , each TF is a different type of molecule and the number of each TF is represented by its predicted occupancy for a cis-regulatory sequence . The number of molecular arrangements is thus W , the number of distinguishable permutations that the TFs can be ordered on the sequence , and the information content of a sequence is then log2W ( Materials and methods ) . We found that on average , strong enhancers have higher information content than both silencers and inactive sequences ( Mann-Whitney U test , p = 1 × 10–23 and 7 × 10–34 , respectively , Figure 3a , Supplementary file 5 ) , confirming that information content captures the effect of both the number and diversity of motifs . Quantitatively , the average silencer and inactive sequence contains 1 . 6 and 1 . 4 bits , respectively , which represents approximately three total motifs for two TFs . Strong enhancers contain on average 2 . 4 bits , representing approximately three total motifs for three TFs or four total motifs for two TFs . To compare the predictive value of our information content metric to the model based on all eight motifs , we trained a logistic regression model and found that information content classifies strong enhancers from silencers with an AUROC of 0 . 634 ± 0 . 008 and an AUPR of 0 . 663 ± 0 . 014 ( Figure 3b and Figure 3—figure supplement 1 ) . This is only slightly worse than the model trained on eight TF occupancies despite an eightfold reduction in the number of features , which is itself comparable to the SVM with 2080 features . The difference between the two logistic regression models suggests that the specific identities of TF motifs make some contribution to the eight TF model , but that most of the signal captured by the SVM can be described with a single metric that does not assign weights to specific motifs . Information content also distinguishes strong enhancers from inactive sequences ( AUROC 0 . 658 ± 0 . 012 , AUPR 0 . 675 ± 0 . 019 , Figure 3b and Figure 3—figure supplement 1 ) . These results indicate that strong enhancers are characterized by higher information content , which reflects both the total number and diversity of motifs . Our results show that except for CRX , none of the lineage-defining motifs occur in a majority of strong enhancers . However , all sequences were tested in reporter constructs with the Rho promoter , which contains an NRL motif and three CRX motifs ( Corbo et al . , 2010; Kwasnieski et al . , 2012 ) . Since NRL is a key co-regulator with CRX in rod photoreceptors , we tested whether strong enhancers generally require NRL , which would be inconsistent with our TF diversity hypothesis . We removed the NRL motif by recloning our MPRA library without the basal Rho promoter . If strong enhancers require an NRL motif for high activity , then only CRX-targeted sequences with NRL motifs will drive reporter expression . If information content ( i . e . total motif content and diversity ) is the primary determinant of strong enhancers , only CRX-targeted sequences with sufficient motif diversity , measured by information content , will drive reporter expression regardless of whether or not NRL motifs are present . We replaced the Rho promoter with a minimal 23 bp polylinker sequence between our libraries and DsRed , and repeated the MPRA ( Figure 1—figure supplement 1 , Supplementary file 3 ) . CRX-targeted sequences were designated as ‘autonomous’ if they retained activity in the absence of the Rho promoter ( log2 ( RNA/DNA ) > 0 , Materials and methods ) . We found that 90% of autonomous sequences are from the enhancer class , while less than 3% of autonomous sequences are from the silencer class ( Figure 4a ) . This confirms that the distinction between silencers and enhancers does not depend on the Rho promoter , which is consistent with our previous finding that CRX-targeted silencers repress other promoters ( Hughes et al . , 2018; White et al . , 2016 ) . However , while most autonomous sequences are enhancers , only 39% of strong enhancers and 9% of weak enhancers act autonomously . Consistent with a role for information content , autonomous strong enhancers have higher information content ( Mann-Whitney U test p = 4 × 10–8 , Figure 4b ) and higher predicted CRX occupancy ( Mann-Whitney U test p = 9 × 10–12 , Figure 4c ) than non-autonomous strong enhancers . We found no evidence that specific lineage-defining motifs are required for autonomous activity , including NRL , which is present in only 25% of autonomous strong enhancers ( Figure 4d ) . Similarly , NRL ChIP-seq binding ( Hao et al . , 2012 ) occurs more often among autonomous strong enhancers ( 41% vs . 19% , Fisher’s exact test p = 2 × 10–14 , odds ratio = 3 . 0 ) , yet NRL binding still only accounts for a minority of these sequences . We thus conclude that strong enhancers require high information content , rather than any specific lineage-defining motifs . Our results indicate that information content distinguishes strong enhancers from silencers and inactive sequences . Information content only takes into account the total number and diversity of motifs in a sequence and not any potential interactions between them . The classification success of information content thus suggests that each TF motif will contribute independently to enhancer activity . We tested this prediction with CRX-targeted sequences where all CRX motifs were abolished by point mutation ( Supplementary file 3 ) . Consistent with our previous work ( White et al . , 2013 ) , mutating CRX motifs causes the activities of both enhancers and silencers to regress toward basal levels ( Pearson’s r = 0 . 608 , Figure 5a ) , indicating that most enhancers and silencers show some dependence on CRX . However , 40% of wild-type strong enhancers show low CRX dependence and remain strong enhancers with their CRX motifs abolished . Although strong enhancers with high and low CRX dependence have similar wild-type information content ( Figure 5b ) , strong enhancers with low CRX dependence have lower predicted CRX occupancy than those with high CRX dependence ( Mann-Whitney U test p = 2 × 10–9 , Figure 5c ) , and also have higher ‘residual’ information content ( i . e . information content without CRX motifs , Mann-Whitney U test p = 1 × 10–7 , Figure 5d ) . Low CRX dependence sequences have an average of 1 . 5 residual bits , which corresponds to three motifs for two TFs , while high CRX dependence sequences have an average of 1 . 0 residual bits , which corresponds to two motifs for two TFs ( Figure 5e ) . Strong enhancers with low and high CRX dependence have similar wild-type information content and similar total predicted occupancy ( Figure 5b and e ) . As a result , sequences with more CRX motifs have fewer motifs for other TFs , suggesting that there is no evolutionary pressure for enhancers to contain additional motifs beyond the minimum amount of information content required to be active . To test this idea , we calculated the minimum number and diversity of motifs necessary to specify a relatively unique location in the genome ( Wunderlich and Mirny , 2009 ) and found that a 164 bp sequence only requires five motifs for three TFs ( Materials and methods ) . These motif requirements can be achieved in two ways with similar information content that differ only in the quantitative number of motifs for each TF . In other words , the number of motifs for any particular TF is not important so long as there is sufficient information content . Taken together , we conclude that each TF motif provides an independent contribution toward specifying strong enhancers .
Many regions in the genome are bound by TFs and bear the epigenetic hallmarks of active cis-regulatory sequences , yet fail to exhibit cis-regulatory activity when tested directly . The discrepancy between measured epigenomic state and cis-regulatory activity indicates that enhancers and silencers consist of more than the minimal sequence features necessary to recruit TFs and chromatin-modifying factors . Our results show that enhancers , silencers , and inactive sequences in developing photoreceptors can be distinguished by their motif content , even though they are indistinguishable by CRX binding or chromatin accessibility . We show that both enhancers and silencers contain more TF motifs than inactive sequences , and that enhancers also contain more diverse sets of motifs for lineage-defining TFs . These differences are captured by our measure of information content . Information content , as a single metric , identifies strong enhancers nearly as well as an unbiased set of 2080 non-redundant 6-mers used for an SVM , indicating that a simple measure of motif number and diversity can capture the key sequence features that distinguish enhancers from other sequences that lie in open chromatin . The results of our information content classifier are consistent with the TF collective model of enhancers ( Junion et al . , 2012; Spitz and Furlong , 2012 ) : globally , active enhancers are specified by the combinatorial action of lineage-defining TFs with little constraint on which motifs must co-occur . We show that CRX-targeted enhancers are distinguished from inactive CRX-targeted sequences by a larger , more diverse collection of TF motifs , and not any specific combination of motifs . This indicates that enhancers are active because they have acquired the necessary number of TF binding motifs , and not because they are defined by a strict regulatory grammar . Sequences with fewer motifs may be bound by CRX and reside within open chromatin , but they lack sufficient TF binding for activity . Such loose constraints would facilitate the de novo emergence of tissue-specific enhancers and silencers over evolution and explain why critical cell type-specific TF interactions , such as CRX and NRL in rod photoreceptors , occur at only a minority of the active enhancers in that cell type ( Hsiau et al . , 2007; Hughes et al . , 2018; White et al . , 2013 ) . Like enhancers , CRX-targeted silencers require higher motif content and are dependent on CRX motifs , but they lack the TF diversity of enhancers . The lack of TF diversity in silencers parallels the architecture of signal-responsive cis-regulatory sequences , which are silencers in the absence of a signal and require multiple activators for induction ( Barolo and Posakony , 2002 ) . Consistent with this , we previously showed using synthetic sequences that high occupancy of CRX alone is sufficient to encode silencers while the addition of a single NRL motif converts synthetic silencers to enhancers , and that genomic sequences with very high CRX motif content repress a basal promoter that lacks NRL motifs ( White et al . , 2016 ) . We found that photoreceptor genes which are de-repressed upon loss of CRX are located near cis-regulatory sequences with high CRX motif content , and that genes near regions that are bound only by CRX are expressed at lower levels than genes near regions co-bound by CRX and NRL ( White et al . , 2016 ) . In the current study , we find that silencers in our MPRA library are more likely to occur near de-repressed photoreceptor genes , while strong enhancers are enriched near genes that lose expression in Crx-/- retina . These findings suggest that the low TF diversity and high CRX motif content that characterize silencers in our MPRA library are also important for silencing in the genome . The contrast in motif diversity between enhancers and silencers that we observe could explain how CRX achieves selective activation and repression of its target genes in multiple cell types and across developmental time points ( Murphy et al . , 2019; Ruzycki et al . , 2018 ) . CRX itself is required for silencing , and we previously showed that some silencers become active enhancers in Crx-/- retina ( White et al . , 2016 ) . The mechanism of CRX-based silencing is unknown , however CRX cooperates with other TFs that can sometimes act as repressors of cell type-specific genes ( Chen et al . , 2005; Peng et al . , 2005; Webber et al . , 2008 ) , while other repressors can directly inhibit activation by CRX or its co-activators ( Dorval et al . , 2006; Hlawatsch et al . , 2013; Mitton et al . , 2003; Sanuki et al . , 2010 ) . In Drosophila photoreceptors , selective silencing of opsin genes is controlled by cell type-specific expression of a repressor , Dve , which acts on the same K50 homeodomain-binding sites as a universally expressed activator , Otd , a homolog of CRX ( Rister et al . , 2015 ) . Other transcriptional activators selectively act as repressors in the same cell type . GATA-1 represses the GATA-2 promoter by displacing CREB-binding protein ( CBP ) , while at other genes GATA-1 binds CBP to activate transcription ( Grass et al . , 2003 ) . Selective repression by GATA-1 is also mediated by chromatin occupancy levels and interaction with co-regulators ( Johnson et al . , 2006 ) , which is consistent with our finding that sequence context enables a TF to both activate and repress genes in the same cell type . Given the central role of CRX in selectively regulating genes in multiple closely related cell types ( Murphy et al . , 2019 ) , we speculate that CRX-targeted silencers may contain sufficient information to act as enhancers in other cell types in which a different set of co-activating TFs are expressed . This hypothesis would be consistent with the finding that many silencers are enhancers in other cell types ( Doni Jayavelu et al . , 2020; Gisselbrecht et al . , 2020; Ngan et al . , 2020 ) . Our work suggests that characterizing sequences by their motif information content offers a way to identify these different classes of cis-regulatory sequences in the genome .
CRX ChIP-seq peaks re-processed by Ruzycki et al . , 2018 were intersected with previously published CRX MPRA libraries ( Hughes et al . , 2018; White et al . , 2013 ) and one unpublished library to select sequences that had not been previously tested by MPRA . These sequences were scanned for instances of CRX motifs using FIMO version 4 . 11 . 2 ( Bailey et al . , 2009 ) , a p-value cutoff of 2 . 3 × 10–3 ( see below ) , and a CRX PWM derived from an electrophoretic mobility shift assay ( Lee et al . , 2010 ) . We centered 2622 sequences on the highest scoring CRX motif . For 677 sequences without a CRX motif , we instead centered them using the Gibbs sampler from ShapeMF ( Github commit abe8421 ) ( Samee et al . , 2019 ) and a motif size of 10 . For sequences unbound in CRX ChIP-seq but in open chromatin , we took ATAC-seq peaks collected in 8-week FACS-purified rods , green cones , and Nrl-/- blue cones ( Hughes et al . , 2017 ) and removed sequences that overlapped with CRX ChIP-seq peaks . The remaining sequences were scanned for instances of CRX motifs using FIMO with a p-value cutoff of 2 . 5 × 10–3 and the CRX PWM . Sequences with a CRX motif were kept and the three ATAC-seq data sets were merged together , intersected with H3K27ac and H3K4me3 ChIP-seq peaks collected in P14 retinas ( Ruzycki et al . , 2018 ) , and centered on the highest scoring CRX motifs . We randomly selected 1004 H3K27ac+H3K4me3- sequences and 541 H3K27ac+H3K4me3+ to reflect the fact that ~35% of CRX ChIP-seq peaks are H3K4me3+ . After synthesis of our library , we discovered 11% of these sequences do not actually overlap H3K27ac ChIP-seq peaks ( 110/1004 of the H3K4me3- group and 60/541 of the H3K4me3+ group ) , but we still included them in the analysis because they contain CRX motifs in ATAC-seq peaks . All data was converted to mm10 coordinates using the UCSC liftOver tool ( Haeussler et al . , 2019 ) and processed using Bedtools version 2 . 27 . 1 ( Quinlan and Hall , 2010 ) . All sequences in our library design were adjusted to 164 bp and screened for instances of EcoRI , SpeI , SphI , and NotI sites . In total , our library contains 4844 genomic sequences ( 2622 CRX ChIP-seq peaks with motifs , 677 CRX ChIP-seq peaks without motifs , 1004 CRX-ATAC+H3K27ac+H3K4me3- CRX motifs , and 541 CRX-ATAC+H3K27ac+H3K4me3+ CRX motifs ) , a variant of each sequence with all CRX motifs mutated , 150 scrambled sequences , and a construct for cloning the basal promoter alone . For sequences centered on CRX motifs , all CRX motifs with a p-value of 2 . 5 × 10–3 or less were mutated by changing the core TAAT to TACT ( Lee et al . , 2010 ) on the appropriate strand , as described previously ( Hughes et al . , 2018; White et al . , 2013 ) . We then re-scanned sequences and mutated any additional motifs inadvertently created . To generate scrambled sequences , we randomly selected 150 CRX ChIP-seq peaks spanning the entire range of GC content in the library . We then scrambled each sequence while preserving dinucleotide content as previously described ( White et al . , 2013 ) . We used FIMO to confirm that none of the scrambled sequences contain CRX motifs . We unintentionally used a FIMO p-value cutoff of 2 . 3 × 10–3 to identify CRX motifs in CRX ChIP-seq peaks , rather than the slightly less stringent 2 . 5 × 10–3 cutoff used with ATAC-seq peaks or mutating CRX motifs . Due to this anomaly , there may be sequences centered using ShapeMF that should have been centered on a CRX motif , and these motifs would not have been mutated because CRX motifs were not mutated in sequences centered using ShapeMF . However , any intact CRX motifs would still be captured in the residual information content of the mutant sequence . We generated two 15 , 000 libraries of 230 bp oligonucleotides ( oligos ) from Agilent Technologies ( Santa Clara , CA ) through a limited licensing agreement . Our library was split across the two oligo pools , ensuring that both the genomic and mutant forms of each sequence were placed in the same oligo pool ( Supplementary files 1 and 2 ) . Both oligo pools contain all 150 scrambled sequences as an internal control . All sequences were assigned three unique barcodes as previously described ( White et al . , 2013 ) . In each oligo pool , the basal promoter alone was assigned 18 unique barcodes . Oligos were synthesized as follows: 5’ priming sequence ( GTAGCGTCTGTCCGT ) /EcoRI site/Library sequence/SpeI site/C/SphI site/Barcode sequence/NotI site/3’ priming sequence ( CAACTACTACTACAG ) . To clone the basal promoter into barcoded oligos without any upstream cis-regulatory sequence , we placed the SpeI site next to the EcoRI site , which allowed us to place the promoter between the EcoRI site and the 3’ barcode . We cloned the synthesized oligos as previously described by our group ( Kwasnieski et al . , 2012White et al . , 2016; White et al . , 2013 ) . Specifically , for each oligo pool , we used 50 femtomoles of template and four cycles of PCR in each of multiple 50 µl reactions ( New England Biolabs [NEB] , Ipswich , MA ) ( NEB Phusion ) using primers MO563 and MO564 ( Supplementary file 6 ) , 2% DMSO , and an annealing temperature of 57°C . PCR amplicons were purified from a 2% agarose gel ( NEB ) , digested with EcoRI-HF and NotI-HF ( NEB ) , and then cloned into the EagI and EcoRI sites of the plasmid pJK03 with multiple 20 µl ligation reactions ( NEB T4 ligase ) . The libraries were transformed into 5-alpha electrocompetent cells ( NEB ) and grown in liquid culture . Next , 2 µg of each library was digested with SphI-HF and SpeI-HF ( NEB ) and then treated with Antarctic phosphatase ( NEB ) . The Rho basal promoter and DsRed reporter gene was amplified from the plasmid pJK01 using primers MO566 and MO567 ( Supplementary file 6 ) . The Polylinker and DsRed reporter gene was amplified from the plasmid pJK03 using primers MO610 and MO567 ( Supplementary file 6 ) . The Polylinker is a short 23 bp multiple cloning site with no known core promoter motifs . Inserts were purified from a 1% agarose gel ( NEB ) , digested with NheI-HF and SphI-HF ( NEB ) , and cloned into the libraries using multiple 20 µl ligations ( NEB T4 ligase ) . The libraries were transformed into 5-alpha electrocompetent cells ( NEB ) and grown in liquid culture . Animal procedures were performed in accordance with a Washington University in St Louis Institutional Animal Care and Use Committee-approved vertebrate animals protocol . Electroporation into retinal explants and RNA extraction was performed as described previously ( Hsiau et al . , 2007; Hughes et al . , 2018; Kwasnieski et al . , 2012; White et al . , 2016; White et al . , 2013 ) . Briefly , retinas were isolated from P0 newborn CD-1 mice and electroporated in a solution with 30 µg library and 30 µg Rho-GFP . Electroporated retinas were cultured for 8 days , at which point they were harvested , washed three times with HBSS ( ThermoFisher Scientific/Gibco , Waltham , MA ) , and stored in TRIzol ( ThermoFisher Scientific/Invitrogen , Waltham , MA ) at –80°C . Five retinas were pooled for each biological replicate and three replicates were performed for each library . RNA was extracted from TRIzol according to manufacturer’s instructions and treated with TURBO DNase ( Invitrogen ) . cDNA was prepared using SuperScript RT III ( Invitrogen ) with oligo dT primers . Barcodes from both the cDNA and the plasmid DNA pool were amplified for sequencing ( described below ) . The resulting products were mixed at equal concentration and sequenced on the Illumina NextSeq platform . We obtained greater than 1300× coverage across all samples . Rho libraries were amplified using primers MO574 and MO575 ( Supplementary file 6 ) for six cycles at an annealing temperature of 66°C followed by 18 cycles with no annealing step ( NEB Phusion ) and then purified with the Monarch PCR kit ( NEB ) . PCR amplicons were digested using MfeI-HF and SphI-HF ( NEB ) and ligated to custom-annealed adaptors with PE2 indexing barcodes and phased P1 barcodes ( Supplementary file 6 ) . The final enrichment PCR used primers MO588 and MO589 ( Supplementary file 6 ) for 20 cycles at an annealing temperature of 66°C ( NEB Phusion ) , followed by purification with the Monarch PCR kit . Polylinker libraries were amplified using primers BC_CRX_Nested_F and BC_CRX_R ( Supplementary file 6 ) for 30 cycles ( NEB Q5 ) at an annealing temperature of 67°C and then purified with the Monarch PCR kit . Illumina adaptors were then added via two further rounds of PCR . First , P1 indexing barcodes were added using forward primers P1_inner_A through P1_inner_D and reverse primer P1_inner_nested_rev ( Supplementary file 6 ) for five cycles at an annealing temperature of 55°C followed by five cycles with no annealing step ( NEB Q5 ) . PE2 indexing barcodes were then added by amplifying 2 µl of the previous reaction with forward primer P1_outer and reverse primers PE2_outer_SIC69 and PE2_outer_SIC70 ( Supplementary file 6 ) for five cycles at an annealing temperature of 66°C followed by five cycles with no annealing step ( NEB Q5 ) and then purified with the Monarch PCR kit . All data processing , statistical analysis , and downstream analyses were performed in Python version 3 . 6 . 5 using Numpy version 1 . 15 . 4 ( Harris et al . , 2020 ) , Scipy version 1 . 1 . 0 ( Virtanen et al . , 2020 ) , and Pandas version 0 . 23 . 4 ( McKinney , 2010 ) , and visualized using Matplotlib version 3 . 0 . 2 ( Hunter , 2007 ) and Logomaker version 0 . 8 ( Tareen and Kinney , 2020 ) . All statistical analysis used two-sided tests unless noted otherwise . Sequencing reads were filtered to ensure that the barcode sequence perfectly matched the expected sequence ( >93% reads in a sample for the Rho libraries , >86% reads for the Polylinker libraries ) . For the Rho libraries , barcodes that had less than 10 raw counts in the DNA sample were considered missing and removed from downstream analysis . Barcodes that had less than five raw counts in any cDNA sample were considered present in the input plasmid pool but below the detection limit and thus set to zero in all samples . Barcode counts were normalized by reads per million ( RPM ) for each sample . Barcode expression was calculated by dividing the cDNA RPM by the DNA RPM . Replicate-specific expression was calculated by averaging the barcodes corresponding to each library sequence . After performing statistical analysis ( see below ) , expression levels were normalized by replicate-specific basal mean expression and then averaged across biological replicates . For the Polylinker assay , the expected lack of expression of many constructs required different processing . Barcodes that had less than 50 raw counts in the DNA sample were removed from downstream analysis . Barcodes were normalized by RPM for each replicate . Barcodes that had less than 8 RPM in any cDNA sample were set to zero in all samples . cDNA RPM were then divided by DNA RPM as above . Within each biological replicate , barcodes were averaged as above but were not normalized to basal expression because there is no basal construct . Expression values were then averaged across biological replicates . Due to the low expression of scrambled sequences and the lack of a basal construct , we were unable to assess data calibration with the same rigor as above . Activity classes were assigned by comparing expression levels to basal promoter expression levels across replicates . The null hypothesis is that the expression of a sequence is the same as basal levels . Expression levels were approximately log-normally distributed , so we computed the log-normal parameters for each sequence and then performed Welch’s t-test . We corrected for multiple hypotheses using the Benjamini-Hochberg FDR procedure . We corrected for multiple hypotheses in each library separately to account for any potential batch effects between libraries . The log2 expression was calculated after adding a pseudocount of 1 × 10–3 to every sequence . Sequences were classified as enhancers if they were twofold above basal and the q-value was below 0 . 05 . Silencers were similarly defined as twofold below basal and q-value less than 0 . 05 . Inactive sequences were defined as within a twofold change and q-value greater than or equal to 0 . 05 . All other sequences were classified as ambiguous and removed from further analysis . We used scrambled sequences to further stratify enhancers into strong and weak enhancers , using the rationale that scrambled sequences give an empirical distribution for the activity of random sequences . We defined strong enhancers as enhancers that are above the 95th percentile of scrambled sequences and all other enhancers as weak enhancers . For the Polylinker assay , we did not have a basal construct as a reference point . Instead , we defined a sequence to have autonomous activity if the average cDNA barcode counts were higher than average DNA barcode counts , and non-autonomous otherwise . The log2 expression was calculated after adding a pseudocount of 1 × 10–2 to every sequence . We obtained RNA-seq data from WT and Crx-/- P21 retinas ( Roger et al . , 2014 ) processed into a counts matrix for each gene by Ruzycki et al . , 2018 . Each sample was normalized by read counts per million and replicates were averaged together . Genes with at least a twofold change between genotypes were considered differentially expressed . We determined which differentially expressed genes are near a member of our library using previously published associations between retinal ATAC-seq peaks and genes ( Murphy et al . , 2019 ) . For de-repressed genes , we determined how often the nearest library member is a silencer; for down-regulated genes , we determined how often the nearest library member is a strong or weak enhancer . We performed motif enrichment analysis using the MEME Suite version 5 . 0 . 4 ( Bailey et al . , 2009 ) . We searched for motifs that were enriched in one group of sequences relative to another group using DREME-py3 with the parameters -mink 6 -maxk 12 -e 0 . 05 and compared the de novo motifs to known motifs using TOMTOM on default parameters . We ran DREME using strong enhancers as positives and silencers as negatives , and vice versa . For TOMTOM , we used version 11 of the full mouse HOCOMOCO database ( Kulakovskiy et al . , 2018 ) with the following additions from the JASPAR human database ( Khan et al . , 2018 ) : NRL ( accession MA0842 . 1 ) , RORB ( accession MA1150 . 1 ) , and RAX ( accession MA0718 . 1 ) . We added these PWMs because they have known roles in the retina , but the mouse PWMs were not in the HOCOMOCO database . We also used the CRX PWM that we used to design the library . Motifs were selected for downstream analysis based on their matches to the de novo motifs , whether the TF had a known role in retinal development , and the quality of the PWM . Because PWMs from TFs of the same family were so similar , we used one TF for each DREME motif , recognizing that these motifs may be bound by other TFs that recognize similar motifs . We did not use any PWMs with a quality of ‘D’ . We excluded DREME motifs without a match to the database from further analysis; most of these resemble dinucleotides . We computed predicted occupancy as previously described ( White et al . , 2013; Zhao et al . , 2009 ) . Briefly , we normalized each letter probability matrix by the most probable letter at each position . We took the negative log of this matrix and multiplied by 2 . 5 , which corresponds to the ideal gas constant times 300 K , to obtain an energy weight matrix . We used a chemical potential μ of 9 for all TFs . At this value , the probability of a site being bound is at least 0 . 5 if the relative KD is at least 0 . 03 of the optimal binding site . We computed the predicted occupancy for every site on the forward and reverse strands and summed them together to get a single value for each TF . To determine if there is a bias in the linear arrangement of motifs , we selected strong enhancers with exactly one site occupied by CRX and exactly one site occupied by a second TF . We counted the number of times the position of the second TF was 5’ and 3’ of the CRX site and then performed a binomial test . We did not observe a statistically significant bias for any TF at an FDR q-value cutoff of 0 . 05 . We also performed this analysis for silencers with exactly one site occupied by CRX and exactly one site occupied by NRL and did not observe a significant difference in the 5’ vs . 3’ bias of strong enhancers vs . silencers ( Fisher’s exact test p = 0 . 17 ) . To capture the effects of TF predicted occupancy and diversity in a single metric , we calculated the motif information content using Boltzmann entropy . Boltzmann’s equation states that the entropy of a system S is related to the number of ways the molecules can be arranged ( microstates ) W via the equation S=kBlogW , where kB is Boltzmann’s constant ( Phillips et al . , 2012 , Chapter 5 ) . The number of microstates is defined as W=N ! ∏iNi ! where N is the total number of particles and Ni are the number of the -th type of particles . In our case , the system is the collection of predicted binding motifs for different TFs in a cis-regulatory sequence . We assume each TF is a different type of molecule because the DNA-binding domain of each TF belongs to a different subfamily . The number of molecular arrangements W represents the number of distinguishable ways that the TFs can be ordered on the sequence . Thus , Ni is the predicted occupancy of the i-th TF and N is the total predicted occupancy of all TFs on the cis-regulatory sequence . Because the predicted occupancies are continuous values , we exploit the definition of the Gamma function , Γ ( N+1 ) =N ! to rewrite W=Γ ( N+1 ) ∏iΓ ( Ni+1 ) . If we assume that each arrangement of motifs is equally likely , then we can write the probability of arrangement w=1 , … , W as pw=1w and rewrite the entropy as S=−log ( 1w ) =−log ( pw ) , where we have dropped Boltzmann’s constant since the connection between molecular arrangements and temperature is not important . Because each arrangement is equally likely , then 1w is also the expected value of pw and we can write the entropy as S=−E[log ( pw ) ]=−∑wpwlog ( pw ) , which is Shannon entropy . By definition , Shannon entropy is also the expected value of the information content: E[I]=−∑wpwlog ( pw ) =∑wpwI ( w ) where the information content I of a particular state is I ( w ) =log ( pw ) . Since we assumed each arrangement is equally likely , then the expected value of the information content is also the information content of each arrangement . Therefore , the information content of a cis-regulatory sequence can be written as I=−log ( pw ) =logW . We use log base 2 to express the information content in bits . With this metric , cis-regulatory sequences with higher predicted TF occupancies generally have higher information content . Sequences with higher TF diversity have higher information content than lower diversity sequences with the same predicted occupancy . Thus , our metric captures the effects of both TF diversity and total TF occupancy . For example , consider hypothetical TFs A , B , and C . If motifs for only one TF are in a sequence , then W is always one and the information content is always zero ( regardless of total occupancy ) . The simplest case for non-zero information content is one motif for A , one motif for B , and zero motifs for C ( 1-1-0 ) . Then W=2 ! 1 ! 1 ! =2 and I=1 bit . If we increase predicted occupancy by adding a motif for A ( 2-1-0 ) , then W=3 ! 2 ! 1 ! =3 and I=1 . 6 bits , which is approximately the information content of silencers and inactive sequences . If we increase predicted occupancy again and add a second motif for B ( 2-2-0 ) , then W=4 ! 2 ! 2 ! =6 and I=2 . 6 bits , which is approximately the information content of strong enhancers . If instead of increasing predicted occupancy , we instead increase diversity by replacing a motif for A with a motif for C ( 1-1-1 ) , then W=3 ! 1 ! 1 ! 1 ! =6 and once again I=2 . 6 bits , which is higher than the lower diversity case ( 2-1-0 ) . According to Wunderlich and Mirny , 2009 , the probability of observing k total motifs for m different TFs in a w bp window is p ( k ) ∼ ( Poisson ( k;λ ) ) , where λ=pmw and p is the probability of finding a spurious motif in the genome . The expected number of windows with k total motifs in a genome of length N is thus E ( k ) =p ( k ) ⋅N . In mammals , N≈109 and Wunderlich and Mirny find that p=0 . 0025 for multicellular eukaryotes . For m=3 TFs and a w=164 bp window ( which is the size of our sequences ) , λ=0 . 123 and E ( 5 ) =1 . 6 meaning that five total motifs for three different TFs specify an approximately unique 164 bp location in a mammalian genome . Five total motifs for three different TFs can be achieved in two ways: three motifs for A , one for B , and one for C ( 3-1-1 ) , or two motifs for A , two for B , and one for C ( 2-2-1 ) . In the case of 3-1-1 , W=5 ! 3 ! 1 ! 1 ! =20 and I=4 . 3 bits . In the case of 2-2-1 , W=5 ! 2 ! 2 ! 1 ! =30 and I=4 . 9 bits . The k-mer SVM was fit using gkmSVM ( Ghandi et al . , 2014 ) . All other machine learning , including cross-validation , logistic regression , and computing ROC and PR curves , was performed using scikit-learn version 0 . 19 . 1 ( Pedregosa et al . , 2011 ) . We wrote custom Python wrappers for gkmSVM to allow for interfacing between the C++ binaries and the rest of our workflow . We ran gkmSVM with the parameters -l 6 -k 6 -m 1 . To estimate model performance , all models were fit with stratified fivefold cross-validation after shuffling the order of sequences . For the TF occupancy logistic regression model , we used L2 regularization . We selected the regularization parameter C by performing grid search with fivefold cross-validation on the values 10–4 , 10–3 , 10–2 , 10–1 , 1 , 101 , 102 , 103 , 104 and selecting the value that maximized the F1 score . The optimal value of C was 0 . 01 , which we used as the regularization strength when assessing the performance of the model with other feature sets . To assess the performance of the logistic regression model , we randomly sampled eight PWMs from the HOCOMOCO database and computed the predicted occupancy of each TF on each sequence . We then fit a new logistic regression model with these features and repeated this procedure 100 times to generate a background distribution of model performances . To generate de novo motifs from the SVM , we generated all 6-mers and scored them against the SVM . We then ran the svmw_emalign . py script from gkmSVM on the k-mer scores with the parameters -n 10 -f 2 -m 4 and a PWM length of 6 , and then used TOMTOM to compare them to the database from our motif analysis . We used our previously published library ( White et al . , 2013 ) as an independent test set for our machine learning models . We defined strong enhancers as ChIP-seq peaks that were above the 95th percentile of all scrambled sequences . There was no basal promoter construct in this library , so instead we defined silencers as ChIP-seq peaks that were at least twofold below the log2 mean of all scrambled sequences . Previously published ChIP-seq data for NRL ( Hao et al . , 2012 ) that was re-processed by Hughes et al . , 2017 and MEF2D ( Andzelm et al . , 2015 ) was used to annotate sequences for in vivo TF binding . We converted peaks to mm10 coordinates using the UCSC liftOver tool and then used Bedtools to intersect peaks with our library . | Different cell types are established by activating and repressing the activity of specific sets of genes , a process controlled by proteins called transcription factors . Transcription factors work by recognizing and binding short stretches of DNA in parts of the genome called cis-regulatory sequences . A cis-regulatory sequence that increases the activity of a gene when bound by transcription factors is called an enhancer , while a sequence that causes a decrease in gene activity is called a silencer . To establish a cell type , a particular transcription factor will act on both enhancers and silencers that control the activity of different genes . For example , the transcription factor cone-rod homeobox ( CRX ) is critical for specifying different types of cells in the retina , and it acts on both enhancers and silencers . In rod photoreceptors , CRX activates rod genes by binding their enhancers , while repressing cone photoreceptor genes by binding their silencers . However , CRX always recognizes and binds to the same DNA sequence , known as its binding site , making it unclear why some cis-regulatory sequences bound to CRX act as silencers , while others act as enhancers . Friedman et al . sought to understand how enhancers and silencers , both bound by CRX , can have different effects on the genes they control . Since both enhancers and silencers contain CRX binding sites , the difference between the two must lie in the sequence of the DNA surrounding these binding sites . Using retinas that have been explanted from mice and kept alive in the laboratory , Friedman et al . tested the activity of thousands of CRX-binding sequences from the mouse genome . This showed that both enhancers and silencers have more copies of CRX-binding sites than sequences of the genome that are inactive . Additionally , the results revealed that enhancers have a diverse collection of binding sites for other transcription factors , while silencers do not . Friedman et al . developed a new metric they called information content , which captures the diverse combinations of different transcription binding sites that cis-regulatory sequences can have . Using this metric , Friedman et al . showed that it is possible to distinguish enhancers from silencers based on their information content . It is critical to understand how the DNA sequences of cis-regulatory regions determine their activity , because mutations in these regions of the genome can cause disease . However , since every person has thousands of benign mutations in cis-regulatory sequences , it is a challenge to identify specific disease-causing mutations , which are relatively rare . One long-term goal of models of enhancers and silencers , such as Friedman et al . ’s information content model , is to understand how mutations can affect cis-regulatory sequences , and , in some cases , lead to disease . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology",
"genetics",
"and",
"genomics"
] | 2021 | Information content differentiates enhancers from silencers in mouse photoreceptors |
Sensory cilia are populated by a select group of signaling proteins that detect environmental stimuli . How these molecules are delivered to the sensory cilium and whether they rely on one another for specific transport remains poorly understood . Here , we investigated whether the visual pigment , rhodopsin , is critical for delivering other signaling proteins to the sensory cilium of photoreceptor cells , the outer segment . Rhodopsin is the most abundant outer segment protein and its proper transport is essential for formation of this organelle , suggesting that such a dependency might exist . Indeed , we demonstrated that guanylate cyclase-1 , producing the cGMP second messenger in photoreceptors , requires rhodopsin for intracellular stability and outer segment delivery . We elucidated this dependency by showing that guanylate cyclase-1 is a novel rhodopsin-binding protein . These findings expand rhodopsin’s role in vision from being a visual pigment and major outer segment building block to directing trafficking of another key signaling protein .
Photoreceptor cells transform information entering the eye as photons into patterns of neuronal electrical activity . This transformation takes place in the sensory cilium organelle , the outer segment . Outer segments are built from a relatively small set of structural and signaling proteins , including components of the classical G protein-coupled receptor ( GPCR ) phototransduction cascade . Such a distinct functional and morphological specialization allow outer segments to serve as a nearly unmatched model system for studying general principles of GPCR signaling ( Arshavsky et al . , 2002 ) and , in more recent years , a model for ciliary trafficking ( Garcia-Gonzalo and Reiter , 2012; Nemet et al . , 2015; Pearring et al . , 2013; Schou et al . , 2015; Wang and Deretic , 2014 ) . Despite our deep understanding of visual signal transduction , little is known about how the outer segment is populated by proteins performing this function . Indeed , nearly all mechanistic studies of outer segment protein trafficking have been devoted to rhodopsin ( Nemet et al . , 2015; Wang and Deretic , 2014 ) , which is a GPCR visual pigment comprising the majority of the outer segment membrane protein mass ( Palczewski , 2006 ) . The mechanisms responsible for outer segment delivery of other transmembrane proteins remain essentially unknown . Some of them contain short outer segment targeting signals , which can be identified through site-specific mutagenesis ( Deretic et al . , 1998; Li et al . , 1996; Pearring et al . , 2014; Salinas et al . , 2013; Sung et al . , 1994; Tam et al . , 2000; 2004 ) . A documented exception is retinal guanylate cyclase 1 ( GC-1 ) , whose exhaustive mutagenesis did not yield a distinct outer segment targeting motif ( Karan et al . , 2011 ) . GC-1 is a critical component of the phototransduction machinery responsible for synthesizing the second messenger , cGMP ( Wen et al . , 2014 ) . GC-1 is the only guanylate cyclase isoform expressed in the outer segments of cones and the predominant isoform in rods ( Baehr et al . , 2007; Yang et al . , 1999 ) . GC-1 knockout in mice is characterized by severe degeneration of cones and abnormal light-response recovery kinetics in rods ( Yang et al . , 1999 ) . Furthermore , a very large number of GC-1 mutations found in human patients causes one of the most severe forms of early onset retinal dystrophy , called Leber’s congenital amaurosis ( Boye , 2014; Kitiratschky et al . , 2008 ) . Many of these mutations are located outside the catalytic site of GC-1 , which raises great interest in understanding the mechanisms of its intracellular processing and trafficking . In this study , we demonstrate that , rather than relying on its own targeting motif , GC-1 is transported to the outer segment in a complex with rhodopsin . We conducted a comprehensive screen of outer segment protein localization in rod photoreceptors of rhodopsin knockout ( Rho-/- ) mice and found that GC-1 was the only protein severely affected by this knockout . Next , we showed that this unique property of GC-1 is explained by its interaction with rhodopsin , which likely initiates in the biosynthetic membranes and supports both intracellular stability and outer segment delivery of this enzyme . These findings explain how GC-1 reaches its specific intracellular destination and also expand the role of rhodopsin in supporting normal vision by showing that it guides trafficking of another key phototransduction protein .
This study was initiated by testing the hypothesis that rhodopsin , by far the most abundant outer segment protein whose proper transport is essential for formation of this organelle , may affect trafficking and abundance of outer segment membrane proteins . This was accomplished by a comprehensive examination of outer segment protein localization in rods of Rho-/- mice . Normal outer segments are cylindrical structures filled with an ordered stack of several hundred membrane discs ( Figure 1A ) . In contrast , Rho-/- rods only develop small ciliary extensions filled with disorganized membrane material ( Figure 1A; Humphries et al . , 1997; Lee et al . , 2006; Lem et al . , 1999 ) . Despite this morphological defect , two outer segment-specific proteins , peripherin and R9AP , have been previously shown to reliably target to this ciliary extension ( Lee et al . , 2006; Pearring et al . , 2014 ) . We broadened this analysis to include the majority of transmembrane outer segment proteins . 10 . 7554/eLife . 12058 . 003Figure 1 . Localization of outer segment membrane proteins in wild-type ( WT ) and Rho-/- retinas . ( A ) Electron micrographs showing the outer segment and connecting cilium in WT and Rho-/- rods ( scale bar 500 nm ) . ( B–K ) Immunofluorescent localization of individual outer segment proteins in WT and Rho-/- retinal cross-sections: ( B ) Rom-1; ( C ) ABCA4; ( D ) guanylate cyclase 2 ( GC-2 ) ; ( E ) cyclic nucleotide gated ( CNG ) α1; ( F ) CNGβ1; ( G ) prominin; ( H ) protocadherin 21 ( PCDH21 ) ; ( I ) peripherin; ( J ) R9AP; and ( K ) GC-1 . ( L ) Double labeling of GC-1 ( green ) and the cone maker , PNA ( magenta ) . Here and in the following figures , the identity of antibodies used in each panel is indicated in ‘Materials and methods’ . Scale bars , 10 μm . Nuclei are stained by Hoechst ( blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12058 . 003 We analyzed ten proteins , whose antibodies have been verified in the corresponding knockout controls . Five of these proteins are components of the phototransduction cascade ( R9AP , GC-1 , GC-2 , CNGα1 , and CNGβ1 ) , two support disc structure ( peripherin and Rom1 ) , one is a membrane lipid flippase ( ATP-binding cassette transporter A4 , ABCA4 ) , and the last two are thought to participate in photoreceptor disc morphogenesis ( protocadherin 21 and prominin ) . All experiments were performed with animals sacrificed on postnatal day 21 when the rudimentary outer segments of Rho-/- rods are fully formed , but photoreceptor degeneration that eventually occurs in these mice remains minimal . Remarkably , nine out of ten proteins were localized specifically to the ciliary extensions of the Rho-/- rods . They included Rom1 , ABCA4 , GC-2 , CNGα1 , CNGβ1 , protocadherin 21 , and prominin ( Figure 1B–H ) , as well as previously reported R9AP and peripherin ( Figure 1I , J ) . A striking exception was GC-1 , which displayed a punctate pattern in the outer segment layer with no distinct signal in rod ciliary extensions ( Figure 1K ) . Further analysis using a cone marker , peanut agglutinin , revealed that the GC-1-positive puncta corresponds to cone outer segments ( Figure 1L; note that cone outer segments in Rho-/- mice are smaller than normal ) . Faint fluorescent signal outside the cone outer segments was indistinguishable from non-specific background in the outer segment layer of GC-1 knockout mice ( Gucy2e-/-; Figure 2 ) . Interestingly , this effect was not reciprocal . As documented in a previous study ( Baehr et al . , 2007 ) , the knockout of GC-1 was not associated with reduction or mislocalization of rhodopsin from rod outer segments ( Figure 2 ) . 10 . 7554/eLife . 12058 . 004Figure 2 . Rhodopsin expression and outer segment localization do not rely on GC-1 . Rhodopsin ( magenta ) and GC-1 ( green ) were co-immunostained in retinal cross-sections from wild-type ( WT ) , GC-1-/- , and Rho-/- mice . Scale bar , 10 μm . Nuclei stained in blue . DOI: http://dx . doi . org/10 . 7554/eLife . 12058 . 004 We then used quantitative Western blotting to measure the amounts of outer segment proteins in the retinas of Rho-/- knockout mice . Availability of suitable antibodies allowed us to analyze eight of the initial ten proteins ( Figure 3 ) . Serial dilutions of retinal lysates from wild-type ( WT ) and Rho-/- mice were run on the same blot ( such as examples in Figure 3A ) and the relative protein amounts were calculated using WT data to generate calibration curves . We found that proteins retaining their normal outer segment localization ( Figure 1 ) were all expressed at 40–80% WT levels ( Figure 3B ) . Considering how small the ciliary extensions of Rho-/- rods are , this amount of protein expression is quite remarkable and suggests a high density of protein packing . 10 . 7554/eLife . 12058 . 005Figure 3 . Quantification of outer segment transmembrane proteins in Rho-/- retinas at P21 . ( A ) Representative Western blots show serial dilutions of wild-type ( WT ) and Rho-/- retinal lysates for three proteins ( guanylate cyclase 1 [GC-1] , GC-2 , and peripherin ) . The fluorescent signal produced by each band in the serial dilution was plotted and used to calculate the amount of each protein in Rho-/- lysate . In these examples , GC-1 was to 10% of its WT content , GC-2 to 38% , and peripherin to 57% . ( B ) Expression levels of outer segment transmembrane proteins in Rho-/- retinal lysates calculated as %WT . A minimum of four independent experiments was performed for each protein . Error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 12058 . 00510 . 7554/eLife . 12058 . 006Figure 3—figure supplement 1 . Transcript levels of GC-1 in the retinas of WT and Rho-/- mice . Quantitative RT-PCR of each transcript was performed on four mice of each genotype at P21 . The relative mRNA expression level in each sample was normalized to the inner retina marker , Thy1 . Error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 12058 . 006 Once again the outlier was GC-1 whose content in Rho-/- retinas was only 10 ± 3% ( SEM , n=5 ) of WT ( Figure 3A , B ) . Considering that ( 1 ) a large fraction of this GC-1 is expressed in cones ( Figure 1L ) ; ( 2 ) cones comprise 3% of mouse photoreceptors; and ( 3 ) cones express more GC-1 than rods ( Dizhoor et al . , 1994 ) , our most conservative estimate is that GC-1 in Rho-/- rods is reduced by at least 95% . This reduction is not caused by a loss of the GC-1 transcript , since qRT-PCR showed that the amount of GC-1 mRNA in Rho-/- retinas was actually twice higher than in WT retinas ( Figure 3—figure supplement 1 ) , perhaps reflecting a feedback mechanism attempting to compensate for the missing GC-1 enzyme . Taken together , the data reported in Figures 1 and 3 demonstrate that GC-1 is unique among outer segment transmembrane proteins in its reliance on rhodopsin for intracellular stability and outer segment localization . We next addressed the mechanistic basis for this phenomenon . We first asked whether the stability and localization of GC-1 in rods require rhodopsin itself or the activity of rhodopsin’s trafficking pathway . The outer segment localization of rhodopsin relies on its C-terminal sequence ( including the VXPX targeting motif ) , which is both necessary and sufficient for rhodopsin delivery ( Deretic et al . , 1998; Li et al . , 1996; Sung et al . , 1994; Tam et al . , 2000 ) . Taking advantage of the fact that adding this sequence to other proteins enables their specific outer segment trafficking ( Tam et al . , 2000; reviewed in Pearring et al . , 2013 ) , we used chimeric proteins to address this question . Two chimeras were generated by exchanging the seven-helical cores and C-terminal sequences between rhodopsin and another GPCR , serotonin receptor Htr6 , whose topology is close to rhodopsin . Although this receptor is not endogenous to rods , we found that both full-length Htr6 and its rhodopsin chimeras robustly express and reliably target to rod outer segments of WT and Rho-/- mice . This is shown in Figure 4 , where these constructs were introduced into rods by the technique of in vivo electroporation . 10 . 7554/eLife . 12058 . 007Figure 4 . Guanylate cyclase 1 ( GC-1 ) rescue in Rho-/- rods requires the seven-helical core structure of rhodopsin . Wild-type ( WT ) and Rho-/- rods were transfected with ( A ) full-length Htr6 , ( B ) seven-helical Htr6 core fused to rhodopsin’s C-terminus , ( C ) seven-helical rhodopsin core fused to the C-terminus of Htr6 , ( D ) full-length rhodopsin . Sections from WT retinas were stained for each recombinant chimera using anti-green fluorescent protein ( GFP ) , anti-myc , or anti-FLAG antibodies ( magenta , each chimera’s tag is depicted in the construct diagram ) . Sections from Rho-/- retinas were co-stained for GFP , myc , or FLAG ( magenta , left panel ) and endogenous GC-1 using anti-GC1 antibodies ( green , middle panel ) . The merged images from Rho-/- sections are shown in the right panel . Scale bar , 10 μm . Nuclei are stained by Hoechst ( blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12058 . 007 Despite its strong expression in outer segments of Rho-/- rods , Htr6 did not restore their GC-1 content ( Figure 4A ) and neither did the chimera containing the Htr6 seven-helical core fused to rhodopsin’s C-terminus ( Figure 4B ) . In contrast , expression of the chimera containing rhodopsin’s core fused to Htr6’s C-terminus did restore GC-1 in the outer segments of transfected rods ( Figure 4C ) , as efficiently as rhodopsin ( Figure 4D ) . Therefore , both intracellular stability and outer segment delivery of GC-1 rely directly on rhodopsin and not on the activity of the ciliary trafficking pathway driven by rhodopsin’s C-terminus . The most straightforward explanation of this finding is that stability and trafficking of GC-1 are being supported by its binding to rhodopsin – a hypothesis tested in the next set of experiments . We investigated whether GC-1 and rhodopsin interact with one another by co-precipitating them from mouse retinal membranes solubilized in a mild detergent , dodecyl maltoside . The experiment in Figure 5A demonstrates that a large fraction of GC-1 can be co-precipitated with rhodopsin using the monoclonal anti-rhodopsin antibody 1D4 . The specificity of this co-precipitation was established by replacing 1D4 with non-immune mouse IgG and by performing the experiment in the presence of the 1D4 epitope-blocking peptide ( Hodges et al . , 1988 ) . Neither rhodopsin nor GC-1 was precipitated under these conditions . We also probed the 1D4 precipitate for the chaperone protein , DnaJB6 ( Figure 5A ) , which was previously shown to link GC-1 to the intraflagellar transport ( IFT ) particle for ciliary transport ( Bhowmick et al . , 2009 ) . Whereas a strong DnaJB6 staining was identified in the retinal lysate , it did not precipitate with the rhodopsin-GC-1 complex . This suggests that , unlike the interaction with rhodopsin , the GC-1 interaction with DnaJB6 is transient . Following a recent report that another rhodopsin-binding protein is peripherin ( Becirovic et al . , 2014 ) , we also probed the rhodopsin precipitate for peripherin . However , no appreciable fraction of peripherin was found in precipitate , both under our experimental conditions and in membranes dissolved in Triton X-100 as in their study . Given that these authors did not show what fraction of total peripherin was precipitating with rhodopsin in their assays , it is hard to fully reconcile these observations , although it is highly unlikely that rhodopsin’s interaction with peripherin could be as prominent as that with GC-1 . 10 . 7554/eLife . 12058 . 008Figure 5 . Guanylate cyclase 1 ( GC-1 ) co-precipitation with rhodopsin from mouse retinal lysate . ( A ) GC-1 and rhodopsin co-precipitation by monoclonal anti-rhodopsin antibody 1D4 . Wild-type ( WT ) mouse retinal lysate ( Input ) was incubated with 1D4 antibody and then bound to protein A/G beads . After the unbound material in flow through ( FT ) was removed , the beads were washed and bound proteins were eluted ( Eluate ) and analyzed by Western blotting for GC-1 , rhodopsin , DnaJB6 , and peripherin . Non-specific protein binding was probed using either non-immune mouse IgG or 1D4 antibody treated with its epitope blocking peptide . ( B ) Co-precipitation of GC-1 and rhodopsin by the 1D4 antibody from retinal membranes solubilized under different detergent conditions . ( C ) Rhodopsin and GC-1 co-precipitation by monoclonal anti-GC-1 antibody 1S4 . Rho+/- mouse retinal lysate ( Input ) was incubated with 1S4 antibody bound to protein A/G beads . After the unbound material in flow through ( FT ) was removed , bound proteins were eluted from the beads ( Eluate ) and analyzed by Western blotting for GC-1 and rhodopsin . Non-specific rhodopsin binding was probed using non-immune mouse IgG . Protein loading on each lane was normalized to input in all panels . DOI: http://dx . doi . org/10 . 7554/eLife . 12058 . 008 Importantly , the efficiency of GC-1 co-precipitation with rhodopsin was highly dependent on the type and concentration of detergent used for membrane solubilization . The co-precipitated fraction of GC-1 diminished when we increased the concentration of dodecyl maltoside , or used Triton X-100 instead ( Figure 5B ) . Such instability of membrane protein complexes in detergent solutions is a common phenomenon ( Prive , 2007 ) , and we believe that this likely hindered identification of GC-1’s interaction with rhodopsin in the past . The reciprocal co-precipitation of rhodopsin by the monoclonal anti-GC-1 antibody 1S4 is shown in Figure 5C . The challenge of this experiment was that rhodopsin in mouse rods is expressed at a 1400-fold molar excess over GC-1 ( Peshenko et al . , 2011b ) . Therefore , the theoretical limit of rhodopsin bound to GC-1 is only 0 . 07% of its total amount , and its detection requires special measures to reduce non-specific rhodopsin binding to the beads . We addressed this issue by employing two strategies . First , we improved the molar ratio between GC-1 and rhodopsin by using retinal lysates from Rho+/- mice , which express twice less rhodopsin without affecting the amounts of other outer segment proteins ( Calvert et al . , 2001; Liang et al . , 2004 ) . Second , we used a minimal amount of beads fully saturated with anti-GC-1 antibodies , just sufficient to precipitate the majority of GC-1 in the lysate . Under these conditions , rhodopsin precipitation with anti-GC-1 antibody exceeded non-specific background with non-immune IgG by over six-fold ( 6 . 5 ± 1 . 7 , n=3 ) . Taken together and combined with our finding that the intracellular stability of GC-1 is critically dependent on the presence of rhodopsin , these data demonstrate that rhodopsin and GC-1 form a complex in photoreceptor cells .
The findings reported in this study expand our understanding of how the photoreceptor’s sensory cilium is populated by its specific membrane proteins . We have found that rhodopsin serves as an interacting partner and a vehicle for ciliary delivery of a key phototransduction protein , GC-1 . This previously unknown function adds to the well-established roles of rhodopsin as a GPCR visual pigment and a major building block of photoreceptor membranes . We further showed that GC-1 is unique in its reliance on rhodopsin , as the other nine proteins tested in this study were expressed in significant amounts and faithfully localized to rod outer segments in the absence of rhodopsin . Our data consolidate a number of previously published observations , including a major puzzle related to GC-1: the lack of a distinct ciliary targeting motif encoded in its sequence . The shortest recombinant fragment of GC-1 which localized specifically to the outer segment was found to be very large and contain both transmembrane and cytoplasmic domains ( Karan et al . , 2011 ) . Our study shows that GC-1 delivery requires rhodopsin and , therefore , can rely on specific targeting information encoded in the rhodopsin molecule . Interestingly , we also found that this information can be replaced by an alternative ciliary targeting sequence from a GPCR not endogenous to photoreceptors . This suggests that the functions of binding/stabilization of GC-1 and ciliary targeting are performed by different parts of the rhodopsin molecule . Our findings also shed new light on the report that both rhodopsin and GC-1 utilize IFT for their ciliary trafficking and co-precipitate with IFT proteins ( Bhowmick et al . , 2009 ) . The authors hypothesized that GC-1 plays a primary role in assembling cargo for the IFT particle bound for ciliary delivery . Our data suggest that it is rhodopsin that drives this complex , at least in photoreceptor cells where these proteins are specifically expressed . Unlike GC-1’s reliance on rhodopsin for its intracellular stability or outer segment trafficking , rhodopsin does not require GC-1 as its expression level and localization remain normal in rods of GC-1 knockout mice ( Baehr et al . , 2007; and this study ) . The outer segment trafficking of cone opsins is not affected by the lack of GC-1 either ( Baehr et al . , 2007; Karan et al . , 2008 ) , although GC-1 knockout cones undergo rapid degeneration , likely because they do not express GC-2 – an enzyme with redundant function . The primary role of rhodopsin in guiding GC-1 to the outer segment is further consistent with rhodopsin directly interacting with IFT20 , a mobile component of the IFT complex responsible for recruiting IFT cargo at the Golgi network ( Crouse et al . , 2014; Keady et al . , 2011 ) . It was also reported that GC-1 trafficking requires participation of chaperone proteins , most importantly DnaJB6 ( Bhowmick et al . , 2009 ) . Our data suggest that GC-1 interaction with DnaJB6 is transient , most likely in route to the outer segment , since we were not able to co-precipitate DnaJB6 with GC-1 from whole retina lysates ( Figure 5 ) . In contrast , the majority of GC-1 co-precipitates with rhodopsin from these same lysates , suggesting that these proteins remain in a complex after being delivered to the outer segment . Although our data do not exclude that the mature GC-1-rhodopsin complex may contain additional protein component ( s ) , our attempts to identify such components by mass spectrometry have not yielded potential candidates . Interestingly , GC-1 was previously shown to stably express in cell culture where it localizes to either ciliary or intracellular membranes ( Bhowmick et al . , 2009; Peshenko et al . , 2015 ) . This strikes at the difference between the composition of cellular components supporting membrane protein stabilization and transport in cell culture models versus functional photoreceptors . The goal of future experiments is to determine whether these protein localization patterns would be affected by co-expressing GC-1 with rhodopsin , thereby gaining further insight into the underlying intracellular trafficking mechanisms . Finally , GC-1 trafficking was reported to depend on the small protein RD3 , which is thought to stabilize both guanylate cyclase isoforms , GC-1 and GC-2 , in biosynthetic membranes ( Azadi et al . , 2010; Zulliger et al . , 2015 ) . In the case of GC-1 , this stabilization would be complementary to that by rhodopsin and potentially could take place at different stages of GC-1 maturation and trafficking in photoreceptors . Another proposed function of RD3 is to inhibit the activity of guanylate cyclase isoforms outside the outer segment in order to prevent undesirable cGMP synthesis in other cellular compartments ( Peshenko et al . , 2011a ) . In summary , this study explains how GC-1 reaches its intracellular destination without containing a dedicated targeting motif , expands our understanding of the role of rhodopsin in photoreceptor biology and extends the diversity of signaling proteins found in GPCR complexes to a member of the guanylate cyclase family . Provided that the cilium is a critical site of GPCR signaling in numerous cell types ( Schou et al . , 2015 ) , it would be interesting to learn whether other ciliary GPCRs share rhodopsin’s ability to stabilize and deliver fellow members of their signaling pathways .
WT C57BL/6J mice were from Jackson Labs ( Bar Harbor , ME ) , WT CD-1 mice were from Charles River , Rho -/- mice ( Lem et al . , 1999 ) were kindly provided by Janis Lem ( Tufts University ) , and fixed eyecups from Gucy2e-/- mice ( Yang et al . , 1999 ) were kindly provided by Shannon Boye ( University of Florida ) . The following antibodies ( with corresponding figures indicated ) were generously provided by: David Garbers , University of Texas Southwestern ( pAb L670 , anti-GC2; Figures 1 , 3 , 5 ) ; Alexander Dizhoor , Salus University ( pAb KHD , anti-RetGC1; Figures 3 , 5 ) ; Wolfgang Baehr , University of Utah ( mAb 1S4 , anti-RetGC1; Figures 1 , 2 , 4 , 5 ) ; Robert Molday , University of British Columbia ( mAb 1D1 PMC , anti-CNGα1; Figure 1 ) ; Steven Pittler , University of Alabama at Birmingham ( pAb , anti-CNGβ1; Figures 1 , 3 ) ; Gabriel Travis , University of California Los Angeles ( pAb , anti-peripherin residues 296–346; Figures 1 , 3 ) ; Jeremy Nathans , Johns Hopkins University ( pAb , anti-protocadherin 21 C-terminus; Figures 1 , 3 ) ; Stefan Heller , Stanford University ( pAb , anti-R9AP residues 144–223; Figures 1 , 3 ) . The polyclonal antibody against Rom-1 was generated in our laboratory ( Gospe et al . , 2011 ) ( Figures 1 , 3 ) . Commercial antibodies were: mAb 1D4 , anti-rhodopsin ( Abcam , Cambridge , MA , Figures 1 , 3 , 5 ) ; pAb , anti-rhodopsin N-terminus ( Sigma , St . Louis , MO; Figure 2 ) ; pAb M-18 , anti-ABCA4 ( Santa Cruz , Dallas , TX; Figure 3 ) ; pAb , anti-ABCA4 C-terminus ( Everest Biotech , Ramona , CA; Figure 1 ) ; mAb 13A4 , anti-prominin ( eBioscience , San Diego , CA; Figure 1 ) ; mAb M2 , anti-FLAG ( Sigma; Figure 4 ) and pAb , anti-FLAG ( Pierce , Grand Island , NY; Figure 4 ) ; pAb , anti-GFP conjugated to Alex Fluor 488 ( Molecular Probes , Grand Island , NY; Figure 4 ) ; pAb 71D10 , anti-Myc-Tag ( Cell Signaling , Danvers , MA; Figure 4 ) ; pAb , anti-DnaJB6 ( Thermo Scientific , Grand Island , NY; Figure 5 ) . Posterior eyecups from C57BL/6J , Rho -/- , or GC1-/- mice were fixed for 1 hr with 4% paraformaldehyde in mouse Ringer’s solution , rinsed three times in Ringer’s , and embedded in 4% UltraPure agarose ( Invitrogen , Grand Island , NY ) . Cross-sections of 100 μm were collected using a vibratome ( Leica Biosystems , Buffalo Grove , IL ) , placed in 24-well plates , and blocked in 5% goat serum and 0 . 5% Triton X-100 in PBS for 1 hr at 22°C . Sections were incubated with primary antibodies in blocking solution overnight at 4°C , rinsed three times , and incubated with goat secondary antibodies conjugated with Alexa Fluor 488 , 568 , or 647 ( Invitrogen ) in blocking solution for 2 hr at 22°C . To stain nuclei , 5 μg/ml Hoechst - ( 33342 , Invitrogen ) was used . To stain mouse cones , 1 μg/ml lectin peptide nucleic acid ( PNA ) conjugated to Alexa Fluor 488 ( Molecular Probes ) was used . Sections were mounted with Immu-Mount ( Thermo Scientific ) and cover-slipped . Images were acquired using a Nikon Eclipse 90i microscope and a C1 confocal scanner controlled by EZ-C1 , version 3 . 10 software . Retinas from C57BL/6J or Rho -/- mice were collected at P21 and sonicated in 250 μl of 2% sodium dodecyl sulfate and 1× cOmplete protease inhibitor mixture ( Roche , Indianapolis , IN ) in phosphate-buffered saline ( PBS ) . Lysates were cleared at 5 , 000 g for 10 min at 22°C . Total protein concentration was measured using the RC DC Protein Assay kit ( Bio-Rad , Hercules , CA ) and serial dilutions of each lysate were subjected to sodium dodecyl sulfate polyacrylamide gel electrophoresis ( SDS-PAGE ) ( with samples not boiled ) . For most proteins , Western blotting was performed using secondary goat or donkey antibodies conjugated with Alexa Fluor 680 or 800 ( Invitrogen ) and bands were visualized and quantified using the Odyssey infrared imaging system ( LiCor Bioscience ) . Bands of PCDH21 and CNGβ1 were visualized using goat or donkey secondary antibodies conjugated with horseradish peroxidase ( HRP ) for enhanced chemiluminescence ( ECL ) detection ( ECL Prime , GE Healthcare , Pittsburgh , PA ) . DNA constructs were electroporated into the retinas of WT CD-1 or Rho-/- neonatal mice ( Matsuda and Cepko , 2004 ) , using a detailed protocol from Gospe et al . , 2011 . The pRho plasmid driving gene expression under the 2 . 2 kb bovine rhodopsin promoter was a gift from Connie Cepko ( Addgene , Cambridge , MA; plasmid # 11156 ) . All DNA constructs were cloned between the 5’ AgeI and 3’ NotI sites in pRho . Full-length mouse rhodopsin and peripherin were cloned from mouse retinal cDNA . A single N-terminal FLAG tag was added to the 5’ end of rhodopsin using overlap extension PCR . The 5-Htr6 serotonin receptor was amplified from a mouse brain cDNA library ( Stratagene , La Jolla , CA ) and the coding sequence for eGFP was fused to its 3’ end . All chimeric constructs were produced using overlap extension PCR . Rhodopsin was split at amino acid 323 , 5-Htr6 was split at amino acid 322 , and the C-terminal 63 amino acids from peripherin were used in the chimera with truncated rhodopsin . Each DNA plasmid ( 4 μg/μl ) was mixed with a construct expressing soluble mCherry ( 2 μg/μl ) for fluorescent identification of electroporated retinal patches . At least four animals yielding consistent results were analyzed for each construct . One C57BL/6J or Rho+/- retina was homogenized in 200 μl of PBS with 1× phosphatase inhibitor cocktail ( PhosSTOP , Roche ) , 1× cOmplete protease inhibitor mixture ( Roche ) , and either n-dodecyl β-D-maltoside or Triton X-100 at desired concentration . Gentle homogenization was performed using a pestle on ice without vortexing or sonication . Lysates were cleared at 100 , 000 g for 20 min at 4°C and 20 μl aliquots were incubated with primary antibodies overnight at 4°C under continuous rotation ( 5 μg of anti-rhodopsin antibody 1D4 , 0 . 2 μg of anti-GC-1 antibody 1S4 , or mouse monoclonal IgG , Santa Cruz ) . For epitope blocking , rhodopsin 1D4 peptide ( AnaSpec , Fremont , CA ) was added to lysate with 1D4 antibody at a final concentration of 2 mM . Protein A/G magnetic beads ( Pierce ) were incubated with the lysate under rotation for 15 min at 22°C; 25 μl of beads were used to precipitate antibodies bound to rhodopsin , while 5 μl of beads were used to precipitate GC-1 . Flow through was collected and beads were washed in 100 μl of the corresponding lysate buffer before being eluted with 20 μl of 2% sodium dodecyl sulfate ( SDS ) in PBS . Finally , 5 μl of 6× sample buffer with 100 mM dithiothreitol ( DTT ) were added to each sample ( input , flow through , eluate ) for SDS-PAGE . Samples were not boiled . Retinas were collected from C57BL/6J or Rho-/- mice at P21 , and RNA was extracted using RNeasy Mini Kit ( Qiagen , Valencia , CA ) and adjusted to 50 ng/μl . cDNA was then synthesized from total RNA using the QuaniTect Reverse Transcriptase kit ( Qiagen ) . GC-1 and Thy1 ( internal control ) primers were designed to PCR across exons 2 and 3 ( GC-1 Fwd ATCCGAGATGGGCCTAGAGT , GC-1 Rev AGCCAGTTCTTCTGCAGCTT , Thy1 Fwd GTCGCTCTCCTGCTCTCAGT , and Thy1 Rev GTTATTCTCATGGCGGCAGT ) . The concentrations of GC-1 and Thy1 in each sample were determined using a standard curve generated by serial dilutions of the corresponding gel-purified PCR products . Real time PCR was performed in a 10 μl reaction volume with 0 . 5 μl of cDNA using the iQ SYBR Green Supermix ( Biorad ) . Thermal cycling and SYBR detection was performed on a CFX96 Real Time System ( Biorad ) . | Our vision begins with light being captured by the rod and cone photoreceptor cells of the retina at the back of the eye . Photoreceptors have large antennae , termed outer segments , which contain specialized proteins that produce electrical signals when stimulated by light . However , it remains mostly unknown how these signaling proteins are delivered specifically to this part of the cell . A light-capturing receptor called rhodopsin is by far the most abundant component of the outer segment and the only one whose transport route through the cell has been well mapped . By examining mice that had been genetically modified to lack rhodopsin , Pearring et al . have now investigated whether the rhodopsin transport pathway also delivers other signaling proteins to the outer segment . These studies revealed that a protein called guanylate cyclase 1 , which makes a messenger molecule inside rod and cone cells , is the only outer segment protein whose delivery to the outer segment and stability inside the cell rely on rhodopsin . Guanylate cyclase literally ‘piggybacks’ on rhodopsin on its route to the outer segment , as these proteins bind to one another . Therefore , rhodopsin is not just a light-sensing receptor protein; it also serves as a “trafficking guide” for another key protein in the same signaling pathway . As Pearring et al . have shown that the majority of outer segment proteins are delivered independently of rhodopsin , future studies will need to search for alternative protein transport pathways in photoreceptor cells . Whether receptor molecules other than rhodopsin stabilize and deliver fellow members of their signaling pathways to specific cell structures also remains to be discovered . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Guanylate cyclase 1 relies on rhodopsin for intracellular stability and ciliary trafficking |
Microtubule dynamics facilitate neurite growth and establish morphology , but the role of minus-end binding proteins in these processes is largely unexplored . CAMSAP homologs associate with microtubule minus-ends , and are important for the stability of epithelial cell adhesions . In this study , we report morphological defects in neurons and neuromuscular defects in mutants of the C . elegans CAMSAP , ptrn-1 . Mechanosensory neurons initially extend wild-type neurites , and subsequently remodel by overextending neurites and retracting synaptic branches and presynaptic varicosities . This neuronal remodeling can be activated by mutations known to alter microtubules , and depends on a functioning DLK-1 MAP kinase pathway . We found that PTRN-1 localizes to both neurites and synapses , and our results suggest that alterations of microtubule structures caused by loss of PTRN-1 function activates a remodeling program leading to changes in neurite morphology . We propose a model whereby minus-end microtubule stabilization mediated by a functional PTRN-1 is necessary for morphological maintenance of neurons .
In animals , stable neuronal morphology is necessary to maintain synaptic architecture . However , morphological alterations are also critical for plastic changes in synaptic strength , pruning of synaptic connections , and occasional responses to injury or other cues . Morphological reorganization of neurons is known to occur as a compensatory mechanism in cases of sight or hearing deprivation ( Karlen et al . , 2006; Barone et al . , 2013 ) , and medium spiny neurons are reported to experience reversible morphology changes in response to mild stress ( Bessa et al . , 2013 ) . Neurons in D . melanogaster and C . elegans , as well as many mammalian peripheral neurons respond to injuries by remodeling their morphology including the growth of neurites , elimination of existing synapses , and establishment of new synapses ( Hilliard , 2009 ) . The study of neuronal remodeling after injury in model organisms has elucidated cellular pathways responsible for morphological changes , however , little is known about the molecular cues that initiate these cascades . The conserved mitogen-activated protein ( MAP ) kinase kinase kinase DLK-1 functions to transmit regenerative signals in C . elegans , and is a component of regenerative signaling in organisms with more complex nervous systems ( Hammarlund et al . , 2009; Xiong et al . , 2010; Shin et al . , 2012 ) . While there is a wealth of information about conserved down-stream signals affected by DLK-1 activity , very little information exits about how DLK-1 itself is activated ( Tedeschi and Bradke , 2013 ) . Recently , calcium spikes were shown to activate C . elegans DLK-1 , but fly and mammalian homologs lack this calcium-binding domain ( Yan and Jin , 2012 ) . It is known that the conserved E3 ubiquitin ligase RPM-1 ( PHR , PAM , HIGHWIRE ) keeps levels of DLK-1 low ( Nakata et al . , 2005; Xiong et al . , 2010 ) . Axonal injury is reported to cause microtubule disassembly and disorganization of higher-order microtubule structures ( Erturk et al . , 2007 ) . Microtubule depolymerization was recently found to activate DLK-1 signaling . However , this study did not investigate in what context ( development , regeneration , or degeneration ) this microtubule-based mechanism might be functioning ( Bounoutas et al . , 2011 ) . Neurons contain both very stable and dynamic microtubule structures . These populations are maintained by the activities of cytoskeletal-associated proteins and by post-translational modifications ( Conde and Caceres , 2009 ) . Proteins known to be involved in stabilizing and organizing microtubule structures include molecular motors , and cross-linkers , as well as microtubule nucleating and severing proteins , and proteins involved in regulating plus-end dynamics ( reviewed in Subramanian and Kapoor , 2012 ) . Post-translational modifications to microtubules include acetylation , tyrosination , glutamylation , glycylation , and phosphorylation . These modifications affect the recruitment of microtubule-associated proteins and denote subcellular populations of microtubule structures such as the tyrosinated microtubules of dynamic growth cones or the acetylated microtubules of stable neurites ( Baas and Black , 1990; Brown et al . , 1993; Janke and Bulinski , 2011; Garnham and Roll-Mecak , 2012 ) . In C . elegans , microtubule acetylation mutants display short microtubules , alterations in microtubule protofilament number , and association of filaments into higher-order structures ( Cueva et al . , 2012; Topalidou et al . , 2012 ) . Furthermore , microtubule de-acetylation was recently shown to be an early signal for the development of growth-cones in neuronal remodeling associated with regeneration ( Cho and Cavalli , 2012 ) . It is currently unclear how destabilizing modifications or depolymerization of microtubule structures could activate neuronal remodeling . Depolymerization events that occur in dynamic microtubule populations clearly do not cause neurite remodeling in healthy neurons , so it is not likely that depolymerization itself or free microtubule monomers are the remodeling signal . One possibility is that the normally stable minus-ends of microtubules initiate remodeling signals . In this study , we report that genetic lesions in ptrn-1 , the C . elegans homolog of the microtubule minus-end binding calmodulin-regulated spectrin-associated ( CAMSAP ) family cause neurite remodeling including overextension of neurites and retraction of collateral presynaptic varicosities and branches . These remodeling events require DLK-1 , a known player in morphology changes associated with degenerative and regenerative signaling in neurons . This is the first report of a CAMSAP family member being involved in remodeling of neurite morphology . Members of the CAMSAP family are known to associate with microtubule minus ends and stabilize microtubule structures ( Goshima et al . , 2007; Meng et al . , 2008; Baines et al . , 2009; Goodwin and Vale , 2010; Tanaka et al . , 2012 ) . CAMSAPs are found in mammalian neuronal tissue ( Baines et al . , 2009; Guo et al . , 2012; Yamamoto et al . , 2009 ) , and genetic variants of human CAMSAP1L1 were recently shown to confer susceptibility to epilepsy ( Guo et al . , 2012 ) . Thus understanding the neuronal consequences of disruptions to CAMSAP function may begin to inform our understanding of disease mechanisms . We found that disruptions in ptrn-1 cause alterations to neuronal morphology , and these changes occur via a DLK-1 neuronal remodeling program . We were able to mimic and enhance ptrn-1 loss-of-function phenotypes with genetic and pharmacological alterations of the microtubule cytoskeleton . Furthermore , mutation of dlk-1 suppressed a microtubule acetylation mutant supporting the idea that altering microtubule structures is an upstream signal for DLK-1 activation . PTRN-1 is present in neurites and at sites of synaptic connectivity , suggesting local microtubule minus-end regulation may affect these mechanisms . In addition to identifying a novel function for CAMSAPs in the maintenance of neuron morphology , our results suggest that alterations to microtubule structures caused by loss of CAMSAP function activates a neuronal remodeling program . We propose a model whereby local minus-end microtubule stabilization mediated by CAMSAPs is necessary for the stabilization of mature neurite morphology .
We use C . elegans touch receptor neurons ( TRNs ) to study establishment and maintenance of neuronal morphology . The laterally positioned TRN neurons each extend a neurite in which touch receptors are distributed at regular intervals ( Zhang et al . , 2004 ) . Together PLM and ALM lateral neurites cover the length of the animal , but they do not overlap ( Figure 1A ) . ALM and PLM chemical synapses are on collateral branches that terminate in presynaptic varicosities . Each PLM mechanosensory neuron forms a single presynaptic varicosity in the ventral nerve cord just posterior to the vulva ( Figure 1A ) . Thus , the lateral TRNs have a highly stereotyped morphology that allows visualization of individual neurons in living animals . 10 . 7554/eLife . 01637 . 003Figure 1 . ptrn-1 mutant neurite overextension and collateral branch retraction phenotypes . ( A ) Schematic diagram of an L4 C . elegans hermaphrodite showing the morphology of the touch receptor neurons . Neurons on the left side are in black , and those on the right side are in gray . ( B ) Morphology of mechanosensory neurons in wild-type and ptrn-1 mutants , as seen by expression of mRFP under the control of the mec-7 promoter . Defects in ptrn-1 animals include PLM neurite overextension and ectopic formation of presynaptic varicosities ( * ) , retraction of PLM collateral presynaptic varicosities and branches ( ** ) . Scale bar = 10 µm . Additional examples and variance in Ptrn-1 phenotypes are presented in Figure 1—figure supplement 1 . ( C ) Quantification of the PLM neurite overextension phenotype at 22°C . Both ptrn-1 alleles ( js1286 and tm5597 ) show similar defects , as do trans-heterozygotes . Also shown is the rescue of the PLM neurite overextension phenotype by expression of PTRN-1a-mCherry under a native promoter and in neurons under a rab-3 promoter , but not under a myo-3 promoter . Non-transgenic siblings were identified as individuals from mothers segregating the transgene , but that lacked detectable mCherry fluorescence . n = 50–60; error bars = SEP . ( D ) Aberrant formation and extension of neurites ( * ) was found in D-type motor neurons . Quantification is in Figure 1—figure supplement 2 . ( E ) Aberrant formation and extension of neurites ( * ) was found in command interneurons . Quantification is in Figure 1—figure supplement 2 . ( F ) Quantification of locomotory defects using L4 animals . Error bars represent standard deviation . N = 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 01637 . 00310 . 7554/eLife . 01637 . 004Figure 1—figure supplement 1 . Variance in the ptrn-1 mutant TRN phenotypes . Midbody of wild-type and mutant L4 animals expressing mRFP in TRNs under the mec-7 promoter ( jsIs973 ) showing the wild-type and ptrn-1 mutant morphology of PLM and PVM , as well as the ALM soma . Scale bar = 10 µm . ( A ) The wild type . ( B ) ptrn-1 ( js1286 ) animals showing PLM neurite overextensions ( * ) , ectopic synaptic varicosities ( ^ ) , loss of collateral branches and associated PLM synaptic varicosities ( ** ) , and secondary posterior-directed ALM processes ( double arrows ) are noted . ( C ) ptrn-1 ( tm5597 ) animals . ( D ) Temperature-dependence of PLM overextension in ptrn-1 mutants . n = 60–80 animals . Error-bars = SEP . ( E ) Variance in penetrance of the overextension defect is less than 10% in ptrn-1 mutants . n=71–94 , error bars = SEP . DOI: http://dx . doi . org/10 . 7554/eLife . 01637 . 00410 . 7554/eLife . 01637 . 005Figure 1—figure supplement 2 . Quantification of other morphology phenotypes in ptrn-1 mutants . ( A ) Quantification of D-type motor neuron defects visualized using an oxIs12 [Punc-47::GFP] transgene . Error bars represent SEP . N=60 . ( B ) Quantification of command interneuron defects visualized using nuIs25 [Pglr-1::GFP] transgene . Error bars represent SEP . N = 65–68 . DOI: http://dx . doi . org/10 . 7554/eLife . 01637 . 005 We screened 1134 haploid genomes looking for maintenance defects in the presynaptic varicosities of PLM mechanosensory neurons . Three candidates from this screen remain uncharacterized , the fourth , js1286 , is the focus of this work . We identified js1286 as a nonsense mutation in the C . elegans CAMSAP , ptrn-1 , that caused a range of partially penetrant phenotypes in PLM and ALM neurons ( Figure 2A , B , Figure 1—figure supplement 1 ) . Defects seen in PLM neurons included loss of collateral branches and presynaptic varicosities , as well as overextended neurites that exited the lateral cord and formed ectopic presynaptic varicosities in the ventral nerve cord anterior to the vulva ( Figure 1B , C ) . We observed the same phenotypes in an independently generated deletion allele , ptrn-1 ( tm5597 ) ( Figure 2A , 1C , Figure 1—figure supplement 1 ) , indicating the observed defects in ptrn-1 are representative of the ptrn-1 null phenotype . We also observed morphological defects in motor neurons and command interneurons , and behavioral defects in locomotion indicating that PTRN-1 functions broadly in the nervous system ( Figure 1D , E , Figure 1—figure supplement 2 ) . 10 . 7554/eLife . 01637 . 006Figure 2 . Gene and protein structure of C . elegans ptrn-1 . ( A ) Exon ( thick regions ) and intron ( thin regions ) organization of the ptrn-1 locus . The position of structural domains is demarcated above and the position of the genetic lesion for each of the ptrn-1 alleles is shown below . ptrn-1b differs from ptrn-1a by the inclusion of 6 bp ( 2 amino acids ) at the exon 10–11 boundary ( * ) . Nucleotide sequences for ptrn-1a and ptrn-1b are listed in Figure 2—figure supplements 1 and 2 , respectively . Omitted from the figure is a distinct ptrn-1 isoform lacking the C-terminal half of the protein , which is not represented in RNA-seq data for which the sole support is a single compound EST yk1268c03 . ( B ) The CAMSAP protein family consists of a calponin-homology domain followed by one or two 30–40 amino acid coiled-coil regions , a 70–85 amino acid coiled-coil region and a CKK microtubule-binding domain . CAMSAP sequences from C . elegans ( this paper ) , D . melanogaster ( AAO41362 . 1 ) , D . rerio ( NP_001159727 . 1 ) , and H . sapiens ( AAI25231 . 1 ) were used for homology analysis . Domains were annotated by running sequences through the NCBI Conserved Domain Database . Coiled-coil regions were defined as those predicted by both the Paircoil2 and COILS programs . DOI: http://dx . doi . org/10 . 7554/eLife . 01637 . 00610 . 7554/eLife . 01637 . 007Figure 2—figure supplement 1 . Nucleotide sequence of ptrn-1a . DOI: http://dx . doi . org/10 . 7554/eLife . 01637 . 00710 . 7554/eLife . 01637 . 008Figure 2—figure supplement 2 . Nucleotide sequence of ptrn-1b . DOI: http://dx . doi . org/10 . 7554/eLife . 01637 . 008 Changes in neuronal morphology could be caused by cell autonomous alterations or by changes in surrounding tissues . We noticed that ptrn-1 animals have uncoordinated locomotion and body positioning , a characteristic shared by mutants with defects in either neurons or muscles . Prior work has focused mostly on the role of CAMSAPs in epithelial tissue; however by expressing a ptrn-1 transcriptional reporter construct , we found that ptrn-1 is broadly expressed in the nervous system ( Figure 3 ) . Driving ptrn-1:mcherry under the pan-neuronal ( rab-3 ) promoter , but not the pan-muscle ( myo-3 ) promoter rescued PLM morphology defects indicating that PTRN-1 functions in neurons to regulate PLM morphology ( Figure 1C ) . 10 . 7554/eLife . 01637 . 009Figure 3 . C . elegans ptrn-1 is broadly expressed in neuronal tissue . ( A ) ptrn-1 expression pattern from a transcriptional fusion between mCherry and a ptrn-1 promoter segment sufficient to rescue Ptrn-1 TRN cellular phenotypes ( see Figure 1C for rescue ) . 0 . 8 kb was chosen as the promoter fragment because of the presence of a 1 . 2 kb Mariner transposable element at this position 5′ of the ptrn-1 ATG . Scale bars = 10 µm . ( B–D ) Views at higher magnification of the head neuronal ganglion ( B ) , ventral nerve cord ( C ) , and pre-anal and lumbar neuronal ganglia ( D ) . mCherry is primarily nuclear localized . DOI: http://dx . doi . org/10 . 7554/eLife . 01637 . 00910 . 7554/eLife . 01637 . 010Figure 3—figure supplement 1 . Close-up view of the ptrn-1 expression pattern . ( A ) ptrn-1 expression pattern from a transcriptional fusion between mCherry and a ptrn-1 promoter segment sufficient to rescue Ptrn-1 TRN cellular phenotypes ( see Figure 1C for rescue ) . View of the head neuronal ganglion . ( B–E ) High-magnification views of individual neuronal cell bodies showing expression of the mCherry reporter construct . Scale bars = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01637 . 010 The ptrn-1 gene represents the only C . elegans ortholog of the CAMSAP family . The domain structure for CAMSAPs consists of an N-terminal calponin-homology domain , a C-terminal microtubule-binding domain and an intervening region of predicted coiled-coil domains ( Figure 2B ) . CAMSAPs regulate microtubule minus-end depolymerization , and are important for the stability of epithelial cell adhesions ( Goodwin and Vale , 2010; Meng et al . , 2008 ) . CAMSAPs localize to microtubule minus ends , and in puncta distributed along higher-order microtubule structures ( Baines et al . , 2009; Goodwin and Vale , 2010; Meng et al . , 2008; Tanaka et al . , 2012 ) . Consistent with these findings , we see that both PTRN-1-cherry and GFP-PTRN-1 constructs exhibit a punctate distribution along neurites and at sites of high synaptic connectivity including the nerve-ring and ventral nerve cord ( Figure 4 ) . In neuronal cell bodies , PTRN-1-mcherry and GFP-PTRN-1 could be found in one or two puncta ( Figure 4A–C , E–G ) , and colocalization of GFP-PTRN-1 with the mechanosensory-specific ß-tubulin MEC-7 was found ( Figure 4H–I ) . These results are consistent with C . elegans PTRN-1 functioning in neurons . 10 . 7554/eLife . 01637 . 011Figure 4 . PTRN-1-localization pattern . Expression of PTRN-1-mcherry under a ptrn-1 0 . 8 kb promoter segment sufficient to rescue Ptrn-1 TRN cellular phenotypes ( A–C ) ( see Figure 1C for rescue data ) . Enlarged views of neurites ( A1 ) , a head ganglion cell body ( A2 ) , and a tail neuron with isolated morphology ( C1 ) are shown . Puncta ( arrows ) , nuclear exclusion ( * ) , and neurite ( double arrow ) are noted . Single-cell resolution showing a N-terminal GFP-PTRN-1 fusion protein under the control of the TRN mec-7 promoter in ALM and PLM neurons ( D–G ) . Puncta ( arrows ) and neurites ( double arrows ) are noted . Scale bars = 10 µm . Whole-mount immunofluorescent staining with anti-GFP ( H ) and anti-MEC-7 ( I ) antibodies . Co-localization of GFP-PTRN-1 with MEC-7 in mechanosensory neurons is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 01637 . 011 We wanted to distinguish between defects that arise during development , and those that are a result of a failure to maintain or stabilize normally patterned neurons . Although ALM neurites extend during embryonic development , the development of PLM neurons occurs in several distinct steps some of which occur during the L1 larval stage ( Figure 5A ) . After extending lateral neurites in the embryo , PLMs form collateral branches early in the L1 larval stage , and subsequently active zone synaptic machinery , vesicles , and mitochondria are recruited successively to nascent collateral presynaptic varicosities . The overall morphological pattern is maintained as the animal quadruples its body length from the L1 to L4 larval stage , with the only substantive change being a ten-fold increase in the size of presynaptic varicosity . In ptrn-1 animals , we noticed that wild-type morphology is lost after the initial L1 patterning , with animals missing presynaptic varicosities and overextending neurites ( Figure 5B ) . A small fraction of ptrn-1 animals had overextended neurites and/or missing presynaptic varicosities at the L1 stage ( Figure 5B ) ; however from this data set , we are unable to determine if these animals originally formed morphologically normal neurites and experienced early retraction/overextension defects prior to the L1 time-point or if they represent a partially penetrant role for ptrn-1 in initial L1 patterning . We also examined individual animals at both the L1 and L4 larval stages to confirm that aberrant morphology phenotypes do occur subsequent to L1 developmental patterning ( Figure 5C ) . Ectopic presynaptic varicosities that formed at the terminus of overextended neurites in L1 larva were also subject to retracting; further supporting the idea that ptrn-1 is required for maintenance , rather than initial establishment of neurite morphology . 10 . 7554/eLife . 01637 . 012Figure 5 . ptrn-1 mutants fail to stabilize mature PLM morphology and synaptic contacts . ( A ) Schematic diagram illustrating wild-type PLM neuron developmental changes during the L1 larva from 0 to 12 hr after hatching , as well as the structure of PLM neurons in L4 larva—note that the figure is not drawn to scale . PLM neurons maintain their relative position within the animal from L1 to L4 , but L4 are approximately four times longer than L1 animals . ( B ) Changes observed in PLM neuronal morphology in wild-type and ptrn-1 ( js1286 ) mutants from a synchronized population of animals analyzed at L1 ( 10 hr ) and L4 ( 36 hr ) stages . Animals were grown at 22°C . PLM neurons show both overextension and ventral targeting of the anterior neurite in ptrn-1 mutants , but rarely in wild type . The ventral targeting posterior neurites display variable sized varicosities that are labeled with presynaptic components . ptrn-1 animals also often exhibit loss of collateral branch and associated presynaptic structures . N = 90–98 neurites . ( C ) Retraction of collateral branch synaptic varicosities and overextended ectopic ventral neurite targeted varicosities in animals imaged as both L1 and L4 larva . N = 11 and 68 neurites . ( B and C ) Black lines and circles represent neurites and presynaptic varicosities respectively . Gray lines and open circles represent variability in ptrn-1 animals . DOI: http://dx . doi . org/10 . 7554/eLife . 01637 . 012 The average size of presynaptic varicosities that persist in ptrn-1 animals not significantly different from wt ( 10 ± 3 μM2 for ptrn-1 compared to 12 ± 3 μM2 for wild-type N = 20; p=0 . 06 ) . While presynaptic varicosity retraction is partially penetrant , once initiated , retraction goes to completion . Thus , maintenance of presynaptic varicosities seems to be subject to a binary switch , and this suggested to us that loss of ptrn-1 might result in initiation or transmission of a retraction signal . The Ptrn-1 phenotypes of neurite overextension and the absence of collateral presynaptic varicosities are reminiscent of defects in C . elegans caused by loss of function mutations in the ubiquitin ligase rpm-1 , a known regulator of neurite remodeling ( Schaefer et al . , 2000; Zhen et al . , 2000 ) . Loss of rpm-1 results in activation of the DLK-1 MAP kinase pathway and rpm-1 phenotypes are DLK-1 dependent ( Nakata et al . , 2005 ) . Because of the phenotypic similarities between rpm-1 and ptrn-1 mutants , we wondered if dlk-1 is required for the remodeling events in ptrn-1 mutants . We found that both rpm-1 and ptrn-1 L4 animals have similar PLM morphological defects , including neurite overextensions that target the ventral nerve cord , and the absence of collateral presynaptic varicosities ( Figure 6A , B ) . By contrast , dlk-1 loss-of-function mutants exhibit grossly normal PLM morphology . Furthermore , like rpm-1 , we found that ptrn-1 presynaptic varicosity retraction and overextension defects are suppressed by loss of dlk-1 , as well as in a triple-mutant background ( Figure 6A , B ) . However , we noticed that in double and triple mutants usually one or both of the varicosities were smaller and elongated ( Figure 6B ) , indicating additive roles for dlk-1 , ptrn-1 and rpm-1 in shaping the synapse . Our data show that ptrn-1 functions upstream of dlk-1 , and indicates that loss of PTRN-1 activates a C . elegans neuronal remodeling program . 10 . 7554/eLife . 01637 . 013Figure 6 . dlk-1 MAP kinase mutants suppress Ptrn-1 overextension phenotypes . ( A ) Single , double , and triple mutant combinations of ptrn-1 ( tm5597 ) , dlk-1 ( km12 ) and rpm-1 ( ok364 ) were grown at 22°C , scored as L4 larvae for PLM neurite morphology defects using jsIs973 ( Pmec-7:mRFP ) , and for the presence of collateral branch presynaptic varicosities using jsIs821 ( Pmec-7::GFP-RAB-3 ) . Since PVM forms small GFP puncta in the ventral nerve cord ( see Figure 1A , B ) , PLM collateral varicosities which retained a small GFP-RAB-3 puncta in the ventral nerve cord cannot be unambiguously distinguished from PLM collaterals completely lacking presynaptic components . n = 200 PLM cells from 100 animals . ( B ) Midbody of L4 animals expressing mRFP in TRNs under Pmec-7 ( jsIs973 ) showing the wild-type and mutant PLM morphology associated with ptrn-1 and dlk-1 . Elongated varicosity in ptrn-1 , rpm-1 , dlk-1 triple mutant identified with double arrows . Scale bar = 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01637 . 013 Remodeling pathways may be activated through a microtubule-based mechanism in ptrn-1 mutants . Recently , microtubule depolymerization was reported to activate DLK-1 , and ß-tubulin mutants have a reduced capacity to regenerate despite the fact that neurite extension appears wild type in these animals ( Bounoutas et al . , 2011; Kirszenblat et al . , 2013 ) . PTRN-1 homologs have been shown to associate with microtubule minus ends and to regulate microtubule dynamics in ex vivo and in vitro model systems ( Baines et al . , 2009; Goodwin and Vale , 2010; Goshima et al . , 2007; Meng et al . , 2008; Tanaka et al . , 2012 ) . It is therefore possible that loss of ptrn-1 alters microtubule structures , initiating a DLK-1-based remodeling program . Because investigating neuronal microtubule dynamics in C . elegans neurons in vivo is technically difficult , we explored a link with microtubule structures by looking for Ptrn-1 phenotypes using known genetic and pharmacological manipulations that perturb the microtubule cytoskeleton . We monitored PLM neurite overextension after exposing worms to the microtubule-poisoning drug colchicine . Colchicine is known to destabilize microtubules ( Gigant et al . , 2009 ) , and while colchicine treatment did not cause regenerative phenotypes in wild-type worms , it caused these phenotypes to be more penetrant in ptrn-1 mutants ( Figure 7A ) . Because the effects of colchicine treatment were additive in combination with ptrn-1 mutants , it is possible that loss of PTRN-1 affects microtubule structures , similar to CAMSAPS in other systems . However , we cannot rule out the possibility that although ptrn-1 mutants are more susceptible to sub-threshold amounts of colchicine-induced perturbation , the effect is independent of direct ptrn-1 alterations in microtubule structures . 10 . 7554/eLife . 01637 . 014Figure 7 . Manipulating the microtubule cytoskeleton leads to terminal neurite extension and ventral nerve cord targeting . ( A ) Wild-type and ptrn-1 ( js1286 ) mutants were grown for two generations on plates containing 1 mM colchicine and scored for PLM neurite overextension ( n = 80 ) . Error bars represent SEP . ( B ) Wild-type , ptrn-1 mutants and mutants with lesions in microtubule acetylases ( paralogs mec-17 and atat-2 ) were scored for PLM neurite overgrowth defects at various temperatures ( n = 55-60 ) . Error bars represent SEP . ( C ) dlk-1 suppresses the PLM neurite overextension phenotypes of mec-17 ( n = 50 ) . Error bars represent SEP . DOI: http://dx . doi . org/10 . 7554/eLife . 01637 . 014 We also found that colchicine treatment suppressed the formation of ALM posterior neurites ( data not shown ) , as has been reported when these structures have occurred in other mutants ( Kirszenblat et al . , 2013; Topalidou et al . , 2012; van Zundert et al . , 2002 ) . While abnormal lateral extension of ALM and PLM neurites occurs in a variety of mutants ( Kirszenblat et al . , 2013; Mohamed et al . , 2012; Topalidou et al . , 2012; Tulgren et al . , 2011 ) , the Ptrn-1 PLM phenotype , where overextended neurites migrate to the ventral nerve cord just anterior to the vulva has only been found in mutants that undergo neurite remodeling . Microtubule acetylation mutants alter microtubules and are reported to have these ALM secondary posterior neurites ( Cueva et al . , 2012; Topalidou et al . , 2012 ) . We asked if disruptions caused by altered microtubule acetylation would be sufficient to cause remodeling phenotypes . We found that the microtubule acetylation mutant mec-17 caused PLM neurite overextension; furthermore , like ptrn-1 mutants , this phenotype was suppressed by growing the worms in cooler temperatures , and in a mec-17 , dlk-1 double mutant ( Figure 7B , C ) . These results indicate that altering the microtubule cytoskeleton can activate a DLK-1 remodeling program . We also saw absence of collateral presynaptic varicosities in mec-17 mutant animals , however this phenotype was additive in mec-17 , dlk-1 double mutants indicating that mec-17 and dlk-1 have parallel presynaptic roles ( data not shown ) . Our work using genetic and pharmacological manipulations of the cytoskeleton supports the idea that altering microtubules activates neuronal remodeling programs , and suggests that the CAMSAP PTRN-1 acts in this pathway . Similar to the function of CAMSAPs in other systems , the neuronal function of CAMSAP proteins may involve the regulation of microtubules or microtubule structures .
Alterations in neuronal morphology are associated with human developmental and neurodegenerative diseases ( Newey et al . , 2005; Lin et al . , 2009; Liu , 2011; Goellner and Aberle , 2012; Yu and Lu , 2012 ) and it is therefore important to understand the cellular mechanisms that underlie these changes . The conserved DLK-1 kinase cascade promotes changes in morphology associated with neurite remodeling ( Hammarlund et al . , 2009 ) , but how this cascade is activated has remained elusive . Our data show that loss of the conserved CAMSAP PTRN-1 induces DLK-mediated remodeling phenotypes . Because CAMSAP family members are known to associate with microtubule minus ends and regulate microtubule structures ( Goodwin and Vale , 2010; Meng et al . , 2008; Tanaka et al . , 2012 ) , we propose a model where microtubule structures act as upstream sensors to neuronal programs that lead to synapse elimination and growth of neurites . How does loss of PTRN-1 initiate neuronal remodeling ? PTRN-1 could itself be a part of the machinery that represses the DLK-1 pathway , or it could be acting indirectly through the sequestration of a remodeling sensor . Because we can enhance defects in ptrn-1 mutants with the microtubule-binding drug colchicine , and mimic Ptrn-1 phenotypes with microtubule acetylation mutants , PTRN-1 is not likely to be the signal itself . We favor a model where perturbations of the microtubule cytoskeleton are responsible for activation of remodeling in ptrn-1 mutants . Interestingly , recent findings indicate that at cooler temperatures , neuronal microtubules incorporate microtubule monomers that have been stabilized through post-translational modification ( Song et al . , 2013 ) . It is possible that the rescue we see of Ptrn-1 defects by growing animals in cooler temperatures is due to this mechanism; by incorporating cold-stable monomers , microtubule structures may be less sensitive to cytoskeletal alterations caused by ptrn-1 or acetylation mutants . Altered association of microtubule filaments into networks may initiate neurite remodeling rather than disruption of individual microtubules . Separation of microtubule filaments from organized cytoskeletal networks has been reported in experiments that decrease expression of PTRN-1 homologs and in microtubule acetylation mutants ( Goodwin and Vale , 2010; Topalidou et al . , 2012 ) . Such alterations would cause the release of factors associated with tethering connections between microtubule filaments , as well as the appearance of free minus ends , either of which could be a signal that a remodeling response is needed . In ptrn-1 mutants , regenerative changes in neuron morphology occurred after developmental patterning of the nervous system . Potential explanations for this observation include that regenerative DLK-1 signaling pathways only become competent to respond to activating cues after development is complete or that the regenerative cue appears later , perhaps because CAMSAPs stabilize microtubule structures present only in mature neurons . One possibility is that CAMSAPs are involved in microtubule-based sequestration of a RPM-1 inhibitor . RPM-1 is known to keep levels of DLK-1 low by targeting it for degradation , and this mechanism must be surmounted in ptrn-1 mutants that have a functional rpm-1 gene . Alternatively , if DLK-1 or a DLK-1 activator is being sequestered , then perhaps a build-up of these stores over time overwhelms the ability of RPM-1 to keep DLK-1 levels low . Neurites contain overlapping microtubules that must be interconnected to fulfill their known functions in trafficking and morphological integrity . These higher-order cytoskeletal networks must be organized in a way that stabilizes microtubule minus ends , in addition to linking individual microtubules together . The minus-end of microtubules are both known for durability and are distributed at regular intervals throughout neurites ( Baas and Black , 1990; Kurachi et al . , 1999; Cueva et al . , 2007; Bellanger et al . , 2012 ) . We found PTRN-1 to be distributed in puncta along neurites , and these structures thus represent a potential reservoir for sensors that promote remodeling . Although these results indicate that neuronal CAMSAPs function similarly to epithelial CAMSAPs by stabilizing microtubule networks , additional work will be required to explore the link between ptrn-1 disruption and specific changes to microtubule-based structures . We also found evidence for PTRN-1 function outside of neuronal remodeling pathways , including roles in neurite polarity and in sculpting the shape of synapses . The appearance of ALM posterior neurites , seen in ptrn-1 mutants , is thought to be a neuron polarity defect ( Hilliard and Bargmann , 2006; Kirszenblat et al . , 2013; Prasad and Clark , 2006 ) , and the cell body puncta seen in animals expressing PTRN-1:mcherry may be microtubule organizing centers necessary for polarized neurite outgrowth . However , the composition of ectopic ALM posterior neurites is incompletely understood . These extensions are reported to occur due to hyper-stabilization of microtubules ( Kirszenblat et al . , 2013 ) , but also appear in mutants where microtubules and cytoskeletal networks are disrupted ( Topalidou et al . , 2012 ) . Mature neurons are known to have stable microtubule structures due to post-translational modifications and interactions with cytoskeletal associated proteins ( Conde and Caceres , 2009; Kawataki et al . , 2008; Song et al . , 2013 ) . The differential effects of colchicine that we see on ALM and PLM phenotypes in ptrn-1 mutants may occur because of the differences in microtubule-based structures present in ectopic ( ALM ) vs native ( PLM ) neurites . Future work will be necessary to probe the role of CAMSAPs in neuron polarity . In ptrn-1 , dlk-1 double mutants , we observed small and sometimes elongated presynaptic varicosities . It was recently reported that the D . melanogaster homolog of DLK-1 has developmental roles in synaptic structure that are independent from roles in neuronal remodeling ( Klinedinst et al . , 2013 ) , and our data support similar conclusions for presynaptic varicosities in C . elegans . Synapses are specialized cell adhesion sites , and proteins involved in cellular adhesions have been found to be involved in maintenance of synaptic structures ( Chang and Balice-Gordon , 2000; Pielage et al . , 2008; Geissler et al . , 2013; Pielarski et al . , 2013 ) . CAMSAPs mediate connections between microtubule minus ends and adhesion proteins at epithelial adherens junctions ( Meng et al . , 2008 ) , and the synaptic-localized CAMSAPs we observed may play a similar bridging role at synaptic adhesion sites . Our data represent the first characterization of CAMSAP function in an intact animal . In C . elegans , we detected CAMSAP expression primarily in neurons . Mammalian CAMSAPs have also been found in neuronal tissue , and CAMSAP1L1 is a genetic trait locus for epilepsy ( Guo et al . , 2012 ) . Abnormal neuronal patterning and seizure-induced morphological remodeling are both thought to contribute to continuing seizures and the development of epilepsies ( Houser et al . , 2012; O’Dell et al . , 2012; Zhao and Overstreet-Wadiche , 2008 ) . We found that mutations in the C . elegans CAMSAP homolog , ptrn-1 initiate remodeling programs that cause aberrant extension and morphological re-patterning of neurites . Our data indicate that cytoskeletal-based activation of regenerative programs may exist in the absence of external injury signals . We speculate that without the stabilizing function of CAMSAP proteins , neuronal remodeling may be more likely to occur , a phenomenon we observed in C . elegans ptrn-1 mutants . We further speculate that the function of CAMSAPs in maintenance of neuronal morphology may be responsible for the association of CAMSAP1L1 with epilepsy . This work represents the first evidence of a role for microtubule minus-ends in neuron remodeling pathways . It identifies CAMSAPs as regulators of neuronal morphology , positions both CAMSAPs and microtubule structures as early regulators of neuronal-remodeling programs , and suggests that CAMSAPS regulate microtubule organization in neurons . Additionally , we found evidence for CAMSAP function in establishing neuron polarity and stabilizing the morphology of the synapse . Future work will be necessary to uncover the specific nature of the regenerative signal activating DLK-1 pathways , and to identify the CAMSAP-interacting proteins that function in neurons .
Strains were maintained at 22°C , unless otherwise specified , on Nematode Growth Medium ( NGM ) agar plates spotted with OP50 E . coli ( Brenner , 1974 ) . Some strains were provided by the Caenorhabditis Genetics Center , which is funded by NIH Office of Research Infrastructure Programs ( P40 OD010440 ) . The tm5597 deletion mutation was provided by the MITANI Lab through the National Bio-Resource Project of the MEXT , Japan . The ptrn-1 ( js1286 ) allele was isolated from an ENU-mutagenesis screen of animals carrying the jsIs973 and jsIs1077 transgenes . js1286 was determined to be a recessive lesion mapping to the X chromosome during outcrossing and positioned to the right of the single nucleotide polymorphism ( SNP ) Ce6-1456 ( X:+17 ) using standard SNP mapping using the HA-8 wild Hawaiian C . elegans strain ( Davis et al . , 2005 ) . Whole genome sequencing by Robert Barstead was performed at Oklahoma Medical Research Foundation and the data set was analyzed using Whole Genomes ( http://seqreport . omrf . org/genome/ ) , a web-based alignment and analysis program . Approximately 30-fold coverage using 100 bp paired end reads revealed only 1 homozygous missense and 1 homozygous non-sense lesion in coding regions right of uCE6-1456 . Fosmid rescue with a 38-kb genomic region ( WRM0636bB07 ) containing F35B3 . 5a ( ptrn-1; a . k . a . cspr-1 and pqn-34 ) suggested the lesion in this gene was the causative lesion resulting in the axon overgrowth phenotype . Oligonucleotides used in plasmid construction are listed in Supplementary file 1 . NM2498 pCFJ151 Prab-3 . The rab-3 promoter was amplified using oligonucleotides 4216 and 4217 , digested with AflII and SbfI and inserted into pCFJ151 ( Frokjaer-Jensen et al . , 2008 ) . NM2703 pCFJ151 Prab-3::mCherry::MCS . mcherry was amplified from pCFJ104 ( Frokjaer-Jensen et al . , 2008 ) using 4218 and 4533 , digested with XhoI and AvrII and inserted into similarly digested NM2498 . NM2704 pCFJ151 Prab-3::mCherry::MCS::unc-10 3’UTR . The 3’ UTR region of unc-10 was amplified using oligonucleotides 4651 and 4652 , digested with BsiWI and SgrAI and inserted into similarly digested NM2703 . NM2705 pCFJ151 Pmyo-3::mCherry::MCS::unc-10 3’UTR . The myo-3 promoter was amplified from pCFJ104 with oligonucleotides 4220 and 4221 , digested with SbfI and NotI and inserted into similarly digested NM2704 replacing the rab-3 promoter . NM2849 Prab-3::ptrn-1::mcherry . A full length ptrn-1a cDNA was amplified from first strand cDNA using oligonucleotides 4741 and 4742 , digested with SbfI and NheI and inserted into NM2704 . NM2925 Pptrn1::ptrn-1::mcherry . The ptrn-1 promoter was amplified using oligonucleotides 4828 and 4829 , digested with SbfI and Not I , and inserted into NM2849 replacing the rab-3 promoter . NM2926 Pptrn-1::mcherry . The ptrn-1 promoter was amplified using oligonucleotides 4828 and 4829 digested with SbfI and NotI and inserted into similarly digested NM2705 . NM2930 Pmyo-3::ptrn-1::mcherry . NM2705 and NM2849 were digested with NotI and SbfI and the myo-3 promoter was ligated in place of the rab-3 promoter . NM2645 pCFJ355K mec-7p cherry gpd2/3 GFP-RAB-3 . The mec-7 promoter driving in mCherry and GFP-RAB-3 in an operon separated by the gpd-2/3 intergenic region in a neomycyin resistant derivative of the X-chromosome MosSCI integration plasmid pCFJ355 ( Frøkjær-Jensen et al . , 2012 ) . This plasmid contains both Cbunc-119 and neomycin selectable markers . An annotated version of the sequence of this plasmid is available at http://thalamus . wustl . edu/nonetlab/ResourcesF/seqinfo . html . jsIs1077 was created by ballistic transformation ( Praitis et al . , 2001 ) of NM2041 pMec-7 GFP-ELKS-1 CbUNC-119 ( Kural et al . , 2009 ) into unc-119 ( ed3 ) and outcrossed into other genetic backgrounds . jsIs1269 was created by MosSCI-mediated insertion of NM2645 into the ttTi14024 Mos insertion site on the X chromosome using the direct integration protocol ( Frøkjær-Jensen et al . , 2012 ) . The insertion was determined to be single copy by long range PCR and restriction digestion . jsEx1284 , jsEx1292 , jsEx1295 , and jsEx1297 were created by standard germline injections ( Mello et al . , 1991 ) of plasmids NM2849 , NM2930 , NM2925 , and NM2926 , respectively along with co-injection marker plasmids NM1090 ( Prab-3::GFP ) ( Mahoney et al . , 2006 ) , pPD118 . 20 ( Pmyo-3::GFP ) , pPD118 . 33 ( Pmyo-2::GFP ) ( Gifts of Andy Fire ) as listed in the strain list . The plasmid of interest was injected at 30 ng/μl and co-markers at 5 ng/μl along with pBluescript KS ( + ) carrier DNA ( 150 ng/μl ) . Animals were anesthetized using either 10 mM sodium azide or immobilized using 0 . 1 micron microspheres and imaged on agarose pads ( Kim et al . , 2013 ) . Observations of fluorescent proteins were made using an Olympus B-MAX microscope with epifluorescence . 3-D images stacks were acquired using a Ziess Axioksop equipped an ASI piezo XYZ-motorized stage , Ludl high speed electronic filter wheels and shutters , and a Hamamatsu Orca-R2 cooled CCD camera all controlled by Volocity software . Typically , images were collected using a 40X Neofluar lens , and 1 μm steps . Stacks of images were flattened using a maximal projection to visualize neuronal structures that crossed multiple individual focal planes . For immunocytochemistry , animals were fixed in 2% formaldehyde and stained with antibodies directed against GFP ( mouse monoclonal , Clonetech ) and MEC-7 ( kindly provided by M Chalfie; Mitani et al . , 1993 ) as previously described ( Schaefer et al . , 2000 ) . To quantify locomotion defects , L4 Animals were transferred to fresh unseeded NMG plates and imaged for 30 s using a Spot camera at 1 frame per second . The velocity of animals was quantified using worm tracker software ( Ramot et al . , 2008 ) . Synchronous populations were generated from eggs laid by adult hermaphrodites over a 30-min time interval . Animals were scored by imaging jsIs973 and jsIs1077 using epifluorescence . Some animals were rescued from the anesthetic by being placed in a drop of M9 , isolated to individual plates containing E . coli , and re-imaged at a later stage of development . Worms were grown on NGM agar plates spotted with E . coli mixed with 1 mM of colchicine . L4 animals from the F2 generation were analyzed using fluorescence microscopy . The following transgenes were created for this analysis:jsEx1284 [Prab-3::ptrn-1:mcherry; Pmyo-2:GFP;Prab-3::GFP , Pmyo-3::GFP]jsEx1292 [Pmyo-3::ptrn-1:mcherry; Pmyo2::GFP]jsEx1297 [Pptrn-1::ptrn-1:mcherry; Pmyo-2:GFP;Prab-3::GFP , Pmyo-3::GFP] , jsEx1295 [Pptrn-1::mcherry; Pmyo-2:GFP;Prab-3::GFP , Pmyo-3::GFP]jsIs1077 IV [Pmec-7::GFP::ELKS-1; Cbunc-119 ( + ) ]jsIs1269 X [Pmec-7::mcherry -gdp2/3 intergenic- GFP:RAB-3; Cbunc-119 ( + ) ]Other transgenes used in this analysis include jsIs821 X [Pmec7::rab-3:GFP; unc-119 ( + ) ] ( Bounoutas et al . , 2009 ) jsIs973 III [Pmec7::RFP; unc-119 ( + ) ] ( Zheng et al . , 2011 ) oxIs12 [Punc-47::GFP; lin-15 ( + ) ] aka osIn12 ( McIntire et al . , 1997 ) , nuIs25 [Pglr-1::GFP; lin-15 ( + ) ] ( Rongo et al . , 1998 ) The following strains were used in this analysis:EG4322 ttTi5605 II; unc-119 ( ed3 ) FX5597 ptrn-1 ( tm5597 ) XNM4192 jsIs973; jsIs1077; ptrn-1 ( js1286 ) NM4406 jsIs973; jsIs1077; ptrn-1 ( tm5597 ) NM3947 jsIs973; jsIs1077NM4353 atat-2 ( ok2415 ) jsIs1269NM4405 jsIs973; mec-17 ( ok2109 ) NM4354 mec-17 ( ok2109 ) ; atat-2 ( ok2415 ) jsIs1269NM4544 dlk-1 ( km12 ) , jsIs973; mec-17 ( ok2109 ) NM4386 jsIs973; jsIs1077; ptrn-1 ( js1286 ) ; jsEx1284NM4450 jsIs973; ptrn-1 ( js1286 ) ; jsEx1292NM4490 dlk-1 ( km12 ) ; jsIs973; rpm-1 ( ok364 ) ; jsIs821 ptrn-1 ( tm5597 ) NM4485 jsIs973; rpm-1 ( ok364 ) ; jsIs821 ptrn-1 ( tm5597 ) NM4483 dlk-1 ( km12 ) ; jsIs973; jsIs821 ptrn-1 ( tm5597 ) NM4482 jsIs973; jsIs821 ptrn-1 ( tm5597 ) NM4484 dlk-1 ( km12 ) ; jsIs973; rpm-1 ( ok364 ) ; jsIs821NM4480 rpm-1 ( ok364 ) ; jsIs973; jsIs821NM4481 dlk-1 ( km12 ) ; jsIs973; jsIs821NM3361 jsIs973; jsIs821NM4470 jsIs973; ptrn-1 ( js1286 ) ; jsEx1297NM4476 jsEx1295; unc-119 ( ed3 ) ; ttTi5605NM4422 ptrn-1 ( tm5597 ) ; oxIs12NM4419 nuIs25; jsIs973; ptrn-1 ( tm5597 ) | Microtubules are tiny tubular structures made from many copies of proteins called tubulins . Microtubules have a number of important roles inside cells: they are part of the cytoskeleton that provides structural support for the cell; they help to pull chromosomes apart during cell division; and they guide the trafficking of proteins and molecules around inside the cell . Most microtubules are relatively unstable , undergoing continuous dis-assembly and re-assembly in response to the needs of the cell . The microtubules in the branches of nerve cells are an exception , remaining relatively stable over time . Now Marcette et al . and , independently , Richardson et al . , have shown that a protein called PTRN-1 has an important role in stabilizing the microtubules in the nerve cells of nematode worms . Marcette et al . became interested in the PTRN-1 protein after conducting a screen of randomly mutated worms to look for those with abnormally developed nerve cells . Although worms with a mutation in the gene encoding the PTRN-1 protein could form nerve cells that looked normal during early development , the pattern of branches on the nerve cells went awry later on . Moreover , the mutant worms lost the swellings that are normally found at the junctions between nerve cells , and they also moved their bodies in an odd way . Engineering the mutant worms to produce the PTRN-1 protein in their nerve cells , but nowhere else , restored normal movement , and experiments with fluorescently tagged PTRN-1 proteins revealed that they are found on the microtubules within the nerve cells . Marcette et al . suggest that the microtubules become less stable when this protein is not present , and that this switches on a repair mechanism that remodels the nerve cells , providing that the active form of a protein called DLK-1 is present . Other mutations that reduce the stability of microtubules also triggered the same remodeling process . Future work will be necessary to uncover exactly what triggers the remodeling process , and to identify the other proteins that are involved in repairing the nerve cells . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2014 | The Caenorhabditis elegans microtubule minus-end binding homolog PTRN-1 stabilizes synapses and neurites |
Alternative splicing ( AS ) programs are primarily controlled by regulatory RNA-binding proteins ( RBPs ) . It has been proposed that a small number of master splicing regulators might control cell-specific splicing networks and that these RBPs could be identified by proximity of their genes to transcriptional super-enhancers . Using this approach we identified RBPMS as a critical splicing regulator in differentiated vascular smooth muscle cells ( SMCs ) . RBPMS is highly down-regulated during phenotypic switching of SMCs from a contractile to a motile and proliferative phenotype and is responsible for 20% of the AS changes during this transition . RBPMS directly regulates AS of numerous components of the actin cytoskeleton and focal adhesion machineries whose activity is critical for SMC function in both phenotypes . RBPMS also regulates splicing of other splicing , post-transcriptional and transcription regulators including the key SMC transcription factor Myocardin , thereby matching many of the criteria of a master regulator of AS in SMCs .
Alternative splicing ( AS ) is an important component of regulated gene expression programmes during cell development and differentiation , usually focusing on different sets of genes than transcriptional control ( Blencowe , 2006 ) . AS programs re-wire protein-protein interaction networks ( Buljan et al . , 2012; Ellis et al . , 2012; Yang et al . , 2016 ) , as well as allowing quantitative regulation by generating mRNA isoforms that are differentially regulated by translation or mRNA decay ( McGlincy and Smith , 2008; Mockenhaupt and Makeyev , 2015 ) . Coordinated cell-specific splicing programs are determined by a combination of cis-acting transcript features and trans-acting factors that compose ‘splicing codes’ ( Barash et al . , 2010; Chen and Manley , 2009; Fu and Ares , 2014 ) . The interactions between the trans component RNA-binding proteins ( RBPs ) and the cis component regulatory elements in target RNAs coordinate the activation and repression of specific splicing events . Many regulatory proteins , including members of the SR and hnRNP protein families , are quite widely expressed , while others are expressed in a narrower range of cell types ( David and Manley , 2008; Fu and Ares , 2014 ) . A further conceptual development of combinatorial models for splicing regulation has been the suggestion that a subset of RBPs act as master regulators of cell-type specific AS networks ( Jangi and Sharp , 2014 ) . The criteria expected of such master regulators include that: ( i ) they are essential for cell-type specification or maintenance , ( ii ) their direct and indirect targets are important for cell-type function , ( iii ) they are likely to regulate the activity of other splicing regulators , ( iv ) they exhibit a wide dynamic range of activity , which is not limited by autoregulation , and ( v ) they are regulated externally from the splicing network , for example by transcriptional control or post-translational modifications . It was further suggested that expression of such splicing master regulators would be driven by transcriptional super-enhancers , providing a possible route to their identification ( Jangi and Sharp , 2014 ) . Super-enhancers are extended clusters of enhancers that are more cell-type-specific than classical enhancers and that drive expression of genes that are essential for cell-type identity , including key transcription factors ( Hnisz et al . , 2013 ) . By extension , RBPs whose expression is driven by super-enhancers are expected to be critical for cell-type identity and might include master regulators of tissue-specific AS networks ( Jangi and Sharp , 2014 ) . Vascular smooth muscle cells ( SMCs ) are important in cardiovascular physiology and pathology ( Bennett et al . , 2016; Fisher , 2010; Owens et al . , 2004 ) . Unlike skeletal and cardiac muscle SMCs exhibit phenotypic plasticity and are not terminally differentiated ( Owens et al . , 2004 ) ( Figure 1A ) . In healthy arteries , vascular SMCs exist in a differentiated contractile state . In response to injury or disease , the SMC phenotype switches towards a more synthetically active , motile and proliferative state ( Fisher , 2010; Owens et al . , 2004 ) . The transcriptional control of SMC phenotypic switching has been intensely studied , but the role of post-transcriptional regulation has been relatively neglected ( Fisher , 2010 ) . For example , some markers of the contractile state , such as h-Caldesmon and meta-Vinculin , arise via AS ( Owens et al . , 2004 ) , but nothing is known about the regulation of these events . A number of known splicing regulators , including PTBP1 , CELF , MBNL , QKI , TRA2B , and SRSF1 , have been implicated in the regulation of individual SMC-specific ASEs , but these proteins are not restricted to differentiated SMCs and most act primarily in the de-differentiated state ( Gooding et al . , 2013; Gooding et al . , 1998; Gromak et al . , 2003; Shukla and Fisher , 2008; van der Veer et al . , 2013; Xie et al . , 2017 ) . Indeed , global profiling confirmed a widespread role of PTBP1 in repressing exons that are used in differentiated mouse aorta SMCs ( Llorian et al . , 2016 ) , but did not identify RBPs that act as direct regulators of the differentiated state . Biochemical identification of such RBPs is hampered by the fact that SMCs rapidly dedifferentiate in cell culture conditions . Here , we used the approach suggested by Jangi and Sharp , to identify candidate AS master regulators as RBP-encoding genes whose cell-specific expression is driven by super-enhancers ( Jangi and Sharp , 2014 ) . We identified RNA-Binding Protein with Multiple Splicing ( RBPMS ) , a protein not previously known to regulate splicing , as a critical regulator of numerous AS events in SMCs . RBPMS is highly expressed in differentiated SMCs , where it promotes AS of genes that are important for SMC function . These include many components of the actin cytoskeleton and focal adhesion machineries , modulation of whose function is key to the transition from contractile to motile phenotypes . RBPMS also targets other splicing regulators , post-transcriptional regulators and the key SMC transcription factor Myocardin where RBPMS promotes inclusion of an exon that is essential for maximal SMC-specific activity . RBPMS therefore meets many of the criteria expected of an AS master regulator in SMCs , and its identification validates the approach of identifying key cell-specific regulators via the super-enhancer-proximity of their genes .
To identify potential SMC master splicing factors we used a catalog of 1542 human RBPs ( Gerstberger et al . , 2014 ) and data-sets of super-enhancers from three human SMC-rich tissues ( aorta , bladder , stomach smooth muscle ) and skeletal muscle ( Hnisz et al . , 2013 ) ( Supplementary file 1 ) . Nine RBP genes were associated with super-enhancers in all SMC tissues but not skeletal muscle ( Figure 1B ) . Using the dbSUPER database ( Khan and Zhang , 2016 ) , which uses more stringent distance constraints for associating genes with superenhancers ( see Materials and methods ) , we identified two candidates , of which only RBPMS was shared with our original nine candidates . Examination of RBPMS expression in human tissues from the Genotype-Tissue Expression ( GTEX ) project ( GTEx Consortium , 2013 ) showed the top eight expressing tissues to be SMC rich , including three arteries ( Figure 1—figure supplement 1A ) . RNA-Seq data from rat aorta SMCs showed that Rbpms levels decreased 3 . 8 fold during de-differentiation from tissue to passage 9 cell culture ( Figure 1—figure supplement 1A–C ) , in parallel with known SMC transcriptional markers ( Acta2 , Cnn1 , Smtn , ( Figure 1—figure supplement 1B , C ) and AS events ( Tpm1 and Actn1 , Figure 1—figure supplement 3A , C ) . Moreover , of the starting candidate RBPs ( Figure 1B ) Rbpms was the most highly expressed of those that were down-regulated from tissue to culture ( Figure 1—figure supplement 2B , C green labels ) . Expression of the Rbpms2 paralog also decreased upon de-differentiation , but its absolute level of expression was >10 fold lower than Rbpms ( Figure 1—figure supplement 2B , C ) . Other RBPs implicated in AS regulation in SMCs ( PTBP1 , MBNL1 , QKI ) either showed more modest changes or higher expression in de-differentiated cells ( Figure 1—figure supplement 2B , C ) . In the PAC1 rat SMC line Rbpms mRNA levels decreased by ~10 fold between differentiated and proliferative states in parallel with SMC marker AS events ( Figure 1D ) and genes ( Figure 1E ) . Rbpms2 was expressed at much lower levels than Rbpms , and did not alter expression between PAC1 cell states ( Figure 1E ) . RBPMS protein decreased to undetectable levels in proliferative PAC1 cells in parallel with smooth muscle actin ( ACTA2 ) ( Figure 1F ) . Immunofluorescence microscopy also showed higher levels of RBPMS in differentiated PAC1 cells where it was predominantly nuclear ( Figure 1G ) , consistent with the hypothesis that it regulates splicing . To further investigate RBPMS expression we cloned cDNAs from PAC1 cells , representing seven distinct mRNA isoforms . These encoded two major protein isoforms ( RBPMSA and RBPMSB ) sharing a common N-terminus and RNA Recognition Motif ( RRM ) domain . RBPMSA and B differed by short C-terminal tails encoded by alternative 3' end exon 7 and exon 8 ( Figure 1C ) , and corresponded in size to the two protein bands seen in western blots ( Figure 1F ) . Other mRNA isoforms differed by inclusion or skipping of exon six and by alternative 3' UTR exon inclusion . RBPMS and RBPMS2 , which are 70% identical , have a single RRM domain that is responsible for both RNA binding and dimerization ( Sagnol et al . , 2014; Teplova et al . , 2016 ) . Optimal RBPMS -binding sites consist of tandem CACs separated by a spacer of ~1–12 nt ( Farazi et al . , 2014; Soufari and Mackereth , 2017 ) . We found significant enrichment of CACN1-12CAC motifs within and upstream of exons that are less included in differentiated compared to cultured rat aorta SMCs ( Figure 1—figure supplement 3D ) . These are locations at which many splicing regulators mediate exon skipping ( Fu and Ares , 2014; Witten and Ule , 2011 ) . Consistent with this , the SMC-specific mutually exclusive exon pairs in Actn1 and Tpm1 ( Gooding and Smith , 2008; Southby et al . , 1999 ) both have conserved clusters of CAC motifs upstream of the exon that is skipped in differentiated SMCs ( see below ) . In summary , the presence of super-enhancers at the RBPMS gene in SMC-rich tissues , its wide dynamic range of expression between differentiated SMCs and other tissues and proliferative SMCs ( Figure 1 , Figure 1—figure supplements 1–3 ) , the nuclear localization of RBPMS , and the presence of potential RBPMS-binding sites adjacent to known SMC-regulated exons are all consistent with the hypothesis that RBPMS might act as a master regulator of AS in differentiated SMCs . To investigate the roles of RBPMS in shaping SMC transcriptomes we manipulated levels of RBPMS expression in differentiated and proliferative PAC1 cells ( Figure 2A , B ) . We used siRNAs to knockdown all Rbpms isoforms in differentiated PAC1 cells , achieving ~75% depletion ( Figure 2B ) . In parallel , proliferative PAC1 cells were transduced with pINDUCER lentiviral vectors ( Meerbrey et al . , 2011 ) to allow Doxycycline inducible over-expression of RBPMSA . No basal RBPMS expression was observed in proliferative cells , but upon induction substantial expression was observed from the FLAG-RBPMSA vector but not from the empty lentiviral vector ( LV ) ( Figure 2B ) . The effects of manipulating RBP levels can sometimes be compensated by related family members ( Mockenhaupt and Makeyev , 2015 ) . However , Rbpms2 levels were not affected by any of the treatments and Rbpms2 knockdown , either alone or in combination with Rbpms , had no effects upon tested AS events ( data not shown ) . RNA samples from Rbpms knockdown and overexpression experiments were prepared for Illumina poly ( A ) RNAseq . Data was analyzed for changes in mRNA abundance using DESEq2 ( Love et al . , 2014 ) ( Figure 2C , Supplementary file 2 ) , and for changes in AS using rMATS ( Shen et al . , 2014 ) ( Figure 2D , Supplementary file 3 ) . In addition to analysis of effects of RBPMS depletion and overexpression , comparison of the differentiated and proliferative PAC1 control samples revealed changes associated with differentiation state . Principal Component Analysis based on mRNA abundance showed clear separation of differentiated and proliferative samples ( PC1 , 82% variance , Figure 2—figure supplement 1A ) . This suggests that at the level of mRNA abundance the differences between differentiated and proliferative PAC1 cells far outweigh any effects of manipulating RBPMS levels . Consistent with this , RBPMS knockdown was associated with ~10 fold fewer changes at the transcript abundance level ( 110 increased , 82 decreased , padj <0 . 05 , fold change >2 fold ) compared to the differentiated vs proliferative control comparison ( 830 increased , 1444 decreased , Figure 2C ) . RBPMS overexpression led to an intermediate number of changes in mRNA abundance levels , but only 29 genes were affected by both overexpression and knockdown , and of these only four genes other than Rbpms were regulated reciprocally . RBPMS knockdown and overexpression had substantial effects at the level of AS affecting all types of AS event ( Figure 2D , Figure 2—figure supplement 1B ) . RBPMS knockdown led to changes in 318 AS events ( FDR < 0 . 05 , |ΔPSI| > 0 . 1 , where PSI is Percent Spliced In ) , which was only 2-fold less than the number of events regulated between control differentiated and proliferative cells . Cassette exons were the largest group of events , with roughly equal numbers of up- and down-regulated exons ( Figure 2D ) . RBPMS overexpression led to a larger number of AS changes ( 4934 regulated events ) , probably resulting from the combination of RBPMS expression in excess of levels usually present in differentiated PAC1 cells ( Figure 2C ) and also because we expressed the more active RBPMSA isoform ( see below ) . Cassette exons affected by RBPMS overexpression were strongly skewed ( 80% ) towards greater exon skipping . A subset of ASEs observed in the RNA-Seq experiments , encompassing RBPMS activated and repressed cassette exons and mutually exclusive exons were validated by RT-PCR ( Figure 2E , F , Figure 2—figure supplements 2 and 3 ) . ΔPSI values determined by RT-PCR and RNA-Seq were in good agreement ( Figure 2—figure supplement 3B ) . As an additional negative control , doxycycline induction of empty lentiviral transduced cells had no effect on RBPMS-regulated AS events ( Figure 2—figure supplement 2A ) . In each of the three comparisons , the overlap between genes regulated at the levels of splicing and mRNA abundance was small: 0 . 8% of all genes regulated at splicing or abundance level for RBPMS knockdown , 3 . 2% for RBPMS overexpression , and 1 . 6% for PAC1 phenotype , ( Figure 2—figure supplement 1C ) . However , there were substantial overlaps of ASEs regulated by RBPMS knockdown , overexpression and PAC1 phenotype ( Figure 3A ) . Twenty percent of ASEs regulated in PAC1 cell differentiation were congruently regulated by RBPMS knockdown , representing 40% of ASEs affected by RBPMS knockdown ( Figure 3A and B left panel ) . The high correlation ( R2 = 0 . 95 ) suggests that for these 127 events changes in RBPMS expression are sufficient to explain their differentiation-specific splicing changes . Similarly , RBPMS overexpression shared 180 events in common with PAC1 differentiation status ( 28% of differentiation specific events , 3 . 7% of overexpression regulated events ) . The ΔPSI correlation for these AS events was lower than for RBPMS knockdown ( R2 = 0 . 86 , Figure 3A and B right panel ) . Sixty seven events regulated by RBPMS knockdown were reciprocally regulated by RBPMS overexpression , of which 52 were shared across all the three comparisons ( 8 . 2% of differentiation specific events ) , as exemplified by events in Ptprf , Piezo1 and Actn1 ( Figure 2E ) . Hierarchical clustering of cassette exons regulated between PAC1 phenotypes across all 12 samples also revealed two clusters of RBPMS-responsive events where knockdown and overexpression were sufficient to reciprocally convert the splicing pattern to that of the other cellular phenotype ( Figure 3E , clusters 1 and 4 , containing RBPMS activated and repressed exons respectively ) . Many AS events in differentiated PAC1 cells do not reach the fully differentiated splicing pattern characteristic of tissue SMCs . However , for some events overexpression of RBPMS led to splicing patterns similar to tissue SMCs . For example the Actn1 SM exon is included to 69% in differentiated PAC1 cells , but to 93–94% in RBPMS overexpressing cells ( Figure 2E ) and aorta tissue ( Figure 1—figure supplement 3C ) . We therefore hypothesized that some ASEs regulated by RBPMS overexpression but not knockdown , might reflect tissue SMC AS patterns that are not usually observed in cultured SMCs . To address this possibility , we used RNA-Seq data monitoring de-differentiation of rat aorta SMCs ( Figure 1—figure supplements 2 and 3 ) . Of 1714 ASES regulated between tissue and passage 9 SMCs , 265 ( 15% ) were also regulated by RBPMS overexpression in PAC1 cells ( Figure 3C , D , R2 = 0 . 68 ) . Strikingly , hierarchical clustering of cassette exons regulated between aorta tissue and passage nine cultured SMCs showed that the RBPMS overexpression sample clustered together with tissue , away from all other samples ( Figure 3F ) . Two clusters of AS events shared very similar splicing patterns in tissue and RBPMSA overexpression , and differed in all remaining samples ( Figure 3F ) . Cluster 1 comprised 28 RBPMS-activated exons , of which 20 were not regulated between PAC1 phenotypes . For example , inclusion of a cassette exon in Fermt2 was observed only in tissue and RBPMS overexpression samples ( Figure 3G ) . Similarly , inclusion of Cald1 exon four in conjunction with a downstream 5’ splice site on exon 3a , producing the hCald1 marker in tissue SMCs ( Figure 3G , Figure 3—figure supplement 1C , D ) , and inclusion of the meta-vinculin exon ( Figure 3—figure supplement 1A ) were also only seen upon RBPMS overexpression . Cluster four contained exons that are skipped in tissue and upon RBPMS overexpression ( e . g . Tsc2 , Figure 3G ) , the majority of which are not regulated in PAC1 differentiated cells . Likewise Tpm1 mutually exclusive exon three is skipped nearly completely upon RBPMS overexpression in a similar pattern to tissue ( Figure 3—figure supplement 1B ) . Upregulation of RBPMS expression therefore appears to be sufficient to promote a subset of splicing patterns usually only observed in differentiated SMCs in vivo . To address whether RBPMS directly regulates target exons we looked for enrichment of its binding motif ( Farazi et al . , 2014 ) adjacent to cassette exons regulated by RBPMS knockdown or overexpression . CACN1-12CAC motifs , the optimal binding motif for RBPMS dimers , were significantly enriched around exons that were activated or repressed by RBPMS , with a similar position-dependent activity as other splicing regulators ( Figure 4A ) . Exons repressed by RBPMS showed strong enrichment of motifs within the exon and the immediate ~80 nt upstream intron flank , while exons activated by RBPMS showed motif enrichment within the downstream intron flank . Consistent with the contribution of RBPMS to the AS changes between PAC1 differentiation states , CACN1-12CAC motifs were also enriched upstream of and within exons that are more skipped in differentiated cells , and downstream of exons that are more included in differentiated cells ( Figure 4A ) . Moreover , in the set of exons activated by RBPMS overexpression , CACN1-12CAC motifs were not only enriched downstream , but also significantly depleted in the repressive locations within and upstream of the exon ( Figure 4A , B ) . This suggests that binding in repressive locations might be dominant over activation . To test whether RBPMS regulates AS events by directly binding to CAC motifs we co-transfected HEK293 cells with RBPMS expression vectors and minigenes of representative activated ( Flnb ) and repressed ( Tpm1 , Actn1 ) exons , with potential RBPMS-binding sites in expected locations for activation or repression ( Figure 4—figure supplements 1B–C and 2B ) . We initially established that transient expression of RBPMSA in HEK293 cells was sufficient to switch AS of endogenous FLNB , TPM1 , MPRIP and ACTN1 towards the SMC splicing pattern ( Figure 4C , G and Figure 4—figure supplement 1A ) . For comparison , we also transfected expression constructs for the RBPMSB isoform and the paralog RBPMS2 . We found that RBPMSB had lower activity for some events ( ACTN1 , Figure 4—figure supplement 1A ) , but that transfected RBPMS2 had similar activity to RBPMSA in all cases . RBPMSA and RBPMS2 also strongly activated inclusion of the Flnb H1 exon in a minigene context while RBPMSB activated to a lower extent ( Figure 4D ) . Tpm1 exon three is the regulated member of a pair of mutually exclusive exons and is repressed in SMCs ( Ellis et al . , 2004; Gooding et al . , 1994 ) . MBNL and PTBP proteins promote this repression but are not sufficient to switch splicing ( Gooding et al . , 2013; Gooding et al . , 1998 ) . In contrast , RBPMSA expression was sufficient to cause a near complete switch from exon inclusion to skipping ( Figure 4H ) . RBPMS2 had lower activity , but RBPMSB was by far the least active protein . Likewise , RBPMSA and RBPMS2 completely switched splicing of Actn1 constructs ( Gromak et al . , 2003; Southby et al . , 1999 ) from the NM to the SM mutually exclusive exon , while RBPMSB was nearly inactive ( Figure 4—figure supplement 2A ) . A construct containing only the Actn1 SM exon was unresponsive to cotransfection , while a construct containing only the NM exon , which has three upstream CAC clusters ( Figure 4—figure supplement 1B ) showed a complete switch from inclusion to skipping upon cotransfection of RBPMSA or RBPMS2 , but not RBPMSB . Thus , for two mutually exclusive events RBPMSA is able to switch splicing to the SMC pattern by repressing the exon that is usually used in non-SMCs , while RBPMSB is less active . We mutated CAC motifs in suspected binding sites to CCC , which disrupts RBPMS binding ( Farazi et al . , 2014 ) . Mutation of 12 CACs downstream of the FlnB H1 exon had no effect on basal splicing , but the H1 exon was completely resistant to activation by RBPMSA ( Figure 4E ) . Likewise , mutation of 9 CAC motifs upstream of Tpm1 exon three had no effect on exon inclusion in the absence of RBPMSA , but completely prevented exon skipping in response to RBPMSA ( Figure 4I , Figure 4—figure supplement 2B ) , while mutations of individual clusters had intermediate effects ( Figure 4—figure supplement 2B ) . The response of both Flnb H1 exon and Tpm1 exon three therefore depends on nearby CAC motifs . To test whether these are binding sites for RBPMS , we used in vitro transcribed RNAs and recombinant RBPMSA and B ( Figure 4F , J and Figure 4—figure supplement 4 ) . Using both electrophoretic mobility shift assay ( EMSA ) and UV crosslinking , both RBPMSA and B were found to bind to the FlnB wild type RNA ( apparent Kd ~0 . 5 μM ) , but not to the mutant RNA ( Figure 4F ) . With Tpm1 , RBPMSA and B bound to the WT RNA as indicated by EMSA assays ( Kd ~0 . 5 μM ) ( Figure 4J ) . Binding was reduced by the mutations that abrogated RBPMSA repression of Tpm1 exon 3 ( Kd >0 . 5 μM ) . Finally , we tested the effects of mutations in RBPMS to disrupt the RNA binding and dimerization interfaces of the RRM ( Sagnol et al . , 2014; Teplova et al . , 2016 ) . Both RNA binding and dimerization mutants of RBPMSA had lost all splicing regulatory activity upon overexpression in HEK293 cells ( Figure 4—figure supplement 3 ) , revealing RBPMS dimer-dependent splicing activity . These data therefore show that RBPMS can inhibit splicing by binding to CAC clusters upstream of exons ( Tpm1 ) and activate splicing by downstream binding ( Flnb ) . To investigate the functional importance of RBPMS-regulated AS , we carried out Gene Ontology ( GO ) analysis of the genes whose splicing or expression levels were regulated by RBPMS ( Figure 5A , Figure 5—figure supplement 1A , Supplementary files 4 and 5 ) , or between SMC differentiation states ( Figure 5—figure supplement 1B , C ) . Splicing events affected by RBPMS-knockdown affected genes involved in processes , components and functions important for SMC biology , such as cytoskeleton , cell projection , cell junction organization and GTPase regulation ( Figure 5A ) . These categories were very similar to those affected by AS during PAC1 cell dedifferentiation ( Figure 5—figure supplement 1B ) . Events regulated by RBPMS knockdown were also enriched within genes that are associated with super-enhancers in SMC tissues ( Figure 5B ) and therefore inferred to be important for SMC identity . In contrast , the relevance of GO terms associated with RBPMSA overexpression to SMC biology was less clear ( Figure 5—figure supplement 1A ) , despite the presence of many events that were reciprocally regulated by RBPMS knockdown . RBPMS overexpression AS targets were also not enriched for aorta super-enhancer-associated genes ( p=0 . 37 ) . This discrepancy might be accounted for by the high level of RBPMS overexpression leading to changes in some AS events that are not physiological targets . At the RNA abundance level enriched GO terms associated with RBPMS knockdown or overexpression did not align with those associated with differentiation , mainly being associated with stress responses ( Supplementary file 5 ) . To further explore the consequences of RBPMS regulation of AS we carried out network analysis using STRING ( Szklarczyk et al . , 2017 ) ( Figure 5C ) . To ensure that all target events were biologically relevant we used only AS events that are coregulated by RBPMS knockdown and PAC1 differentiation state ( Figure 3A , E ) or by RBPMS overexpression and in aorta tissue ( Figure 3C , F ) . We also restricted the output network to high confidence interactions . RBPMS targets comprised a network ( Figure 5C ) focused on functions associated with cell-substrate adhesion ( yellow nodes ) and the actin cytoskeleton ( blue ) . Six of the network proteins ( SORBS1 , CALD1 , PDLIM5 , PDLIM7 , ACTN1 and ARHGEF7 ) are also components of the consensus integrin adhesome ( Horton et al . , 2015 ) , which mechanically connects , and mediates signalling between , the actin cytoskeleton and the extracellular matrix . Underlining the importance of this network to SMC function , many of the genes are themselves super-enhancer-associated in one or more smooth muscle tissues ( bold text , Figure 5C ) . Actomyosin activity , and its mechanical connection to the extracellular matrix are important for both the contractile and motile states of SMCs ( Min et al . , 2012 ) . It appears that RBPMS plays an important role in modulating the activity of this protein network to suit the needs of the two cell states . Despite these coordinated changes in AS , we were unable to observe obvious changes in PAC1 cell phenotype after under the overexpression and knockdown treatments used for RNA analysis . We reasoned that more sustained perturbation of RBPMS levels might be required to allow turnover of protein isoforms before phenotypes would become apparent . We therefore prolonged RBPMS knockdown to a total of 120 hr ( Figure 5D–F and Figure 5—figure supplements 2 and 3 ) . Effective knockdown of RBPMS protein and RNA , and effects upon known RBPMS-regulated AS events were all confirmed ( Figure 5—figure supplement 3 ) . Under these conditions we observed a modest reduction in levels of smooth muscle actin ( Acta2 ) protein and RNA ( Figure 5—figure supplement 3 ) , which had not been seen with shorter knockdown . Strikingly , we observed substantial changes in cell morphology and actin organization of differentiated PAC1 cells upon prolonged RBPMS depletion . Compared to control cells , PAC1 cells treated with RBPMS siRNA ( D KD ) displayed less alignment of actin fibers , as indicated by the lower actin anisotropy score ( ~0 . 6 fold ) , accompanied by larger nuclear ( ~1 . 7 fold ) and cell sizes ( ~2 . 6 fold ) ( Figure 5D , F ) . Remarkably , the RBPMS knockdown differentiated cells closely resembled the proliferative cells in these characteristics ( Figure 5D , E ) . Therefore , the control of a functionally coherent set of targets that are important for SMC morphology and function is consistent with the hypothesis that RBPMS is a master regulator of AS in SMCs . Among the direct targets of master splicing regulators are expected to be events controlling the activity of other AS regulators leading to further indirect AS changes . Consistent with this , we noted that RBPMS affected splicing of Mbnl1 and Mbnl2 . RBPMS promoted skipping of the 36 nt and 95 nt alternative exons of Mbnl1 ( here referred to as exons 7 and 8 ) , as indicated by both knockdown and overexpression ( Figure 6A , B ) . In Mbnl2 the 36 nt exon was fully skipped in all conditions but the 95 nt exon was repressed by RBPMS ( Figure 6C ) . Consistent with the RNA-Seq data , we obtained Mbnl1 and Mbnl2 cDNAs from PAC1 cells that varied by inclusion of the 36 and 95 nt exons . Corresponding shifts in MBNL1 , but not MBNL2 , protein isoforms could be observed by western blot ( Figure 6D ) . These events affect the unstructured C-termini of MBNL proteins and have been shown to affect their splicing activity ( Sznajder et al . , 2016; Tabaglio et al . , 2018; Tran et al . , 2011 ) , suggesting that RBPMS might indirectly affect some AS events by modulating MBNL activity . To test this , we made expression constructs of rat MBNL1 isoforms with and without exons 7 and 8 ( FL , Δ7 , Δ8 , Δ7Δ8 ) and MBNL2 with and without exon 8 ( FL and Δ8 ) , obtained as cDNAs from PAC1 cells . When transfected into HEK293 cells , full length MBNL1 and 2 caused a shift from use of the downstream to an upstream 5' splice site ( 5'SS ) on NCOR2 exon 47 ( Figure 6E ) , an event that is differentially regulated by MBNL isoforms ( Sznajder et al . , 2016; Tran et al . , 2011 ) . The shorter MBNL isoforms showed lower activity in shifting towards the upstream 5'SS , although the difference was not statistically significant for the MBNL1 shorter isoforms ( Figure 6E ) . In proliferative PAC1 cells , NCOR2 exon 47 mainly uses the upstream 5'SS , and knockdown of MBNL1 and 2 caused a significant shift to the downstream 5'SS ( Figure 6F ) . Overexpression of RBPMSA , also caused a small shift to the downstream 5'SS ( Figure 6F , also detected by rMATs , ΔPSI = 13% , FDR = 2×10−6 ) , but had no effect when MBNL1 and MBNL2 were knocked down . These results are consistent with MBNL1 and 2 being direct regulators of Ncor2 splicing , with RBPMS acting indirectly by promoting production of less active MBNL isoforms . Another post-transcriptional regulator affected by RBPMS is LSM14B , which is involved in cytoplasmic regulation of mRNA stability and translation ( Brandmann et al . , 2018 ) but also shuttles to the nucleus ( Kırlı et al . , 2015 ) . RBPMSA overexpression promoted skipping of Lsm14b exon 6 , a pattern which is also seen in tissue SMCs ( Figure 6G , H ) , and knockdown was also seen to affect this event ( Figure 6G , H ) , with corresponding changes in LSM14B protein ( Figure 6I ) . Exon six contains the only predicted nuclear localization signal ( RPPRRR ) in LSM14B ( Figure 6H ) lying between the LSM and FDF domains . This suggests that RBPMS mediated AS might prevent nuclear shuttling and function of LSM14B in mRNA turnover . Myocardin ( MYOCD ) is a key transcription factor in SMCs and cardiac muscle ( Li et al . , 2003 ) . Skipping of Myocd exon 2a in cardiac muscle and proliferative SMCs produces a canonical mRNA encoding full length MYOCD ( Figure 7A ) ( Creemers et al . , 2006; van der Veer et al . , 2013 ) . Inclusion of exon 2a in differentiated SMCs introduces an in frame stop codon , and the N-terminally truncated Myocd isoform produced using a downstream AUG codon lacks the MEF2 interacting domain and is more potent in activating SMC-specific promoters and SMC differentiation ( Creemers et al . , 2006; Imamura et al . , 2010; van der Veer et al . , 2013 ) . Significant changes in Myocd exon 2a splicing were not detected by rMATS , but manual inspection of RNA-Seq data and RT-PCR confirmed that Myocd exon 2a is more included in differentiated than proliferative PAC1 SMCs ( Figure 7B ) and its inclusion decreases upon RBPMS knockdown in differentiated PAC1 cells ( Figure 7B , Figure 2—figure supplement 2B ) . Effects of lentiviral RBPMS-A overexpression were less clear ( PSI -Dox = 37 . 7 ± 5 . 1 , PSI + Dox = 47 . 1 ± 11 . 9 ) , possibly related to the very low Myocd expression in the transduced lines . A conserved 200 nt region downstream of exon 2a contains two clusters of CAC motifs ( Figure 7C ) , suggesting that RBPMS directly activates exon 2a splicing . To better understand the regulation of the Myocd exon 2a , we created a minigene of exon 2a and its flanking intronic regions . In transfected PAC1 cells exon 2a in the minigene was included to a basal level of ~30% and RBPMS-A or RBPMS-B significantly increased exon 2a inclusion ( Figure 7D ) . The Myocd minigene was also tested in HEK293 cells . As expected , exon 2a was fully skipped in the non-smooth muscle cell line , but was highly responsive to RBPMSA and B with inclusion increasing to ~75% and 37% , respectively ( Figure 7E ) . To test the role of the downstream CAC clusters we mutated each cluster ( CAC to CCC mutations ) . Mutation of cluster 1 ( mCAC ) had a modest effect on activation by RBPMSA , although RBPMSB activity was severely impaired . Mutation of cluster two impaired activity of both RBPMS isoforms , while the combined mutations abolished all activation by RBPMS ( Figure 7E ) . Insertion of a previously defined RBPMS site ( Ube2v1 from Farazi et al . , 2014 ) into the double mutant minigene restored activation by RBPMS-A , although RBPMS-B had minimal activity in this context ( Figure 7G ) . To confirm binding of RBPMS to the CAC motifs , EMSAs and UV crosslinking were carried out with in vitro transcribed RNAs containing the wild-type and mutant CAC clusters ( Figure 7F ) . RBPMS A and B were both able to bind to the wild-type RNA at similar affinities as indicated by EMSA ( apparent Kd ~0 . 5 μM ) ( Figure 7F upper panels ) . Binding was modestly reduced by mutation of the first cluster ( Kd ~0 . 5–2 . 0 μM ) , more severely affected by mutation of the second cluster ( Kd ~2 . 0 μM ) , and eliminated by combined mutation of both clusters . UV crosslinking of RBPMS was inefficient with only a very faint signal evident with wild type , but not mutant , probes ( Figure 7F lower panels ) . These data therefore indicate that RBPMS directly activates Myocd exon 2a inclusion via downstream CAC clusters . Myocd exon 2a is repressed by binding of the RBP QKI to the 5’ end of the exon ( van der Veer et al . , 2013 ) . QKI is expressed more highly in proliferative SMCs ( Llorian et al . , 2016; van der Veer et al . , 2013 ) suggesting that RBPMS and QKI could act antagonistically on AS events during phenotypic switching . To test for functional antagonism , we co-transfected the Myocd minigene with the two RBPs in HEK293 cells ( Figure 7H ) . QKI strongly antagonized RBPMS activitation of exon 2a inclusion , even at low concentrations , showing it to be the dominant regulator ( Figure 7H ) . Thus , splicing of Myocd exon 2a , and thereby the transcriptional activity of Myocd , is under the antagonistic control of RBPs that are preferentially expressed in differentiated ( RBPMS ) or proliferative ( QKI ) SMCs .
By focusing on RBPs whose expression is driven by super-enhancers ( Jangi and Sharp , 2014 ) we identified RBPMS as a key regulator of the differentiated SMC AS program , with many of the criteria expected of a master regulator: i ) it is highly up-regulated in differentiated SMCs ( Figure 1 , Figure 1—figure supplements 1–3 ) . Indeed , single cell RNA-Seq identified Rbpms as part of a transcriptome signature of contractile mouse aorta SMCs cells ( Dobnikar et al . , 2018 ) ; ii ) changes in RBPMS activity appear to be solely responsible for 20% of the AS changes between differentiated and proliferative PAC1 cells ( Figure 3 ) ; iii ) RBPMS target splicing events are enriched in functionally coherent groups of genes affecting cell-substrate adhesion and the actin cytoskeleton , which are important for SMC cell phenotype-specific function ( Figure 5 ) ; iv ) it regulates splicing and activity of other post-transcriptional regulators in SMCs ( Figure 6 ) , and v ) it regulates splicing of the key SMC transcription factor MYOCD ( Figure 7 ) to an isoform that promotes the contractile phenotype ( van der Veer et al . , 2013 ) . RBPMS is reported to be a transcriptional co-activator ( Fu et al . , 2015; Sun et al . , 2006 ) . However , we did not observe changes in expression levels of SMC marker genes upon RBPMS knockdown or overexpression and changes in RNA abundance were outnumbered by regulated AS events ( Figure 2 ) . As a splicing regulator RBPMS can only affect actively transcribed genes , so it is unlikely to be sufficient to initially drive SMC differentiation . Notably , we identified RBPMS using super-enhancers mapped in adult human tissues ( Hnisz et al . , 2013 ) . Combined with the ability of RBPMS to promote AS patterns characteristic of fully differentiated tissue SMCs , such as in Cald1 , Fermt2 and Tns1 ( Figure 3F , G ) , this suggests that RBPMS plays a key role in maintaining a mature adult SMC phenotype . Consistent with a role in promoting a fully differentiated state , in other cell types RBPMS has anti-proliferative tumor-suppressive activity ( Fu et al . , 2015; Hou et al . , 2018; Rastgoo et al . , 2018 ) . RBPMS and RBPMS2 ( referred to as Hermes in Xenopus ) are present across vertebrates , while the related proteins Drosophila Couch Potato and C . elegans MEC-8 bind to similar RNA sequences ( Soufari and Mackereth , 2017 ) . RBPMS and RBPMS2 can localize to the cytoplasm and nucleus , but apart from transcriptional co-regulation ( Fu et al . , 2015; Sun et al . , 2006 ) most attention has been paid to cytoplasmic roles in mRNA stability ( Rambout et al . , 2016 ) , transport ( Hörnberg et al . , 2013 ) and localization in cytoplasmic granules ( Farazi et al . , 2014; Furukawa et al . , 2015; Hörnberg et al . , 2013 ) . RBPMS2 interacts with eukaryote elongation factor-2 ( eEF2 ) in gastrointestinal SMCs , suggesting translational control ( Sagnol et al . , 2014 ) . MEC-8 was reported to regulate splicing of Unc-52 in C . elegans ( Lundquist et al . , 1996 ) , but otherwise RBPMS family members have not been reported to regulate splicing . PAR-CLIP with over-expressed RBPMS in HEK293 cells revealed its preferred binding site , but accompanying mRNA-Seq did not identify regulated AS events associated with PAR-CLIP peaks ( Farazi et al . , 2014 ) . Using rMATS to re-analyze the RNA-Seq data of Farazi et al , we found a small number of regulated AS events , including the Flnb H1 exon ( Figure 4C–F ) . We readily detected specific binding of RBPMS to target RNAs by EMSA , but UV crosslinking was inefficient , even with purified protein and in vitro transcribed RNA ( Figures 4F , J and 7F ) . This suggests that CLIP might significantly under-sample authentic RBPMS-binding sites . Nevertheless , the functional binding of RBPMS to Flnb , Tpm1 and Myocd RNAs ( Figures 4 and 7 ) , the strong enrichment of dual CAC motifs with RBPMS-regulated exons ( Figure 4A ) , the associated requirement for RBPMS RNA binding and dimerization ( Figure 4—figure supplement 3 ) , and distinct positional signatures for activation and repression of splicing ( Figure 4A , B ) , indicate that RBPMS acts widely to directly regulate splicing in differentiated SMCs . While earlier reports have shown that RBPMS can act in other steps in gene expression , our data provide the strongest evidence to date for a widespread molecular function of RBPMS as a splicing regulator . Cell-specific splicing programs are usually driven by more than one regulatory RBPs . A number of splicing regulatory RBPs are known to promote the proliferative SMC phenotype , including QKI , PTBP1 , and SRSF1 ( Llorian et al . , 2016; van der Veer et al . , 2013; Xie et al . , 2017 ) . The extent to which these proteins coordinately regulate the same target ASEs remains to be established . RBPMS and QKI antagonistically regulate at least two targets in addition to Myocd ( Figure 7 ) . The Flnb H1 exon is activated by RBPMS ( Figure 2—figure supplement 3A , Figure 4 ) and repressed by QKI ( Li et al . , 2018 ) , while the penultimate exon of Smtn is repressed by RBPMS ( Figure 2—figure supplement 3A ) but activated by QKI ( Llorian et al . , 2016 ) . Moreover , QKI binding motifs are significantly associated with exons regulated during SMC dedifferentiation , and with exons directly regulated by RBPMS overexpression or knockdown ( data not shown ) . It is therefore an interesting possibility that RBPMS and QKI might target a common set of ASEs perhaps acting as antagonistic master regulators of differentiated and proliferative SMC phenotypes . The two RBPs show reciprocal regulation of expression levels during SMC dedifferentiation ( Figure 1—figure supplement 2B , C ) , which combined with antagonistic activities could lead to switch like changes in many ASEs . For Myocd splicing , QKI appears to have dominant activity , driving skipping of exon 2a even when RBPMS is present at higher levels ( Figure 7G ) , so full inclusion of exon 2a requires the presence of RBPMS and absence of QKI . The logic of this regulatory input can explain inclusion of Myocd exon 2a in SMC and skipping in cardiac muscle and is consistent with the observation that RBPMS is super-enhancer associated both in vascular SMCs and heart left ventricle , while QKI is super-enhancer associated in left ventricle but not differentiated SMCs . Notably , QKI also controls a significant fraction of ASEs regulated during myogenic differentiation of skeletal muscle cells , but it promotes differentiated myotube AS patterns ( Hall et al . , 2013 ) , in contrast to its promotion of dedifferentiated splicing patterns in SMCs . RBPMS was originally characterized as a human gene that encoded multiple isoforms of an RBP ( Shimamoto et al . , 1996 ) , but differential activity of common RBPMS isoforms has not previously been reported . Human RBPMSA and B have similar activity for co-regulation of AP1 transcriptional activity ( Fu et al . , 2015 ) although a third human-specific isoform had lower activity . We found differential activity of RBPMSA and B upon some ASEs ( Figures 4 and 7 ) . In general , RBPMSA has higher activity particularly for repressed targets ( Figure 4F , Figure 4—figure supplements 1A and 2A ) . The differential activity was seen with similar levels of overall expression , although we could not rule out the possibility of variation in nuclear levels as GFP-tagged RBPMS was predominantly cytoplasmic . Nevertheless , some ASEs were differentially responsive to RBPMSA or B while others responded equally ( e . g . compare MPRIP with ACTN1 , Figure 4—figure supplement 1A ) . The differential activity upon Tpm1 splicing could not be accounted for by differences in RNA binding by RBPMSA or B ( Figure 4F , H ) . Therefore , it is possible that RBPMS repressive function depends on other interactions mediated by the 20 amino acid RBPMSA C-terminal ( Figure 1C ) . While the RRM domain is sufficient for RNA binding and dimerization in vitro ( Sagnol et al . , 2014; Teplova et al . , 2016 ) , previous studies have shown the extended C terminal region downstream of the RRM domain to be involved in several aspects of RBPMS/RBPMS2 function including granular localization in retinal ganglion cells ( Hörnberg et al . , 2013 ) and interaction with cFos in HEK293 cells ( Fu et al . , 2015 ) . In vitro assays suggested that the C-terminal region increases RNA-binding affinity and possibly the oligomeric state of RBPMSA ( Farazi et al . , 2014 ) and the C-terminal 34 amino acids of Xenopus RBPMS2 is required for binding to Nanos1 RNA in vivo ( Aguero et al . , 2016 ) . In preliminary studies we have also found the C-terminal region to be essential for splicing regulation ( data not shown ) . Future work will aim to address the mechanisms of RBPMS splicing activation and repression as well as isoform-specific differential activity . SMC phenotypic switching involves interconversion between a contractile phenotype and a motile , secretory , proliferative state ( Owens et al . , 2004 ) . The actin cytoskeleton and its connections to the extracellular matrix ( ECM ) via focal adhesions are central to the function of both cell states , but with contrasting outcomes: tissue-wide contraction or independent movement of individual cells . RBPMS-mediated AS plays a major role in remodelling the actin cytoskeleton , the integrin adhesome ( Horton et al . , 2015 ) and ECM components in the two cell states . This is reflected in the similar GO terms shared by RBPMS-regulated AS events and the entire PAC1 cell AS program and in the morphological changes , including actin fiber reorganization , observed upon RBPMS knockdown in PAC1 cells ( Figure 5 , Figure 5—figure supplements 1 and 2 ) . The importance of many of the RBPMS target genes to SMC function is further indicated by their proximity to super-enhancers in SMC tissues ( Figure 5B , C ) . Indeed , three targets - ACTN1 , FLNB and TNS1 - are all super-enhancer associated , interact directly with both actin and integrins and are components of distinct axes of the consensus integin adhesome ( Horton et al . , 2015 ) . ACTN1 is a major hub in the network of proteins affected by RBPMS ( Figure 5C ) . The AS event in ACTN1 produces a functional Ca2+-binding domain in the motile isoform , but lack of Ca2+ binding in the differentiated isoform stabilizes ACTN1 containing structures in contractile cells ( Waites et al . , 1992 ) . Many other RBPMS-regulated AS events have not previously been characterized . Focal adhesion complexes and the integrin adhesome are mechanosensitive complexes that connect the cytoskeleton and ECM , and are hubs of regulatory tyrosine phosphorylation signalling . We found a small network of RBPMS-regulated AS events in the receptor tyrosine phosphatase PTPRF ( Figure 2E ) and two interacting proteins PPFIA1 and PPFIBP1 ( Figure 2—figure supplement 3A ) . Another focal adhesion associated target is PIEZO1 ( Figure 2E ) , a mechanosensitive ion channel that is important in SMCs during arterial remodelling in hypertension ( Retailleau et al . , 2015 ) . An RBPMS repressed exon lies within the conserved Piezo domain immediately adjacent to the mechanosensing ‘beam’ ( Liang and Howard , 2018 ) . ECM components affected by RBPMS include fibronectin ( FN1 ) , which interacts directly with integrins , and HSPG2 ( Figure 2—figure supplement 3A ) . Notably , HSPG2 ( also known as Perlecan ) is the basement membrane heparan sulfate proteoglycan that is the identified splicing target of MEC-8 in C . elegans ( Lundquist et al . , 1996 ) . In addition to direct regulation of numerous functionally related targets , by directly targeting transcriptional and post-transcriptional regulators RBPMS has the potential for more widespread action ( Figures 6 and 7 ) . RBPMS-regulated events in MBNL1 and 2 modulate their splicing regulatory activity ( Sznajder et al . , 2016; Tabaglio et al . , 2018; Tran et al . , 2011 ) . Changes in secondary AS targets are challenging to observe in a short duration overexpression experiment . Nevertheless , the modest change in NCOR2 splicing appears to be attributable to an RBPMS-induced switch to less active MBNL isoforms ( Figure 6F ) . The regulated event in LSM14B ( Figure 6G–I ) also has the potential to regulate mRNA stability , or an as yet uncharacterized nuclear role of LSM14B ( Kırlı et al . , 2015 ) . We also identified the SMC transcription factor as a direct target of RBPMS ( Figure 7 ) . A similar role has been shown for the proposed myogenic AS master regulator RBM24 ( Jangi and Sharp , 2014 ) , which stabilizes mRNA of the transcription factor Myogenin by binding to its 3'UTR ( Jin et al . , 2010 ) . Similarly , by activating inclusion of Myocd exon 2a ( Figure 7 ) , RBPMS directs production of a Myocardin isoform that more potently promotes the differentiated SMC phenotype ( van der Veer et al . , 2013 ) . Additional effects upon transcription could also be conferred by RBPMS activation of the FLNB H1 exon ( Figure 2—figure supplement 3A , Figure 4 ) . FLNB is primarily an actin binding and adhesion protein , but inclusion of the H1 hinge domain allows nuclear localization and antagonism of the transcription factor FOXC1 in epithelial cells ( Li et al . , 2018 ) . FOXC1 and FOXC2 are expressed at higher levels in adult arteries than any other human tissue ( GTEx Consortium , 2013 ) . The modulation by RBPMS of FLNB isoforms therefore provides another route for indirect transcriptome regulation . The importance of Filamin RNA processing in SMCs by adenosine to inosine editing of FLNA was also recently highlighted by the cardiovascular phenotypes arising from disruption of this editing ( Jain et al . , 2018 ) . A number of splicing regulators also influence miRNA processing ( Michlewski and Cáceres , 2019 ) , so it is an interesting possibility that RBPMS might affect processing of SMC miRNAs such as miR143-145 ( Boettger et al . , 2009; Cordes et al . , 2009 ) . RBPMS2 plays an important role in SMCs of the digestive tract . The RBPMS2 gene is associated with super-enhancers in stomach smooth muscle ( Supplementary file 1 ) and is expressed early in visceral SMC development and at lower levels in mature cells ( Notarnicola et al . , 2012 ) . Ectopic RBPMS2 overexpression led to loss of differentiated contractile function via translational upregulation of Noggin ( Notarnicola et al . , 2012; Sagnol et al . , 2014 ) . This contrasts with our observations that RBPMS exclusively promotes differentiated SMC AS patterns . RBPMS2 is expressed at low levels in PAC1 and primary aorta SMCs ( Figure 1 ) and its knockdown was without effect . Nevertheless , ectopic expression of RBPMS2 in PAC1 or HEK293 cells promoted differentiated SMC AS patterns in a similar manner to RBPMSA ( Figure 4 , Figure 4—figure supplements 1 and 2 ) , suggesting that RBPMSA and RBPMS2 have intrinsically similar molecular activities . The reasons for the apparent discrepancy between the promotion by RBPMS2 of differentiated SMC splicing patterns , but de-differentiated visceral SMC phenotypes , remain to be resolved . Possible explanations include variations in cell-specific signalling pathways , pre-mRNA and mRNA targets , interacting protein partners , post-translational modifications and subcellular localization , all of which could differentially modulate RBPMS and RBPMS2 activity in different SMC types . In conclusion , our data vindicate the proposal that tissue-specific AS master regulators might be identified by the association of their genes with superenhancers ( Jangi and Sharp , 2014 ) , paving the way for the identification of further such regulators in other tissues . While our data suggest that RBPMS has a critical role in SMCs , it is likely to play important roles in other cell types where its expression is also super-enhancer driven , including cardiac muscle and embryonic stem cells . Our approach aimed to identify AS master regulators common to diverse smooth muscle types ( vascular , bladder , stomach ) . However , SMCs show a great deal of diversity ( Fisher , 2010 ) , even within single blood vessels ( Cheung et al . , 2012 ) . The splicing regulator Tra2 β is responsible for some splicing differences between tonic and phasic SMCs ( Shukla and Fisher , 2008 ) , and it is possible that other RBPs might act as master regulators of some of these specialized SMC types . Our future studies aim to understand the mechanisms of splicing regulation by RBPMS , the role of the RBPMS-regulated splicing program in controlling different aspects of SMC phenotype and the potential role of subversion of this program in cardiovascular diseases . While the changes in actin organization observed upon RBPMS knockdown were striking ( Figure 5 ) , it will be important to directly test various functional readouts of SMC physiology related to actin function and cell adhesion , including cell motility and agonist-induced cell contraction . VSMCs can be generated by controlled differentiation from human embryonic stem cells ( Cheung et al . , 2012 ) . In combination with the use of genomic safe harbors to allow inducible protein or shRNA expression , these cells provide an ideal model system in which to test the effects of RBPMS on VSMC differentiation and function . In the longer term , conditional knockout mouse models will provide invaluable insights into the role of RBPMS in vivo .
Locations of human super-enhancers ( genome build Hg19 ) were taken from the data sets UCSD_Aorta , UCSD_Bladder , BI_Stomach_Smooth_Muscle and BI_Skeletal_Muscle in Hnisz et al . ( 2013 ) . Associated genes were obtained using the UCSD Table Browser ( GREAT version 3 . 0 . 0 ) with the Association rule: ‘Basal +extension: 5000 bp upstream , 1000 bp downstream , 1000000 bp max extension , curated regulatory domains included’ . The list of super-enhancer associated genes from the dbSUPER database consists of genes assigned more stringently to super-enhancers within a 50 kb window or where experimental verification was available ( Khan and Zhang , 2016 ) . We used the 1542 human RBPs from Gerstberger et al . ( 2014 ) to identify potential master AS regulators within each set of super-enhancer proximal genes ( Supplementary file 1 ) . Coding sequences of rat Rbpms isoforms were PCR amplified from differentiated PAC1 cell cDNA and cloned into XhoI/EcoRI sites of the pEGFP-C1 vector ( Clontech ) and into EcoRI/XhoI sites of the pCI-neo-3x-FLAG vector ( Rideau et al . , 2006 ) to generate N-terminal Venus and 3xFLAG tagged in vivo overexpression constructs . The two major Rbpms isoforms identified were RBPMS A ( XM_006253240 . 2/XP_006253302 . 1 ) and RBPMS B ( NM_001271244 . 1/NP_001258173 . 1 ) . RNA binding ( K100E ) and dimerization ( R38Q and R38A/E39A ) mutants , previously described in Farazi et al . ( 2014 ) , were generated by site-directed mutagenesis of RBPMS A . QKI construct has been described in a previous study ( Llorian et al . , 2016 ) . Splicing reporters of Tpm1 exon three and Actn1 exon NM and SM were described in Gooding et al . ( 2013 ) ; Gromak et al . ( 2003 ) . Myocd exon 2a and Flnb exon H1 splicing reporters were obtained by PCR amplification of the target exons and respective flanking intron regions from genomic PAC1 DNA , approximately 250 bp upstream and downstream for Myocd and 500 bp for Flnb . PCR products were subsequently cloned into XhoI/EcoRV and NotI/SphI sites of pCAGGs-EGFP vector , which contains a GFP expression cassette with an intron inserted into its second codon ( Wollerton et al . , 2004 ) . Point mutations of the CAC motifs in the splicing reporters were generated by PCR using oligonucleotides that contained A to C mutations . Intronic regions containing CACs from Tpm1 , Flnb and Myocd were PCR amplified and cloned into HindIII/EcoRI sites of pGEM4Z ( Promega ) for in vitro transcription . DNA constructs were confirmed by sequencing . All the oligonucleotides used for cloning and mutagenesis are found in Supplementary file 6 . All cell lines were tested for mycoplasma contamination by RNA capture ELISA , and tested negative . Rat PAC1 pulmonary artery SMCs ( Rothman et al . , 1992 ) were grown to a more differentiated or proliferative state as described in Llorian et al . ( 2016 ) . HEK293 and HEK293T cells were cultured following standard procedures . Rbpms siRNA mediated knockdown in PAC1 cells was performed as in Llorian et al . ( 2016 ) . Briefly , 105 differentiated PAC1 cells were seeded in a six well plate . After 24 hr , cells were transfected using oligofectamine reagent ( Invitrogen ) and 90 pmols of Stealth siRNAs from Thermo Fisher Scientific ( siRNA1: RSS363828 , GGCGGCAAAGCCGAGAAGGAGAACA ) . A second treatment was performed after 24 hr using lipofectamine2000 ( Thermo Fisher ) and siRNA at the same concentration of the first treatment . Total RNA and protein were harvested 48 hr after the second knockdown . C2 scrambled siRNA was used as a control in the knockdown experiments ( Dharmacon , C2 custom siRNA , AAGGUCCGGCUCCCCCAAAUG ) . To assess morphological changes in PAC1 cells upon RBPMS knockdown , the last siRNA treatment with lipofectamine2000 was repeated 48 hr after the second knockdown . PAC1 cells were then assessed 48 hr after the last treatment ( 120 hr after first siRNA treatment ) . For Mbnl1 and Mbnl2 siRNA knockdown , the Mbnl1 THH2 siRNA ( CACGGAAUGUAAAUUUGCAUU ) and Mbnl2 specific siRNA ( Dharmacon , GAAGAGUAAUUGCCUGCUUUU ) were used ( Gooding et al . , 2013 ) . 3xFLAG N-terminally tagged rat RBPMSA cDNA was cloned into pInducer22 ( Meerbrey et al . , 2011 ) using the Gateway system . Generation of stable PAC1 cell lines with pInducer22 vector only ( LV ) or pInducer22-RBPMSA was done as follows . Lentiviral particles were produced in HEK293T cells by transient transfection using 30 µl of Mirus TransIT-lenti ( MIR6604 ) , 7 µg 3xFLAG tagged RBPMSA cDNA and 0 . 75 µg of the packaging plasmids gag , pol , tat and VSV-G transfecting 2 × 106 cells per 10 cm dish . After 24 hr the medium was transferred to 4°C and replaced with fresh medium . After a further 24 hr the medium was removed , pooled with the first batch , spun at 1000 g for 5 min and filtered through 0 . 45 micron PVDF filter . Lentiviral particles were diluted 1:2 with fresh DMEM medium containing Glutamax , 10% FBS and 16 µg/ml polybrene and used to replace the medium on PAC1 cells plated 24 hr earlier at 104/35 mm well , setting up two wells for pInducer22 vector only and six wells for RBPMSA . Fresh medium was added 24 hr later and the populations amplified as necessary . To induce expression of 3xFLAG RBPMSA the cells were plated at 4 × 105 cells per 35 mm well ±1 µg/ml doxycycline harvesting RNA and protein 24 hr later . For transient transfections of HEK293 cells with splicing reporters and effectors , lipofectamine 2000 reagent was used and cells harvested 48 hr after transfection . To verify knockdown and transfection efficiency , total cell lysates were obtained by directly adding protein loading buffer to the cells . Lysates were run on a SDS-PAGE , followed by western blot against RBPMS and loading controls . See Supplementary file 6 for information on the antibodies used in this study . To monitor changes in splicing and mRNA abundance , RNA was extracted using TRI reagent ( Sigma ) according to manufacturer’s instructions , DNase treated with Turbo DNA-free kit ( Thermo Fisher ) and cDNA synthesized , as described below , followed by PCR and QIAxcel or qRT-PCR analysis . cDNA was prepared using 1 μg total RNA , oligo ( dT ) or gene-specific oligonucleotides and SuperScript II ( Life technologies ) or AMV RT ( Promega ) as described in manufacturer’s protocol . qRT-PCR reactions were prepared with 50 ng of cDNA , oligonucleotides for detection of mRNA abundance and SYBER Green JumpStart Taq Ready Mix ( Sigma ) . Three-step protocol runs were carried out in a Rotor-Gene Q instrument ( QIAGEN ) . Analysis was performed in the Rotor-Gene Q Series Software 1 . 7 using the Comparative Quantitative analysis . To normalize the relative expression values , two housekeeper genes were included in each experiment ( Gapdh and Rpl32 ) and their geometric mean used for normalization . Expression values were acquired from biological triplicates . PCRs with 50 ng of the prepared cDNAs were carried out to detect the different mRNA splicing isoforms of the reporters using the oligonucleotides detailed in Supplementary file 6 . For visualization and quantification of the PSI values , PCR products were resolved in a Qiaxcel Advanced System ( QIAGEN ) and PSI calculated within the QIAxcel ScreenGel software . A minus RT cDNA , representative of each triplicate , and a no template PCR reactions were also included in all the experiments ( data not shown ) . Statistical significance was tested as previously described for gene expression ( paired Student t-test for lentiviral experiments ) . PSI values are shown as mean ( % ) ± standard deviation ( sd ) . For lentiviral RBPMSA overexpression , ΔPSI values determined by RT-PCR were derived from at least three independent transductions , and no events responded to doxycycline treatment in cells transduced with the empty pINDUCER22 vector ( Figure 2—figure supplement 2A ) . Statistical significance was tested by a two-tailed Student’s t test , paired for lentiviral experiments and unpaired for all the others ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . For immunodetection of RBPMS in PAC1 cells , differentiated and proliferative cells were grown on coverslips , fixed with 4% paraformaldehyde ( PFA ) for 5 min , rinsed with phosphate buffer saline ( PBS ) and permeabilized with 0 . 5% NP-40 for 2 min followed by PBS washes . Coverslips were incubated with blocking buffer ( 1% BSA in PBS ) for 1 hr and incubated with RBPMS primary antibody diluted in blocking buffer for another hour . Coverslips were rinsed and secondary antibody in blocking buffer applied to the coverslips which were incubated for 1 hr . Coverslips were washed and mounted on ProLong Diamond Antifade with DAPI ( Thermo Fisher Scientific ) . For staining of actin fibers , cell were incubated with Alexa Fluor 488 phalloidin ( Invitrogen , A12379 ) , which was added at the secondary antibody incubation step at a 1:1000 dilution . All the steps were carried out at room temperature . Images were acquired from a fluorescence microscope ( Zeiss Ax10 , 40X ) attached to CCD AxioCam and analyzed on AxioVision ( v4 . 8 . 2 ) . All the imaging analyses were carried out using the ImageJ free software ( https://imagej . nih . gov/ij/download . html ) . To calculate the actin anisotropy , the FibrilTool plugin in ImageJ was used ( Boudaoud et al . , 2014 ) . Phalloidin stained images were divided into four regions of interest and analysis performed for each area . To calculate nucleus size , DAPI stained images were analyzed using the following ImageJ commands: Adjust color threshold/Make Binary/Analyze Particles . The area values ( Pixels ) and count of nucleus in the field were then obtained for each field . To calculate the average of the cell size , first the fraction of the field occupied by the ACTIN staining was calculated using ImageJ ( Adjust color threshold/Make Binary/Measure ) . The percentage of the area of ACTIN staining was divided by the number of cells in the respective field to obtain the average of the cell size . Statistical significance was tested using Mann-Whitney-Wilcoxon Test ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . Statistical analyses and data visualization were carried out in RStudio ( http://www . rstudio . com/ ) . Total RNA from three biological replicates of RBPMS knockdown in differentiated PAC1 and three populations of RBPMS inducible overexpression in proliferative PAC1 cells were extracted for RNAseq . Cells were lysed with TRI-reagent and total RNA purified by Direct-zol purification column ( Zymo Research ) followed by DNase treatment . RNAseq libraries of polyA selected RNAs were prepared with NEBNext ultra II RNA library prep kit for Illumina . Both RNA and RNAseq libraries were checked for their qualities . Barcoded RNAseq libraries were then multiplexed across two lanes of an Illumina HiSeq4000 platform for sequencing on a 150 bp paired-end mode , providing around 60 million reads per sample . To investigate AS changes in vascular smooth muscle dedifferentiation , rat aortas were isolated from 8 to 12 weeks old Wistar rats . Aortas were briefly treated for 30 min at 37°C with 3 mg/ml collagenase ( Sigma C-0130 ) to help in cleaning away the adventitia . The tissue was finely chopped and either used directly to make tissue RNA or enzymatically dispersed to single cells . This was achieved by treating the tissue pieces with 5 ml 1 mg/ml elastase ( Worthington Biochemical Corporation LS002292 ) for 30 min at 37°C and then 5 ml collagenase added for a further 1–2 hr . Cells were washed and counted and plated at 4 × 105 cells/ml in M199 media containing 10%FBS , 2 mM Glutamine and 100 U/ml Penicillin-Streptomycin in a suitable dish according to the cell number . To promote the proliferative state , SMCs were 1:2 passaged switching to DMEM media containing Glutamax and 10% FBS and harvested at passage 9 . For RNAseq , total RNA was harvested from three replicas , each a pool of 5 rats , from rat aorta tissue ( T ) , enzymatically dispersed single cultured SMCs ( SC ) , passage 0 ( P0 ) and passage 9 ( P9 ) . Total RNA extraction was carried out with Tri-reagent ( Sigma ) . Libraries for mRNAseq were prepared using Ribozero and TrueSeq kits and sequencing performed on a HiSeq2000 platform in a paired-end mode . Read trimming and adapter removal were performed using Trimmomatic version 0 . 36 ( Bolger et al . , 2014 ) . Reads were aligned using STAR version 2 . 5 . 2a ( Dobin et al . , 2013 ) to the Rat genome Rnor_6 . 0 release-89 obtained from Ensembl and RSEM package version 1 . 2 . 31 ( Li and Dewey , 2011 ) was used to obtain gene level counts . mRNA abundance analysis , was carried out with DESeq2 package ( version 1 . 18 . 1 ) ( Love et al . , 2014 ) within R version 3 . 4 . 1 ( https://www . r-project . org/ ) . Genes were considered to be differential expressed with p-adj less than 0 . 05 in the paired analysis ( Supplementary file 2 ) . rMATS v3 . 2 . 5 ( Shen et al . , 2014 ) was used for detection of differential alternative splicing . rMATS analysis was carried out allowing for new splicing event discovery using the flag novelSS 1 . rMATS calculates the inclusion levels of the alternative spliced exons and classifies them into five categories of AS events ( Figure 2—figure supplement 1B ) : skipped exons ( SE ) , mutually exclusive exons ( MXE ) , alternative 5’ and 3’ splice sites ( A5SS and A3SS ) and retained intron ( RI ) . Before further analysis , the results from reads on target and junction counts were filtered to include only events with a total of read counts above 50 across the triplicates in at least one of the conditions compared . Removal of events with low counts discarded false positive events . Only ASE with an FDR less than 0 . 05 were considered significant and a minimal inclusion level difference of 10% imposed to significant AS events . Finally , to identify specific AS events , unique IDs were created using the AS type , gene name and the chromosomal coordinates of the regulated and flanking exons ( Supplementary file 3 ) . For visualization of differentially spliced exons , sashimi plots were generated using rmats2sashimiplot ( Gohr and Irimia , 2019 ) . The sashimi plots show the RNAseq coverage reads mapping to the exon-exon junctions and PSI values from rMATS . Twenty eight ASE identified by rMATS in the RBPMS knockdown or overexpression were also validated by RT-PCR in the same manner as described in the qRT-PCR and RT-PCR section . ΔPSI predicted from the RNAseq analysis and the ΔPSI observed in the RT-PCR were then tested for a Pearson correlation in RStudio ( http://www . rstudio . com/ ) . For comparison and visualization of the overlap between the genes with different mRNA abundance and the genes differentially spliced in the RBPMS knockdown , RBPMS overexpression and the PAC1 dedifferentiation , proportional Venn diagrams were made using BioVenn ( Hulsen et al . , 2008 ) . Venn diagrams were also generated for the visualization of the common AS events across RBPMS knockdown , overexpression and the PAC1 or aorta tissue dedifferentiation datasets . RBPMS motif enrichment analyses in the regulated cassette exons ( SE ) of RBPMS knockdown and overexpression , PAC1 and Aorta tissue dedifferentiation were performed using the toolkit MATT ( Gohr and Irimia , 2019 ) . Cassette exons in transcripts identified by rMATS with significant changes ( FDR < 0 . 05 and |ΔPSI| > 10% ) were used to test enrichment or depletion of CACN1-12CAC RBPMS recognition element ( Farazi et al . , 2014 ) , against a background set of unregulated exons defined as events with FDR > 0 . 1 and |ΔPSI| < 5% . 250 bp of the flanking intronic regions were examined for RBPMS motif signals . The motif enrichment scores were first obtained using the test_regexp_enrich command of the Matt suite in the quant mode with statistical significance determined using a permutation test with 50 , 000 iterations . This module inherently divides the examined regions ( exons or 250 bp of the introns ) into thirds and provides positional information for the occurrence of the RBPMS motif . RNA maps for distribution of the RBPMS motif were generated using the rna_maps command in the Matt suite . Here the unregulated set of exons was randomly downsampled to include a final background of 2000 events only . Additionally , the program was instructed to scan only 135 bp of the cassette exon and 250 bp at either end of the flanking introns . A sliding window of 31 was used for scanning the motif and the statistically significant regions ( p<0 . 05 ) for enrichment or depletion were identified by the permutation tests specifying 1000 iterations . Enrichment for gene ontology terms in the differentially abundant and spliced genes were obtained from Gorilla ( Eden et al . , 2009 ) . Two unranked lists of genes , target and background lists , were used for the GO analysis . The target list contained either the significant differential abundant genes ( p-adj <0 . 05 and log2 Fold Change greater than one or less than −1 ) or the significant differentially spliced genes ( FDR < 0 . 05 and ΔPSI threshold of 10% ) . Background gene lists were created for the PAC1 experiments and aorta tissue by selecting the genes whose expression were higher than 1 TPM in either of the conditions analyzed . For visualization , only the top five enriched GO terms of each category ( biological process , cell component and molecular function ) were shown in Figure 5A and Figure 5—figure supplement 1 . The complete list of enriched terms in differentially spliced and abundant genes can be found in Supplementary files 4 and 5 . A Protein-protein interaction network for the genes differentially spliced by RBPMS was constructed using the STRING v10 . 5 database ( Szklarczyk et al . , 2017 ) . RBPMS-regulated genes were obtained by merging two lists: i ) overlap of genes concordantly differentially spliced in the RBPMS knockdown and PAC1 experiments and , ii ) overlap of genes concordantly differentially spliced in the RBPMSA overexpression and the aorta tissue datasets that were shown to be regulated in the same range in both conditions . A cut off of ΔPSI greater than 10% was also applied to the RBPMS-regulated gene list , similar to the GO analysis . The human database was chosen for the analysis and the following parameters applied to the PPI network: confidence as the meaning of the network edges , experiments and database as the interaction sources and high confidence ( 0 . 700 ) as the minimum required interaction score . STRING functional enrichments , using the whole genome as statistical background , were also included for visualization . Human super-enhancer associated genes from Supplementary file 1 were highlighted . Rat RBPMS A and B with a 3xFLAG N terminal tag were cloned into the BamHI/XhoI sites of the expression vector pET21d , for expression of recombinant RBPMS containing a T7 N-terminal tag and a His6 C-terminal tag in E . coli . Recombinant RBPMS A protein was purified using Blue Sepharose six and HisTrap HP columns whereas RBPMS B was purified only through the latter , since low binding was observed to Blue Sepharose 6 . The identity of purified recombinant proteins was confirmed by western blot ( Figure 4—figure supplement 4 ) and mass mapping by mass spectrometry . α32P-UTP labelled RNA probes were in vitro transcribed using SP6 RNA polymerase . For EMSAs , a titration of the recombinant RBPMS A and B ( 0 , 0 . 125 , 0 . 5 and 2 μM ) was incubated with 10 fmol of in vitro transcribed RNA in binding buffer ( 10 mM Hepes pH 7 . 2 , 3 mM MgCl2 , 5% glycerol , 1 mM DTT , 40 mM KCl ) for 25 min at 30°C . After incubation , samples were run on a 4% polyacrylamide gel . For UV-crosslinking experiments , the same binding incubation was performed followed by UV-crosslink on ice in a Stratalinker with 1920 mJ . Binding reactions were then incubated with RNase A1 and T1 at 0 . 28 mg/ml and 0 . 8 U/ml respectively , for 10 min at 37°C . Prior to loading the samples into a 20% denaturing polyacrylamide gel , SDS buffer was added to the samples which were then heated for 5 min at 90°C . Analysis and quantification of RNAseq , RT-PCR and imaging experiments were described in their respective sections with further information of the tests used in the different experiments present in the figure legends . Graphics were generated in RStudio ( http://www . rstudio . com/ ) . mRNAseq of RBPMS ( knockdown and overexpression ) and Aorta tissue dedifferentiation data from this study have been deposited in NCBI Gene Expression Omnibus ( GEO ) repository under GEO accession GSE127800 , accession number GSE127799 and GSE127794 , respectively . | All the cells in our body contain the same genetic information , but they only switch on the genes that they need to fulfill their specific role in the organism . Genetic sequences known as enhancers can turn on the genes that are required by a particular cell to perform its tasks . Once a gene is activated , its sequence is faithfully copied into a molecule of RNA which contains segments that code for a protein . A molecular machine then processes the RNA molecule and splices together the coding segments . RNA binding proteins can also regulate this mechanism , and help to splice the coding sections in different ways depending on the type of cell . The process , known as alternative RNA splicing , therefore creates different RNA templates from the same gene . This gives rise to related but different proteins , each suited to the needs of the particular cell in which they are made . However , in some cell types , exactly how this happens has not yet been well documented . For example , in cells that line blood vessels – known as vascular smooth muscle cells – the RNA binding proteins that drive alternative splicing have not been identified . To find these proteins , Nakagaki-Silva et al . used catalogs of DNA regions called super-enhancers as clues . These sequences strongly activate certain genes in a tissue-specific manner , effectively acting as labels for genes important for a given cell type . In vascular smooth muscle cells , if a super-enhancer switches on a gene that codes for a RNA-binding protein , this protein is probably crucial for the cell to work properly . The approach highlighted a protein called RBPMS , and showed that it controlled alternative RNA splicing of many genes important in smooth muscle cells . This may suggest that when RBPMS regulation is disrupted , certain diseases of the heart and blood vessels could emerge . Finally , the results by Nakagaki-Silva et al . demonstrate that super-enhancers can signpost genes important in regulating splicing or other key processes in particular cell types . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2019 | Identification of RBPMS as a mammalian smooth muscle master splicing regulator via proximity of its gene with super-enhancers |
Long-range cis-regulatory elements such as enhancers coordinate cell-specific transcriptional programmes by engaging in DNA looping interactions with target promoters . Deciphering the interplay between the promoter connectivity and activity of cis-regulatory elements during lineage commitment is crucial for understanding developmental transcriptional control . Here , we use Promoter Capture Hi-C to generate a high-resolution atlas of chromosomal interactions involving ~22 , 000 gene promoters in human pluripotent and lineage-committed cells , identifying putative target genes for known and predicted enhancer elements . We reveal extensive dynamics of cis-regulatory contacts upon lineage commitment , including the acquisition and loss of promoter interactions . This spatial rewiring occurs preferentially with predicted changes in the activity of cis-regulatory elements and is associated with changes in target gene expression . Our results provide a global and integrated view of promoter interactome dynamics during lineage commitment of human pluripotent cells .
Cell fate decisions are associated with profound changes in chromatin organisation , which underlie the activation of lineage-specific and the silencing of lineage-inappropriate genes ( Buecker and Wysocka , 2012; Bulger and Groudine , 2010; Calo and Wysocka , 2013; Hallikas et al . , 2006; Ong and Corces , 2012 ) . Cis-regulatory elements such as transcriptional enhancers play a key role in this process by integrating regulatory inputs from intrinsic and extracellular cues , and by mediating the recruitment of core activator and repressor complexes ( Pennacchio et al . , 2013; Shlyueva et al . , 2014; Spitz and Furlong , 2012 ) . The definition of chromatin signatures has enabled the genome-wide identification of enhancer elements across multiple human cell types ( ENCODE Project Consortium , 2012; Heintzman et al . , 2007 , 2009; Pennacchio et al . , 2006; Rada-Iglesias et al . , 2011; Kundaje et al . , 2015 ) . Chromatin states can also provide a robust readout of cis-regulatory activity associated with poised and active enhancers and have been used to show that widespread changes in enhancer position and activity occur upon cell fate decisions such as the lineage commitment of pluripotent cells ( Creyghton et al . , 2010; Rada-Iglesias et al . , 2011; Zentner et al . , 2011 ) . Cis-regulatory elements are often considerable distances away from their target gene promoters and may not control their nearest genes ( Carvajal et al . , 2001; Jeong et al . , 2006; Marinić et al . , 2013; Pennacchio et al . , 2006; Ruf et al . , 2011; Sagai et al . , 2005; Spitz et al . , 2003 ) . It is generally accepted that this long-range action is facilitated by DNA-looping interactions ( Pennacchio et al . , 2013; Shlyueva et al . , 2014 ) . However , specific determinants of chromosomal interactions are still not fully understood , which presents challenges for the high-confidence prediction of regulatory interactions from sequence and epigenetic information ( Mora et al . , 2016; Roy et al . , 2015; Shlyueva et al . , 2014; Whalen et al . , 2016 ) . As a result , the target genes of most cis-regulatory elements remain unknown . Furthermore , while it is generally accepted that many genes are controlled by multiple regulatory elements ( Barolo , 2012; Miguel-Escalada et al . , 2015 ) our understanding of multi-modular gene regulation remains limited , particularly in the context of mammalian development and stem cell differentiation . Over the last decade , chromosome conformation capture ( 3C ) and derived methods have enabled the biochemical mapping of looping interactions to offer new insights into their architecture across different cell types ( Dekker et al . , 2013; de Laat and Duboule , 2013; Schmitt et al . , 2016 ) . In particular , Hi-C has allowed genome-wide characterisation of higher order chromatin dynamics during differentiation at the level of contact domains , including A/B compartments and topologically associated domains ( TADs ) ( Fraser et al . , 2015; Dixon et al . , 2015 ) . The complexity of Hi-C samples creates challenges for a comprehensive identification of individual enhancer-promoter loops using this technology . However , analyses focusing on candidate-interacting regions or those bound by specific proteins ( such as cohesin and RNA polymerase II ) have made it possible to detect subsets of promoter-enhancer interactions at high resolution ( Heidari et al . , 2014; Li et al . , 2015; Sanyal et al . , 2012 ) and delineate their dynamics during cell differentiation and reprogramming . These studies provided evidence of interactions associated with transcriptional changes upon lineage commitment ( Denholtz et al . , 2013; Kieffer-Kwon et al . , 2013; Phillips-Cremins et al . , 2013; Zhang et al . , 2013 ) , as well as revealed interactions formed in anticipation of changes in gene activity ( Apostolou et al . , 2013; Ghavi-Helm et al . , 2014; Wei et al . , 2013 ) . However , despite these advances , the global and unbiased high-resolution mapping of promoter cis-regulatory interactions that form and remodel during development and stem cell differentiation is still lacking . This hampers an integrated understanding of the cis-regulatory logic underlying transcriptional decisions during lineage commitment . Recently , we developed Promoter Capture Hi-C that uses sequence capture to enrich Hi-C libraries for interactions involving the promoters of most annotated genes , providing a global view on promoter interactions that is independent of the activity of interacting regions and identity of proteins recruited to them ( Mifsud et al . , 2015; Schoenfelder et al . , 2015a ) . Here , we use PCHi-C in human embryonic stem cells ( ESCs ) and ESC-derived neuroectodermal cells ( NECs ) ( Bajpai et al . , 2009 ) to create a high-resolution resource of promoter contacts and their dynamics during early lineage commitment in the context of extensive chromatin changes that occur at the interacting cis-regulatory regions as the cells differentiate ( Rada-Iglesias et al . , 2011 ) . Our large-scale dataset links thousands of known and predicted enhancer elements with their putative target genes in human pluripotent and early lineage-committed cells , including those known to drive tissue-restricted reporter gene expression in transgene assays . We integrate the promoter interacting regions of each gene to define cis-regulatory units ( CRUs ) that provide a view of multi-modular gene regulation . We show that CRUs undergo extensive reorganisation during lineage commitment that involves both the ‘rewiring’ ( acquisition or loss ) of promoter interactions , as well as chromatin state changes at pre-existing interactions . Importantly , we demonstrate that this reorganisation is associated with changes in target gene expression , thereby providing insights into the transcriptional control of early human development .
We used PCHi-C to profile the interactomes of 21 , 841 promoters in human ESCs and NECs ( Figure 1A ) . We generated NECs using an established protocol ( Rada-Iglesias et al . , 2011 ) ( Figure 1—figure supplement 1A ) and confirmed efficient differentiation by flow cytometry analysis and RNA-sequencing ( Figure 1—figure supplement 1B , C ) . PCHi-C data normalisation and signal detection using the CHiCAGO pipeline ( Cairns et al . , 2016 ) identified 75 , 795 significant cis-interactions between promoters and other genomic regions in ESCs and 75 , 624 in NECs . In addition , approximately 300 significant trans-interactions were detected in each cell type . As examples of this rich dataset , high-confidence interactions are shown for the SOX2 and PAX6 promoters ( Figure 1B and Figure 1—figure supplement 2A ) . These examples illustrate the multiple promoter-contacts observed , alongside the conventional Hi-C profiles additionally generated in this study that reveal higher-order genome topology over the same region . Overall , PCHi-C samples showed an 11 to 15-fold enrichment for promoter-containing interactions over conventional Hi-C . This data resource provides a global , high-resolution atlas of chromosomal interactions in human pluripotent and lineage-committed cells . Processed datasets have been made available through Open Science Framework ( http://osf . io/sdbg4 ) , and raw sequencing reads have been deposited to Gene Expression Omnibus ( accession GSE86821 ) . 10 . 7554/eLife . 21926 . 003Figure 1 . A resource of high-resolution promoter interactions in human embryonic stem cells ( ESCs ) and ESC-derived neuroectodermal cells ( NECs ) . ( A ) Overview of the experimental design . Human embryonic stem cells ( ESCs ) and ESC-derived neuroectodermal progenitors ( 1 ) were analysed with Promoter Capture Hi-C to profile interactions involving 21 , 841 promoter-containing HindIII fragments ( 2 ) . Signal detection with the CHiCAGO pipeline revealed ~75 , 000 high-confidence promoter interactions in each cell type ( 3 ) . These data were integrated with histone modification and gene expression profiles in the same cells ( 4 ) to study chromatin and interaction dynamics during lineage commitment . Characterisation of ESCs and NECs is shown in Figure 1—figure supplement 1 . ( B ) Genome browser representation of the SOX2 promoter interactome in ESCs ( upper ) and NECs ( lower ) . Significant interactions are shown as purple arcs , with one end of the interaction within the SOX2 promoter and the other end at a promoter-interacting region ( PIR ) . ChIP-seq ( H3K27me3 , H3K27ac , H3K4me1 , H3K4me3; from [Rada-Iglesias et al . , 2011] ) and mRNA-seq tracks are shown . Chromatin states for each genomic region were defined by ChromHMM ( Ernst and Kellis , 2012 ) using ChIP-seq data ( active chromatin , green; poised chromatin , orange; Polycomb-associated chromatin , red; intermediate , yellow; background , grey ) . Conventional Hi-C heatmaps of contact frequencies reveal chromatin topology over this region . As an additional example , the PAX6 promoter interactome is shown in Figure 1—figure supplement 2 . Read count interaction profiles for SOX2 and PAX6 are shown in Figure 1—figure supplement 4 . ( C ) PIRs are significantly enriched in regions that contain histone marks associated with putative regulatory functions , compared with promoter distance-matched control regions ( permutation test p-value<0 . 01 for each mark ) ( ESCs , left; NECs , right ) . Blue bars show the number of overlaps observed in detected PIRs , and grey bars show the mean number of overlaps observed in distance-matched random regions over 100 permutations . Error bars show 95% confidence intervals across permutations . ( D ) Promoters and their associated PIRs show significant concordance in chromatin states . Heatmaps show the log2 odds ratios for the co-occurrence of each combination of promoter and PIR chromatin state compared with that expected at random . p-Values are from Pearson’s χ2 test on the corresponding contingency tables . Clustering of chromatin states and additional examples of promoter interactomes are shown in Figure 1—figure supplement 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 21926 . 00310 . 7554/eLife . 21926 . 004Figure 1—figure supplement 1 . Characterisation of ESCs and NECs . ( A ) Phase contrast images of undifferentiated ESC colonies ( left ) and day 7 NEC spheres ( right ) . ( B ) Flow cytometry analysis of ESCs ( blue ) and NECs ( red ) using lineage-specific cell surface markers . CD56 is expressed by ESCs and NECs; EPCAM ( CD326 ) is expressed by ESCs but not NECs ( Gifford et al . , 2013 ) . Percent positive cells in each quadrant is shown . ( C ) Genome browser representations of RNA-seq data from our study and from ( Rada-Iglesias et al . , 2011 ) shows expression levels of the ESC-specific genes POU5F1 and NANOG; of SOX2 and CDH2 , which are expressed by both ESCs and NECs , and of the NEC-specific genes ZIC4 and PAX6 . DOI: http://dx . doi . org/10 . 7554/eLife . 21926 . 00410 . 7554/eLife . 21926 . 005Figure 1—figure supplement 2 . PAX6 promoter interactome and CTCF enrichment at PIRs . ( A ) Genome browser representation of the PAX6 promoter interactome in ESCs ( upper ) and NECs ( lower ) . Significant interactions are shown as purple arcs . ChIP-seq ( H3K27me3 , H3K27ac , H3K4me1 , H3K4me3; from [Rada-Iglesias et al . , 2011] ) and mRNA-seq tracks are shown . Chromatin states for each genomic region were defined by ChromHMM ( Ernst and Kellis , 2012 ) using ChIP-seq data ( active chromatin , green; poised chromatin , orange; Polycomb-associated chromatin , red; intermediate , yellow; background , grey ) . Conventional Hi-C heatmaps of contact frequencies are shown for ESCs and NECs and reveal chromatin topology over this region . ( B ) PIRs in ESCs are significantly enriched in regions that contain CTCF binding sites , compared with promoter distance-matched random control regions ( permutation test p-value<0 . 01 ) . The blue bar shows the number of overlaps observed in detected PIRs , and the grey bar shows the mean number of overlaps observed in distance-matched random regions over 100 permutations . Error bars show 95% confidence intervals across permutations . DOI: http://dx . doi . org/10 . 7554/eLife . 21926 . 00510 . 7554/eLife . 21926 . 006Figure 1—figure supplement 3 . Integrated view of chromatin states and PCHi–C data . ( A ) Definition of 16 ChromHMM emission states based on the relative presence or absence of signals from H3K27me3 , H3K4me1 , H3K27ac and H3K4me3 ChIP-seq data . Similar states were pooled to form six categories defined as active , poised , Polycomb , mixed , intermediate and background ( see Materials and methods for details ) . ( B ) Integration of promoter interactions with chromatin states in ESCs ( upper tracks ) and NECs ( lower tracks ) for six different promoters . Purple arcs indicate significant interactions between the baited promoters and PIRs . Open circles denote background interaction signals and filled circles identify significant interactions ( i . e . those exceeding the defined threshold CHiCAGO score , see Materials and methods ) . Filled circles have been coloured according to the chromatin state of the PIR ( active , green; poised , orange; Polycomb-associated , red; intermediate , yellow; background , grey ) . The arrows correspond to the transcriptional state of the baited gene in ESCs and NECs ( sharp end , active; blunt end , inactive ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21926 . 00610 . 7554/eLife . 21926 . 007Figure 1—figure supplement 4 . Read-count interaction profiles for baited promoters presented in Figures 1–3 . Plots show the read counts corresponding to the interactions of baited promoter fragments ( grey line ) with other HindIII fragments . Significant interactions detected by CHiCAGO ( score ≥12 ) are shown in red , and sub-threshold interactions ( score ≥11 in the cell type shown , and score ≥12 in the other cell type ) are shown in blue . ( A ) SOX2 , from Figure 1B . ( B ) PAX6 , from Figure 1—figure supplement 2A . ( C ) POU3F2 , from Figure 2A . ( D ) SNAI2 , GLI2 , PRDM1 and POU3F1 , from Figure 3E–H . DOI: http://dx . doi . org/10 . 7554/eLife . 21926 . 007 To gain insight into the chromatin properties of the promoter-containing interactions , we integrated the PCHi-C data with published genome-wide histone modification profiles in ESCs and NECs ( Rada-Iglesias et al . , 2011 ) . In both cell types , promoter-interacting regions ( PIRs ) were significantly enriched for histone marks that are associated with regulatory functions ( Figure 1C ) , including H3 lysine 4 monomethylation ( H3K4me1 ) and H3 lysine 27 acetylation ( H3K27ac ) , which can identify enhancers in human cell types , as well as H3 lysine four trimethylation ( H3K4me3 ) , which is associated with transcriptional activation , and H3 lysine 27 trimethylation ( H3K27me3 ) , which is associated with Polycomb-mediated transcriptional repression ( Di Croce and Helin , 2013; Heintzman et al . , 2007 , 2009 ) . PIRs were also significantly enriched for sites bound by the architectural protein CTCF ( Figure 1—figure supplement 2B; based on ENCODE data available for ESCs only ) , consistent with previous observations in other cell types ( Jin et al . , 2013; Phillips-Cremins et al . , 2013; Sanyal et al . , 2012 ) . We used ChromHMM ( Ernst and Kellis , 2012 ) to integrate these histone marks and to define four combinatorial chromatin states in both ESCs and NECs , as follows: active ( characterised by H3K4me3 and/or H3K27ac ) ; Polycomb-associated ( H3K27me3 ) ; poised ( H3K4 methylation and H3K27me3 ) ; and background ( no detectable signal for the tested histone modifications ) ( Figure 1—figure supplement 3A; see Materials and methods for details ) . Overall , we detected just under 20 , 000 PIRs in each cell type that harboured either active ( 12 , 250 in ESCs and 7510 in NECs ) , Polycomb-associated ( 3505 in ESCs and 5856 in NECs ) or poised ( 2274 in ESCs and 4262 in NECs ) chromatin state signatures , connecting a large set of putative regulatory sequences in human pluripotent and lineage committed cells to their target promoters . In addition , 25 , 727 PIRs in ESCs and 20 , 016 PIRs in NECs were in the background state . The chromatin states of several example promoters , including those for the POU5F1 , PRDM14 and CHD7 genes , together with each of their respective PIRs , are shown in Figure 1—figure supplement 3B . When analysing the whole dataset , we found a significant concordance between the chromatin states at promoters and their PIRs ( Figure 1D ) , which is in line with previous studies in other human cell types ( Jin et al . , 2013; Mifsud et al . , 2015; Sanyal et al . , 2012 ) and provides validation of our dataset . Notably , poised and Polycomb-associated promoters showed similar interaction preferences for PIRs in either of these two states ( Figure 1D ) . This finding suggests that poised and Polycomb-associated regions are broadly interchangeable in terms of their interaction affinities , which is consistent with a key role for Polycomb-group proteins in mediating interactions in the poised state ( Schoenfelder et al . , 2015b ) . Taken together , these data provide a comprehensive resource that links many thousands of known and predicted regulatory elements with their putative target genes and will enable the investigation of regulatory contacts during human lineage commitment . The enrichment of PIRs for specific chromatin regulatory features points to the presence of functional enhancer elements at these regions that could potentially provide inputs to the promoters they contact . To assess the enhancer activity of the identified PIRs , we examined whether they were known to efficiently drive reporter gene expression in embryonic day 11 . 5 mouse embryos based on information from the VISTA Enhancer Browser ( Visel et al . , 2007 ) . As an initial example , we focused on the 39 PIRs detected in NECs that interact with the promoter of the neural transcription factor POU3F2 . Strikingly , four out of the five POU3F2 PIRs tested experimentally in VISTA transgenic assays showed reporter activity exclusively in neural tissues , and one PIR was inactive ( Figure 2A ) . Furthermore , the mRNA expression pattern of mouse Pou3f2 broadly matched the combined tissue-restricted pattern of the tested human POU3F2 PIRs ( Figure 2A ) . These results highlight how PCHi-C can contribute to our understanding of the cis-regulatory networks for key developmental genes . 10 . 7554/eLife . 21926 . 008Figure 2 . Promoter-interacting regions can function as tissue-restricted developmental enhancers and identify associated target genes . ( A ) A genome browser representation of the POU3F2 promoter interactome in NECs . Genome coordinates are shown underneath . Chromatin states are indicated ( active chromatin , green; poised chromatin , orange; Polycomb-associated chromatin , red; intermediate , yellow; background , grey ) . Significant interactions are shown as coloured arcs . Five of the identified POU3F2 PIRs have been tested experimentally using a transgenic reporter assay as part of the VISTA Enhancer Browser ( Visel et al . , 2007 ) . Of those five , four regions ( indicated by blue arcs ) can drive tissue-restricted LacZ expression in E11 . 5 mouse embryos . Representative images of X-gal stained mouse embryos are shown for each sequence . These show neural-restricted enhancer activity within the forebrain , midbrain , hindbrain and neural tube , which are tissues derived from NECs . The mRNA expression pattern of Pou3f2 in an E10 . 5 mouse embryo ( EMAGE gene expression database; EMAGE:1689; [Richardson et al . , 2014] ) broadly matches the combined tissue-restricted pattern of its enhancers . One experimentally tested PIR ( indicated by brown arc ) is inactive at this developmental stage in mouse embryos . ( B ) PIRs identified in NECs are enriched for sequences that can drive reporter gene activity in neural tissues and other neuroectodermal derivatives ( see also Figure 2—figure supplement 1A ) . The barplot shows the distribution of tissue-specific reporter expression patterns for all experimentally tested PIRs identified in ESCs ( n = 219 ) and NECs ( n = 267 ) . Embryo reporter assays and enhancer activity patterns are from the VISTA Enhancer Browser ( Visel et al . , 2007 ) . The number of PIRs active within a particular tissue is shown above each bar . PIRs with an active chromatin state in NECs showed an even more pronounced enrichment for enhancer activity in neural tissues ( Figure 2—figure supplement 1B ) . ( C ) Representative images of X-gal stained mouse embryos from the VISTA Enhancer Browser ( Visel et al . , 2007 ) reveal neural-restricted reporter gene activity for six example NEC PIRs . Shown underneath is the gene promoter assignment for the associated enhancer in VISTA and in our PCHi-C dataset . DOI: http://dx . doi . org/10 . 7554/eLife . 21926 . 00810 . 7554/eLife . 21926 . 009Figure 2—figure supplement 1 . Active PIRs are enriched for enhancers with neural-specific activity . ( A ) In ESCs and NECs , PIRs are significantly enriched for VISTA-validated enhancer regions , compared to promoter distance-matched regions ( p-value<0 . 001 for all six permutation tests based on 100 permutations ) . Black bars show the number of overlaps observed in PIRs , and light grey bars show the mean number of overlaps observed across the random samples . Error bars refer to 95% confidence intervals . ( B ) The barplot shows the distribution of tissue-specific reporter expression patterns for all experimentally tested PIRs with an active chromatin state in ESCs ( n = 46 ) and NECs ( n = 49 ) . Embryo reporter assays and tissue expression patterns are from the VISTA Enhancer Browser ( Visel et al . , 2007 ) . The number of PIRs active within a particular tissue is shown above each bar . DOI: http://dx . doi . org/10 . 7554/eLife . 21926 . 009 We next examined all putative enhancer elements and their PCHi-C-identified promoter targets . Overall , 219 PIRs in ESCs and 267 PIRs in NECs overlapped VISTA-annotated human elements ( Supplementary file 2 ) . Notably , we found that NEC PIRs were strongly enriched for sequences that could drive reporter-gene activity in neural tissues and in other neuroectodermal derivatives , such as the neural tube and cranial structures ( z-score = 11 , Figure 2B and Figure 2—figure supplement 1A ) . In contrast , ESC PIRs were enriched for sequences active in neural ( z-score = 6 . 7 ) and non-neural tissues ( z-score = 4 . 5 ) at similar levels ( Figure 2B and Figure 2—figure supplement 1A ) . Interestingly , the enrichment of PIRs with neural enhancer activity in NECs relative to their enrichment in ESCs was even more pronounced when we focused on PIRs in an active chromatin state ( Figure 2—figure supplement 1B ) . Collectively , these results validate the function of several hundreds of PIRs as cell-type-specific developmental enhancers . We next sought to link enhancers documented in the VISTA Enhancer Browser to their putative target genes on the basis of PCHi-C data . We detected the interactions of 267 VISTA-annotated human enhancers with 277 target gene promoters in NECs ( Supplementary file 2 ) . Of these , 122 PIRs ( 46% ) interacted with their nearest gene , which is consistent with their current annotation in the VISTA Enhancer Browser . The remaining PIRs , however , did not interact with their nearest gene in NECs , but engaged with more distal promoters ( Supplementary file 2 ) . Figure 2C shows PCHi-C-based reassignment of enhancer targets for several examples of key neural regulators including SOX2 , SOX4 , and FZD3 ( Figure 2C ) , and the full results are listed in Supplementary file 2 . Taken together , these findings provide a functional validation of the detected human PIRs , and identify the putative promoter targets of multiple known enhancers . We found interacting promoters to engage a median of four PIRs ( Figure 3A ) , consistent with findings in other human cell types ( Jin et al . , 2013; Sanyal et al . , 2012 ) . To obtain an integrated view of promoter interactions , we considered PIRs connected to each promoter to jointly form a ‘cis-regulatory unit’ ( CRU , Figure 3B ) . Focusing on protein-coding genes , and considering all promoters associated with at least one PIR , we defined 9008 CRUs in ESCs and 9361 in NECs , and studied their localisation , chromatin properties and dynamics during cell lineage commitment . 10 . 7554/eLife . 21926 . 010Figure 3 . Characterisation of cis-regulatory units ( CRUs ) . ( A ) Boxplot shows the distributions of the number of PIRs per interacting promoter in ESCs ( n = 17955 ) and NECs ( n = 18146 ) . Promoters with no detected PIRs are not shown ( 4121 in ESCs; 3930 in NECs ) . The number of interactions per promoter showed only a minor dependence on transcriptional activity and promoter chromatin state ( Figure 3—figure supplement 1A , B ) . ( B ) A schematic illustrating the concept of a CRU as a collection of all PIRs together with their associated promoter . Note that it cannot be ruled out that some PIRs may provide alternative rather than concurrent interactions . ( C ) Boxplot shows the distributions of CRU span in ESCs and NECs . We observed a moderate dependence between the span and the number of PIRs ( Figure 3—figure supplement 1C ) . ( D ) CRUs are preferentially contained within an individual TAD . Line graph shows the percentage of CRUs with different proportions of interactions that reside within an individual TAD ( purple ) and the summary statistics ( mean and 95% confidence error bars ) obtained for 1000 random samples , keeping the same CRU structure ( grey ) . There is a significant tendency for CRUs to be contained entirely within a TAD ( * denotes permutation test p-value<0 . 001 ) . In addition , fewer CRUs span entirely over a TAD boundary ( * denotes permutation test p-value<0 . 001 ) . The 1000 random samples were generated by permutations of CRUs across all promoter fragments , retaining the same overall CRU structure . Error bars show 95% confidence intervals . Data shown are for ESCs ( n = 9008 CRUs ) ; data for NECs are shown in Figure 3—figure supplement 1D . We found that CRUs crossing TAD and IN boundaries generally contained a higher number of PIRs ( Figure 3—figure supplement 1F , G ) . E–H ) Genome browser representations of CRUs in ESCs . Examples include the SNAI2 CRU ( E ) and GLI2 CRU ( F ) , which both fit entirely within a TAD and INs; PRDM1 CRU ( G ) , which fits entirely within a TAD but extends beyond INs , and POU3F1 CRU ( H ) , which extends over a TAD boundary and also beyond an IN . ( I ) CRUs are preferentially contained within INs , but interactions can extend beyond IN boundaries . The line graph shows the percentage of CRUs with different proportions of interactions that reside within an individual IN in ESCs ( coordinates obtained from [Ji et al . , 2016] ) . There is a significant tendency for CRUs ( purple ) to be contained entirely within an IN , compared to random ( grey ) ( * denotes p-value<0 . 001 from a permutation test done with 1000 random samples ) . In addition , fewer CRUs span entirely beyond an IN ( * denotes p-value<0 . 001 from a permutation test done with 1000 random samples ) . Error bars show 95% confidence intervals . Promoters outside of a defined IN were excluded from the analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 21926 . 01010 . 7554/eLife . 21926 . 011Figure 3—figure supplement 1 . Additional CRU characterisation . ( A ) Boxplots showing the number of PIRs per promoter in ESCs ( left ) and NECs ( right ) , separated by gene expression quartiles . ( B ) Number of PIRs per promoter in ESCs ( left ) and NECs ( right ) , separated by promoter chromatin state . ( C ) Number of PIRs per CRU in ESCs ( left ) and NECs ( right ) , separated by CRU span length . These variables are significantly positively correlated in both cell types according to Spearman’s association test ( ρ = 0 . 063 , correlation test p<0 . 001 in ESCs and ρ = 0 . 111 , correlation test p<0 . 001 in NEC ) . ( D ) CRUs are preferentially contained within an individual TAD in NECs ( n = 9361 CRUs ) . Line graph shows the percentage of CRUs with different proportions of interactions that reside within an individual TAD ( blue ) and the summary statistics ( mean and 95% confidence error bars ) obtained for 1000 random samples , keeping same CRU structure ( grey ) . There is a significant tendency for CRUs to be contained entirely within a TAD ( * denotes permutation test p-value<0 . 001 ) . In addition , fewer CRUs span entirely over a TAD boundary ( * denotes permutation test p-value<0 . 001 ) . The 1000 random samples were generated by permutations of CRUs across all promoter fragments , retaining the same overall CRU structure . Error bars show 95% confidence intervals . ( E ) Percentage of CRUs that cross TAD boundaries in ESCs and NECs . Orange bars show results for ESCs and NECs; grey bars show results for 1000 random permutations of CRUs across all promoter fragments , keeping the overall CRU structure . Error bars show 95% confidence intervals . For ESCs , we show the results obtained with TADs called by two different methods ( left , Dixon et al . , 2012; right , this study ) . Promoters overlapping TAD boundaries were excluded from this analysis , as were interactions where both ends were not included in any TAD . CRUs in both ESCs and NECs cross TAD boundaries less frequently than expected by random as the difference in height between the orange and grey bars is larger than the 95% confidence interval ( permutation test , p-value<0 . 001 ) . ( F ) Number of PIRs per CRU in ESCs ( left ) and NECs ( right ) , separated by whether the CRU crosses a TAD boundary ( ‘Inter-TAD’ ) or is contained within an individual TAD ( ‘Intra-TAD’ ) . In both cell types , CRUs crossing TAD boundaries have significantly more interactions than CRUs fully contained within a TAD ( ESCs - p-value<0 . 001 , NECs - p-value<0 . 001 , according to two-sided Wilcoxon rank sum tests ) . ( G ) Number of PIRs per CRU in ESCs , separated by whether the CRU crosses an IN boundary or is contained within an IN . CRUs crossing IN boundaries have significantly more interactions than CRUs fully contained within INs ( p-value<0 . 001 according to two-sided Wilcoxon rank sum test ) . ( H ) Strength of TAD boundaries , separated by whether they are crossed or not crossed by promoter interactions in ESCs ( left ) and NECs ( right ) . ( I ) The line graph shows the percentage of CRUs with different proportions of interactions that reside within the largest span of each overlapping IN set ( coordinates calculated from [Ji et al . , 2016] ) . Fewer than 50% of CRUs , but significantly more than expected at random , were fully contained with the extremities of IN boundaries ( * denotes permutation test p-value<0 . 001 ) . In addition , fewer CRUs span entirely over an IN boundary ( * denotes permutation test p-value<0 . 001 ) . The 1000 random samples were generated by permutations of CRUs across all promoter fragments , retaining the same overall CRU structure . Error bars show 95% confidence intervals . Promoters outside of a defined IN were excluded from the analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 21926 . 011 CRUs spanned a median of ~230 kb ( with a range of 1 kb-200Mb ) in both cell types ( Figure 3C ) . We assessed their localisation with respect to higher order features of chromosome architecture including TADs and Insulated Neighborhoods ( INs ) ( Dixon et al . , 2012; Ji et al . , 2016; Nora et al . , 2012; Sexton et al . , 2012 ) . We defined TADs in ESCs and NECs using Hi-C data for these cells generated as part of this study ( see Materials and methods ) . Overall , ~75% of CRUs were fully contained within a TAD in ESCs and NECs , which was significantly higher than expected by random ( Figure 3D–G and Figure 3—figure supplement 1D , E ) . In the remaining ~25% of CRUs , either some or all PIRs localised outside of the promoter-harbouring TAD ( Figure 3H and Figure 3—figure supplement 1D ) . We found that TAD boundaries crossed by promoter interactions were generally weaker than non-crossed boundaries ( Wilcoxon test p-value=1 . 8e-14; Figure 3—figure supplement 1H ) . However , the ranges of strength scores for ‘crossed’ and ‘non-crossed’ TAD boundaries were highly overlapping , and even some of the strongest boundaries were penetrable to interactions ( Figure 3—figure supplement 1H ) . For INs , we used the published genomic coordinates ( available for ESCs only ) that were defined on the basis of cohesin ChIA-PET and CTCF-binding data ( Ji et al . , 2016 ) . Just under 30% of CRUs were fully contained within IN boundaries , and this proportion increased to ~45% when considering the largest span of each overlapping set of INs as a single unit ( Figure 3I and Figure 3—figure supplement 1I ) . These numbers significantly exceeded the proportions expected at random ( Figure 3I and Figure 3—figure supplement 1I ) , but at the same time , also provided abundant examples of IN-spanning CRUs ( Figure 3G , H ) . Taken together , these results suggest that CRUs are partially constrained by , but not fully contained within , higher order topological structures such as TADs and INs . To investigate the potential regulatory features of CRUs , we first characterised their chromatin properties by considering the proportion of PIRs in each chromatin state within a CRU . Applying hierarchical clustering based on this property , we obtained eight distinct clusters of CRUs in both ESCs ( Figure 4A ) and NECs ( Figure 4—figure supplement 1 ) , corresponding to different combinations of PIR chromatin states within CRUs . We found that CRUs within three prevalent clusters contained PIRs in one predominant , non-background , chromatin state ( clusters 1–3; Figure 4A and Figure 4—figure supplement 1A ) . In contrast to these ‘uniform’ CRUs , 18% of CRUs in ESCs and 24% in NECs contained combinations of PIRs in active , poised and Polycomb-associated chromatin states ( clusters 4–7; Figure 4A and Figure 4—figure supplement 1 ) . Finally , CRUs in cluster 8 contained PIRs exclusively in the background state ( Figure 4A and Figure 4–figure supplement 1 ) . Examples of genes in ESCs assigned to the different CRU clusters are shown in Figure 4B and Figure 4—figure supplement 2 . Notably , the chromatin state of each promoter generally matched that of the most prevalent CRU chromatin state ( Figure 4A and Figure 4—figure supplement 1 ) . Overall , this classification provides a framework for exploring CRU properties . 10 . 7554/eLife . 21926 . 012Figure 4 . Clustering of CRUs according to chromatin state of each PIR in ESCs . ( A ) CRUs from ESCs were clustered hierarchically according to the distribution and fractions of their PIRs that correspond to each chromatin state . Boxplots show the distribution of PIR fractions for each chromatin state ( Act , active; Pois , poised; PcG , Polycomb-associated; Bg , background ) . Heatmaps show the log2 odds ratios of observing each promoter state associated with a CRU in each cluster ( p<0 . 001 , χ2 test on the contingency table ) . Data for NECs are shown in Figure 4—figure supplement 1 . ( B ) Genome browser representations of CRUs in ESCs . U2AF1 CRU from cluster 1 and TBX3 CRU from cluster 2 , each exemplify cases where non-background PIRs within a CRU are associated with a uniform chromatin state . SUV39H2 CRU from cluster 6 and CDKN2B CRU from cluster 7 , each exemplify cases where PIRs within a CRU are associated with multiple chromatin states . Interaction arcs are coloured according to PIR chromatin state ( active , green; poised , orange; Polycomb-associated , red; background , grey ) . See Figure 4—figure supplement 2 for additional examples , and Figure 4—figure supplement 3 for read count interaction profiles . DOI: http://dx . doi . org/10 . 7554/eLife . 21926 . 01210 . 7554/eLife . 21926 . 013Figure 4—figure supplement 1 . Clustering of CRUs according to chromatin state of each PIR in NECs . CRUs were hierarchically clustered according to their fractions of PIRs from different chromatin states in NECs . Boxplots depict the distribution of PIR fractions for each chromatin state ( Act , active; Pois , poised; PcG , Polycomb-associated; Bg , background ) . Heatmaps coupled to boxplots show log2 odds-ratios of promoter states ( p-value<0 . 001 , χ2 test on the contingency tables ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21926 . 01310 . 7554/eLife . 21926 . 014Figure 4—figure supplement 2 . Additional examples of CRUs in ESCs . Genome browser representations of CRUs in ESCs . Interaction arcs are coloured according to PIR chromatin state ( active , green; poised , orange; Polycomb-associated , red; background , grey ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21926 . 01410 . 7554/eLife . 21926 . 015Figure 4—figure supplement 3 . Read-count interaction profiles for baited promoters presented in Figures 4 and 5 . Plots show the read counts corresponding to the interactions of baited promoter fragments ( grey line ) with other HindIII fragments . Significant interactions detected by CHiCAGO ( score ≥12 ) are shown in red , and sub-threshold interactions ( score ≥11 in the cell type shown and score ≥12 in the other cell type ) are shown in blue . ( A ) U2AF1 , TBX3 , SUV39H2 and CDKN2B from Figure 4B . ( B ) CARS , PAX9 , CNPY1 , RXRG and AGAP2 from Figure 4—figure supplement 2 . ( C ) RGMB , MAP2 and KCNE3 from Figure 5D–F . DOI: http://dx . doi . org/10 . 7554/eLife . 21926 . 015 We set out to investigate CRU chromatin state transitions on ESC to NEC differentiation . For this analysis , and in each cell type , we classified CRUs into either single-state active ( containing active , and possibly also background-state , PIRs ) , single-state repressed ( containing poised and/or Polycomb-associated , and possibly also background-state PIRs ) , or background ( containing only background-state PIRs ) . CRUs containing a combination of both active and repressed ( Polycomb-associated/poised ) PIRs were classified as dual-state . We found that 65% of the single-state CRUs in ESCs remained single-state CRUs in NECs , although approximately half of them switched their state ( Figure 5A ) . In addition , similar proportions of CRUs lost ( 11% ) and acquired ( 13% ) a dual-state configuration on ESC to NEC differentiation . These findings demonstrate that considerable reorganisation of CRUs occurs during lineage commitment . 10 . 7554/eLife . 21926 . 016Figure 5 . CRU state transitions occur during ESC differentiation and are associated with changes in gene transcription . ( A ) Pie chart summarising CRU state transitions that occur upon ESC to NEC differentiation . The number of CRUs within each transition category are shown . Transitions that involve dual-state to single-state , and single-state to dual-state , are further subdivided into whether the single-state is classified as active or repressed ( Polycomb-associated or poised ) . ( B–C ) Heatmaps show the log2 odds ratios for CRU state transitions and associated changes in gene expression . ( B ) Single-state transitions showing a non-random segregation withgene expression changes ( p-value=0 . 0031 , Fisher’s exact test ) ; ( C ) dual-state transitions showing a non-random segregation with gene expression changes ( p-value=0 . 0014 , Fisher’s exact test ) . Number of CRUs within each transition category are shown . Genes differentially expressed between ESCs and NECs were identified using DESeq2 ( FDR < 0 . 05 and a log2 fold change of >1 . 5 ) . Repressed state includes Polycomb-associated and poised states . ( D–F ) Genome browser representations of CRU state transitions that occur upon the differentiation of ESCs ( top image ) to NECs ( lower image ) . ( D ) RGMB provides an example of a CRU transitioning from a single repressed to a single active state and an associated increase in RGMB transcription . ( E ) MAP2 provides an example of a dual-state to a single active state CRU transition and an associated increase in MAP2 transcription . ( F ) KCNE3 provides an example of a single active state to a dual-state CRU transition and an associated decrease in KCNE3 transcription . Arcs are coloured according to PIR chromatin state ( active , green; poised , orange; Polycomb-associated , red; background , grey ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21926 . 016 CRU state transitions associated significantly with changes in the expression of their target genes ( p-value<0 . 005 , Fisher’s exact test; Figure 5B , C ) . In particular , genes that were transcriptionally upregulated upon ESC differentiation preferentially gained an active single-state in NECs , either through switching the chromatin state of a single-state CRU ( Figure 5B ) or through resolving a dual-state CRU ( Figure 5C ) . Examples of CRUs undergoing each scenario include RGMB and MAP2 , which are transcriptionally upregulated in NECs ( Figure 5D , E ) . Pronounced chromatin changes were also detected at the CRUs of genes downregulated upon differentiation , including a loss of the active single-state and/or a transition to the repressed single-state ( Figure 5B , C; example shown in Figure 5F ) . Taken together , these results suggest that the modulation of CRU chromatin state is associated with transcriptional changes upon ESC differentiation . This modulation might potentially underlie many transcriptional changes in early lineage commitment . To investigate the underlying processes that drive changes in CRU organisation during cell lineage commitment , we studied the dynamics of promoter interactions and chromatin states at the individual PIRs . We refer to changes in PIR connectivity as ‘rewiring’ , and to chromatin state changes at PIRs as ‘recolouring’ , and note that they do not need to be mutually exclusive ( Figure 6A ) . To distinguish between interactions that are rewired and retained on ESC differentiation at high confidence , we applied additional filters to the PCHi-C data , resulting in 1153 rewired ( present in only one cell type ) and 1258 retained ( present in both ESCs and NECs ) interactions ( see Materials and methods for details ) . 10 . 7554/eLife . 21926 . 017Figure 6 . Interaction Dynamics: ‘recolouring’ versus ‘rewiring’ . ( A ) A schematic of interaction dynamics during cell differentiation . In a ‘recolouring’ interaction ( left ) the PIR undergoes a change in chromatin colour ( reflecting a change in chromatin state ) between the two cell types . In a ‘rewiring’ interaction ( right ) , an interaction is gained or lost upon cell differentiation . In a 'rewiring + recolouring' interaction , the loss or gain of an interaction is concomitant with changes in chromatin colour at the respective PIR . ( B ) Heatmap of log2 odds ratios showing the association between different PIR chromatin state transitions ( recolouring ) and PIR interaction dynamics ( rewiring ) on ESC to NEC differentiation ( p-value<0 . 001 , Fisher’s exact test ) . ( C–D ) Genome browser representations of interaction dynamics upon ESC to NEC differentiation . Note that only one interaction is shown for each example . Arcs are coloured according to PIR chromatin state ( active , green; Polycomb-associated , red ) . ( C ) Rewiring and recolouring: upon differentiation , NR2F1 gains an interaction with a PIR that is active in NECs , but repressed in ESCs . ( D ) Rewiring: the JAG1 promoter gains an interaction with an active PIR in NECs . Additional examples are shown in Figure 6—figure supplement 1A , B . ( E ) Pie chart summarising the different scenarios in which an interaction is retained upon ESC to NEC differentiation . ( F ) Recolouring: the IRX3 promoter retains an interaction , but the PIR changes from repressed ( in ESCs ) to active ( in NECs ) . ( G ) Boxplots revealing the transcriptional changes as a function of active PIR dynamics during recolouring ( left ) and rewiring ( right ) events . In either scenario , there was a significant association between the acquisition and loss of an active state and changes in gene expression ( p-values<0 . 001 for both recolouring and rewiring according to one-sided Wilcoxon rank sum tests ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21926 . 01710 . 7554/eLife . 21926 . 018Figure 6—figure supplement 1 . Interaction dynamics involving recolouring and rewiring . Genome browser representations of interaction dynamics upon ESC to NEC differentiation . Note that only one interaction is shown for each example . Arcs are coloured according to PIR chromatin state ( active , green; background , grey ) . ( A ) Rewiring and recolouring: ZSCAN2 loses an interaction with a background PIR that was previously classified as active in ESCs . ( B ) Rewiring: HAPLN3 promoter loses an interaction with an active PIR . ( C ) Recolouring: RAB3B promoter retains an interaction , but the PIR changes from active to background . DOI: http://dx . doi . org/10 . 7554/eLife . 21926 . 018 Importantly , we found that the co-occurrence of rewiring and recolouring interactions on ESC differentiation was significantly more common than expected at random ( Figure 6B , p-value<0 . 001 , Fisher’s exact test ) . Specifically , new interactions that were gained by NECs preferentially acquired the active state , or transitioned from the background to repressed state ( Figure 6B , bottom row ) . Interactions that were lost on ESC differentiation were enriched for PIRs that transitioned from the active to poised/Polycomb-associated states , as well as for those switching to the background state ( Figure 6B , middle row ) . Notably , the vast majority of rewiring events ( 99 . 7% ) were not associated with larger-scale A/B compartment dynamics ( not shown ) . Together , these observations indicate that lineage commitment associates with concerted changes in the connectivity and chromatin state of regulatory regions . Interactions at NR2F1 ( Figure 6C ) and ZSCAN2 ( Figure 6—figure supplement 1A ) exemplify the preferential co-occurrence of rewiring and recolouring events , with interactions present in the cell type in which the respective PIR is in the active state . However , we also found examples of PIR rewiring that showed unchanged chromatin states in both cell types , such as those at the JAG1 and HAPL3 genes ( Figure 6D and Figure 6—figure supplement 1B ) . Finally , we observed that 25% of PIRs that were retained in both cell types undergo chromatin state recolouring ( Figure 6E , red segment ) . These regions included , for example , PIRs associated with the IRX3 and RAB3B promoters ( Figure 6F and Figure 6—figure supplement 1C ) . We asked how rewiring and recolouring events at PIRs contribute to gene expression dynamics on ESC to NEC differentiation . We found that the loss or gain of interactions with active-state PIRs associated significantly with changes in gene expression ( Figure 6G ) , suggesting their functional contribution to transcriptional control . Notably , gene expression changes were detected at retained and recoloured PIRs ( Figure 6G , left panel ) , and also when active-state PIRs were lost or gained through rewiring ( Figure 6G , right panel ) . Taken together , these findings demonstrate that chromatin state changes and rewiring of interactions at PIRs contribute jointly to transcriptional regulation . Furthermore , our results show that promoter interaction dynamics preferentially co-occur with chromatin state dynamics upon cell lineage commitment .
Our study presents an atlas of promoter interactions in human pluripotent and early lineage-committed cells , and offers new insights into the association between genome organisation and gene regulation . The high resolution of PCHi-C has enabled us to detect individual promoter-associated loops at a single restriction enzyme fragment resolution . We find that promoter-interacting regions in both cell types harbour multitudes of previously known and putative enhancer elements , which we link with their physically associated target genes . While the identified connections are predictive of regulatory relationships , it is important to note that the current data are correlative and will require functional validation using targeted genetic approaches and reporter assays . In addition to active enhancers , we find extensive promoter connectivity to regions associated with Polycomb-mediated repression and poising , reinforcing the role of Polycomb-group proteins in controlling chromosomal topology at transcriptionally inactive genes ( Entrevan et al . , 2016; Li et al . , 2015; Schoenfelder et al . , 2015b; Vieux-Rochas et al . , 2015 ) . Consistent with previous observations ( Sanyal et al . , 2012 ) , we also detect large numbers of interactions between promoters of both active and inactive genes , and regions devoid of chromatin features . It is possible that such interactions are structural , rather than play gene regulatory roles . However , a regulatory function for some ‘unmarked’ PIRs also cannot be ruled out as recent mutagenesis experiments have identified functional elements that lack 'classic' chromatin annotations ( Pradeepa et al . , 2016; Rajagopal et al . , 2016 ) . The high-resolution promoter-interaction information has enabled the identification of cis-regulatory units ( CRUs ) as sets of interactions connected to the same promoter . Taking the view of CRUs , we consider jointly the dynamics of chromatin states and interactions as ESCs differentiate , and assess the potential contribution of these processes to changes in gene expression during development . We observe that CRUs reconfigure extensively upon cell differentiation . These include the CRUs of ESC- and NEC-specific genes , for which CRU reconfiguration associates with transcriptional changes upon differentiation , as well as the CRUs of genes that are not expressed in either cell type , consistent with the model of progressive chromatin changes at lineage-inappropriate genes during lineage commitment ( Spivakov and Fisher , 2007 ) . Previous studies on candidate loci have proposed that cell-state changes are associated with two predominant models of enhancer–promoter interaction dynamic that have been termed instructive and permissive ( de Laat and Duboule , 2013 ) . Instructive interactions are established de novo upon cell state change and are concomitant with target gene activity . In contrast , permissive interactions are already in place before the gene activation occurs and may therefore contribute to enhancer priming . Examples of instructive and permissive interactions have been described in pluripotent cell reprogramming and differentiation ( Apostolou et al . , 2013; Denholtz et al . , 2013; Phillips-Cremins et al . , 2013; Wei et al . , 2013; Zhang et al . , 2013 ) , but little was known on a global scale about which model of enhancer-promoter interactions is predominant during lineage commitment . Here , we show that developmental changes at CRUs involve both the rewiring of ‘instructive’ interactions and the recolouring of the chromatin state of ‘permissive’ interacting regions . Notably , we find that these two processes tend to occur hand in hand , with the strongest association occurring between cell-type-specific promoter interactions and the active state of the respective PIRs . Importantly , CRU chromatin dynamics ( at both rewired and preformed interactions ) associates with consistent changes in gene expression , suggesting that both mechanisms are functionally important in mediating lineage-specific transcriptional programmes . The exact determinants of ‘permissive’ versus ‘instructive’ interactions remain to be elucidated and may depend on the identity of cis-acting factors recruited to the regulatory regions , as well as on local chromatin environments . The CRU view provides an opportunity to consider multi-modular gene regulation in early human development that has hitherto been studied on a limited number of genes , predominantly in model organisms ( Barolo , 2012; Cannavò et al . , 2016; Hong et al . , 2008 ) . The ‘single-state’ architecture that we detect at the majority of CRUs is in line with observations of ‘shadow enhancers’ with overlapping activities in Drosophila ( Hong et al . , 2008 ) . It has been suggested that this cis-regulatory organisation ensures the robustness of gene regulation and can buffer the effects of deleterious sequence variation , as well as providing opportunities for evolutionary innovation ( Barolo , 2012; Cannavò et al . , 2016; Hong et al . , 2008; Perry et al . , 2010 ) . ‘Dual-state’ CRUs , although representing a relative minority of the CRUs we analysed , offer additional insights into signal integration at promoters . Specifically , the fact that the chromatin state of the promoter largely associates with the predominant chromatin state of the connected PIRs suggests that promoters may integrate signals from remote elements based on ‘majority vote logic’ . Mechanistically , this logic may be a consequence of largely independent enhancer action ( potentially at both single-state and dual-state CRUs ) that is consistent with the ‘hit-and-run’ model of transcriptional regulation ( Schaffner , 1988; Varala et al . , 2015 ) , and provides a flexible way to fine-tune the expression of multi-enhancer genes ( Guerrero et al . , 2010; Lagha et al . , 2012 ) . However , this model also does not preclude the possibility that promoter chromatin states at ‘dual-state’ CRUs undergo a continuous turnover depending on the state of the PIR they contact . In this case , the observed ‘majority-vote’ promoter chromatin states would correspond to the predominant state detected at the population level . Theoretically , the generally independent enhancer action also enables activation signals from individual elements to quantitatively ‘add up’ ( at least to some extent ) to promote stronger transcriptional outputs ( Arnold et al . , 2013; Bothma et al . , 2015; Lam et al . , 2015; Spivakov , 2014 ) . Our observation that the resolution of dual-state CRUs toward a uniformly active state generally results in increased expression ( and vice versa ) supports this model . However , analyses in Drosophila have identified exceptions to additive enhancer activity ( Bothma et al . , 2015 ) and have provided examples of enhancers that activate more than one promoter in a coordinated fashion ( Fukaya et al . , 2016 ) , which is not immediately expected from the ‘hit-and-run’ looping model . Finally , there is also a possibility that multiple enhancers are jointly engaged in ‘chromatin hubs’ with promoters , rather than acting individually ( Hanscombe et al . , 1991; Jiang et al . , 2016; Patrinos et al . , 2004; Tolhuis et al . , 2002; Wijgerde et al . , 1995 ) . These mechanistic questions go beyond the capabilities of Hi-C-based analyses of cell populations , and as such it is possible that multiple promoter interactions detected within a CRU take place either concurrently or simultaneously . The emerging studies at the single-molecule level ( such as [Bartman et al . , 2016; Fukaya et al . , 2016] ) will likely shed further light on the molecular mechanisms that underpin the principles of CRU organisation . Promoter – enhancer interactions are vitally important for gene regulation and their disruption may lead to pronounced developmental abnormalities ( Epstein , 2009 ) . The high-resolution resource of the promoter-interaction landscape in pluripotent and early lineage-committed cells presented here , therefore , provides a stepping stone to understanding the logic of gene regulation and its aberrations during human embryogenesis .
ESCs ( H9/WA09; obtained from WiCell ( Madison , WI ) ; RRID:CVCL_9773 ) were cultured at 37°C in 5% CO2 in air in Pluripro media and matrix ( Cell Guidance Systems ( Cambridge , UK ) ) . Authentication of ESCs was achieved by confirming the expression of pluripotency genes and protein markers , and by SNP analysis of sequencing data . ESCs were routinely verified as mycoplasma-free using a PCR-based assay ( Sigma ( St . Louis , MO ) ) . The H9/WA09 line is not on the list of commonly misidentified cell lines ( International Cell Line Authentication Committee ) . ESCs were differentiated into NECs using a previously described protocol ( Rada-Iglesias et al . , 2011 ) and samples were harvested on day 7 . Following dissociation with accutase , ESCs and NECs were stained on ice for 45 min with CD326-AF647 ( BioLegend ( London , UK ) , Cat# 324212 , RRID:AB_756086; 5 µL per million cells ) and CD56-PE ( BD Biosciences ( San Jose , CA ) , Cat# 345812 , RRID:AB_2629216; 20 µL per million cells ) antibodies in 100 µl PBS containing 2% FBS . After washing , DAPI was included at a final concentration of 5 µl/mL for live/dead cell discrimination , and flow cytometry analysis was performed using a BD LSRFortessa with subsequent data analysis using FlowJo V10 . 1 . Hi-C and Promoter CHi-C libraries were generated essentially as described ( Mifsud et al . , 2015; Schoenfelder et al . , 2015a ) , with minor modifications . 3 to 4 × 107 cells ( ESCs or NECs ) were fixed in 2% formaldehyde ( Agar Scientific ( Stansted , UK ) ) for 10 min , after which the reaction was quenched with ice-cold glycine ( 0 . 125 M final concentration ) . Cells were collected by centrifugation ( 400 x g for 10 min at 4°C ) , and washed once with PBS ( 50 ml ) . After another centrifugation step ( 400 x g for 10 min at 4°C ) , the supernatant was completely removed and the cell pellets were immediately frozen in liquid nitrogen and stored at −80°C . After thawing , the cell pellets were incubated in 50 ml ice-cold lysis buffer ( 10 mM Tris-HCl pH 8 , 10 mM NaCl , 0 . 2% Igepal CA-630 , protease inhibitor cocktail ( Roche ( Basel , Switzerland ) ) for 30 min on ice . After centrifugation to pellet the cell nuclei ( 650 x g for 5 min at 4°C ) , nuclei were washed once with 1 . 25 x NEBuffer 2 . The nuclei were then resuspended in 1 . 25 x NEBuffer 2 , SDS was added ( 0 . 3% final concentration ) and the nuclei were incubated at 37°C for 1 hr with agitation ( 950 rpm ) . Triton X-100 was added to a final concentration of 1 . 7% and the nuclei were incubated at 37°C for 1 hr with agitation ( 950 rpm ) . Restriction digest was performed overnight at 37°C with agitation ( 950 rpm ) with HindIII ( NEB; 1500 units per 7 million cells ) . Using biotin-14-dATP ( Life Technologies ( Carlsbad , CA ) ) , dCTP , dGTP and dTTP ( all at a final concentration of 30 µM ) , the HindIII restriction sites were then filled in with Klenow ( NEB ( Ipswich , MA ) ) for 75 min at 37°C . After addition of SDS ( 1 . 42% final concentration ) and incubation at 65°C with agitation ( 950 rpm ) for 20 min , ligation was performed for 4 hr at 16°C ( 50 units T4 DNA ligase ( Life Technologies ) per 7 million cells starting material ) in a total volume of 8 . 2 ml ligation buffer ( 50 mM Tris-HCl , 10 mM MgCl2 , 1 mM ATP , 10 mM DTT , 100 µg/ml BSA , 0 . 9% Triton X-100 ) per 7 million cells starting material . After ligation , reverse crosslinking ( 65°C overnight in the presence of Proteinase K ( Roche ) ) was followed by RNase A ( Roche ) treatment and two sequential phenol/chloroform extractions . After DNA precipitation ( sodium acetate 3 M pH 5 . 2 ( 1/10 volume ) and ethanol ( 2 . 5 x volumes ) ) overnight at −20°C , the DNA was spun down ( centrifugation 3200 x g for 30 min at 4°C ) . The pellets were resuspended in 400 µl TLE ( 10 mM Tris-HCl pH 8 . 0; 0 . 1 mM EDTA ) , and transferred to 1 . 5 ml eppendorf tubes . After another phenol/chloroform extraction and DNA precipitation overnight at −20°C , the pellets were washed three times with 70% ethanol , and the DNA concentration was determined using Quant-iT Pico Green ( Life Technologies ) . The efficiency of biotin incorporation was assayed by amplifying a 3C ligation product ( primers available upon request ) , followed by digest with HindIII or NheI . To remove biotin from non-ligated fragment ends , 40 µg of Hi-C library DNA were incubated with T4 DNA polymerase ( NEB ) for 4 hr at 20°C , followed by phenol/chloroform purification and DNA precipitation overnight at −20°C . After a wash with 70% ethanol , sonication was carried out to generate DNA fragments with a size peak around 400 bp ( Covaris E220 settings: duty factor: 10%; peak incident power: 140W; cycles per burst: 200; time: 55 s ) . After end repair ( T4 DNA polymerase , T4 DNA polynucleotide kinase , Klenow ( all NEB ) in the presence of dNTPs in ligation buffer ( NEB ) ) for 30 min at room temperature , the DNA was purified ( Qiagen ( Hilden , Germany ) PCR purification kit ) . dATP was added with Klenow exo- ( NEB ) for 30 min at 37°C , after which the enzyme was heat-inactivated ( 20 min at 65°C ) . A double size selection using AMPure XP beads ( Beckman Coulter , Brea , CA ) was performed: first , the ratio of AMPure XP beads solution volume to DNA sample volume was adjusted to 0 . 6:1 . After incubation for 15 min at room temperature , the sample was transferred to a magnetic separator ( DynaMag-2 magnet; Life Technologies ) , and the supernatant was transferred to a new eppendorf tube , while the beads were discarded . The ratio of AMPure XP beads solution volume to DNA sample volume was then adjusted to 0 . 9:1 final . After incubation for 15 min at room temperature , the sample was transferred to a magnet ( DynaMag-2 magnet; Life Technologies ) . Following two washes with 70% ethanol , the DNA was eluted in 100 µl of TLE ( 10 mM Tris-HCl pH 8 . 0; 0 . 1 mM EDTA ) . Biotinylated ligation products were isolated using MyOne Streptavidin C1 Dynabeads ( Life Technologies ) on a DynaMag-2 magnet ( Life Technologies ) in binding buffer ( 5 mM Tris pH8 , 0 . 5 mM EDTA , 1 M NaCl ) for 30 min at room temperature . After two washes in binding buffer and one wash in ligation buffer ( NEB ) , PE adapters ( Illumina , San Diego , CA ) were ligated onto Hi-C ligation products bound to streptavidin beads for 2 hr at room temperature ( T4 DNA ligase NEB , in ligation buffer , slowly rotating ) . After washing twice with wash buffer ( 5 mM Tris , 0 . 5 mM EDTA , 1 M NaCl , 0 . 05% Tween-20 ) and then once with binding buffer , the DNA-bound beads were resuspended in a final volume of 90 µl NEBuffer 2 . Bead-bound Hi-C DNA was amplified with seven PCR amplification cycles using PE PCR 1 . 0 and PE PCR 2 . 0 primers ( Illumina ) . After PCR amplification , the Hi-C libraries were purified with AMPure XP beads ( Beckman Coulter ) . The concentration of the Hi-C libraries was determined by Bioanalyzer profiles ( Agilent Technologies , Santa Clara , CA ) and qPCR ( Kapa Biosystems ( Wilmington , MA ) ) , and the Hi-C libraries were paired-end sequenced ( HiSeq 1000 , Illumina ) at the Babraham Institute Sequencing Facility . For Promoter Capture Hi-C , 500 ng of Hi-C library DNA was resuspended in 3 . 6 µl H2O , and custom hybridization blockers ( Agilent Technologies ) were added to the Hi-C DNA . Hybridization buffers and the custom-made RNA capture bait system ( Agilent Technologies; designed as previously described ( Mifsud et al . , 2015 ) : 37 , 608 biotinylated RNAs targeting the ends of 22 , 076 promoter-containing HindIII restriction fragments ) were prepared according to the manufacturer’s instructions ( SureSelect Target Enrichment , Agilent Technologies ) . The Hi-C library DNA was denatured for 5 min at 95°C , and then incubated with hybridization buffer and the RNA capture bait system at 65°C . After 24 hr incubation at 65°C , biotin/streptavidin pulldown ( MyOne Streptavidin T1 Dynabeads; Life Technologies ) and washes were performed according to the SureSelect Target enrichment protocol ( Agilent Technologies ) . After the final wash , the beads were resuspended in 30 µl NEBuffer 2 . After a post-capture PCR ( four amplification cycles using Illumina PE PCR 1 . 0 and PE PCR 2 . 0 primers ) , the Promoter CHi-C libraries were purified with AMPure XP beads ( Beckman Coulter ) . The concentration of the Promoter CHi-C libraries was determined by Bioanalyzer profiles ( Agilent Technologies ) and qPCR ( Kapa Biosystems ) , and the Promoter CHi-C libraries were paired-end sequenced ( HiSeq 1000 , Illumina ) at the Babraham Institute Sequencing Facility . Raw sequencing reads were processed using the HiCUP pipeline ( Wingett et al . , 2015 ) , which mapped sequencing read pairs against the human genome ( GRCh37 ) , filtered out experimental artifacts such as circularized reads and re-ligations , and removed all duplicate read pairs . The aligned Hi-C data were analysed using HOMER v4 . 7 ( http://homer . salk . edu/homer/ ) ( Heinz et al . , 2010 ) . Coverage- and distance-related correction factors of the binned data were computed at 25 kb and 250 kb resolutions , based on the iterative correction algorithm ( Imakaev et al . , 2012 ) . TADs were identified based on directionality indices ( Dixon et al . , 2012 ) of Hi-C interactions 1 Mb upstream and downstream from a 25 kb sliding window every 5 kb steps , which were then smoothed using a running average over a ± 25 kb window . TADs were called between pairs of consecutive local maxima ( TAD start ) and minima ( TAD end ) of the smoothed directionality indices with a standard score difference ( TAD ∆Z score ) above 2 . 0 , and the TAD ends were extended outward to the genomic bins with no directionality bias . These TAD definitions were used to compute the fraction of significant PCHi-C interactions falling inside TADs , alongside TADs reported by Dixon et al . , 2015 . To assess the strength of the TAD boundaries crossed by promoter interactions , we defined a TAD boundary strength score ( TADB ∆Z score ) as the difference between the smoothed directionality index values at the local maximum ( end of the preceding TAD ) and the local minimum ( start of the following TAD ) . Defined this way , TADB ∆Z scores ( unlike the TAD ∆Z scores ) do not depend on the stringency of the opposite boundary of the respective TAD . A/B compartments were called by computing the principal components of the distance- and coverage-corrected interaction profile correlation matrix at 250 kb resolution ( Lieberman-Aiden et al . , 2009 ) . Positive values of the principal component were aligned with H3K4me3 ChIP-seq signals for H9 human ESCs ( Rada-Iglesias et al . , 2011 ) . For chromosomes 4 and X , we used the second principal component instead of the first , as the first component described the preferential contact pattern within chromosome arms rather than compartments . The principal component values ranged from −42 to 42 . To quantify the compartment changes of significant interactions , each side of the interaction was classified as A or B compartment if the principal component of its 250 kb bin was above 10 or below −10 , respectively . Interactions falling within the 250 bins that had the principal component scores between −10 and 10 were considered as falling outside either compartment . Interactions were called at the level of individual HindIII fragments using version 0 . 1 . 4 of the CHiCAGO pipeline ( Cairns et al . , 2016 ) based on two biological replicates for each cell type that were normalised and combined as part of the pipeline . CHiCAGO incorporates a convolution background model , which emcompasses the ‘Brownian’ ( real , but expected interactions ) and ‘technical’ ( assay and sequencing artefacts ) components , and a weighted multiple testing correction procedure trained on interaction distance . CHiCAGO interaction scores correspond to –log-transformed , weighted p-values for each fragment read pair . A score threshold of 12 was used ( equivalent to a threshold of 5 in Chicago v1 . 0 . 0+ due to a soft-thresholding procedure introduced in this version ) . This threshold was chosen empirically based on balancing the enrichment for chromatin marks at PIRs with the overall number of detected interactions . Additionally , interactions with scores between 11 and 12 were included in the analysis if they scored above 12 in the other cell type . Total RNA was extracted from ESCs and NECs using an RNeasy Mini Kit ( Qiagen ) . Indexed mRNA-seq libraries were constructed from 500 ng total RNA using the Tru-Seq RNA Library Prep Kit v2 ( Illumina ) . Library fragment size and concentration was determined using an Agilent Bioanalyzer 2100 and KAPA Library Quantification Kit ( KAPA Biosystems ) . Samples were sequenced on an Illumina HiSeq as single-end libraries at the Babraham Institute Sequencing Facility . Reads were trimmed using trim galore ( http://www . bioinformatics . babraham . ac . uk/projects/trim_galore/ ) with default parameters to remove the standard Illumina adapter sequence . Reads were mapped to the GRCh37 assembly using tophat ( Trapnell et al . , 2009 ) . BAM files were imported to Seqmonk ( http://www . bioinformatics . babraham . ac . uk/projects/seqmonk/ ) . Raw read counts per transcript were calculated using the RNA-seq quantitation pipeline on the Ensembl v70 gene set using non-directional counts . Differential analysis of gene expression was performed using the default settings in DESeq2 ( Love et al . , 2014 ) without independent filtering of the results . Differentially expressed genes were called at padj < 0 . 05 and log2 fold change above 1 . 5 or below −1 . 5 . The histone modification ChIP-seq data ( H3K4me1 , H3K4me3 , H3K27ac and H3K27me3 ) for ESCs and NECs were from Rada-Iglesias et al . , 2011 , available in Gene Expression Omnibus under accession number GSE24447 . Data were converted to GRCh37 using liftOver ( Kent et al . , 2002 ) . CTCF ChIP-seq data were from ENCODE ( ENCODE Project Consortium , 2012 ) . Chromatin segmentations were performed on the basis of multiple histone modification ChIP datasets using a Hidden Markov Model-based method implemented in ChromHMM ( version 1 . 10 ) ( Ernst and Kellis , 2012 ) with default settings . The segmentation was carried out jointly through providing ‘concatenated’ data for both cell types as input . The resulting 16 states were curated into four broad chromatin states based on analysing their enrichment for different histone marks ( Figure 1—figure supplement 3A ) as follows . States 1–6 characterised by the presence of H3K4me3 and/or H3K27ac , and the absence of H3K27me3 , were labelled ‘active’; states 7–9 showing a combination of H3K4 methylation and H3K27me3 were labelled ‘poised’; state 10 showing H3K27me3 and no H3K4 methylation was labelled ‘Polycomb-associated’; states 14–16 showing no detectable signal for the four tested histone modifications were labelled ‘background’ . In addition , two more curated states were defined , but not considered further: states 11–12 were characterised by a ‘mixed’ pattern of both H3K27ac and H3K27me3 , which likely arose from a technical issue such as heterogeneity within the samples; state 13 characterised by H3K4me1 alone was classified as ‘intermediate enhancers’ , but the fraction of PIRs bearing this signature ( ~1% ) was too small to analyse them as an individual category . HindIII fragments in the human genome ( including baits and PIRs ) were then classified according to the chromatin states detected within them . When more than one chromatin state was present , classification was resolved in the following manner: ( i ) any functional state ( e . g . active , poised , Polycomb-associated ) was prioritised above background; ( ii ) active , poised and Polycomb-associated states were prioritised above intermediate; ( iii ) poised state was prioritised above the Polycomb-associated state; ( iv ) active state together with any inactive state ( i . e . poised or Polycomb-associated ) was labelled as mixed . Based on these heuristics , we assigned a single chromatin state ( including the background state ) to 81% of PIRs in both cell types . Transgenic reporter assays for enhancer activity are described within the VISTA Enhancer Browser ( Visel et al . , 2007 ) . The enhancer sequences from VISTA were overlapped with PIRs , and their putative target genes were defined according to the PCHi-C detected promoter-PIR interactions . HindIII fragments were mapped to TADs defined as described above and INs obtained from Ji et al . , 2016 . Baited fragments overlapping TAD boundaries , and those mapping outside INs , were excluded from respective analyses . For each CRU , the percentage of interactions that map within the same TAD or IN was calculated and these values were collected into 12 bins , The first and last bins contained the values of 0% and 100% , respectively , and the remaining bins contained all other values split into 10% intervals . These results were compared to 1000 random permutations of CRUs across all promoter fragments performed in a manner retaining the overall CRU structure . Each CRU was categorised according to the fraction of PIRs in the active , poised and Polycomb-associated state . These fractions were used for hierarchical clustering based on Euclidian distances ( method='Euclidian’ in dist function in R ) with the average agglomeration method ( method='average’ in hclust function in R ) . PIRs assigned intermediate or mixed chromatin states did not contribute to the clustering procedure , in the latter case because the states of the regulatory elements interacting with target promoters within these PIRs are not identifiable . CRUs containing only mixed or intermediate PIRs were not included in the analysis . False-negative rates associated with stringent signal thresholds drive down the observed overlap between conditions and may overestimate the proportion of cell-type-specific interactions . Therefore , we applied additional criteria to identify high-confidence subsets of rewired and retained interactions based on replicate-level CHiCAGO interaction calls . First , we required that rewired interactions have scores above 12 in both biological replicates of the same cell type , and below 12 in both replicates of the other cell type . We then binned the interactions satisfying these criteria into five groups of equal size according to their interaction scores in the merged samples . Interactions belonging to the top bin in one cell type and the bottom bin in the other cell type were considered as rewired . Interactions with scores above 12 in the two replicates of both cell types were considered as retained . Applying these criteria and filtering out interactions with PIRs in the mixed and intermediate states , we obtained high-confidence sets of 1258 retained and 1153 rewired interactions that were used in the analysis . Sequencing data have been deposited in Gene Expression Omnibus ( GEO ) with accession number GSE86821 . Processed data including interaction peak calls in the WashU Genome Browser text format and RNA-seq raw read counts were deposited in the same GEO repository . CHiCAGO objects containing all detected interactions , ChromHMM segmentation data , DESeq2-processed RNA-seq data and the defitions of TADs have been made available through the Open Science Framework ( http://osf . io/sdbg4 ) . | Virtually every cell in the body contains the same set of DNA , which encodes thousands of genes . The activities of these genes vary between different types of cells and at different points in time . As a result , our DNA contains a complex array of molecular switches that instruct genes to switch on and off at the right time and in the right cells . These molecular switches , termed regulatory elements , are often a long way away from the genes that they control , and this can make it difficult to find out which switch controls which genes . DNA is made up of several different building blocks known as bases and the order of these bases encodes specific information about the gene . Every human cell contains approximately two meters of DNA , which is highly folded in the cell nucleus . This three-dimensional folding allows regions that are far apart on the DNA thread to physically contact each other . To reach the genes they control , regulatory elements form loops on the DNA that are near-impossible to predict from looking at the sequence of bases alone . Mapping the locations of these loops can reveal the hidden circuitry within our DNA and help us to understand how unwanted changes ( mutations ) within regulatory elements may cause disease . Freire-Pritchett , Schoenfelder et al . set out to reveal how loops between genes and their regulatory elements change as the stem cells specialise into immature brain cells . The experiments show that the pattern of DNA loops is extensively altered after the stem cells specialise into brain cells , that is , some loops are lost and new ones form . Furthermore , the regulatory elements themselves were often toggled between “on” and “off” states . These two processes tend to occur together , so that new loops are formed at the same time as the switch is activated . Freire-Pritchett , Schoenfelder et al . also show that individual genes are often connected to many different regulatory elements . The next step is to understand how these multiple connections work together to coordinate gene activity , and whether this information could be used to control how stem cells specialise . This knowledge may lead to the development of stem cell-based therapies for stroke , Parkinson’s disease and other conditions . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"chromosomes",
"and",
"gene",
"expression",
"tools",
"and",
"resources"
] | 2017 | Global reorganisation of cis-regulatory units upon lineage commitment of human embryonic stem cells |
Innate immunity is the first immunological defence against pathogens . During virus infection detection of nucleic acids is crucial for the inflammatory response . Here we identify DNA-dependent protein kinase ( DNA-PK ) as a DNA sensor that activates innate immunity . We show that DNA-PK acts as a pattern recognition receptor , binding cytoplasmic DNA and triggering the transcription of type I interferon ( IFN ) , cytokine and chemokine genes in a manner dependent on IFN regulatory factor 3 ( IRF-3 ) , TANK-binding kinase 1 ( TBK1 ) and stimulator of interferon genes ( STING ) . Both cells and mice lacking DNA-PKcs show attenuated cytokine responses to both DNA and DNA viruses but not to RNA or RNA virus infection . DNA-PK has well-established functions in the DNA repair and V ( D ) J recombination , hence loss of DNA-PK leads to severe combined immunodeficiency ( SCID ) . However , we now define a novel anti-microbial function for DNA-PK , a finding with implications for host defence , vaccine development and autoimmunity .
The innate immune response is mediated by the production of cytokines , including type I interferons ( IFNs ) , and chemokines following the detection of pathogen-specific molecules by host cells ( Akira et al . , 2006; Medzhitov , 2007 ) . Detection of nucleic acids is crucial in triggering the innate immune to pathogens , particularly in response to viruses ( Pichlmair and Reis e Sousa , 2007 ) . Various double and single stranded RNA substrates are recognised directly by the DExD/H box RNA helicases retinoic acid-inducible gene I ( RIG-I ) and melanoma differentiation-associated gene-5 ( MDA-5 ) in the cytoplasm and by the endosomal toll-like receptors ( TLRs ) ( Pichlmair and Reis e Sousa , 2007 ) . The engagement of such RNA receptors leads to the rapid transcription of genes encoding anti-viral proteins via the activation of transcription factors belonging to the interferon regulatory factor ( IRF ) and nuclear factor-kappa B ( NF-κB ) families . It has been recognised for some time that intracellular DNA can activate a similar IRF-3-dependent innate immune response ( Stetson and Medzhitov , 2006 ) and it is known that this signalling pathway depends on both the IRF-3-activating kinase , TANK-binding kinase 1 ( TBK1 ) ( Ishii et al . , 2006 , 2008 ) and the adaptor protein stimulator of IFN genes ( STING—also known as MITA , ERIS and TMEM173 ) ( Ishikawa et al . , 2009 ) . Only more recently , however , have some candidate receptors for these pathways been identified . DNA-dependent activator of IFN-regulatory factors ( DAI ) ( Takaoka et al . , 2007 ) , RNA polymerase III ( RNA-Pol III ) ( Ablasser et al . , 2009; Chiu et al . , 2009 ) , IFN inducible gene 16 ( Unterholzner et al . , 2010 ) and DDX41 ( Zhang et al . , 2011b ) have been described as cytoplasmic DNA sensors that activate IRF-3 . Nonetheless , the in vivo relevance of these sensors remains unknown and the normal immune response to DNA stimulation of Dai−/− and Ips1−/− cells ( Ishii et al . , 2008; Wang et al . , 2008 ) and sensing of plasmodium DNA independent of these receptors ( Sharma et al . , 2011 ) indicates other IRF-3-activating cytoplasmic DNA sensors exist . In addition , although TBK1 and STING are essential for activation of IRF-3 following DNA stimulation , the molecular details of this signalling pathway are poorly understood ( Paludan et al . , 2011; Barber , 2011 ) . The presence of unidentified DNA sensors in fibroblasts is especially pertinent to virus infection since these cells are often a primary target of virus infection in vivo . These cells should have sentinel innate immune receptors in place to detect the presence of foreign nucleic acid and respond by producing IFN , cytokines and chemokines to initiate the anti-viral state in the surrounding tissue as well as to attract immune cells to the site of infection . In this study we identify DNA-dependent protein kinase ( DNA-PK ) as a novel DNA sensor in fibroblasts where it is present at high levels enabling it to respond to incoming infection without the need for prior stimulation . DNA-PK is a heterotrimeric protein complex consisting of three proteins , Ku70 , Ku80 ( also known as Ku86 ) and the catalytic subunit DNA-PKcs ( encoded by the xrcc6 , xrcc5 and prkcd genes respectively ) . Ku70 and Ku80 themselves form a heterodimer and the absence of one subunit de-stabilises the expression of the other ( Nussenzweig et al . , 1996; Gu et al . , 1997 ) . Both the Ku heterodimer ( Walker et al . , 2001 ) and DNA-PKcs ( Hammarsten and Chu , 1998 ) can bind directly to DNA but , in the absence of Ku the affinity of DNA-PKcs for DNA is greatly reduced ( Yaneva et al . , 1997 ) . DNA-PK has a well described role in the nucleus where it is necessary for non-homologous end joining ( NHEJ ) and so has a key role in repairing double-strand DNA breaks ( Lieber et al . , 2003 ) . DNA-PK has also been detected in the cytoplasm by immunofluorescence and cell fractionation ( Huston et al . , 2008; Balazs et al . , 2012 ) , although prior to this study no function has been assigned to DNA-PK in this localisation . Here we found that , in the cytoplasm , DNA-PK signals via IRF-3 to activate an anti-microbial innate immune response to DNA mediated by the production of IFN , cytokines and chemokines . We show that DNA-PK co-localises with sites of viral DNA replication during VACV infection and the innate immune response to DNA and to infection with vaccinia virus ( VACV ) and herpes simplex virus ( HSV-1 ) was impaired in both cells and mice which lack components of DNA-PK .
To identify novel cytoplasmic DNA sensors , we transfected biotinylated dsDNA composed of a concatenated 45-bp oligonucleotide ( called immunostimulatory DNA , ISD , ( Stetson and Medzhitov , 2006 ) ) into human embryonic kidney ( HEK ) 293T cells and isolated DNA/protein complexes from the cytoplasm by affinity purification ( Figure 1A ) . The three abundant proteins that bound specifically to DNA , and not to biotinylated lipoprotein pam-3-cys , were unequivocally identified as Ku70 , Ku80 and DNA-PKcs ( Figure 1A ) . We confirmed by subcellular fractionation that DNA-PK is present in the cytoplasm of resting cells ( Figure 1B ) , as reported previously ( Huston et al . , 2008; Balazs et al . , 2012 ) and that this cytoplasmic localisation was not due to nuclear contamination ( Figure 1B ) . The DNA pull-down was reproduced from murine fibroblasts ( Figure 1C ) , showing the cross-species conservation of this interaction . Additionally , in murine embryonic fibroblasts ( MEFs ) lacking Ku80 ( Xrcc5−/− ) , the remaining DNA-PK components did not interact with cytoplasmic DNA under the conditions tested , whereas in Prkdc−/− MEFs , Ku80 was still recruited ( Figure 1C ) . Therefore , the association between DNA-PKcs and cytoplasmic DNA is enhanced by Ku . 10 . 7554/eLife . 00047 . 003Figure 1 . DNA-PK binds DNA in the cytoplasm . ( A ) Biotinylated DNA or biotinylated pam-3-cys were transfected into HEK 293T cells and 30 min later DNA was affinity purified from the cytoplasmic fraction before analysis of DNA/protein complexes by SDS-PAGE . The three major protein bands visible by coomassie blue staining were excised from the gel and identified as Ku70 , Ku80 and DNA-PKcs by mass spectrometry . ( B ) Cells were untreated or transfected with DNA and 2 hr later proteins were extracted from the nucleus and cytoplasm . 50 μg of protein from each fraction ( representing 10% of total cytoplasmic protein by volume and 5% of the total nuclear protein ) was analysed by immunoblotting for DNA-PKcs , Ku70 and Ku80 . β-tubulin and histone-H3 were used as controls to indicate successful fractionation . ( C ) Ku80 is required for efficient binding to DNA . MEFs of indicated genotypes were transfected with DNA biotinylated at the 3′ end or biotinylated pam-3-cys ( Pam ) and lysed 1 hr later . Following affinity purification ( AP ) of biotinylated ligands with streptavidin agarose , proteins were analysed by SDS-PAGE and immunoblotting for DNA-PKcs and Ku80 . WCL; whole cell lysate . DOI: http://dx . doi . org/10 . 7554/eLife . 00047 . 003 In fibroblasts , the transcription of genes encoding chemokines and cytokines was induced in response to DNA in a length and dose-dependent manner , was sensitive to DNase treatment and required DNA to be transfected into the cell ( Figure 2A and data not shown ) . These observations are consistent with other studies ( Karayel et al . , 2009; Unterholzner et al . , 2010 ) . Next we tested DNA from various sources , including vaccinia virus ( VACV ) and Escherichia coli , for its ability to bind DNA-PK in the cytoplasm and to stimulate Cxcl10 transcription . Cxcl10 is strongly induced by intracellular nucleic acids ( Ishii et al . , 2006 ) via the IRF-3- and NF-κB-binding sites in its promoter ( Spurrell et al . , 2005 ) . Each of these different DNA species associated with DNA-PK in MEFs , and this association correlated with Cxcl10 induction ( Figure 2B , C—black bars ) , implying DNA sequence-independence of this response . A similar correlation between DNA binding and Il6 induction was also observed ( data not shown ) . 10 . 7554/eLife . 00047 . 004Figure 2 . The innate immune response to DNA requires DNA-PK in fibroblasts . ( A ) ISD DNA of different lengths was transfected into MEFs and the transcription of Cxcl10 was assayed by qPCR 6 hr later . ( B ) Double stranded oligonucleotides ( bio-ISD ) , concatenated ISD DNA ( bio-concatamers ) , genomic vaccinia virus DNA ( bio-VACV ) , genomic E . coli DNA ( bio-E . coli ) , poly ( dA:dT ) or the RNA analogue poly ( I:C ) were biotinylated and transfected into HEK293 cells . Following affinity purification of proteins from cytoplasmic extracts using streptavidin agarose , the bound proteins were analysed by SDS-PAGE and immunoblotting . AP; affinity purification . ( C ) Primary MEFs of the indicated genotype were transfected with 10 μg/ml of the same ( non-biotinylated ) nucleic acids as in ( A ) followed by qRT-PCR analysis measuring induction of Cxcl10 mRNA 6 hr later . ( D ) Wild type and Prkdc−/− transformed MEFs were transfected with DNA ( 10 μg/ml , left panel ) or stimulated with LPS ( 100 ng/ml , right panel ) and the level of transcription of cxcl10 was measured at the indicated times post stimulation . ( E ) Levels of Cxcl10 and Ifnβ were measured by ELISA from the supernatants of primary wild type and Prkdc−/− MEFs at passage 1 , 24 hr after transfection with DNA or poly ( I:C ) . ( F ) , ( G ) Primary wild type and Prkdc−/− MEFs at passage 1 were transfected with DNA or poly ( I:C ) and the level of induction of ( F ) Ifnb and Il-6 and ( G ) ccl4 and ccl5 mRNA was measured by qRT-PCR 6 hr later . ( H ) MEFs expressing Ku80 or lacking Ku80 were transfected with DNA and the transcription of Cxcl10 or Il6 was measured by qPCR 6 hr later . ( I ) Primary murine skin fibroblasts ( MSF ) from wild type adult mice or those lacking both Ku genes were transfected with DNA or poly ( I:C ) and the level of Ifnb induction was measured 6 hr later by qRT-PCR . ( J ) Xrcc5+/+/Trp53−/− and Xrcc5−/−/Trp53−/− MEFs were transfected with an expression plasmid encoding Ku80 or an empty vector ( EV ) control and Cxcl10 production was measured 24 hr later by ELISA . *** p<0 . 001 , ** p<0 . 01 , * p<0 . 05 , n ≥ 3 , error bars ± SEM , ns; non-stimulated . DOI: http://dx . doi . org/10 . 7554/eLife . 00047 . 004 Strikingly , in MEFs lacking DNA-PKcs ( Prkdc−/− ) there was a significant impairment in Cxcl10 transcription in response to double stranded concatenated ISD DNA ( Stetson and Medzhitov , 2006 ) ( from now on referred to as DNA ) , viral and bacterial DNA and poly ( dA:dT ) ( Figure 2C , white bars ) . In contrast , the response to the dsRNA analogue , poly ( I:C ) , was similar in the presence or absence of DNA-PKcs indicating the signalling defect is specific to DNA ( Figure 2C ) . Time-course experiments indicated that the defect in DNA stimulation of Prkdc−/− cells was consistent at all times tested ( Figure 2D ) and that these cells are not simply responding to DNA with different kinetics . The data presented in Figure 2C , D were carried out with MEFs derived from separate strains of Prkdc−/− mice , indicating that this phenotype was not confined to a single MEF line . As a control we also tested the response to LPS ( Figure 2D , right panel ) and found the Prkdc−/− cells responded like the wild type cells , with typically rapid kinetics , to this stimulus . In addition , the secretion of Cxcl10 and Ifnβ as well as the transcription of ifnb , il6 , and the chemokines Ccl4 and Ccl5 , was also consistently impaired in multiple preparations of passage 1 Prkdc−/− MEFs in response to DNA but not RNA ( Figure 2E–G ) . It is notable that , in MEFs , the transcription of other type 1 IFNs , type III IFNs and anti-inflammatory cytokines such as Il4 and Il10 was not observed in response to DNA stimulation ( data not shown ) . Defects in the production of cytokines in response to DNA were found in transformed MEFs lacking Ku80 ( Xrcc5−/− ) ( Figure 2H ) or Ku70 ( Xrcc6−/− ) ( not shown ) and the transcription of Ifnb in response to DNA was also impaired in primary adult murine skin fibroblasts lacking both Ku genes ( Figure 2I ) indicating that this phenotype requires both Ku and DNA-PKcs and is not restricted to embryonic cells . Finally , the re-expression of Ku80 in Xrcc5−/− cells restored their DNA-dependent production of Cxcl10 to wild-type levels ( Figure 2J ) . Collectively , these data indicate that DNA-PK acts as a DNA sensor by binding foreign DNA in the cytoplasm and activating a host innate immune response . We delimitated the signalling pathway downstream of DNA-PK by using various MEFs lacking specific signalling components . In Irf3−/−cells the increase in transcription of Ifnb , Cxcl10 , Il6 and Isg54 ( an IRF-3-dependent gene ( Navarro et al . , 1998 ) ) in response to DNA was lost ( Figure 3A ) . Equally , the up-regulation of Il6 and Cxcl10 mRNA was lost in Tbk1−/− MEFs ( Figure 3B ) and primary Tmem173−/− MEFs ( data not shown ) . Similar results have been reported previously ( Stetson and Medzhitov , 2006; Ishii et al . , 2006; Ishikawa et al . , 2009 ) and in a similar manner confirmed that cytoplasmic DNA sensing in fibroblasts is independent of TLRs ( Ishii et al . , 2006 ) , IPS-1 ( Kumar et al . , 2006 ) ( and hence both dsRNA-sensing pathways ( Kumar et al . , 2006 ) and RNA-pol III-dependent DNA sensing ( Ablasser et al . , 2009; Chiu et al . , 2009 ) ) , and DAI ( Ishii et al . , 2008; Wang et al . , 2008 ) , because there was no defect in DNA-dependent cytokine production in Myd88−/−/Ticam1−/− , Mavs−/− or Zbp1−/− MEFs ( data not shown ) . 10 . 7554/eLife . 00047 . 005Figure 3 . IRF-3 and TBK1 are required for the stimulation of multiple genes in response to DNA . ( A ) Primary wild type or Irf3−/− MEFs were transfected with DNA and the level of induction of Ifnb , Il6 , Cxcl10 and Isg54 mRNAs were measured by qRT-PCR 6 hr later . ( B ) Immortalised wild type or Tbk1−/− MEFs were transfected with DNA or poly ( I:C ) and the level of induction of Il-6 and Cxcl10 was measured by qRT-PCR 6 hr later . ( B ) *** p<0 . 001 , ** p<0 . 01 , n = 3 , error bars ± SEM , ns; non-stimulated . DOI: http://dx . doi . org/10 . 7554/eLife . 00047 . 005 To establish whether DNA-PK acts upstream of IRF-3 activation , we monitored IRF-3 translocation in response to DNA and RNA stimulation . In Prkdc−/− MEFs IRF-3 translocation in response to DNA , but not to RNA , was abrogated ( Figure 4A ) and we obtained similar results in Xrcc5−/− MEFs ( Figure 4B ) . In contrast , translocation into the nucleus of the p65 component of NF-κB , was unaffected by absence of the Prkdc gene ( Figure 4C ) . To confirm the activity of IRF-3 and NF-κB at the transcriptional level we examined induction of the IFN-inducible gene Isg54 ( Navarro et al . , 1998 ) , which is entirely dependent on IRF-3 activity following DNA stimulation ( Figure 3A ) . We also assayed expression of Nfkbia , a known NF-κB-dependent gene that encodes the IκBα protein ( Rupec et al . , 1999 ) . Induction of Isg54 , but not Nfkbia , was impaired in Prkdc−/− MEFs in response to DNA but not RNA transfection ( Figure 4D ) . These data confirm the DNA-specific defect in IRF-3 , but not NF-κB , activation in cells lacking DNA-PKcs and demonstrate that DNA-PK acts as a DNA sensor upstream of the IRF-3-dependent innate immune response . Our data also imply that the existence of an additional DNA sensing pathway in MEFs that is independent of DNA-PK and capable of activating NF-κB . 10 . 7554/eLife . 00047 . 006Figure 4 . DNA-PK activates IRF-3-dependent , NF-κB–independent signalling . ( A ) The localisation of endogenous IRF-3 was analysed by immunofluorescence 1 hr after transfection of primary wild type or Prkdc−/− MEFs with DNA or poly ( I:C ) ( left panels ) and quantified by scoring cells with nuclear staining ( right panels , n = 3 , counts of at least 50 nuclei per slide in randomised fields of view ) . ( B ) As ( A ) but with Xrcc5+/+/Trp53−/− and Xrcc5−/−/Trp53−/− MEFs . ( C ) Analysis of p65 translocation in wild type and Prkdc−/− MEFs carried out as in ( A ) . ( D ) Primary wild type or Prkdc−/− MEFs were transfected with DNA or poly ( I:C ) and the level of induction of Isg54 and Nfkbia were measured by qRT-PCR 6 hr later . ** p<0 . 01 , * p<0 . 05 , n = 3 , error bars ± SEM . ns; non-stimulated . Scale bar; 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00047 . 006 Previously , DNA-PK was shown to interact directly with IRF-3 ( Karpova et al . , 2002 ) . We confirmed this interaction by co-immunoprecipitation of both proteins ( data not shown ) . In the study of Karpova et al . ( 2002 ) DNA-PKcs was shown to phosphorylate IRF-3 at its N terminus , so enhancing its nuclear retention ( Karpova et al . , 2002 ) . To test whether phosphorylation of IRF-3 is necessary for the DNA-sensing function of DNA-PK , two approaches were taken . First , we tested PrkdcSCID MEFs , which express a kinase-dead form of the DNA-PKcs protein lacking its C terminal 83 amino-acids ( Blunt et al . , 1996; Guimarães-Costa et al . , 2009 ) , for their ability to produce Ifnb , Cxcl10 and Il-6 in response to DNA or RNA . However , PrkdcSCID MEFs showed no defect in induction of Ifnb ( Figure 5A ) , Il-6 ( Figure 5B ) , and Cxcl10 ( Figure 5C ) in response to either DNA or RNA stimulation . Second , we used the DNA-PKcs-specific inhibitor NU7026 ( Veuger et al . , 2003 ) to inhibit DNA-PKcs kinase activity in wild type MEFs . Cells treated with NU7026 showed no impairment in production of Cxcl10 in response to DNA transfection ( Figure 5D ) . Hence , the kinase activity of DNA-PKcs is not required for activation of IRF-3 in response to DNA . 10 . 7554/eLife . 00047 . 007Figure 5 . DNA-PKcs kinase activity is dispensable for the innate immune response to DNA . Primary fibroblasts from Balb/c or PrkdcSCID mice were transfected with DNA or poly ( I:C ) or infected with MVA or NDV and the level of induction of ( A ) Ifnb , ( B ) Cxcl10 , or ( C ) Il-6 was measured 6 hr later by qRT-PCR . ( D ) Fibroblasts were incubated with the indicated dose of DNA-PKcs kinase inhibitor , Nu7026 , or carrier control and then stimulated with 10 μg/ml DNA . Cxcl10 was measured by ELISA in the supernatants 24 hr following stimulation . n = 3 , error bars ± SEM . ( E ) . Hek293 Trex cells were stably transfected with FLAG-tagged STING under the control of a doxycycline-inducible promoter . STING expression was induced by addition of doxycycline ( Dox , 2 μg/ml ) for 24 hr and cells were stimulated by transfection with 5 μg/ml DNA for 6 hr . Protein lysates were then immunoblotted with the indicated antibodies . ( F ) STING-293Trex cells were induced to express STING by addition of doxycycline ( Dox , 2 μg/ml ) for 24 hr and stimulated with 5 μg/ml DNA for the indicated times . STING was then immunoprecipitated and whole cell lystes ( WCL ) or precipitated proteins ( IP ) were immunoblotted using the indicated antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 00047 . 007 The finding that the kinase activity of DNA-PKcs was not necessary for DNA sensing prompted an investigation of other interactions with signalling components . To do this we developed an inducible system to study the DNA sensing signalling pathway . HEK293 cells do not activate IRF3 in response to DNA stimulation due to a defect in STING , but by introducing an inducible version of STING into these cells , and then inducing STING expression with doxycycline , the cells activate IRF-3 as shown by its phosphorylation ( Figure 5E ) and produce cytokines ( not shown ) in response to double stranded ISD DNA concatamers . Using this system , immunoprecipitation showed that prior to DNA stimulation , STING exists in a complex with Ku70 but this interaction is abrogated following transfection of DNA ( Figure 5F ) . These data suggest that DNA-PK interacts with the STING-dependent signalling pathway and this changes upon DNA activation . We tested the relevance of this novel DNA-PK-dependent DNA sensing mechanism to virus infection using VACV strain modified virus Ankara ( MVA ) . MVA activates innate immunity via TLR-dependent and independent pathways ( Delaloye et al . , 2009 ) and , during infection , VACV DNA accumulates in cytoplasmic virus factories in association with many virus proteins ( Moss , 2007 ) . We reasoned that such large aggregates of foreign DNA , present in a cellular compartment where DNA does not normally reside , would present an excellent target for an innate immune DNA-sensing mechanism . By 6 hr post infection , both Ku80 and DNA-PKcs had accumulated in these viral factories ( Figure 6A ) together with the IRF-3-activating kinase TBK1 ( Figure 6B , left panel ) , consistent with its role in this sensing pathway , although IRF-3 was mostly nuclear , reflecting its activation by virus infection ( Figure 6B , right panel ) . 10 . 7554/eLife . 00047 . 008Figure 6 . DNA-PK and TBK1 localise to sites of VACV DNA replication in infected cells . HeLa cells were untreated or infected with MVA ( m . o . i . = 5 ) for 6 hr . Cells were then fixed and stained with antibodies against ( A ) Ku80 or DNA-PKcs , and ( B ) TBK1 or IRF-3 . Cytoplasmic viral factories formed after MVA infection are visualised with DAPI ( blue ) . Sites of co-localisation of DNA-PKcs or Ku80 with viral factories are indicated by white arrows . Scale bars; 10 μm . ni; non-infected . DOI: http://dx . doi . org/10 . 7554/eLife . 00047 . 008 In fibroblasts , MVA induces an IRF-3-dependent response ( Figure 7A ) which is independent of TLR signalling , RNA sensing ( and hence RNA-pol III-dependent DNA sensing ) and DAI ( data not shown ) , indicating that viral genomic DNA is a major target for the host response to VACV infection . We found the production of Cxcl10 and Il-6 was strongly impaired in cells lacking Prkdc−/− or Xrcc5−/− during MVA infection ( Figure 7B , C ) whereas the response of these cells to infection by Newcastle disease virus ( NDV , an RNA virus ) remained intact ( Figure 7B , C ) . Isg54 , but not Nfkbia , transcription was impaired in response to MVA infection in Prkdc−/− MEFs , whilst the response of both genes to NDV was equivalent ( Figure 7D ) , directly indicating that the DNA-PKcs-dependent activation of IRF-3 is important in the response to DNA virus infection . Additionally , in primary Prkdc−/− MEFs , there was greater MVA protein synthesis 4–8 hr post infection ( Figure 7E ) , not only re-enforcing the function of DNA-PKcs in the anti-viral response but also showing the failure to respond to DNA in Prkdc−/− MEFs is not due to a failure to infect these cells or synthesise viral macromolecules . These data provide direct evidence that the innate immune response to infection by a DNA virus is regulated by DNA-PK and confirm the role of this complex in cytoplasmic DNA sensing in the context of infection . 10 . 7554/eLife . 00047 . 009Figure 7 . DNA-PK contributes to the IRF-3-dependent innate immune response to MVA . ( A ) Ifnb and Cxcl10 transcription was measured 6 hr following MVA infection of primary WT and Irf3−/− fibroblasts at an m . o . i . of 5 . ( B ) Wild type and Prkdc−/− MEFs were infected with MVA or NDV and the level of induction of Cxcl10 and Il-6 were measured 6 hr later by qRT-PCR . ( C ) As ( A ) but with immortalised Xrcc5+/+/Trp53−/− and Xrcc5−/−/Trp53−/− MEFs . ( D ) The induction of Isg54 and Nfkbia mRNA was measured by qRT-PCR 6 hr after MVA or NDV infection of wild type and Prkdc−/− cells . ( E ) Expression of VACV proteins , analysed by immunoblotting with a rabbit polyclonal anti-VACV serum , at the indicated times following infection of primary Prkdc+/+ and Prkdc−/− MEFs with MVA ( m . o . i . = 5 ) . *** p<0 . 001 , ** p<0 . 01 , * p<0 . 05 , n = 3 , error bars ± SEM , ni; non-infected . DOI: http://dx . doi . org/10 . 7554/eLife . 00047 . 009 To confirm that DNA-PK-dependent DNA sensing contributes to the innate immune response in vivo , we transfected nucleic acids directly into the ear pinnae of mice and assayed the induction of innate immune transcriptional responses by qPCR analysis of extracted RNA 12 hr later . DNA transfection induced both Ifnb and Il6 and this was significantly reduced in Prkdc−/− mice . By contrast , both WT and Prkdc−/− mice exhibited equivalent responses to poly ( I:C ) ( Figure 8A ) . Hence , DNA-PKcs plays a key role in potentiating the innate immune response to ectopic DNA in vivo . We next infected WT and Prkdc−/− mice with either the DNA virus MVA or with influenza virus ( an RNA virus ) , and assayed expression of Ifnb and Il6 12 hr later . The absence of DNA-PK severely impaired induction of Ifnb and Il6 in response to MVA but not to influenza virus ( Figure 8B ) . Similarly , Prkdc−/− mice exhibited severe impairment of Il6 induction in response to HSV-1 ( Figure 8C ) : no Ifnb induction could be detected in this experiment . Overall these data confirm that the innate immune response to DNA viruses immediately after infection is significantly dependent upon DNA-PK , despite the presence of other DNA sensors reported hitherto . 10 . 7554/eLife . 00047 . 010Figure 8 . DNA-PKcs contributes to the innate immune response to MVA , HSV-1 and DNA in vivo . ( A ) Groups of five Prkdc+/+ and Prkdc−/− mice were injected intradermally into the ear pinna with cationic lipids complexed with 1 μg DNA or poly ( I:C ) . RNA was extracted from the tissue 12 hr later and Ifnb and Il6 transcription was measured by qPCR . ( B ) Groups of five Prkdc+/+ and Prkdc−/− mice were infected intradermally with 106 pfu of MVA or influenza virus strain A/PR/8/34 ( PR8 ) and 12 hr later Ifnb and Il6 transcription was measured by qPCR from RNA extracted from the local site of infection . ( C ) As ( B ) but with HSV-1 strain S17 . Note that levels of ifnb could not be measured above background in this experiment . *p<0 . 1 , ** p<0 . 01 , n = 5 , error bars ± SEM , ni; non-infected . DOI: http://dx . doi . org/10 . 7554/eLife . 00047 . 010
The identification of DNA-PK as a DNA sensor advances understanding of the innate immune response to infection and expands the current repertoire of DNA sensing mechanisms . In this study we show that the three proteins which constitute the DNA-PK complex; Ku70 , Ku80 and DNA-PKcs , bind in significant amounts to DNA transfected into the cytoplasm of resting cells leading to an IRF-3-dependent innate immune response . Consistent with this , we found DNA-PK and TBK1 localised to sites of DNA replication during virus infection . Previous studies have ruled out a role for DNA-PK in the production of Ifnb in response to DNA in monocytic cells ( Stetson and Medzhitov , 2006 ) and HEK293 cells ( Zhang et al . , 2011a ) . However , although DNA-PK components are expressed at high levels in a wide range of tissues and cell types , they are absent in primary macrophages ( ( Moll et al . , 1999 ) and data not shown ) , and we now show that in fibroblasts and in mice the innate immune response to DNA and DNA viruses is dependent on DNA-PKcs . The in vivo deficiency in DNA sensing in the absence of DNA-PKcs is observed despite the presumed presence of other DNA sensors . The observation that this defect is not extended to RNA , LPS or to RNA viruses shows that DNA-PK loss does not confer a general defect on intracellular innate immune signalling , and the same is true in cultured Prkdc−/− MEFs which respond normally to RNA . It was noticed that the defect in the innate immune response to DNA was greater in cells lacking DNA-PKcs than in cells lacking Ku , even though Ku can still bind DNA in the absence of DNA-PKcs . This suggests that signalling can progress with DNA-PKcs interacting with DNA in the absence of Ku but that Ku enhances the signalling process by increasing the affinity of the protein complex for DNA . The phenotypic defect in DNA sensing was consistent in cells from four separate genotypes , ruling out the possibility that it was caused by a second site mutation . Consistent with this reintroduction of Ku80 into Xrcc5−/− cells restored DNA sensing ( Figure 2J ) . The well-studied PrkdcSCID mutation kills the kinase activity of DNA-PKcs via the introduction of a premature stop codon that results in expression of a truncated protein ( Blunt et al . , 1996 ) . Although kinase activity is essential for DNA-PKcs to function in DNA repair , and therefore V ( D ) J recombination , this mutation did not affect the ability of DNA-PK to function as an innate immune DNA sensor . The functional significance of the interaction of DNA-PK with IRF-3 and its subsequent N-terminal phosphorylation reported by Karpova et al . ( 2002 ) remains unknown . However , we show here that Ku interacts with STING in resting cells and that this interaction is abrogated upon DNA stimulation ( Figure 5F ) . In the future , further work is necessary to understand how TBK-1 and STING contribute to the full activation of IRF-3 following binding of DNA to DNA-PK to allow IRF-3 translocation to occur . DNA can act as a powerful immunostimulatory agent in many contexts . DNA vaccination relies on DNA sensing to invoke a powerful innate immune response that , in turn , assists the adaptive response ( Ishii et al . , 2008 ) . Understanding how to optimise such vaccines , therefore , relies on understanding the mechanisms of detection of DNA by the immune system . Furthermore , one of the most common vaccine adjuvants , alum , acts by stimulating the release of DNA from neutrophils ( Zhang et al . , 2011b ) . Neutrophil extracellular traps ( NETs ) consist of webs of DNA with globular proteins , are released by a specific form of cell death ( Fuchs et al . , 2007 ) and function as antimicrobial traps , thereby contributing to the innate immune response ( Brinkmann et al . , 2004 ) . Outside the context of infection , DNA can act as a damage-associated molecular pattern ( DAMP ) , accelerating inflammatory responses following its release from dying or damaged cells directly contributing to the pathogenesis of various diseases such as atherosclerosis ( Oka et al . , 2012 ) and deep vein thrombosis ( Brill et al . , 2012 ) . The ability of DNA to act as a DAMP may also link nucleic acid sensing to several autoimmune disorders . In general , autoimmune conditions are characterised by immune responses against host molecules and tissues and are frequently associated with inflammation . Although the causative factors of many such conditions are incompletely understood , it is clear that deregulation of immune signalling may lead to autoinflammation and autoimmunity in some instances ( Rioux and Abbas , 2005 ) . Indeed , the accumulation of DNA and its subsequent detection by DNA sensing pathways can result in the initiation of autoimmune diseases . Mice which lack either of two enzymes responsible for degradation of DNA , 3′ repair exonuclease 1 ( Trex1 ) and DNaseII , develop spontaneous autoimmune disorders associated with the initiation of IRF-3-dependent cytosolic DNA sensing ( Kawane et al . , 2006; Stetson et al . , 2008; Okabe et al . , 2008 ) . The absence of DNaseII results in chronic polyarthritis which is thought to be a result of the inflammation caused by a lack of clearance of DNA from macrophages ( Kawane et al . , 2006; Okabe et al . , 2008 ) . Trex1 deficiency or mutation , on the other hand , is associated with systemic lupus erythematosus ( SLE ) , chilblain lupus and the human disease Aicardi-Goutières syndrome ( AGS ) ( Lee-Kirsch et al . , 2007a , 2007b; Stetson et al . , 2008 ) and in this case inflammation is initiated from non-haematopoietic cells via a STING-dependent pathway ( Gall et al . , 2012 ) . Interestingly , defects in the clearance of NETs have also been suggested to contribute to the initiation of SLE ( Hakkim et al . , 2010 ) . Furthermore , the up-regulation of the inflammasome-activating DNA sensor absent in melanoma 2 ( AIM2 ) in a mouse model of lupus and in patients with SLE-associated nephritis ( Roberts et al . , 2009; Kimkong et al . , 2009 ) and the presence of anti-DNA-PK and anti-RNA-Pol III antibodies in patients with SLE and systemic sclerosis ( Cavazzana et al . , 2008 , 2009 ) makes the link between DNA sensing and autoimmune disorders worthy of further investigation . The identification of several candidate DNA sensors in multiple cells types in recent years ( Hornung and Latz , 2010 ) suggests the evolution of redundancy in this system . This redundancy is not surprising given the tendency of pathogens to evolve escape-mechanisms for evading host immune mediators ( Versteeg and García-Sastre , 2010; Bardoel and Strijp , 2011 ) in turn inducing the host to evolve further pathogen recognition mechanisms . In the context of DNA sensing this is exemplified by the relatively recent evolution of the PYHIN domain proteins , such AIM2 and IFI16 ( Schattgen and Fitzgerald , 2011 ) , which are a mammalian addition to the ancient innate immune system . Furthermore , a function of DAI was recently uncovered by identifying murine cytomegalovirus protein , vIRA , which interacts with DAI and inhibits its ability to initiate DNA-induced necroptosis ( Upton et al . , 2012 ) . The biological function for DAI may therefore be in the initiation of a cell death pathway , rather than an IRF-3 dependent inflammatory response . This indicates there are at least three outcomes to cytosolic DNA sensing , the induction of cytokine expression via IRF-3 and NF-kB activation , the secretion of IL-1β via the AIM 2 inflammasome and the induction of necroptosis by DAI . What is not clear though , is how these different responses contribute to the overall immune response to infection by DNA pathogens and to what extent they are cell and tissue-type dependent . Further work is necessary to uncover the relative contributions of these different DNA sensing mechanisms in specific cell types and to different DNA structures as well as to understand how these sensors co-ordinate with STING and other adaptor proteins to activate TBK1 and IRF-3 ( Paludan et al . , 2011; Barber , 2011 ) . Overall these findings provide a novel function for DNA-PK in the innate immune response , beyond its roles in DNA repair and V ( D ) J recombination , and increase our understanding of the innate immune response to cytoplasmic DNA .
Prkdc+/− mice on a 129 background were a kind gift from Dr Fred Alt ( Gao et al . , 1998 ) and PrkdcSCID and Balb/c mice were from Harlan laboratories . Animals were maintained as required under UK Home Office regulations . Groups of 5 age and sex matched mice were injected intradermally with 1 µg DNA or poly ( I:C ) ( Invivogen , San Diego , CA ) pre-incubated with 2 μl Lipofectamine2000 in Optimem ( Life Technologies , Grand Island , NY ) or injected intradermally with 106 plaque forming units ( pfu ) of MVA , HSV-1 or influenza virus A/PR/8/34 in PBS . Primary mouse embryonic fibroblasts ( MEFs ) were isolated from E13 . 5 embryos derived from time-mated pregnant mice using standard protocols . HEK293T and HEK293 Trex cells ( Life Technologies , Grand Island , NY ) were maintained in DMEM containing 10% FBS with the addition of blasticidin ( 5 μg/ml ) and zeocin ( 100 μg/ml ) for the selection and maintenance of the inducible STING-expressing cell line ( STING-293Trex ) . MEFs and murine skin fibroblasts ( MSFs ) from various genetic backgrounds were maintained in DMEM containing 15% FBS , 100 U/ml penicillin and 100 μg/ml streptomycin . Primary Prkdc+/+ and Prkdc−/− MEFs were prepared in house on multiple occasions or supplied as a kind gift by Dr Brian Hemmings and were used only at passage 1 . Transformed Prkdc−/− MEFs ( used solely for experiments leading to the data presented in Figure 2D ) were a kind gift from Professor Penelope Jeggo . HeLa cells were maintained in RPM . I containing 10% FBS and 2 mM L-glutamine . Transfections were carried out with Fugene6 ( Roche , Penzburg , Germany ) . Double stranded oligonucleotide DNA , ISD ( sense sequence , TACAGATCTACTAGTGATCTATGACTGATCTGTACATGATCTACA ) was phosphorylated at the 5′ end by incubation with T4 polynucleotide kinase ( New England Biolabs , Ipswich , MA ) for 30 min at 37°C and then ligated with T4 DNA ligase ( Promega , Madison , WI ) for 16 hr at 15°C . Concatenation was confirmed by agarose gel electrophoresis . This DNA was used at 10 μg/ml for transfection unless stated otherwise . 3′ biotinylated oligonucleotides were purchased from IDT DNA Technologies . Other nucleic acids were biotinylated using the Photoprobe biotinylation kit ( Vector Labs , Burlingame , CA ) following the manufacturer's instructions . Concatenated oligonucleotide DNA ( sequence as above ) that was biotinylated at the 3′ end was transfected into cells using PEI ( Sigma-Aldrich , St Louis , MI ) . After 30 min , cells were lysed in buffer containing 10 mM Tris–Cl , pH 8 , 0 . 1% NP40 , 10 mM MgCl2 and the cytoplasmic fraction was isolated by centrifugation at 1500g for 3 min . Streptavidin agarose ( Thermo Scientific , Rockford , IL ) , 30 μl , was incubated with the lysate for 1 hr at 4°C and then washed three times in PBS . Purified proteins were analysed by SDS-PAGE and immunoblotting or stained by coomassie-blue and identified by liquid chromatography and tandem mass spectrometry ( LC-MS/MS ) at the Centre for Systems Biology at Imperial College London . VACV strain MVA was purified from cytoplasmic extracts of infected BHK-21 cells by sedimentation through a cushion of 36% ( wt/vol ) sucrose and was titrated by plaque assay on chicken embryo fibroblasts . NDV and influenza virus strain A/PR/8/34 were kind gifts from Prof Wendy Barclay . HSV-1 strain S17 , a gift from Dr Colin Crump , was grown in Vero cells and purified on ficoll gradients . These viruses were used for infections for the indicated times and at indicated doses . For immunoblotting , cell lysates were separated by polyacrylamide gel electrophoresis and transferred onto Immobilon P membranes ( GE Healthcare , Little Chalfont , UK ) . The membranes were blocked in 5% non-fat milk in TBS containing 0 . 1% Tween 20 for 1 hr at room temperature . Membranes were probed with antibodies against Ku70 ( Abcam , Cambridge , UK ) , Ku80 ( Santa Cruz Biotech , Santa Cruz , CA ) , DNA-PKcs ( Millipore , Billerica , MA ) , tubulin ( Millipore , Billerica , MA ) , histone H3 ( Millipore , Billerica , MA ) , hIRF-3 ( Santa Cruz Biotech , Santa Cruz , CA ) , p-IRF-3 ( serine 396 , Abcam , Cambridge , UK ) , HMGB1 ( Abcam , Cambridge , UK ) , FLAG ( Sigma-Aldrich , St Louis , MI ) or VACV strain Western Reserve ( Law et al . , 2006 ) and bound immunoglobulin was detected with horse-radish peroxidase-linked secondary antibodies ( Agilent , Santa Clara , CA ) . Ku80 , IRF-3 ( Life Technologies , Grand Island , NY ) or control antibodies were used for immunoprecipitation from HeLa cell lysates expressing IRF-3 . FLAG-agarose matrix ( Sigma-Aldrich , St Louis , MI ) was used for immunoprecipitation of STING from STING-293Trex cells . For immunofluorescence , cells were seeded onto 15 mm glass coverslips , infected with MVA at 5 pfu per cell for 6 hr or transfected and fixed with 4% paraformaldehyde . Cells were permeabilised with PBS containing 0 . 2% Triton-X100 and blocked with 5% non-fat milk in PBS containing 0 . 1% Tween 20 for 1 hr at 20°C . Incubation with primary antibodies against Ku70 ( Abcam , Cambridge , UK ) , DNA-PKcs ( Millipore , Billerica , MA ) , mIRF-3 ( Life Technologies , Grand Island , NY ) hIRF-3 , p65 ( both Santa Cruz Biotech , Santa Cruz , CA ) or TBK1 ( Cell Signalling , Danvers , MA ) diluted in PBS with 1% milk , for 1 hr at 20°C was followed by detection with alexa-fluor-conjugated secondary antibodies ( Life Technologies , Grand Island , NY ) . Cells were counterstained with DAPI and mounted with Mowiol . Images were obtained with a Zeiss Pascal 510 confocal microscope and processed with Zeiss LSM software ( Zeiss , Oberkochen , Germany ) . For quantification of translocation or IRF-3 or NF-κB into the nucleus 50 cells were counted in random fields of view , in biological triplicates , for each condition and scored for the presence of nuclear staining . Levels of Cxcl10 and Ifnβ in cell supernatants were measured using ELISA kits ( R&D systems , Minneapolis , MN or PBL , Piscataway , NJ , respectively ) according to the manufacturer's instructions . Total cellular RNA was extracted using an RNeasy kit ( Qiagen , Hilden , Germany ) . cDNA synthesis was carried out with Superscript III Reverse Transcriptase ( Life Technologies , Grand Island , NY ) using 500 ng of template RNA . qPCR was performed on a 7900HT series thermocycler ( Life Technologies , Grand Island , NY ) with Fast SYBR Green Master Mix ( Life Technologies , Grand Island , NY ) . HPRT was used as the reference gene in all assays . Data were analysed with RQ manager 1 . 2 software ( Life Technologies , Grand Island , NY ) and presented as a fold increase relative to time zero . Primers for qPCR were as follows:Cxcl10 For 5′ ACTGCATCCATATCGATGAC 3′ , Cxcl10 Rev 5′ TTCATCGTGGCAATGATCTC 3′ , Ifnβ For 5′ CATCAACTATAAGCAGCTCCA 3′ , Ifnβ Rev 5′ TTCAAGTGGAGAGCAGTTGAG 3′Ccl5 For 5′ ACGTCAAGGAGTATTTCTACAC 3′ , Ccl5 Rev 5′ GATGTATTCTTGAACCCACT 3′ , Il-6 For 5′ GTAGCTATGGTACTCCAGAAGAC 3′ , Il-6 Rev 5′ GTAGCTATGGTACTCCAGAAGAC 3′ , Cxcl2 For 5′ GAGCTTGAGTGTGACGCCCCC 3′ , Cxcl2 Rev 5′ GTTAGCCTTGCCTTTGTTCAG 3′ , Ccl3 For 5′ ACTGCCTGCTGCTTCTCCTA 3′ , Ccl3 Rev 5′ TTGGAGTCAGCGCAGATCTG 3′ , Ccl4 For 5′ GCCCTCTCTCTCCTCTTGCT 3′ , Ccl4 Rev 5′ CTGGTCTCATAGTAATCCATC 3′ , Ccl2 For 5′ CTTCTGGGCCTGCTGTTCA 3′ , Ccl2 Rev 5′ CCAGCCTACTCATTGGGATCA3′ , Ifnγ For 5′ TCAAGTGGCATAGATGTGGAAGAA3′Ifnγ Rev 5′ TGGCTCTGCAGGATTTTCATG 3′ , Il-4 For 5′ CATGCACGGAGATGGATG 3′ , Il-4 Rev 5′ ACCTTGGAAGCCCTACAGAC 3′ , Il-10 For 5′ TCCTTAATGCAGGACTTTAAGGGTTACTTG 3′ , Il-10 Rev 5′ GACACCTTGGTCTTGGAGCTTATTAAAATC 3′ , HPRT For 5′ GTTGGATACAGGCCAGACTTTGTTG 3′ , HPRT Rev 5′ GATTCAACTTGCGCTCATCTTAGGC 3′ , Nfkbia For 5′ CTGCAGGCCACCAACTACAA3′ , Nfkbia Rev 5′ CAGCACCCAAAGTCACCAAGT 3′Isg54 For 5′ ATGAAGACGGTGCTGAATACTAGTGA 3′Isg54 Rev 5′ TGGTGAGGGCTTTCTTTTTCC 3′ Statistical analysis was carried out using student's t-test with Welch's correction where necessary . | For multicellular organisms , the innate immune system is the first immunological defence against infection , rapidly recognizing and responding to the presence of any pathogen . Many different cell types contribute to the innate immunity , including fibroblasts , epithelial cells , dendritic cells and macrophages . Once alerted to injury or infection , these cells release proteins called cytokines , interferons and chemokines into the blood or directly into tissue . These proteins act as messengers and interact with receptors on the surfaces of other cells in the immune system , stimulating them to join the battle against the infection . Detecting nucleic acids such as DNA is an important part of recognizing pathogens and infectious agents , particularly viruses , and activating the innate immune system . However , while the presence of DNA in the cytoplasm is known to initiate an innate immune response , we do not fully understand how this foreign DNA is sensed , or how the innate immune system is activated once foreign DNA has been detected . Here Ferguson et al . report that a well-known complex of three proteins , collectively called DNA-dependent protein kinase , is able to activate an innate immune response when it detects foreign DNA . This enzyme , called DNA-PK for short , is best known for its ability to repair broken DNA inside the nucleus . Now Ferguson et al . have found that it is also present at high levels within fibroblasts , cells that are often primary targets of viral infection , and they go on to explain how the detection of DNA by DNA-PK triggers a sequence of events that leads to the innate immune response being activated . These events include the transcription of type I interferon , chemokines and cytokines in a manner that depends on the presence IRF-3 , a transcription factor that has a central role in the response of the immune system to viral infection . By identifying a role for DNA-PK in the cytoplasm as a DNA sensor , the work of Ferguson et al . increases our understanding of innate immunity . It may also , in the future , lead to an improved understanding of autoimmunity , and might also assist in the development of more immunogenic vaccines based on DNA or microbes that contain DNA . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease",
"immunology",
"and",
"inflammation"
] | 2012 | DNA-PK is a DNA sensor for IRF-3-dependent innate immunity |
Animals discriminate stimuli , learn their predictive value and use this knowledge to modify their behavior . In Drosophila , the mushroom body ( MB ) plays a key role in these processes . Sensory stimuli are sparsely represented by ∼2000 Kenyon cells , which converge onto 34 output neurons ( MBONs ) of 21 types . We studied the role of MBONs in several associative learning tasks and in sleep regulation , revealing the extent to which information flow is segregated into distinct channels and suggesting possible roles for the multi-layered MBON network . We also show that optogenetic activation of MBONs can , depending on cell type , induce repulsion or attraction in flies . The behavioral effects of MBON perturbation are combinatorial , suggesting that the MBON ensemble collectively represents valence . We propose that local , stimulus-specific dopaminergic modulation selectively alters the balance within the MBON network for those stimuli . Our results suggest that valence encoded by the MBON ensemble biases memory-based action selection .
To survive in a dynamic environment , an animal must discover and remember the outcomes associated with the stimuli it encounters . It then needs to choose adaptive behaviors , such as approaching cues that predict food and avoiding cues that predict danger . The neural computations involved in using such memory-based valuation of sensory cues to guide action selection require at least three processes: ( 1 ) sensory processing to represent the identity of environmental stimuli and distinguish among them; ( 2 ) an adaptive mechanism to assign valence—positive or negative survival value—to a sensory stimulus , store that information , and recall it when that same stimulus is encountered again; and ( 3 ) decision mechanisms that receive and integrate information about the valence of learned stimuli and then bias behavioral output . To understand such decision-making processes , one approach is to locate the sites of synaptic plasticity underlying memory formation , identify the postsynaptic neurons that transmit stored information to the downstream circuit and discover how their altered activities bias behavior . The mushroom body ( MB ) is the main center of associative memory in insect brains ( de Belle and Heisenberg , 1994 , Heisenberg et al . , 1985; Dubnau et al . , 2001; McGuire et al . , 2001 ) . While the MB processes several modalities of sensory information and regulates locomotion and sleep ( Martin et al . , 1998; Liu et al . , 1999; Joiner et al . , 2006; Pitman et al . , 2006; Zhang et al . , 2007; Hong et al . , 2008; Vogt et al . , 2014 ) , MB function has been most extensively studied in the context of olfactory memory—specifically , associating olfactory stimuli with environmental conditions in order to guide behavior . In Drosophila , olfactory information is delivered to the MB by projection neurons from each of ∼50 antennal lobe glomeruli ( Marin et al . , 2002; Wong et al . , 2002; Jefferis et al . , 2007; Lin et al . , 2007; Vosshall and Stocker , 2007; Yu et al . , 2010 ) . Connections between the projection neurons and the ∼2000 Kenyon cells ( KCs ) , the neurons whose parallel axonal fibers form the MB lobes ( Crittenden et al . , 1998; Aso et al . , 2009 ) , are not stereotyped ( Figure 1A ) ( Murthy et al . , 2008; Caron et al . , 2013 ) ; that is , individual flies show distinct wiring patterns between projection neurons and KCs . Sparse activity of the KCs represents the identity of odors ( Laurent and Naraghi , 1994; Perez-Orive et al . , 2002; Turner et al . , 2008 ) . The output of the MB is conveyed to the rest of the brain by a remarkably small number of neurons—34 cells of 21 cell types per brain hemisphere ( Figure 1B , Table 1 ) ( Aso et al . , 2014 ) . 10 . 7554/eLife . 04580 . 003Figure 1 . Circuit diagrams of the mushroom body . ( A ) The innervation patterns of extrinsic neurons define 15 compartments in the MB lobes and one compartment in the core of distal pedunculus ( pedc ) ; the compartments are represented by rectangles that are color-coded based on the neurotransmitter used by the mushroom body output neurons ( MBONs ) having dendrites in that compartment ( green , glutamate; blue , GABA; orange , acetylcholine ) . Projection neurons ( far left , colored arrows ) from the antennal lobe convey olfactory sensory information to the MB calyx where they synapse on the dendrites of Kenyon cells ( KCs ) . The parallel axon fibers of the KCs ( gray lines ) form the lobes ( α/β , α′/β′ and γ ) where KCs terminate onto the dendrites of the MBONs . Each of the seven types of KCs innervates a specific layer within a given lobe . The dendrites of individual MBON types and the terminals of dopaminergic neurons ( DANs ) intersect the longitudinal axis of KC axon-bundles in specific compartments along the lobes . MBONs using the same transmitter are spatially co-localized in the lobes ( See Figure 1B ) . Innervation areas of PPL1 and PAM cluster DANs axons in the MB lobes are indicated by the rectangles outlined in dashed lines . Activation of subsets of DANs can convey punishment or reward , respectively inducing aversive or appetitive memory when activation is paired with odor presentation . The size of arrows indicates magnitude of memory induced by DAN activation . See text for references . ( B ) Schematic representation of the 21 cell types of MBONs in the lobes and one cell type of MBON in the calyx based on the data presented in the accompanying manuscript ( Aso et al . , 2014 ) : circles , cell bodies; semicircles , dendrites; arrowheads , axon terminals; color-coding is by neurotransmitter as in panel ( A ) Three MBON cell-types ( GABAergic MBON-γ1pedc>α/β , glutamatergic MBON-γ4>γ1γ2 and MBON-β1>α; marked as 11 , 5 and 6 respectively ) send axons into the MB lobes . Axons of MBON-γ4>γ1γ2 project from γ4 to γ1 and γ2 , and thus have the potential to affect activity of MBON-γ1pedc>α/β . From γ1 , the axon of MBON-γ1pedc>α/β projects to compartments in the α/β lobes including β1 , where dendrites of MBON-β1>α arborize . Axons of both MBON-γ1pedc>α/β and MBON-β1>α project to the compartments in the α lobe . Therefore activity of MBONs in the α lobe can be regulated by these layered inter-compartmental connections . These three types of MBONs ( 11 , 5 and 6 ) do not project back to their own dendrites . Therefore , the organization of the MBONs can be viewed as forming a multilayered feed-forward network ( Aso et al . , 2014 ) . MBONs project to a small number of brain areas: the crepine ( CRE; a region surrounding the horizontal/medial lobes ) , the superior medial protocerebrum ( SMP ) , superior intermediate protocerebrum ( SIP ) and superior lateral protocerebrum ( SLP ) and the lateral horn ( LH ) . The size of the arrowhead reflects the relative number of termini in each area . The MBONs are numbered and listed in Table 1 . See the accompanying manuscript ( Aso et al . , 2014 ) and Table 1 for details . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 00310 . 7554/eLife . 04580 . 004Table 1 . The list of MBON cell typesDOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 004cell type namecell body clusterputative transmitterNumber of cellsshort names ( number in Figure 1B ) names in previous publicationsMBON-γ5β′2aM4/M6 clusterglutamate1MBON-01MB-M6MBON-β2β′2a1MBON-02MBON-β′2mp1MBON-03MB-M4MBON-β′2mp_bilateral1MBON-04MB-M4 ? MBON-γ4>γ1γ2MV2 cluster1MBON-05MBON-β1>α1MBON-06MB-MV2MBON-α12MBON-07MBON-γ3γ3 clusterGABA1MBON-08MBON-γ3β′11MBON-09MBON-β′18MBON-10MBON-γ1pedc>α/β1MBON-11MB-MVP2MBON-γ2α′1V3/V4 clusteracetycholine2MBON-12MBON-α′21MBON-13MB-V4MBON-α32MBON-14MB-V3MBON-α'1V2 cluster2MBON-15MBON-α′3ap1MBON-16MB-V2α′MBON-α′3m2MBON-17MB-V2α′MBON-α2sc1MBON-18MB-V2αMBON-α2p3p2MBON-19MBON-γ1γ2N . D . 1MBON-20MBON-γ4γ51MBON-21MBON-calyx1MBON-22MB-CP1The name of each cell type is given . For MBONs , we use MBON plus the name of MB lobe compartment ( s ) in which their dendrites arborize ( for example , MBON-α1 for the MB output neurons that have dendrites in the α1 compartment; MBON-γ5β′2a for the MB output neurons that have dendrites in the γ5 compartment and the anterior layer of the β′2 compartment ) . For the three classes of MBONs which also have axon terminals in the MB lobes , the ‘>’ symbol is followed by the compartments ( or lobes ) in which the axons terminate ( for example , MBON-β1>α for the MB output neuron with dendrites in β1 and synaptic terminals in the α lobe . ) . Subsets of MBONs have their cell bodies clustered , presumably reflecting their common developmental origin; these clusters are indicated . The neurotransmitter used by each MBON ( where known ) is indicated as is the number of cells of that type found per brain hemisphere ( data from the accompanying manuscript; ( Aso et al . , 2014 ) . We also provide a short cell type name based on simple numbering; the number of an MBON corresponds to number shown in Figure 1B . Many of these cell types have been previous described under alternative names , as indicated ( Tanaka et al . , 2008 ) . The information flow from the KCs to the MB output neurons ( MBONs ) has been proposed to transform the representation of odor identity to more abstract information , such as the valence of an odor based on prior experience ( See discussion in Aso et al . , 2014 ) . In contrast to KCs , MBONs have broadly tuned odor responses; any given odor results in a response in most MBONs , although the magnitude of the response varies among MBON cell types ( Hige et al . , unpublished ) ( Cassenaer and Laurent , 2012 ) . Unlike the stereotyped response to odors of the olfactory projection neurons that deliver odor information to the MB , the odor tuning of the MBONs is modified by plasticity and varies significantly between individual flies , suggesting that MBONs change their response to odors based on experience ( Hige et al . , unpublished ) . For olfactory associative memory in Drosophila , multiple lines of evidence are consistent with a model in which dopamine-dependent plasticity in the presynaptic terminals of KCs alters the strength of synapses onto MBON dendrites . This is thought to provide a mechanism by which the response of MBONs to a specific odor could represent that odor's predictive value . D1-like dopamine receptors and components of the cAMP signaling pathway , such as the Ca2+/Calmodulin-responsive adenylate cyclase encoded by the rutabaga gene , are required specifically in the KCs for memory formation ( Livingstone et al . , 1984; Zars et al . , 2000; Schwaerzel et al . , 2003; Kim et al . , 2007; McGuire et al . , 2003; Gervasi et al . , 2010; Qin et al . , 2012 ) and rutabaga was shown to be required for the establishment of the differences in MBON odor tuning between individuals ( Hige et al . , unpublished ) . Reward and punishment recruit distinct sets of dopaminergic neurons ( DANs ) that project to specific regions in the MB lobes ( Mao and Davis , 2009; Burke et al . , 2012; Liu et al . , 2012 ) . Moreover , exogenous activation of these DANs can substitute for reinforcing stimuli to induce either appetitive or aversive memory , depending on DAN cell type ( Figure 1A ) ( Yamagata et al . , in press , Perisse et al . , 2013; Schroll et al . , 2006; Claridge-Chang et al . , 2009; Aso et al . , 2010 , 2012; Liu et al . , 2012; Burke et al . , 2012 ) . In sum , while the identity of the learned odor is likely encoded by the small subset of KCs activated by that odor , whether dopamine-mediated modulation assigns positive or negative valence to that odor would be determined by where in the MB lobes KC-MBON synapses are modulated and thus which MBON cell types alter their response to the learned odor . Combining the above observations with our comprehensive anatomical characterization of MB inputs and outputs ( Aso et al . , 2014 ) lays the groundwork for testing models of how the MB functions as a whole . We suggest that each of the 15 MB compartments—regions along the MB lobes defined by the arborization patterns of MBONs and DANs ( see Figure 1 ) —functions as an elemental valuation system that receives reward or punishment signals and translates the pattern of KC activity to a MBON output that serves to bias behavior by altering either attraction or aversion . This view implies that multiple independent valuation modules for positive or negative experiences coexist in the MB lobes , raising the question of how the outputs across all the modules are integrated to result in a coherent , adaptive biasing of behavior . Although several MBON cell types have been shown to play a role in associative odor memory ( Sejourne et al . , 2011; Pai et al . , 2013; Placais et al . , 2013 ) , the functions of most MBONs have not been studied . Based on our anatomical analyses ( Aso et al . , 2014 ) , we believe that just 34 MBONs of 21 types provide the sole output pathways from the MB lobes . To gain mechanistic insight into how the ensemble of MBONs biases behavior , we would first like to know the nature of the information conveyed by individual MBONs and the extent to which their functions are specialized or segregated into different information channels . Then we need to discover how the activities of individual MBONs contribute to influence the behavior exerted by the complete population of MBONs . Thus , in order to understand how memory is translated into changes in behavior , we need to have experimental access to a comprehensive set of MBONs and investigate how the outputs from different MBONs bias behavior , singly and in combination . In the accompanying paper ( Aso et al . , 2014 ) , we describe the detailed anatomy of the DANs and MBONs ( Figure 1 ) and the generation of intersectional split-GAL4 driver lines to facilitate their study . All but one of the 21 MBON cell types consists of only one or two cells per hemisphere ( Table 1 ) . Dendrites of MBONs that use the same neurotransmitter—GABA , glutamate or acetylcholine—are spatially clustered in the MB lobes . Intriguingly , this spatial clustering resembles the innervation patterns of modulatory input by two clusters of dopaminergic neurons , PPL1 and PAM . MBONs have their axonal terminals in a small number of brain regions , but their projection patterns also suggest pathways for relaying signals between compartments of the MB lobes; three MBONs send direct projections to the MB lobes and several other MBONs appear to target the dendrites of specific DANs . Our split-GAL4 drivers give us the capability to express genetically encoded effectors in identified MBONs to modify their function . In this study , we examine the roles of specific MBONs in various learning and memory tasks as well as in the regulation of locomotion and sleep . We also studied whether direct activation of specific MBONs are sufficient to elicit approach or avoidance . Our results indicate that the ensemble of MBONs does not directly specify particular motor patterns . Instead , MBONs collectively bias behavior by conveying the valence of learned sensory stimuli , irrespective of the modality of the stimulus or the specific reward or punishment used during conditioning .
A powerful strategy to discover if a neuronal population plays a role in a particular behavior is to observe the effects of inactivating or activating those neurons . A genetic driver can be used to express an exogenous protein that either promotes or blocks neuronal activity . By repeating such manipulations with a large collection of drivers , each specific for a different set of neurons , one can in principle discover cell types required for a particular behavior . This is analogous to a screen to identify genes that when mutated disrupt a cellular function , but it is the activity of cells—rather than that of genes—being altered . This approach has been widely used in Drosophila ( Reviewed in Venken et al . , 2011; Griffith , 2012 ) . There are many challenges in carrying out such an approach . As in all biological systems , we expect extensive resiliency to perturbation . Such robustness might mask the effects of manipulating the activity of a small population of neurons , making them undetectable above the normal variation between animals . In addition to these inherent limitations , the genetic tools at our disposal have often been inadequate . In this study , we have used improved genetic tools and employed several strategies to mitigate their remaining limitations , as detailed below and in ‘Materials and methods’ . By assaying many different behaviors with the same genetic reagents , we were better able to evaluate the specificity of the behavioral effects we observed . Our experimental design placed an emphasis on avoiding false positives . Extensive analysis was restricted to MBON lines showing a phenotype in the initial screens and our scoring criteria were conservative . Consequently , we are likely to have missed detecting some cell types with effects on the behavior under assay . We sought to first determine the nature of the information conveyed by MBONs . In the most widely used olfactory conditioning assay , memory is assessed after training by allowing flies to distribute between two arms of a T-maze: one arm perfused with a control odor and the other arm with an odor that had been previously associated with punishment or reward . If the valence of the learned odor were encoded by the altered activities of specific MBONs , artificial activation of those MBONs in untrained flies in the absence of odor presentation would be expected to result in avoidance or approach behavior that mimicked the conditioned odor response . To test this hypothesis , we used a circular arena in which groups of flies expressing the red-shifted channelrhodopsin CsChrimson ( Klapoetke et al . , 2014 ) in MBONs were allowed to freely distribute between dark quadrants and quadrants with activating red light ( Figure 2A; see ‘Materials and methods’ for details ) . Activating sensory neurons for CO2 or bitter taste in this manner induced strong avoidance ( Figure 2B ) , consistent with previous reports ( Suh et al . , 2007 ) . By testing our collection of MBON split-GAL4 drivers in this assay , we found cell types whose activation resulted in avoidance of the red light and others whose activation led to attraction ( Figure 2C ) . Behavioral valence was highly correlated with MBON transmitter type: all MBONs eliciting aversion were glutamatergic and all the MBONs eliciting attraction were either GABAergic or cholinergic . We selected two split-GAL4 drivers for each neurotransmitter type for further analysis , choosing lines that gave robust phenotypes in the initial screening and that showed highly specific expression patterns based on direct assessment of CsChrimson expression ( Figure 3A–F ) . The phenotypes of these lines were reproducible ( Figure 3G–I ) and none of the drivers showed significant preference to the red light in the absence of the CsChrimson effector ( Figure 3G ) , confirming that phototaxis was limited at the wavelength and intensity of light used . These results demonstrate that activation of MBONs is itself sufficient to elicit either approach or avoidance , depending on the cell type . 10 . 7554/eLife . 04580 . 005Figure 2 . Screening MBONs for behavioral valence . ( A ) Behavioral valence assay . Approximately twenty female flies were placed in a 10 cm circular arena in a dark chamber and allowed to distribute themselves freely among the four quadrants . From 30–60 s , two of the quadrants ( Q2&3 ) were illuminated with 617 nm peaked red LED lights to activate CsChrimson-containing neurons; from 90–120 s , the other two quadrants ( Q1 & 4 ) were illuminated instead . No quadrants were illuminated from 0–30 s or from 60–90 s . The flies were video recorded and their locations in the arena were used to calculate behavioral preferences . ( B ) For these time-domain plots , the quadrant preference at each point in time was calculated as [ ( number of flies in Q2&3 ) - ( number of flies in Q1&4 ) ]/ ( total number of flies ) . Flies expressing CsChrimson in receptor neurons for CO2 or for bitter taste ( using the indicated drivers and 20xUAS-CsChrimson-mVenus in attP18 ) avoided the illuminated quadrants , whereas the control genotype ( empty driver , pBDPGAL4U in attP2/20xUAS-CsChrimson-mVenus in attP18 ) showed a very slight preference for illuminated quadrants . Lighter colored areas around lines indicate the standard error of the mean . ( C ) Screening of MBON drivers for behavioral valence . The red-light preference index ( PI ) was defined as: [ ( number of flies in illuminated quadrants ) - ( number of flies in non-illuminated quadrants ) ]/ ( total number of flies ) . For each 2-min experiment , the overall PI was calculated by averaging the PIs from the final 5 s of each of the two red-lights-on conditions ( namely , 55–60 s and 115–120 s ) . MBON cell types expressed in each driver are shown in the matrix below the graph; collectively all the drivers covered 20 of the 22 MBON types . MBON-γ3 and MBON-γ3β′1 are listed together , because these MBON cell types are always labeled together in our split-GAL4 lines . Driver lines shown in bold were selected for more detailed experiments . MBONs have been grouped by cell body cluster , neurotransmitter and color-coded as indicated ( see Table 1 ) . The bottom and top of each box represents the first and third quartile , and the horizontal line dividing the box is the median . The whiskers represent the 10th and 90th percentiles . The gray rectangle spanning the horizontal axis indicates the interquartile range of the control . Relative expression levels in individual cell types were estimated by confocal microscopy of brains stained for CsChrimson-mVenus and are shown in gray scale ( See ‘Materials and methods’; Figure 3 and Figure 3—figure supplements 1 , 2 show confocal images ) . Statistical tests are described in methods: * , p < 0 . 05; ** , p < 0 . 01; *** , p < 0 . 001; n = 11–29 for GAL4/CsChrimson; n = 67 for control . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 00510 . 7554/eLife . 04580 . 006Figure 3 . MBONs for attraction and repulsion . ( A–F ) Anatomy of MBONs in six selected drivers . Top: Diagrams of the cell types that show expression with each driver , color-coded by neurotransmitter . The name of the split-GAL4 driver line and the cell type ( s ) in which it expresses are given at the bottom of each panel . The positions of their dendrites in the MB lobes are indicated; the arrows show the projections of their axons . Bottom: Confocal images of a single brain hemisphere for each line; MB compartments occupied by the dendrites of the MBONs in each line are labeled and the arrowheads indicate the sites of their synaptic terminals . See Figure 1 legend for abbreviations . Full confocal stacks of these images are available at www . janelia . org/split-gal4 . ( A′–F′ ) Expression patterns of the same drivers shown in a complete brain and ventral nerve cord: green , expression of CsChrimson-mVenus; magenta , neuropil reference stain ( nc82 antibody ) . Frontal views of maximum intensity projections are shown . See the accompanying manuscript ( Aso et al . , 2014 ) for a detailed description of the morphology of each cell type . ( G ) The preference index ( PI ) for each experimental group ( split-GAL4/CsChrimson ) was compared with split-GAL4/+ and pBDPGAL4U/CsChrimson controls . Bars and error bars indicate mean and standard error of the mean respectively . Asterisk indicates significance: * , p < 0 . 05; ** , p < 0 . 01; *** , p < 0 . 001 . ( H–I ) Time course of the preference index . See legend to Figure 2 for more explanation . Data for six drivers and control are displayed in two panels for clarity , color-coded as indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 00610 . 7554/eLife . 04580 . 007Figure 3—figure supplement 1 . Expression patterns of split-GAL4s with UAS-CsChrimson . ( A–I ) Anatomy of MBONs in additional split-GAL4 drivers . See legend to Figure 3A–F for explanation . The neurotransmitter used by MBON-calyx has not been determined . Full confocal stacks of these images are available at www . janelia . org/split-gal4 . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 00710 . 7554/eLife . 04580 . 008Figure 3—figure supplement 2 . Expression patterns of split-GAL4s with UAS-CsChrimson . ( A–J ) Anatomy of MBONs in additional split-GAL4 drivers . See legend to Figure 3A–F for explanation . Full confocal stacks of these images are available at www . janelia . org/split-gal4 . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 008 Although activation of individual cell types ( see Figure 3D , E ) can result in robust phenotypes , some of the strongest effects were observed with drivers that express in combinations of MBONs . For example , MB011B and MB052B , which drive expression in groups of three or five cell types that use the same transmitter ( the glutamatergic M4/M6 and cholinergic V2 clusters , respectively; Table 1 ) , caused strong responses , while activation with drivers for smaller subsets or individual cell types within these groups had much reduced effects ( Figure 2C ) . Although activation of the single cell types MBON-γ1pedc>α/β or MBON-γ4>γ1γ2 had a significant effect , these neurons send axonal projections to other compartments in the MB lobes , giving them the potential to directly influence the activity of additional MBON cell types . Given that multiple MBONs can independently contribute to behavioral valence as measured by attraction vs repulsion , we sought to determine how conflicting or consonant information from multiple MBONs is integrated to bias behavior . We combined two split-GAL4 drivers for cell types with either similar or opposite effects in the same fly ( Figure 4 , Figure 4—figure supplement 1; ‘Materials and methods’ ) . When combining drivers eliciting similar responses , flies generally showed a stronger response than to either driver alone . Conversely , co-activation of MBON cell types with opposing effects resulted in intermediate responses . Together , these data are consistent with a simple combinatorial model of valence integration . 10 . 7554/eLife . 04580 . 009Figure 4 . Additive effects of MBONs for attraction and repulsion . PIs of individual drivers and combinations of drivers are shown when tested by the protocol shown in Figure 2A . The expression patterns of the combination drivers are shown in Figure 4—figure supplement 1 . Bar graphs are color coded by the transmitter of the MBONs . To facilitate comparisons , data for some genotypes are shown in more than one panel . Asterisk indicates significance: * , p < 0 . 05; ** , p < 0 . 01; *** , p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 00910 . 7554/eLife . 04580 . 010Figure 4—figure supplement 1 . Expression patterns of split-GAL4 combinations . ( A–I ) Expression patterns of combined split-GAL4 drivers . See legend to Figure 3A–F for explanation . Combination drivers carry four hemi-drivers , the two found in each of the two indicated split-GAL4 lines . In the cases shown , all the MBON cell types seen in each of the parent split-GAL4 lines are seen in the combination lines . However , the combination lines can also form dimers of hemi-drivers not found in either parent and additional off-target cells are observed in most combination lines . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 010 How do MBONs bias behavior to cause approach or avoidance ? A conditioned response to a learned odor is unlikely to be achieved by eliciting a predetermined motor pattern . To carry out appropriate changes in speed and direction , a fly needs to evaluate both the valence of the learned odor and its own trajectory relative to the location of the odor source . To assess which aspects of locomotion are modified by MBON activation to generate a bias between approach and avoidance , we analyzed the behavior of flies in a 10 mm-wide graded lighting choice zone centered on the border between dark and illuminated quadrants . We tracked the trajectories of individual flies ( Figure 5A–C ) and calculated the fraction of flies crossing the choice zone from the ‘light-off’ to the ‘light-on’ area ( and vice versa ) as well as the fraction of flies changing direction in the choice zone thus returning to the area they came from . 10 . 7554/eLife . 04580 . 011Figure 5 . MBONs bias choice at the border . ( A ) Pseudo-colored light intensity in the behavioral arena . The choice zone was defined as ± 5 mm from the light ON/OFF border ( green box ) . ( B ) Gradient of light intensity across the light ON/OFF border axis shown . ( C ) An example of MB434B + MB011B/UAS-CsChrimson fly that entered the choice zone from the light-off side at time = 0 and then stopped and quickly turned around to exit the choice zone , going back into the light-off side . Eleven images taken at 0 . 1 s intervals have been superimposed . Entry angle to the choice zone was defined as diagramed ( top left ) . Speed and angular speed was calculated based on the difference in the position of a fly in successive frames of 30 frames per second video recordings . ( D ) The fraction of trajectories entering the choice zone from the dark side and then exiting to the illuminated side is plotted for the indicated drivers in combination with CsChrimson ( top ) . The control genotype was the empty driver , pBDPGAL4U , in attP2 in combination with 20xUAS-CsChrimson-mVenus in attP18 . Only flies that entered the choice zone at an entry angle of between 45 and 135° ( facing to the light ON/OFF border ) and had moved more than 5 mm in the 1 s prior to entering the choice zone were analyzed . The error bars show the 95% confidence interval . Between 79 and 410 trajectories were analyzed per genotype . Compared to the control , MB434B + MB011B , MB434B and MB011B showed a significantly lower fraction of trajectories that exit to the light-on side ( more avoidance of light ) , whereas significantly higher fraction of trajectories of MB112C + MB077B flies exit to the light-on side ( more attraction to the light ) ; multiple comparisons with the Dunn-Sidak correction: *** , p < 0 . 001 . Similarly , the fraction of trajectories entering the choice zone from the illuminated side and then changing direction so as to also exit to the illuminated side is plotted for the indicated drivers in combination with CsChrimson ( bottom ) . Between 43 and 280 trajectories were analyzed per genotype . ( E ) Representative trajectories are shown for the indicated genotypes . The trajectories are color-coded to indicate the position of the fly in the trajectory as a function of time after entering the choice zone . The triangle shows the position of the fly at time = 0 , when flies entered into the choice zone ( indicated by the white line at −5 mm from the light ON/OFF border ) . The gray scale background in the panels and pseudo-color scale on the right indicate the intensity of CsChrimson activating light . ( F ) Preference index to the CsChrimson-activating light ( Figure 4 ) was plotted against the fraction of trajectories that exit from the choice zone to the illuminated side irrespective of side of entry: [ ( number of trajectories enter to the choice zone from dark side and then exit to illuminated side ) + ( number of trajectories enter to the choice zone from illuminated side and then exit to illuminated side ) ] divided by total number of trajectories that entered into the choice zone . They were highly correlated ( Spearman's rank-order correlation: Pearson r = 0 . 91; R square = 0 . 83; p < 0 . 001 ) . Genotypes are the same as in panel D and are shown with same color code . ( G ) Preference index to the CsChrimson light ( Figure 4 ) was plotted against the mean walking speed change in the illuminated quadrants compared to dark quadrants ( see Figure 5—figure supplement 1 ) . There was no significant correlation ( Spearman's rank-order correlation: Pearson r = 0 . 13; R square = 0 . 01; p = 0 . 72 ) . Genotypes are the same as in panel D and are shown with same color code . ( H ) Preference index to the CsChrimson light ( Figure 4 ) was plotted against mean angular speed in the illuminated quadrants compared to dark quadrants ( see Figure 5—figure supplement 1 ) . There was no significant correlation ( Spearman's rank-order correlation: Pearson r = 0 . 12; R square = 0 . 001; p = 0 . 75 ) . Genotypes are the same as in panel D and are shown with same color code . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 01110 . 7554/eLife . 04580 . 012Figure 5—figure supplement 1 . MBON activation has only small effects on walking and angular speeds . Walking speed ( A ) and angular speed ( B ) was calculated for individual trajectories , then averaged separately for all flies on dark and all flies on illuminated quadrants during the two 30 s light-on segments of one video ( see Figure 2A ) . Approximately 20 flies were analyzed in each video . Bars and error bars show mean and standard error of the mean for videos ( n = 11-43 for experimental groups; n = 66 for the pBDPGAL4U empty driver control ) . For comparison to the MBONs , values obtained with GAL4 drivers for other cell-types are shown: NP225 , broad expression in antennal lobe projection neurons ( Tanaka et al . , 2004 ) ; Gr66a-GAL4 , bitter taste receptor neurons ( Dunipace et al . , 2001 ) ; Ir76b-GAL4 , co-receptor for ionotorpic receptors ( Benton et al . , 2009; Silbering et al . , 2011 ) ; Gr63a-GAL4 , CO2 receptor neurons ( Dunipace et al . , 2001 ) ; Or13a-GAL4 and Or49a-GAL4 , olfactory receptor neurons ( Fishilevich and Vosshall , 2005 ) . Asterisk indicates significance ( Kruskal–Wallis one-way analysis of variance followed by Dunns post-test ) for comparison between light-off ( left bar ) and light-on ( right bars ) for the same genotype: * , p < 0 . 05; *** , p < 0 . 001 . We found a weak correlation ( Spearman's rank-order correlation: Pearson r = −0 . 60; R square = 0 . 36; p = 0 . 06 ) between absolute mean walking speed in illuminated quadrants and preference index to red light ( Figure 4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 012 Most control flies entering the choice zone from either the light-off or light-on side continued moving forward , crossing into the other side ( Figure 5D , E; Video 1 ) . Flies expressing CsChrimson in either GABAergic ( MB112C or MB083 ) or cholinergic ( MB077B or MB052B ) MBONs behaved similarly to control flies when entering the choice zone from the light-off area . When they entered from the light-on side , these flies showed a slight tendency to turn around in the choice zone ( Figure 5D ) consistent with their preference for illuminated areas ( Figure 2C ) . Flies expressing CsChrimson in a combination of GABAergic and cholinergic MBONs ( MB112C plus MB077B ) , which displayed the strongest preference for lighted areas ( Figure 4 ) , also showed the highest rates of exiting to the light-on side when entering the choice zone from the light-off area ( Figure 5D ) . On the other hand , flies expressing CsChrimson in glutamatergic MBONs ( MB434B , MB011B or a combination of them ) frequently turned around in the choice zone when entering from the light-off side while crossing into the light-off area when entering the choice zone from the illuminated side ( Figure 5D , E; Video 2 and Video 3 ) , behaviors that are consistent with these flies' avoidance of illuminated areas ( Figure 2C ) . 10 . 7554/eLife . 04580 . 013Video 1 . Choice behaviors of control flies . A representative 12 trajectories for the control genotype ( pBDPGAL4U/CsChrimson ) showing flies that entered the choice zone from the light-off side at the 1-s time point and their subsequent behavior for 3 s . See legend of Figure 5E for more explanation . Time is shown in the upper right corner . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 01310 . 7554/eLife . 04580 . 014Video 2 . Choice behaviors of MB434B/CsChrimson flies . A representative 12 trajectories for MB434B/CsChrimson flies showing flies that entered the choice zone from the light-off side at the 1-s time point and their subsequent behavior for 3 s . When two glutamatergic MBONs ( MBON-γ4>γ1γ2 and MBON-β1>α ) were activated using MB434B/CsChrimson , flies tended to avoid entering light-on quadrants . The motor patterns they used for avoiding the light-on quadrants were not stereotyped , as illustrated by the randomly selected examples shown in this video . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 01410 . 7554/eLife . 04580 . 015Video 3 . Choice behaviors of MB434B + MB011B/CsChrimson flies . A representative 12 trajectories for MB434B + MB011B/CsChrimson flies showing flies that entered the choice zone from the light-off side at the 1-s time point and their subsequent behavior for 3 s . When five glutamatergic MBONs ( MBON-γ4>γ1γ2 , MBON-β1>α , MBON-γ5β′2a , MBON-β′2mp and MBON-β′2mp_bilateral ) were activated using MB434B + MB011B/CsChrimson , the avoidance response is observed in a slightly higher fraction of flies than in Video 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 015 We next calculated the fraction of flies exiting the choice zone into the light-on side irrespective of its entry direction and found this ‘choice probability’ to be highly correlated with preference index ( Figure 5F ) . In contrast , we found little correlation between preference index and either mean walking speed or mean angular velocity , an indicator of turning probability , of flies in illuminated quadrants ( Figure 5G , H ) . Flies are able to adjust these parameters to execute avoidance behaviors in other contexts; we found that activation of some chemosensory neurons and projection neuron combinations that repelled flies significantly altered both walking and angular speed ( Figure 5—figure supplement 1; Video 4 and Video 5 ) . Our results indicate that MBON activity biases the direction that flies turn in the choice zone , thereby biasing the direction in which they exit that zone . We did not observe a stereotyped turning behavior in the choice zones; more specifically , the time between entering the choice zone and making a turn as well as the precise direction of the turn varied ( Figure 5E; Videos 2 and 3 ) . Moreover , flies displayed apparently normal behavior within the uniform illumination of lighted quadrants ( Video 2 and 3 ) , showing no apparent increase in their speed and turning rates . These observations support the view that MBONs represent valence , abstract information that serves to bias—rather than direct—specific motor patterns . 10 . 7554/eLife . 04580 . 016Video 4 . Choice behaviors of Gr66a-GAL4/CsChrimson flies . A representative 12 trajectories for Gr66a-GAL4/CsChrimson flies showing flies that entered the choice zone from the light-off side at the 1-s time point and their subsequent behavior for 3 s . When the Gr66a sensory neurons for bitter taste were activated using Gr66a-GAL4/CsChrimson , flies increased their velocity when they entered the illuminated area , which facilitated their escape from light-on quadrants ( see Figure 2B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 01610 . 7554/eLife . 04580 . 017Video 5 . Choice behaviors of NP225-GAL4/CsChrimson flies . A representative 12 trajectories for NP225-GAL4/CsChrimson flies showing flies that entered the choice zone from the light-off side at the 1-s time point and their subsequent behavior for 3 s . When this broad set of antennal lobe projection neurons was activated using NP225-GAL4/CsChrimson , flies showed a stereotyped backward walking behavior when they approached the illuminated area; a response was observed at lower light intensities than with the other driver lines . Flies that were already in the illuminated quadrants when the light turned on showed continuous rotation that typically lasted for entire light-on period ( 30 s ) and often extended a few seconds after the red light was turned off ( not shown ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 017 We asked which MBONs were required for aversive and appetitive odor memory , using a well-established discriminative olfactory learning paradigm ( Figure 6A ) ( Tempel et al . , 1983; Tully and Quinn , 1985; Schwaerzel et al . , 2003; Gerber et al . , 2004; Davis , 2005 ) . In this paradigm , flies are exposed to an odor together with an unconditioned stimulus ( US ) of either an electric shock punishment or a sugar reward , and then to another odor without the US . The ‘trained’ flies are tested at a later time to determine if they exhibit a differential response to the two odors , which is taken as indication of memory formation . 10 . 7554/eLife . 04580 . 018Figure 6 . Requirement of MBONs for 2-hr aversive odor memory . ( A ) Schematic of the T-maze apparatus . In one group of flies , 4-methylcyclohexanol ( MCH ) was paired with electric shocks for 60 s . After a 60 s pause , 3-octanol ( OCT ) was delivered without electric shock . For the reciprocal group , OCT was paired with electric shock , and MCH was delivered without shock . 2 hr after training , flies were given a choice between the two odors in a T-maze and their distribution was used to calculate a performance index ( PI ) . The PI corresponds to the mean of the [ ( number of flies in the OCT tube minus number of flies in the MCH tube ) /total number of flies when OCT was paired with electric shock] and [ ( number of flies in the MCH tube minus number of flies in the OCT tube ) /total number of flies when MCH was paired with electric shock] . The set of 23 driver lines was first screened using pJFRC100-20XUAS-TTS-Shibire-ts1-p10 ( VK00005 ) as the Shits effector; lines that showed a strongly decreased PI in this initial screen were retested with the UAS-Shi x1 effector . ( B ) Results of secondary screening of MBONs for 2-hr aversive memory using UAS-Shi x1 . MBONs have been grouped by neurotransmitter and color-coded . The bottom and top of each box represents the first and third quartile , and the horizontal line dividing the box is the median . The whiskers represent the 10th and 90th percentiles . The gray rectangle spanning the horizontal axis indicates the interquartile range of the control . Statistical tests are described in methods: ** , p < 0 . 01 . ( C ) 2-hr aversive odor memory at the restrictive temperature . In addition to confirming the result obtained with MB112C in the screening assays , a second driver line for MBON-γ1pedc>α/β , ( MB085C ) was tested and found to have a similar impairment . Blocking dopaminergic input to the MB compartments where the dendrites of MBON-γ1pedc>α/β reside , using MB438B , also impairs 2-hr aversive odor memory . UAS-Shi x1 was used as the effector . See Figure 6—figure supplement 1 for the expression patterns of the MB112C , MB085C and MB438B drivers . Bars and error bars show mean and standard error of the mean ( SEM ) . Statistical tests are described in methods: * , p < 0 . 05; *** , p < 0 . 001 . ( D ) 2-hr aversive odor memory in MB112C/Shi and MB438B/Shi flies is not impaired at the permissive temperature . UAS-Shi x1 was used as the effector . ( E ) Avoidance of MCH and OCT in untrained MB112C/Shi and MB438B/Shi flies is not impaired at the restrictive temperature , indicating that these genotypes can detect and respond to the odors . UAS-Shi x1 was used as the effector . ( F ) Shock avoidance in untrained MB112C/Shi and MB438B/Shi flies is not impaired at the restrictive temperature . UAS-Shi x1 was used as the effector . ( G ) Activation of the PPL1-γ1pedc dopaminergic neurons can substitute for electric shock as the unconditioned stimulus ( US ) . Flies trained using thermoactivation of MB438B/dTrpA1 as the US showed robust 3 min aversive memory . ( H ) Diagram of a circuit module for aversive odor memory . MBON-γ1pedc>α/β is shown in blue . PPL1-γ1pedc conveys the US of electric shock ( pink arrow ) . ( I ) Morphology of MBON-γ1pedc>α/β . A frontal view of maximum intensity projection of an image of a single cell generated by stochastic labeling using the multicolor flip-out technique ( MCFO; Nern et al . , in prep . ) is shown . MBON-γ1pedc>α/β has dendrites in γ1 and the core of pedunclulus ( pedc ) , where α/β KCs project . Its axon bilaterally innervates the α/β lobes and the pedc , and neighboring neuropils ( SMP , SIP and CRE ) ; see Figure 6—figure supplement 1 for more anatomical details . The orange dashed line indicates the plane of the cross section shown in ( K ) of the γ1 and the pedc . ( J ) Morphology of PPL1-γ1pedc . A frontal view of maximum intensity projection of an image of a single cell generated by MCFO is shown . PPL1-γ1pedc extends dendrites unilaterally in the SMP , SIP and CRE , and bilaterally innervates the γ1 and the pedc . The orange dashed line indicates the plane of the cross section of γ1 and the pedc shown in ( K ) . ( K ) Double labeling of MBON-γ1pedc>α/β ( green; R12G04-LexA , pJFRC216-13XLexAop2-IVS-myr::smGFP-V5 [su ( Hw ) attP8] and PPL1-γ1pedc ( magenta; MB320C , pJFRC200-10XUAS-IVS-myr::smGFP-HA ( attP18 ) ) . A coronal section is shown . The dendrites of MBON-γ1pedc>α/β and the terminals of PPL1-γ1pedc precisely overlap in γ1 and the pedc . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 01810 . 7554/eLife . 04580 . 019Figure 6—figure supplement 1 . Anatomy of MBON-γ1pedc>α/β and PPL1-γ1pedc , an MBON-DAN pair essential for aversive memory . ( A ) Expression patterns of MB112C and MB085C ( MBON-γ1pedc>α/β ) and MB438B ( PPL1-γ1pedc; MB-MP1 ) as assessed with pJFRC2-10XUAS-IVS-mCD8::GFP ( VK00005 ) . The scale bar is 20 µm . Full confocal stacks of these images are available at www . janelia . org/split-gal4 . ( B ) Confocal image of a single brain hemisphere showing the expression pattern of MB438B . ( C ) A cell body of MBON-γ1pedc>α/β ( green; pJFRC225-5XUAS-IVS-myr::smGFP-FLAG ( VK00005 ) ) is immunoreactive to antibody to GABA ( magenta ) . This result confirms the conclusion that this cell is GABAergic , as reported in the accompanying manuscript ( Aso et al . , submitted ) , based on anti-GAD1 staining . ( D ) Morphology of MBON-γ1pedc>α/β . ( E ) Terminals of MBON-γ1pedc>α/β in the α/β lobes and the core of distal pedunculus ( green; pJFRC67-3XUAS-Syt-smHA [su ( Hw ) attP1] ) . Mushroom body lobes were visualized with nc82 ( magenta ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 019 In the first set of experiments we assayed memory 2 hr after training , using a set of 23 split-GAL4 lines to transiently block neuronal activity in different subsets of MBON cell types . Two memory processes that are thought to rely on different molecular and circuit mechanisms , anesthesia resistant memory and anesthesia sensitive memory , contribute to memory at this retention time ( Dudai et al . , 1976; Folkers et al . , 1993; Isabel et al . , 2004; Krashes and Waddell , 2008; Aso et al . , 2010; Pitman et al . , 2011; Knapek et al . , 2011 ) . We blocked neuronal function throughout the training , retention and test periods . Thus we expect to detect impairments in any phase of memory processing , including formation , consolidation and retrieval . We first screened the lines using a strong Shibirets1 effector ( pJFRC100-20XUAS-TTS-Shibire-ts1-p10 in VK00005 ) . Because some lines had phenotypes at the permissive temperature , presumably due to the effector's high level of expression , we retested the nine lines that showed a reduction in memory performance with a weaker Shibirets1 effector ( UAS-Shi x1; see Figure 6 legend and ‘Materials and methods’ ) . With these parameters , only one line , MB112C , also showed significant memory impairment ( Figure 6B ) . The experimental MB112C flies showed significantly lower memory performance than genetic control groups at the restrictive temperature ( Figure 6C ) , but not at the permissive temperature ( Figure 6D ) . These flies displayed normal shock and odor avoidance at the restrictive temperature ( Figure 6E , F ) , indicating that the observed memory impairment was not due to a defect in sensory or motor pathways . MB112C drives expression in MBON-γ1pedc>α/β ( Figure 6—figure supplement 1A ) . We confirmed the requirement for this cell type using a second driver , MB085C , not included in the original 23 lines screened ( Figure 6C; Figure 6—figure supplement 1A ) . One of the PPL1 cluster dopaminergic neurons , PPL1-γ1pedc ( also known as MB-MP1 ) , innervates the same MB compartments as MBON-γ1pedc>α/β . Blocking PPL1-γ1pedc activity , using the MB438B split-GAL4 driver , also impaired aversive memory ( Figure 6C–F ) . Conversely , activation of PPL1-γ1pedc with the temperature gated cation channel dTrpA1 ( Hamada et al . , 2008 ) substituted for electric shock as the unconditioned stimulus ( US ) , inducing robust aversive odor memory ( Figure 6G ) . This confirms a conclusion reached using less specific enhancer trap GAL4 drivers ( Aso et al . , 2010 , 2012 ) . Consistent with these observations , restoring expression of the D1-like dopamine receptor specifically in the γ Kenyon cells has been shown to rescue the aversive odor memory defect of a receptor mutant ( Qin et al . , 2012 ) . MBON-γ1pedc>α/β is immunoreactive to GABA ( Figure 6—figure supplement 1C ) and is one of only three MBON cell types with axon terminals within the MB lobes ( Figure 6H; Figure 6—figure supplement 1D , E ) ( Aso et al . , 2014 ) . We visualized the single-cell morphologies of MBON-γ1pedc>α/β ( Figure 6I ) and PPL1-γ1pedc ( Figure 6J ) and using two-color labeling confirmed that the axon terminals of the DAN ( one or two cells per brain hemisphere ) precisely overlap with the dendrites of the MBON ( a single cell per brain hemisphere ) in the MB pedunculus and γ1 compartment ( Figure 6K ) . This establishes two essential components of a circuit for 2-hr aversive odor memory . A larger set of MBONs was involved in 2-hr appetitive olfactory memory ( Figure 7 ) . Inactivation of the neurons represented in 12 out of the 23 split-GAL4 lines tested showed an effect in the initial screening with the strong Shibirets1 effector; when retested with the weaker effector , eight of these lines produced a significant impairment ( Figure 7 , Figure 7—Figure supplement 1 ) . All eight lines displayed normal memory at the permissive temperature ( Figure 7C ) and normal attraction to sugar at the restrictive temperature ( Figure 7D ) . Five of these lines ( MB082C , MB093C , MB018B , MB051B and MB077B ) express in subsets of the so-called V3/V4 cluster of cholinergic MBONs ( MBON-γ2α′1 , MBON-α′2 and MBON-α3; Table 1 ) , establishing a role for this group of MBONs in 2-hr appetitive memory . MB310C labels the glutamatergic MBON-α1 . MB011B labels three types of M4/M6 cluster glutamatergic MBONs ( MBON-γ5β′2a , MBON-β′2mp and MBON-β′2mp_bilateral; Table 1 ) , suggesting a role for one or more of these cell types . Two lines that express in subsets of MB011B cell types , MB210B and MB002B , showed a memory defect in the primary screening but failed to reach statistical significance when retested with the weaker effector . Similarly , we observed that activating CsChrimson with MB011B , but not with MB210B or MB002B , produced a significant aversive effect ( see Figure 2C ) . These data are consistent with an additive role of these MBONs on behavior , which is further supported by the anatomical observation that the axon terminals of some of these MBONs converge to the same areas outside the MB ( Figure 7E , F ) ( Aso et al . , 2014 ) . 10 . 7554/eLife . 04580 . 020Figure 7 . Requirement of MBONs for 2-hr appetitive odor memory . ( A ) Flies were trained and tested in a similar way as in Figure 6A , except that flies were starved for 28–40 hr prior to experiments and trained with a reward consisting of a tube covered with sugar absorbed filter paper . ( B ) Results of secondary screening of MBONs for 2-hr appetitive memory using UAS-Shi x1 . MBONs have been grouped by neurotransmitter and color-coded . The bottom and top of each box represents the first and third quartile , and the horizontal line dividing the box is the median . The whiskers represent the 10th and 90th percentiles . The gray rectangle spanning the horizontal axis indicates the interquartile range of the control . Statistical tests are described in methods: * , p < 0 . 05; ** , p < 0 . 01; *** , p < 0 . 001 . The relative expression levels produced by the split-GAL4 driver lines in each cell type ( indicated by the gray scale ) are based on imaging with pJFRC2-10XUAS-IVS-mCD8::GFP ( VK00005 ) . Images of the expression patterns of selected drivers are shown in Figure 7—figure supplement 1 . MB082C and MB093C label the cholinergic MBON-α3; however , these lines also show weak labeling in MBON-α′2 . As blocking MBON-α′2 alone also gave a significant phenotype ( MB018B/Shi ) , we cannot rule out the formal possibility that the phenotype observed with MB082C and MB093C results from blocking MBON-α′2 . MB050B labels MBON-α2 and MBON-α′1; however , because MB052B , which labels the same cell types in addition to others , did not give a phenotype , the requirement for these cell types is not resolved . ( C ) 2 hr appetitive odor memory at permissive temperature . ( D ) Sugar attraction in untrained flies at restrictive temperature . ( E ) Rendering of MBONs with outline of MB lobes and brain . MBONs grouped by square brackets represent cases where the available set of driver lines do not allow assigning an effect to a single cell type , but only to that set of MBONs . MBONs in parentheses represent cases where the data implicating them are only suggestive . ( F ) Diagram of MBONs for appetitive odor memory . MBONs are shown in lighter colors when data implicating them are only suggestive . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 02010 . 7554/eLife . 04580 . 021Figure 7—figure supplement 1 . Expression patterns of split-GAL4s that caused appetitive odor memory phenotypes . ( A–G ) Expression patterns of the indicated split-GAL4 driver lines as assessed with pJFRC2-10XUAS-IVS-mCD8::GFP ( VK00005 ) . Full confocal stacks of these images are available at www . janelia . org/split-gal4 . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 021 The MB has been proposed to play diverse roles in visual behaviors including context generalization and sensory preconditioning between olfactory and visual cues ( Liu et al . , 1999; Zhang et al . , 2013 , 2007; Brembs , 2009; van Swinderen et al . , 2009 ) . Using visual learning assays in which flies are trained to associate a color ( blue or green ) with either an electric shock punishment or a sugar reward ( Figure 8A and Figure 9A ) , Vogt et al . , 2014 demonstrated that γ-lobe KCs are required for immediate visual associative memory and that activation of specific subsets of PPL1 and PAM cluster DANs can substitute as the US for electric shock or sugar , respectively . Here we use the same assays to ask which MBONs are required for visual memory . We used the strong Shits effector ( pJFRC100-20XUAS-TTS-Shibire-ts1-p10 in VK00005 ) to silence MBONs ( Materials and methods ) . 10 . 7554/eLife . 04580 . 022Figure 8 . Requirement of MBONs for aversive visual memory . ( A ) Diagram of the training protocol used for aversive visual conditioning . Groups of 30–40 flies were trained in a circular arena with green or blue LED light and electric shock in a reciprocal manner , as in olfactory conditioning . Training was repeated four times and memory was tested immediately after the last training session . The PI corresponds to the mean of the [ ( number of flies in the blue quadrant minus number of flies in green quadrant ) /total number of flies when blue was paired with electric shock] and [ ( number of flies in the green quadrant minus number of flies in the blue tube ) /total number of flies when green was paired with electric shock]; See ‘Materials and methods’ for more detail . ( B ) Results for the requirement of MBONs in aversive visual conditioning . pJFRC100-20XUAS-TTS-Shibire-ts1-p10 ( VK00005 ) was used to block synaptic transmission . Flies were trained and tested at the restrictive temperature . MBONs have been grouped by neurotransmitter and color-coded . The bottom and top of each box represents the first and third quartile , and the horizontal line dividing the box is the median . The whiskers represent the 10th and 90th percentiles . The gray rectangle spanning the horizontal axis indicates the interquartile range of the control . Statistical tests are described in methods: * , p < 0 . 05; ** , p < 0 . 01 . The two lines marked with gray asterisks failed one of the control assays: MB434B/Shi did not differ significantly from MB434B/+ and MB298B had altered color preference . One line , MB112C , with significantly impaired memory passed all control assays ( see panel C ) . ( C ) Blocking MBON-γ1pedc>α/β with either of two split-GAL4 lines ( MB262B and MB112C ) caused significant visual memory impairment compared to GAL4/+ and +/Shi controls . Blocking the output of PPL1-γ1pedc dopaminergic neurons ( MB438B ) also resulted in a strong defect . Split-GAL4/Shi flies were not impaired at the permissive temperature compared to +/Shi . Bars and error bars show mean and standard error of the mean . ( D ) Electric shock avoidance in untrained flies was not impaired at restrictive temperature compared to +/Shi . Bars and error bars show mean and standard error of the mean . ( E ) Diagram of the MB circuit for aversive visual memory . The GABAergic MBON-γ1pedc>α/β is required for conditioned color preference . The dopaminergic PPL1- γ1pedc neurons terminate in the same compartments where the dendrites of MBON-γ1pedc>α/β are found; they likely convey the electric shock punishment signal . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 02210 . 7554/eLife . 04580 . 023Figure 8—figure supplement 1 . Expression patterns of split-GAL4s that caused aversive visual memory phenotypes . ( A–C ) Expression patterns of the indicated split-GAL4 drivers as assessed with pJFRC2-10XUAS-IVS-mCD8::GFP ( VK00005 ) . Full confocal stacks of these images are available at www . janelia . org/split-gal4 . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 02310 . 7554/eLife . 04580 . 024Figure 9 . Requirement of MBONs for appetitive visual memory . ( A ) Diagram of the training protocol used for appetitive visual conditioning . Starved flies were trained with sugar and the PI calculated in the same manner as in aversive visual conditioning ( see Figure 8A ) . Flies were trained and tested at the restrictive temperature . ( B ) Screening for MBONs required for aversive visual conditioning . pJFRC100-20XUAS-TTS-Shibire-ts1-p10 ( VK00005 ) was used to block synaptic transmission . MBONs have been grouped by neurotransmitter and color-coded . The bottom and top of each box represents the first and third quartile , and the line inside the box is the median . The whiskers represent the 10th and 90th percentiles . The gray rectangle spanning the horizontal axis indicates the interquartile range of the control . Statistical tests are described in methods: * , p < 0 . 05 . Gray asterisks indicate lines that were not significantly different from GAL4/+ controls ( MB083C , MB082C , MB018B and MB242A ) or had altered color preference ( MB298B ) . The relative expression levels produced by the split-GAL4 driver lines in each cell type ( indicated in a gray scale ) are based on imaging with pJFRC2-10XUAS-IVS-mCD8::GFP ( VK00005 ) . Images of the expression patterns of selected drivers are shown in Figure 9—figure supplement 1 . ( C ) Visual appetitive memory was impaired at restrictive temperature by blocking the subsets of glutamate or acetylcholine MBONs labeled by MB434B , MB011B , MB210B , MB542B and MB052B compared to +/Shi and GAL4/+ controls . ( D ) No appetitive visual memory impairment was found at the permissive temperature compared to +/Shi . ( E ) Sugar attraction in untrained flies was not impaired at the restrictive temperature compared to +/Shi . ( F ) Rendering of MBONs implicated in appetitive visual memory . MBONs grouped by square brackets represent cases where the available set of driver lines do not allow assigning an effect to a single cell type , but only to the bracketed set . ( G ) Circuit diagram for appetitive visual memory . Glutamatergic MBONs ( MBON-β′2mp and MBON-γ5β′2a ) mediate visual appetitive memory . PAM neurons innervating these compartments of the MB lobes ( see Figure 1A ) are thought to convey the sucrose reward signal ( Vogt et al . , 2014 ) . MBON-γ4>γ1γ2 feeds forward onto the dendrites of MBON-γ1pedc>α/β , a neuron required for aversive odor and aversive visual memory . Cholinergic V2 cluster neurons also appear to play a role . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 02410 . 7554/eLife . 04580 . 025Figure 9—figure supplement 1 . Expression patterns of split-GAL4s that caused appetitive visual memory phenotypes . ( A–E ) Expression patterns of the indicated split-GAL4 drivers as assessed with pJFRC2-10XUAS-IVS-mCD8::GFP ( VK00005 ) . Full confocal stacks of these images are available at www . janelia . org/split-gal4 . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 025 For aversive visual memory ( Figure 8B ) , we identified one driver line , MB112C , that labels MBON-γ1pedc>α/β . We confirmed the requirement for this cell type with two additional split-GAL4 lines , MB262B ( Figure 8C; Figure 8—figure supplement 1 ) and MB085C ( data not shown ) . This is the same MBON we found to be important for olfactory aversive memory ( Figure 6 ) and so we asked if the same DAN as in olfactory memory was likewise required . Indeed , we found that MB438B , a driver line for PPL1-γ1pedc , showed a significant impairment ( Figure 8C; Figure 8—figure supplement 1 ) . The experimental MB112C , MB262B and MB438B lines showed normal memory at the permissive temperature ( Figure 8C ) and normal shock avoidance at the restrictive temperature ( Figure 8D ) . Our finding that the same MBON and its modulatory DAN were required for both visual and olfactory aversive memory suggests that this local MB circuit is important for aversive memory in general , rather than specifically for a particular sensory modality . For appetitive visual memory , we identified five drivers with significant impairments: MB434B , MB011B , MB210B , MB542B and MB052B ( Figure 9B , C; Figure 9—figure supplement 1 ) . These lines showed normal memory at the restrictive temperature in the absence of the effector ( Figure 9C ) and at the permissive temperature with the effector ( Figure 9D ) , as well as normal sugar attraction at restrictive temperature ( Figure 9E ) . The anatomy of the cell types in these driver lines is illustrated in Figure 9F , G . Unlike in the case of aversive memory , the cell types required for appetitive memory differed somewhat between visual and olfactory modalities . Nevertheless , these modalities both employed the M4/M6 cluster glutamatergic MBONs labeled in MB011B and MB210B ( MBON-γ5β′2a , MBON-β′2mp and MBON-β′2mp_bilateral; Table 1 ) . Moreover , the observation that multiple cholinergic and glutamatergic MBONs play a role in appetitive memory , but not aversive memory , was shared across modalities . Repeated pairing of an odor and an electric shock , with inter-trial rest intervals , results in protein synthesis dependent aversive long-term memory ( LTM ) ( Tully et al . , 1994 ) . The molecular and cellular mechanisms underlying aversive LTM are known to differ from those responsible for memories with shorter retention times ( Yin et al . , 1994; Pascual and Preat , 2001; Dubnau et al . , 2003; Comas et al . , 2004; Isabel et al . , 2004; Yu et al . , 2006; Blum et al . , 2009; Akalal et al . , 2010 , 2011; Trannoy et al . , 2011; Huang et al . , 2012; Placais et al . , 2012 ) . We assessed the requirement of MBONs in the retrieval phase of aversive LTM by training flies at permissive temperature and blocking their activity only during the memory test at 24 hr after training ( Figure 10A ) . Inactivation of MB052B , which broadly labels cholinergic V2 cluster MBONs ( Figure 3F; Table 1 ) , resulted in nearly complete loss of aversive odor LTM recall ( Figure 10C ) . Our results confirm a previously reported requirement for the V2 cluster MBONs in long-term aversive odor memory recall ( Sejourne et al . , 2011 ) . We also assayed five lines that express in subsets of the MBONs found in MB052B . While some of them showed lower memory scores , none showed a statistically-significant memory impairment ( Figure 10B ) . Therefore , our results suggest that V2 cluster MBONs ( Figure 10D ) function as a group in the retrieval of aversive odor LTM . Consistent with this implied combinatorial action of V2 cluster MBONs , activating subsets of these MBONs with CsChrimson resulted in weak attraction that only reached statistical significance when MB052B was used as the driver ( Figure 2C ) . In calcium imaging experiments , V2 cluster MBONs were reported to reduce their response to an odor that had been learned to be aversive ( Sejourne et al . , 2011 ) ; this sign of plasticity is consistent with our observation that activation of these neurons elicited attraction . 10 . 7554/eLife . 04580 . 026Figure 10 . Requirement of long-term aversive odor memory . ( A ) Diagram of the conditioning protocol . Training sessions with electric shocks were repeated five times with 15 min intervals between training , all at the permissive temperature ( 25°C ) . Following training , flies were kept for 24 hr at 18°C , then shifted to the restrictive temperature ( 32°C ) 30 min prior to test . The PI corresponds to the mean of the [ ( number of flies in the OCT tube minus number of flies in the MCH tube ) /total number of flies when OCT was paired with electric shock] and [ ( number of flies in the MCH tube minus number of flies in the OCT tube ) /total number of flies when MCH was paired with electric shock] . ( B ) Results of the primary screening for MBONs required for long-term aversive memory retrieval . pJFRC100-20XUAS-TTS-Shibire-ts1-p10 ( VK00005 ) was used to block synaptic transmission . The bottom and top of each box represents the first and third quartile , and the line inside the box is the median . The bottom and top of each box represents the first and third quartile , and the horizontal line dividing the box is the median . The whiskers represent the 10th and 90th percentiles . The gray rectangle spanning the horizontal axis indicates the interquartile range of the control . Statistical tests are described in methods: **: p < 0 . 01 . The relative expression levels produced by the split-GAL4 driver lines in each cell type ( indicated by the gray scale ) are based on imaging with pJFRC2-10XUAS-IVS-mCD8::GFP ( VK00005 ) . Two lines , MB052B and MB082C showed significant memory impairment . MB082C drives expression in MBON-α3 and to a lesser extent in MBON-α′2 . In subsequent control experiments , MB082C/Shits showed significant impairment compared to MB082C/+ and +/Shits ( data not shown ) . However , we were unable to attribute this aversive LTM phenotype to MBON-α3 , because blocking this cell type by other drivers did not result in a consistent phenotype ( MB093C , Figure 10B; G0239 , ( Placais et al . , 2013 ) ) . The requirement for MBON-α3 for aversive odor LTM has been reported , but it is unclear whether these neurons are required for canonical LTM after spaced training ( Pai et al . , 2013 ) or instead for so-called ‘fasting LTM’ , a memory that mildly fasted flies can form after a single cycle of conditioning ( Hirano and Saitoe , 2013; Placais and Preat , 2013; Placais et al . , 2013 ) . We note that these same two split-GAL4 driver lines for MBON-α3 ( MB082C and MB093C ) also showed discordant results in the ethanol memory assay ( see below ) , suggesting some underlying difference between the two lines , perhaps in off-targeted expressions or genetic background . ( C ) MB052B/Shi showed memory impairment compared to MB052B/+ and +/Shi at the restrictive temperature but not at the permissive temperature . Statistical tests are described in methods: *** , p < 0 . 001 . ( D ) Diagram of the output neurons that are required for long-term aversive odor memory . MB052B labels five cell types of cholinergic MBONs from the vertical lobes , four of them project to the lateral horn . Lines labeling only a subset of these MBONs gave only small effects when blocked . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 026 Flies are able to associate an odor with the intoxicating properties of ethanol ( Figure 11A ) . Briefly , flies are exposed to two consecutive odors; the second of which is paired with a mildly intoxicating concentration of ethanol vapor . Flies are later tested for their odor preference for the paired vs unpaired odor ( Figure 11A ) ( Kaun et al . , 2011 ) . Flies avoid the odor they experienced at the time of intoxication when tested 30 min after training , but show a long-lasting preference for that odor when tested 24 or more hours later ( control data in Figure 11B , C ) ( Kaun et al . , 2011 ) . Blocking KC synaptic output has been shown to interfere with this memory ( Kaun et al . , 2011 ) , indicating a role for the MB . 10 . 7554/eLife . 04580 . 027Figure 11 . Requirement of MBONs for appetitive odor-ethanol intoxication memory . ( A ) Diagram of odor conditioning using ethanol vapor . Conditioned ethanol preference is measured by presenting two odors in sequence , one in the presence of an intoxicating dose of ethanol , three times for 10 min with 50-min breaks between exposures , followed by testing for preference between the two odors in a Y-maze in the absence of ethanol either 30 min or 24 hr after training . Wild-type flies find the ethanol-paired odor to be aversive when tested 30 min after training , but appetitive 24 hr after training . ( B ) Results of assays for the requirement of MBONs in ethanol-induced appetitive odor memory . pJFRC100-20XUAS-TTS-Shibire-ts1-p10 ( VK00005 ) was used to block synaptic transmission during training and test; flies were kept at the permissive temperature for the 24 hr between training and test . MBONs have been grouped by neurotransmitter and color-coded . The bottom and top of each box represents the first and third quartile , and the horizontal line dividing the box is the median . The whiskers represent the 10th and 90th percentiles . The gray rectangle spanning the horizontal axis indicates the interquartile range of the control . Statistical tests are described in methods: * , p < 0 . 05; ** , p < 0 . 01; *** , p < 0 . 001 . The relative expression levels produced by the split-GAL4 driver lines in each cell type ( indicated by the gray scale ) are based on imaging with pJFRC2-10XUAS-IVS-mCD8::GFP ( VK00005 ) . Images of the expression patterns of selected drivers are shown in Figure 11—figure supplement 1 . ( C ) 30-min ethanol-induced odor memory . ( D ) 24-hr ethanol-induced appetitive odor memory compared to +/Shi and GAL4/+ genetic controls . ( E ) Blocking transmission in MBON-α′2 with MB091C resulted in a failure to switch from aversive to appetitive memory for ethanol intoxication , confirming the result seen with MB018B . ( F ) Renderings of the MBONs implicated in 24 hr ethanol odor memory . MBONs grouped by square brackets represent cases where the available set of driver lines do not allow assigning an effect to a single cell type , but only to that set of MBONs . MBONs cell types in parentheses indicate cases where the data implicating them are only suggestive . ( G ) Schematic of potential circuits for appetitive ethanol memory . Cholinergic outputs from the MBON-α′2 and MBON-γ2α′1 , and glutamatergic MBON-γ4>γ1γ2 , MBON-γ5β′2a and MBON-β′2mp are implicated in appetitive ethanol memory . MBONs for which the data shows inconsistency across driver lines ( MBON-α3 and MBON-γ2α′1 ) are shown in a lighter color; additional experiments will be required to resolve the role of these MBONs . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 02710 . 7554/eLife . 04580 . 028Figure 11—figure supplement 1 . Expression patterns of split-GAL4s that caused odor-ethanol intoxication memory phenotypes . ( A–I ) Expression patterns of selected split-GAL4 drivers as assessed with pJFRC2-10XUAS-IVS-mCD8::GFP ( VK00005 ) . Full confocal stacks of these images are available at www . janelia . org/split-gal4 . ( J ) Confocal image of a single brain hemisphere of MB018B for the MBON-α′2 . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 028 To test the role of MBONs in appetitive ethanol memory , we blocked MBON function during training and memory retrieval and assayed for changes in performance . When tested at 24 hr , eight driver lines showed significant memory impairment compared to genetic controls ( Figure 11B , D; Figure 11—figure supplement 1 ) . These eight lines were not significantly different from controls in 30-min aversive memory , although two exhibited a trend towards decreased 30-min memory ( Figure 11C ) ; the ability to form memories at 30 min establishes that flies of these genotypes can sense the odors and learn to associate them with ethanol . Our results indicate that MBON-γ4>γ1γ2 ( MB434B and MB298B ) and MBON-α′2 ( MB018B ) , whose involvement we confirmed with a second line not part of the original screening set ( MB091C; Figure 11E ) , are preferentially required for 24 hr memory . Our data also suggest the involvement of MBON-γ2α′1 , where one driver line ( MB077B ) had a strong effect , while the second driver ( MB051B ) had a weaker effect that did not reach statistical significance ( Figure 11D ) . Our results raise the possibility of the involvement of MBON-α3 . However , the two lines we have for this cell type ( MB082C and MB093C ) gave discordant results and MB082C shows significant expression in MBON-α′2 , a cell type that has a large effect on appetitive ethanol memory . Finally , blocking M4/M6 cluster MBONs from γ5 and β′2 ( MB011B and MB210B ) significantly affected 24 hr memory , but also appeared to decrease 30 min memory . The MBONs required for 24 hr appetitive ethanol memory ( Figure 11F , G ) are partially overlapping with those required for other forms of appetitive memory , again with involvement of multiple glutamatergic and cholinergic MBONs . Silencing MBON-α′2 , a cell type composed of just one cell in each hemisphere ( Figure 11—figure supplement 1F , G ) ( Aso et al . , 2014 ) , resulted in persistence of the aversive memory after 24 hr , when control flies show appetitive memory ( see MB018B in Figure 11D and MB091C in Figure 12E ) . These results suggest that , while memories for the aversive and rewarding effects of ethanol intoxication are formed simultaneously , they are expressed at different times through independent MB circuits . Moreover , maintenance of the aversive memory upon inactivation of MBON-α′2 argues against a passive dissipation of the aversive memory with time , and implies an active process in the conversion of aversive to appetitive memory . 10 . 7554/eLife . 04580 . 029Figure 12 . MBONs bi-directionally regulate sleep . ( A ) Top: Schematic of the experimental apparatus used for assaying sleep . Single flies are placed in a tube and activity is measured by counting the number of times the fly crosses an infrared beam . Bottom: Diagram of the experimental assay . Sleep was measured at 21 . 5°C for 3 days to establish the baseline sleep profile . Flies were then shifted to 28 . 5°C for 2 days to increase activity of the targeted cells by activating the dTrpA1 channel , and then returned to 21 . 5°C after activation to assay recovery . The effect of MBON activation on sleep amount is quantified as percentage change in sleep . Negative and positive values indicate decreased and increased sleep , respectively . ( B ) Box plots of change in sleep induced by dTrpA1 [UAS-dTrpA1 ( attP16 ) ; ( Hamada et al . , 2008 ) ] mediated activation of the neurons targeted by each of the indicated split-GAL4 driver lines . MBONs are grouped by neurotransmitter and color-coded as indicated . The bottom and top of each box represents the first and third quartile , and the horizontal line dividing the box is the median . The whiskers represent the 10th and 90th percentiles . The gray rectangle spanning the horizontal axis indicates the interquartile range of the control . Statistical tests are described in methods: * , p < 0 . 05; ** , p < 0 . 01; *** , p < 0 . 001 . The matrix below the box plots shows the relative levels of expression ( indicated by the gray scale ) in the cell types observed with each driver line using pJFRC200-10XUAS-IVS-myr::smGFP-HA ( attP18 ) . Images of the expression patterns of selected drivers are shown in Figure 12—figure supplement 1 . ( C ) Sleep profiles of four split-GAL4 lines are shown . Each plot shows a 4-day period starting with subjective dawn . Sleep duration ( min/30 min ) on day 3 ( 21 . 5°C; permissive temperature ) , days 4 , 5 ( 28 . 5°C; non-permissive temperature ) and day 6 ( 21 . 5°C; permissive temperature ) are plotted ( colored line , split-GAL4 line in combination with pJFRC124-20XUAS-IVS-dTrpA1 ( attP18 ) ; black line , represents +/dTrpA1; gray line , split-GAL4/+ ) . ( D ) Sleep phenotypes were replicated with pJFRC124-20XUAS-IVS-dTrpA1 ( attP18 ) for the eight drivers shown , but not for MB242A ( not shown ) . Corresponding split-GAL4/+ flies showed normal sleep . ( E , F ) Renderings of MBONs responsible for the decreasing ( E ) or increasing ( F ) sleep . MBONs are color-coded based on their putative transmitter . ( G ) Diagram of MBONs responsible for the sleep regulation . MBONs are color-coded based on their putative transmitter as indicated . The wake promoting glutamatergic MBON-γ5β′2a , MBON-β′2mp and MBON-β′2mp_bilateral converge with the sleep promoting cholinergic MBON-γ2α′1 and GABAergic MBON-γ3 and MBON-γ3β′1 in the SMP and CRE . The wake-promoting glutamatergic MBON-γ4>γ1γ2 terminates in the dendritic region of MBON-γ2α′1 . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 02910 . 7554/eLife . 04580 . 030Figure 12—figure supplement 1 . Expression patterns of split-GAL4s that caused sleep phenotypes . ( A–L ) Expression patterns of selected split-GAL4 lines are shown , as assessed with pJFRC200-10XUAS-IVS-myr::smGFP-HA ( attP18 ) . Full confocal stacks of these images are available at www . janelia . org/split-gal4 . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 030 The involvement of the MB in regulating sleep was first established by demonstrating that blocking synaptic output from KCs can either increase or decrease sleep , depending on the GAL4 driver used ( Joiner et al . , 2006; Pitman et al . , 2006 ) . Sleep in Drosophila is under circadian and homeostatic regulation and defined as sustained periods of inactivity ( 5 min or more ) that coincide with increased sensory thresholds , altered brain activity and stereotyped body posture ( Hendricks et al . , 2000; Shaw et al . , 2000; Hendricks and Sehgal , 2004; Ganguly-Fitzgerald et al . , 2006; Parisky et al . , 2008; Donlea et al . , 2011 ) . Little is known about how the MB's sleep regulating functions are executed . As a first step in elucidating these mechanisms , we asked if activation of specific MBONs changed sleep pattern and amount . Sleep was measured before , during , and after heat-gated activation of MBONs ( Figure 12A ) by expression of the temperature-gated dTrpA1 channel . We identified five MBON drivers that suppressed sleep and seven MBON drivers that increased sleep ( Figure 12B ) . All five sleep-suppressing drivers express in glutamatergic MBONs ( MBON-γ5β′2a , MBON-β′2mp , MBON-β′2a_bilateral and MBON-γ4>γ1γ2 ) , while sleep-promoting drivers express either in GABAergic ( MBON-γ3β′1 and MBON-γ3 ) or cholinergic MBONs ( MBON-γ2α′1 ) ; the neurotransmitter for MBON-calyx has not been determined ( Figure 12B ) . Despite the 48 hr period of dTrpA1 activation , the effects of activation were reversible and the temperature shift had only minor effects on sleep in genetic control groups ( Figure 12C ) . We retested the split-GAL4 drivers that showed a significant effect in the primary screening using a different dTrpA1 effector inserted at another genomic location , allowing us to assess expression more accurately since we had access to a reporter construct inserted at that site ( Figure 12—figure supplement 1 ) ; these assays confirmed the initial results for all but one of the split-GAL4 driver lines ( Figure 12D ) . None of the lines showed significant effects on general locomotion as assessed by video tracking ( see ‘Materials and methods’ ) . We identified distinct MBONs that either decrease ( Figure 12E ) or increase sleep ( Figure 12F ) . The subset of MBONs that promoted sleep was similar to the subset in which CsChrimson activation was attractive; conversely , the MBON subset that promoted wakefulness was similar to the subset in which CsChrimson activation was aversive . The projection patterns of MBONs provide insight into how these bidirectional signals might be integrated in the fly brain ( Figure 12G ) . For example , the axons of the sleep promoting cholinergic MBON-γ2α′1 and wake promoting glutamatergic MBON-β′2mp project their termini to the same location in the brain ( Video 6 ) . These circuit arrangements of sleep- and wake-promoting neurons may facilitate the transition between the sleep and wake behavioral states by providing opposing inputs to shared downstream targets . 10 . 7554/eLife . 04580 . 031Video 6 . Convergence of glutamatergic and cholinergic MBONs . Rendering of two-color labeling for the glutamatergic MBON-β′2mp and weakly labeled MBON-γ5β′2a ( MB074C; green ) and cholinergic MBON-γ2α′1 ( R25D01-LexAp65 in attP40; magenta ) . A small area dorsal to the MB medial lobes ( an area that includes part of the CRE and the SMP ) is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 031
We demonstrated that optogenetic activation of MBONs in untrained flies can induce approach or avoidance . The ability of the MBONs to induce changes in behavior in the absence of odors suggests that MBONs can bias behavior directly . This observation is consistent with a recent study showing that flies are able to associate artificial activation of a random set of KCs—instead of an odor stimulus—with electric shock , and avoid reactivation of the same set of KCs in the absence of odors ( Vasmer et al . , 2014 ) , a result that recapitulates a finding in the potentially analogous piriform cortex of rodents ( Choi et al . , 2011 ) . We found that the sign of the response to MBON activation was highly correlated with neurotransmitter type; all the MBONs whose activation resulted in avoidance were glutamatergic , whereas all the attractive MBONs were cholinergic or GABAergic . By tracking flies as they encounter a border between darkness and CsChrimson-activating light , we showed that activation of an MBON can bias walking direction . Although activation of glutamatergic MBONs repelled flies , the avoidance behaviors were not stereotyped; flies showed a variety of motor patterns when avoiding the red light . This observation implies that MBONs are unlikely to function as command neurons to drive a specific motor pattern , as has been observed , for example , in recently identified descending neurons that induce only stereotyped backward walking ( Bath et al . , 2014; Bidaye et al . , 2014 ) . Rather , we view fly locomotion as a goal-directed system that uses changes in MBON activity as an internal guide for taxis . For example , walking in a direction that increases the relative activity of aversive-encoding MBONs , which would occur as a fly approaches an odorant it had previously learned to associate with punishment ( or when a fly expressing CsChrimson in an avoidance-inducing MBON approaches the CsChrimson activating light as in Figure 5E ) , signals the locomotive system to turn around and walk the other direction . Detailed studies of locomotor circuitry will be required to determine the mechanisms of executing such taxic behaviors and should help elucidate how MBON inputs guide this system . In this view , the MBON population functions as neither a purely motor nor a purely sensory signal . From the motor perspective , as described above , MBONs bias locomotive outcomes rather than dictate a stereotyped low-level motor program . From the sensory perspective , we have shown that the same MBONs can be required for experience-dependent behavioral plasticity irrespective of whether a conditioned stimulus is a color or an odor , and irrespective of the specific identity of the odor . Taken together with the fact that MBONs lie immediately downstream of the sites of memory formation , these observations support our proposal that MBONs convey that a stimulus has a particular value—but not the identity of the stimulus itself . This contrasts with sensory neurons whose activity can also induce approach or avoidance , but which do convey the stimulus per se . In mammals , neural representations of abstract variables such as ‘value’ , ‘risk’ and ‘confidence’ are thought to participate in cognition leading to action selection ( for example , see Kepecs et al . , 2008; Kiani and Shadlen , 2009; Levy et al . , 2010; Yanike and Ferrera , 2014 ) . From the point of view of this framework , the MBON population representing the value of a learned stimulus and informing locomotion might be operationally viewed as a cognitive primitive . Co-activating multiple MBON cell types revealed that the effects of activating different MBONs appear to be additive; that is , activating MBONs with the same sign of action increases the strength of the behavioral response , whereas activating MBONs of opposite sign reduces the behavioral response . Thus , groups of MBONs , rather than individual MBONs , likely act collectively to bias behavioral responses . Consistent with the idea of a distributed MBON population code , all 19 MBON cell types imaged so far show a calcium response to any given odor ( Hige et al . , unpublished , Sejourne et al . , 2011; Pai et al . , 2013; Placais et al . , 2013 ) . If it is the ensemble activity of a large number of MBONs that determines memory-guided behavior , how can local modulation of only one or a few MB compartments by dopamine lead to a strong behavioral response ? Activation of a single DAN such as PPL1-γ1pedc that innervates a highly localized region of the MB can induce robust aversive memory , yet the odor associated with the punishment will activate MBONs from all compartments , including MBONs that can drive approach as well as those that drive avoidance . We propose that , in response to a novel odor stimulus , the activities of MBONs representing opposing valences may initially be ‘balanced’ , so that they do not impose a significant bias . Behavior would then be governed simply by any innate preference a fly might have to that odor , using neuronal circuits not involving the MB . Now suppose an outcome associated with that stimulus is learned . Such learning involves compartment-specific , dopamine-dependent plasticity of the KC-MBON synapses activated by that stimulus . If that occurs , the subsequent ensemble response of the MBONs to that stimulus would no longer be in balance and an attraction to , or avoidance of , that stimulus would result . Consistent with this idea , eliminating MB function by disrupting KCs , which are nearly 10% of neurons in the central brain , had surprisingly minor effects on odor preference ( de Belle and Heisenberg , 1994 , Heisenberg et al . , 1985; McGuire et al . , 2001 ) . Figure 14 shows a conceptual model of how this could be implemented at the level of neuronal circuits . 10 . 7554/eLife . 04580 . 033Figure 14 . A simplified circuit model for the encoding of learned valence by MBONs ensemble and feedforward network . ( A ) In response to an odor , a sparse ensemble of Kenyon cells provides excitatory synaptic input to MBONs ( black arrows; Hige et al . , unpublished ) ( Cassenaer and Laurent , 2007 ) . Glutamatergic ( Glu; green ) , GABAergic ( GABA; blue ) and cholinergic ( Ach; orange ) MBONs all receive KC input; the names of MBONs are based on the lobe compartment where their dendrites arborize . CsChrimson activation of some glutamatergic MBONs can be repulsive , whereas activation of GABA or cholinergic MBONs can be attractive ( color coded as indicated by the scale at bottom right ) ( Figure 2C ) . We often only observed significant behavioral effect with combinations of cell types ( indicated by dashed lines grouping multiple cell types ) . MBON-γ1pedc>α/β , and glutamatergic MBONs , MBON-β1>α and MBON-γ4>γ1γ2 , have synaptic terminals inside the compartments of MB lobes ( Aso et al . , 2014 ) . While the microcircuits within each MB compartment remain to be elucidated , our light level anatomical studies have enumerated the cell types present in each compartment . The α2 and α3 compartments contain the dendrites of cholinergic MBONs and are targeted by both the GABAergic MBON-γ1pedc>α/β and the glutamatergic MBON-β1>α . These MBONs cover only a small fraction of volume in their target compartments ( see Figure 6—figure supplement 1 ) ; since they could contact only a fraction of Kenyon cells , we propose that they target MBONs directly . Here we hypothesize that glutamate is inhibitory to cholinergic and GABAergic MBONs , and GABA is inhibitory to glutamatergic MBONs ( inhibitory connections are indicated by circles ) . The thickness of lines and size of their endings are meant to indicate activity levels . ( B ) PPL1- γ1pedc DANs play a major role to mediate punishment signals to the MB for formation of aversive memory together with minor contribution from other DANs including PPL1-γ2α′1 ( Aso et al . , 2012 ) . If an aversive memory is formed by the simultaneous presentation of an odor and punishment and results in a synaptic depression of KC terminals by dopamine ( represented by red dashed circles ) , the response of the GABAergic MBONs to the CS+ would be depressed ( see text ) . Reduced GABAergic inhibitory input would then increase the CS+ response of glutamatergic MBONs , whereas the CS+ response of cholinergic MBONs would be reduced because of dis-inhibition of the inhibitory glutamatergic MBONs . The end result is enhanced activity of aversion-mediating glutamatergic MBONs together with the reduced activity of attraction-mediating GABAergic and cholinergic MBONs in response to CS+ . The CS- is represented by a different set of KCs whose synaptic connections to the MBONs would not be expected to be modified by training and so the responses of the ensemble of MBONs to the CS- would remain balanced . The change in response of the MBONs to the CS+ , relative to their unchanged response to the CS- , biases choice toward the CS+ ( see diagram in panel D ) . This model is consistent with the essential role of MBON-γ1pedc>α/β in aversive memory . Also , cholinergic MBONs in V2 cluster have been shown to reduce their response to an odor after olfactory conditioning with electric shock ( Séjourné et al . , Nat . Neurosci . , 2011 ) , although we detected their requirement only for long-term memory , but not for 2 hr memory . This model predicts a role for glutamatergic MBONs , but we did not observe significant effect by blocking subsets of glutamtergic MBONs ( Figures 6 and 8 ) . Thus , to test this model , it will be necessary to block broader sets of glutamatergic MBON cell types by using combinations of split-GAL4 drivers . ( C ) In contrast to aversive memory in which one type of DAN ( PPL1-γ1pedc ) plays a major role in memory formation , reward signals are mediated by a distributed set of PAM cluster DANs that innervate the compartments of glutamatergic MBONs ( Yamagata et al . , in press ) ( Burke et al . , 2012; Liu et al . , 2012; Perisse et al . , 2013 ) . If an appetitive memory is formed by synaptic depression of KC terminals in response to dopamine release , the response of the glutamatergic MBONs to the CS + would be depressed . The resultant reduced glutamatergic inhibitory input to the GABAergic and cholinergic MBONs would increase their response to the CS+ . In turn , increased GABAergic input to glutamatergic MBONs may further amplify and stabilize the initial effects of plasticity . The end result would be reduced activity of aversion-mediating glutamatergic MBONs together with the increased activity of attraction-mediating GABAergic and cholinergic MBONs in response to CS+ . This model is consistent with requirement of glutamatergic MBONs: blocking the M4/M6 cluster MBONs ( MBON-γ5β′2a , MBON-β′2mp and MBON-β′2mp_bilateral; Table 1 ) by MB011B resulted in memory impairment for all three appetitive memory assays ( Figures 7 , 9 and 11 ) . Blocking the MBON-γ4>γ1γ2 and MBON-β1>α by MB434B resulted in memory impairment in two of three assays ( Figures 9 and 11 ) . While we did not detect a requirement for the GABAergic MBON-γ1ped>α/β in appetitive memory , previous study have shown that dopamine input to the γ1 and pedc suppresses expression of appetitive memory in fed flies ( Krashes et al . , 2009 ) , indicating some role of MBON-γ1ped>α/β in appetitive memory . Blocking cholinergic MBONs in the V3/V4 cluster ( MBON-γ2α′1 , MBON-α′2 and MBON-α3 ) resulted in memory impairment in appetitive odor memories ( Figures 7 and 11 ) ( Placais et al . , 2013 ) but not in appetitive visual memory ( Figure 9 ) . Some of driver lines for the cholinergic MBONs in the V2 cluster showed impairment of appetitive memory in all three assays ( Figures 7 , 9 and 11 ) , although our data did not allow mapping to the resolution of specific cell types due to inconsistent results obtained using other lines . ( D ) In the matrix shown , the number of circles represents the activity levels of MBONs in response to the CS+ odor ( the odor that is paired with the unconditioned stimulus during conditioning ) and the CS− odor ( a control odor ) . In untrained flies , the activities of MBONs for opposing effects are balanced . Dopamine modulation breaks this balance to bias the choice between CS+ and CS− . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 033 Recent studies of dopamine signaling have implicated distinct sites of memory formation within the MB lobes ( summarized in Figure 1A ) ( Schroll et al . , 2006; Claridge-Chang et al . , 2009; Aso et al . , 2010 , 2012; Burke et al . , 2012; Liu et al . , 2012; Perisse et al . , 2013 ) . Consistent with this large body of work , we found that one type of PPL1 cluster DAN , PPL1-γ1pedc , played a central role in formation of aversive memory in both olfactory ( Figure 6; see also Aso et al . , 2012; Aso et al . , 2010 ) and visual learning ( Figure 8 ) paradigms . This DAN also mediates aversive reinforcement of bitter taste ( Das et al . , 2014 ) . For appetitive memory , PAM cluster DANs that innervate other regions of the MB lobes , in particular the compartments of glutamatergic MBONs , are sufficient to induce appetitive memory ( Yamagata et al . , in press ) ( Burke et al . , 2012; Liu et al . , 2012 ) ( Perisse et al . , 2013 ) . These results strongly suggest that the synaptic plasticity underlying appetitive and aversive memory generally occurs in different compartments of the MB lobes . The sign of preference we observed in response to CsChrimson activation of particular MBONs was , in general , opposite to that of the memory induced by dopaminergic input to the corresponding MB compartments . For example , activation of MBON-γ1pedc>α/β and MBON-γ2α′1 attracted flies ( Figure 2 ) , whereas DAN input to these regions induced aversive memory ( Figure 6G ) ( Aso et al . , 2010 , 2012 ) . Conversely , activation of glutamatergic MBONs repelled flies ( Figure 2 ) , while DAN input to the corresponding regions is known to induce appetitive memory ( Yamagata et al . , in press ) ( Burke et al . , 2012; Liu et al . , 2012 ) ( Perisse et al . , 2013 ) . These results are most easily explained if dopamine modulation led to synaptic depression of the outputs of the KCs representing the CS + stimulus . Consistent with this mechanism , the PE1 MBONs in honeybees ( Okada et al . , 2007 ) as well as the V2 cluster MBONs in Drosophila ( Sejourne et al . , 2011 ) reduce their response to a learned odor and depression of KC-MBON synapses has been shown for octopamine modulation in the locust MB ( Cassenaer and Laurent , 2012 ) . Moreover , long-term synaptic depression is known to occur in the granular cell synapses to Purkinje cells in the vertebrate cerebellum ( Ito et al . , 1982 ) , a local neuronal circuit with many analogies to the MB ( Schurmann , 1974; Laurent , 2002; Farris , 2011 ) . Other mechanisms are also possible and multiple mechanisms are likely to be used . For example , dopamine may modulate terminals of KCs to potentiate release of an inhibitory cotransmitter such as short neuropeptide F , which has been demonstrated to be functional in KCs ( Knapek et al . , 2013 ) and hyperpolarizes cells expressing the sNPF receptor ( Shang et al . , 2013 ) . RNA profiling of MBONs should provide insights into the molecular composition of synapses between KCs and MBONs . It is also noteworthy that the effect of dopamine can be dependent on the activity status of Kenyon cells; activation of PPL1-γ1pedc together with odor presentation induces memory , while its activation without an odor has been reported to erase memory ( Berry et al . , 2012; Placais et al . , 2012 ) . In the vertebrate basal ganglia , dopamine dependent synaptic plasticity important for aversive and appetitive learning is known to result in both synaptic potentiation and synaptic depression ( Shen et al . , 2008 ) . In this study , we looked at the effects of selectively and specifically manipulating the activities of a comprehensive set of MBONs on several behaviors . As a consequence , we gained some insights into the extent to which the relative importance of particular MBONs differed between behaviors ( Figure 13 ) . Most obvious was the segregation between appetitive and aversive behaviors . For example , we found that blocking MBON-γ1pedc>α/β impaired both short-term aversive odor and visual memory , suggesting a general role in aversive memory independent of modality . Conversely , a subset of glutamatergic MBONs was required in all three appetitive memory assays . It still remains to be demonstrated that the outputs of these MBONs are required transiently during memory retrieval . Nevertheless , CsChrimson activation experiments demonstrate that activation of these MBONs can directly and transiently induce attraction and avoidance behaviors . In the cases described above , the DANs and MBONs mediating a particular behavior innervate the same regions of MB lobes . We also found cases where the DANs and MBONs required for a behavior do not innervate the same compartments of the MB lobes . For example , even though several cholinergic MBONs are required for appetitive memory ( Placais et al . , 2013 ) , the compartments with cholinergic MBONs do not receive inputs from reward-mediating PAM cluster DANs , but instead from PPL1 cluster DANs that have been shown to be dispensable for odor-sugar memory ( Schwaerzel et al . , 2003 ) ( Figure 1A ) . What accounts for this mismatch ? Perhaps these cholinergic MBONs' primarily function is in memory consolidation rather than retrieval . But the fact that CsChrimson activation of the cholinergic MBON-γ2α′1 and V2 cluster MBONs resulted in attraction , strongly suggests that at least some of the cholinergic MBONs have a role in directly mediating the conditioned response . Indeed , previous studies found a requirement for cholinergic MBONs ( the V2 cluster and MBON-α3 ) during memory retrieval ( Sejourne et al . , 2011; Pai et al . , 2013; Placais et al . , 2013 ) . One attractive model is that requirement of cholinergic MBONs originates from the transfer of information between disparate regions of the MB lobes through the inter-compartmental MBONs connections within the lobes or by way of connections outside the MB , like those described in the next two sections . The multilayered arrangement of MBONs ( see Figure 17 of the accompanying manuscript ) ( Aso et al . , 2014 ) provides a circuit mechanism that enables local modulation in one compartment to affect the response of MBONs in other compartments . Once local modulation breaks the balance between MBONs , these inter-compartmental connections could amplify the differential level of activity of MBONs for opposing effects ( Figure 14 ) . For example , the avoidance-mediating MBON-γ4>γ1γ2 targets the compartments of attraction-mediating MBON-γ2α′1 and MBON-γ1pedc>α/β ( Figure 14 ) . This network topology might also provide a fly with the ability to modify its sensory associations in response to a changing environment ( see Discussion in the accompanying manuscript ) ( Aso et al . , 2014 ) . Consider the α lobe . Previous studies and our results indicate that circuits in the α lobe play key roles in long-term aversive and appetitive memory ( Figure 10 ) ( Pascual and Preat , 2001; Isabel et al . , 2004; Yu et al . , 2006; Blum et al . , 2009; Sejourne et al . , 2011; Cervantes-Sandoval et al . , 2013; Pai et al . , 2013; Placais et al . , 2013 ) . The α lobe is targeted by MBONs from other compartments and comprises the last layer in the layered output model of the MB ( Figure 1B; see also Figure 17 of the accompanying manuscript ) ( Aso et al . , 2014 ) . The GABAergic MBON-γ1pedc>α/β and the glutamatergic MBON-β1>α both project to α2 and α3 , where their axonal termini lie in close apposition ( see Figure 17 of the accompanying manuscript ) ( Aso et al . , 2014 ) . DAN input to the compartments housing the dendrites of these feedforward MBONs induces aversive and appetitive memory , respectively ( this work ) ( Perisse et al . , 2013; Yamagata et al . , in press ) . As pointed out in ( Aso et al . , 2014 ) , this circuit structure is well-suited to deal with conflicts between long-lasting memory traces and the need to adapt to survive in a dynamic environment where the meaning of a given sensory input may change . To test the proposed role of the layered arrangement of MBONs in resolving conflicts between old memories and new sensory inputs , we will also need behavioral paradigms that , unlike the simple associative learning tasks used in our current study , assess the neuronal requirements for memory extinction and reversal learning . The neuronal circuits that are downstream of the MBONs and that might read the ensemble of MBON activity remain to be discovered . However , the anatomy of the MBONs suggests that , at least in some cases , summation and canceling effects may result from convergence of MBON terminals on common targets ( Aso et al . , 2014 ) . For example , the terminals of the sleep-promoting cholinergic MBON-γ2α′1 overlap with terminals of wake-promoting glutamatergic MBONs ( γ5β′2a , β′2mp and β′2mp bilateral ) in a confined area in CRE and SMP . In addition , some MBONs appear to terminate on the dendrites of DANs innervating other compartments , forming feedback loops . Using these mechanisms , local modulation in a specific compartment could broadly impact the ensemble of MBON activity and how it is interpreted . Testing these and other models for the roles of the MBON network , both within the MB lobes and in the surrounding neuropils , will be facilitated by an EM-level connectome to confirm the synaptic connections we have inferred based on light microscopy . We will also need physiological assays to confirm the sign of synaptic connections and to measure plasticity . For example , we do not know the sign of action of glutamate in the targets of glutamatergic MBONs , as this depends on the receptor expressed by the target cells ( Xia et al . , 2005; Jan and Jan , 1976; Liu and Wilson , 2013 ) . In this regard , we note that previous studies demonstrated a role for NMDA receptors in olfactory memory ( Wu et al . , 2007; Miyashita et al . , 2012 ) . Neurons that are thought to mediate innate response to odors—a subset of projection neurons from the antennal lobes and output neurons from the lateral horn—also project to these same convergence zones ( Figure 15; Figure 15—figure supplement 1 ) . We propose that these convergence zones serve as network nodes where behavioral output is selected in the light of both the innate and learned valences of stimuli . What are the neurons downstream to these convergence zones ? One obvious possibility is neurons that project to the fan-shaped body of the central complex whose dendrites are known to widely arborize in these same areas ( Hanesch et al . , 1989; Young and Armstrong , 2010a; Ito et al . , 2013; Yu et al . , 2013 ) . It would make sense for the MB to provide input to the central complex , a brain region involved in coordinating motor patterns ( Strauss , 2002 ) . Figure 15 provides a diagrammatic summary of these proposed circuits . 10 . 7554/eLife . 04580 . 034Figure 15 . Convergence zone of MBON terminals as network nodes to integrate innate and learned valences . MB lobes are consisted of three groups of compartments based on the putative transmitter of MBONs ( glutamate , GABA and acetylcholine; color-coded as indicated ) , which are interconnected inside the lobes and send converging outputs to the lateral horn , CRE , SMP , SIP and SLP . These regions also receive input from the antennal lobes and the lateral horn , some of which appear to be in close apposition to the terminals of MBONs and likely target common downstream neurons ( see Figure 15—figure supplement 1 ) . Therefore , these convergence zones are well positioned to function as integration sites for selecting adaptive behaviors based on both innate and learned valences . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 03410 . 7554/eLife . 04580 . 035Figure 15—figure supplement 1 . Convergence of olfactory pathways . Double labeling of MBONs , antennal lobe ( AL ) projection neurons , and lateral horn ( LH ) output neurons ( LHONs ) reveals convergence of olfactory pathways in a subset of the SMP , SIP and SLP that surrounds the vertical lobes of the MB , a brain area resembling the ‘ring neuropils’ originally identified in the hymenopteran brain ( Abel et al . , 2001; Tanaka et al . , 2012 ) . These eight examples show sub-stack projections of confocal image data; we estimate that the terminals of the two types of neurons in each case are separated by less than a micron , as judged by examination of the original stacks . Only neurons of interest are shown for clarity; several of these GAL4 and LexA drivers label many other cell types . The reporter constructs pJFRC200-10XUAS-IVS-myr::smGFP-HA ( attP18 ) and pJFRC216-13XLexAop2-IVS-myr::smGFP-V5 ( su ( Hw ) attP8 ) were used . ( A–B ) One of clonal lineages in the antennal lobe , ALlv1 , contains an atypical class of AL projection neurons , the ALlv1-PNs , which do not project to the MB calyx , but directly project to the SMP , SIP and SLP in addition to the lateral horn ( Yu et al . , 2013 ) . ALlv1-PNs that are visualized in R35F02-LexA terminate in a specific region of the SMP ( arrowhead ) , where MBON-γ2α′1 ( A; MB077C ) and MBON-α′2 ( B; MB091C ) also terminate . ( C ) The lateral horn output neurons ( LHONs ) in R47G10-LexA have dendrites in the dorsal region of the LH and axons that project to a confined region in the SIP , where ALlv1-PNs in R59G08-GAL4 also project ( arrowhead ) . ( D ) The GAL4 enhancer trap line NP225 ( Tanaka et al . , 2004 ) expresses in mlPN3 neurons ( Wong et al . , 2002; Tanaka et al . , 2012 ) , another class of atypical AL projection neurons that do not project to the MB calyx , and in several types of projection neurons that terminate in the calyx and the LH ( iACT PNs ) . The axons of mlPN3 neurons ( magenta ) project to the LH and the anterior SMP and SIP , where the ALlv1-PNs in R31G11-LexA also terminate ( arrowheads ) . ( E ) Axons of LHONs in R47G10-LexA and MBON-α′2 ( MB090C ) converge in the SIP ( arrowhead ) . MBON-γ2α′1 is also labeled in MB090C , but its terminals do not contact those of LHONs . The dendrites of MBON-α′2 and MBON-γ2α′1 in the MB lobes are indicated . ( F ) Axons of MBON-β2β′2a ( MB074C ) also converge with those of LHONs in R47G10-LexA in the SIP ( arrowhead ) , whereas the two other MBON cell types in MB074C , MBON-β′2mp and MBON-γ5β′2a , do not project to this region . The dendrites of MBON-β2β′2a and MBON-β′2mp in the MB lobes are indicated . ( G ) LHONs in R13E04-LexA has dendrites in the ventral region of the LH and axons projecting to a confined region ( arrowheads ) spanning the SMP and SIP border , where MBON-α′2 ( MB018B ) also has terminals . ( H ) Axons of LHONs in R13E04-LexA and mlPN3 ( from GAL4 line NP2331; ( Tanaka et al . , 2012 ) ) converge in the SMP and SIP ( arrowheads ) . NP2331 also expresses in olfactory receptor neurons . ( I ) Axons of MBON-α2p3p ( MB062C ) and dendrites of one of PPL1 cluster dopaminergic neurons ( PPL1-FB; R40E08-LexA ) ( Mao and Davis , 2009 ) overlap in the SMP . ( J ) Axons of MBON-α2p3p ( R71D08-GAL4 ) and dendrites of the fan-shaped body neurons ( ExF/2; R84C10-LexA ) ( Young and Armstrong , 2010b ) overlap in the SMP . ExF/2 neurons are also included in 104y and implicated in visual memory ( Liu et al . , 2006 ) . R71D08-GAL4 also expresses in other MBONs and LHONs . MB062C in ( I ) use R71D08 enhancer for intersection . DOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 035 Using inactivation to uncover the roles of specific cell types is inherently limited by redundancy and resiliency within the underlying neural circuits . For example , consider the MBONs from the α/β lobes . The output of the α/β Kenyon cells is known to be required for retrieval of aversive memory ( Isabel et al . , 2004; Krashes et al . , 2007; McGuire et al . , 2001; Cervantes-Sandoval et al . , 2013 ) . Our anatomical and behavioral results show that MBON-γ1pedc>α/β , a cell type we found to be critical for aversive memory , has terminals largely confined inside the α/β lobes , well-positioned to regulate a total 6 types of MBONs from the α/β lobes ( Figure 1B ) . Yet we did not detect a requirement for any of these MBONs in short term aversive memory when tested individually . The ability to detect phenotypes also depends on the strength of the effector; for example , four glutamatergic MBON drivers showed aversive memory impairment in initial screening assays with a strong inhibitor of synaptic function , but we were not able to confirm these effects using a weaker effector ( Figure 6B ) . The failure to see effects when inactivating individual cell types is most easily explained by combinatorial roles and redundancy between MBONs . We note that this high level of resiliency is very reminiscent of observations made with genetic networks , where less than half of gene knockouts of evolutionarily conserved Drosophila genes result in a detectable phenotype ( Ashburner et al . , 1999 ) . Whether or not we detect a requirement for a particular MBON in a particular learning paradigm is likely to depend on which DANs are recruited by the US used in that paradigm as well as the degree of redundancy in the MBON representation of valence . It will be informative to test systematically whether blocking combinations of MBONs , which did not show significant behavioral effect when blocked separately , results in significant memory impairment . It will also be important in future experiments to employ imaging and electrophysiological methods , in which the activities of individual neurons , and the consequences of plasticity , can be observed without being obscured by redundancy . The MBs are implicated in functions beyond processing of associative memory ( Martin et al . , 1998; Liu et al . , 1999; Joiner et al . , 2006; Pitman et al . , 2006; Zhang et al . , 2007; Hong et al . , 2008 ) . MBONs that influence approach to , or avoidance of , a learned stimulus may also have roles in innate preference behaviors for temperature and hunger-dependent CO2 avoidance ( Hong et al . , 2008; Bang et al . , 2011; Bracker et al . , 2013 ) . Moreover , we expect the behavioral repertoire that MBONs govern to go beyond simple approach and avoidance; the MB is known to play a role in experience-dependent regulation of proboscis extension ( Masek and Scott , 2010 ) and courtship ( McBride et al . , 1999 ) as well as regulation of sleep ( Joiner et al . , 2006; Pitman et al . , 2006 ) and post-mating behaviors such as oviposition ( Fleischmann et al . , 2001; Azanchi et al . , 2013 ) . Intriguingly , we found that MBONs whose activation was repulsive promoted wakefulness , whereas MBONs whose activation was attractive promoted sleep; it would make sense for flies to be awake and attentive in an adverse environment . Other internal states , in addition to sleep , are likely to affect the decision to carry out a particular memory-guided behavior; for example , the state of satiety has been shown to regulate memory expression ( Krashes et al . , 2009 ) . The diverse influences of MBONs on behavior can be most easily explained if we assume that the activity of the ensemble of MBON conveys an abstract representation of both valence and internal state . In this view , the ensemble of MBONs may represent internal states along axes such as pleasant-unpleasant or aroused-not aroused . It is upon these axes that primitive forms of emotion are thought to have evolved ( Anderson and Adolphs , 2014 ) .
The construction and characterization of the split-GAL4 lines are described in detail in ( Aso et al . , 2014 ) . The MBON cell types are listed in Table 1 and diagrammed in Figure 1B . The split-GAL4 lines used in this study are described in Table 2; expression patterns , using the most relevant UAS-reporter , are shown in the Figures . pBDPGAL4U ( attP2 ) , an enhancerless GAL4 construct ( Pfeiffer et al . , 2010 ) , was used as a control driver line in behavioral assays . 10 . 7554/eLife . 04580 . 036Table 2 . Split GAL4 drivers for MBONsDOI: http://dx . doi . org/10 . 7554/eLife . 04580 . 036DriverCell typesp65ADZp DNAZpGdbd DNAMB002B ( MBON-γ5β′2a ) , MBON-β′2mpR12C11R14C08MB011BMBON-γ5β′2a , MBON-β′2mp , MBON-β′2mp_bilateralR14C08R15B01MB018BMBON-α′2R20G03R19F09MB027BMBON-α′3m , MBON-α′3apR24H08R53F03MB050BMBON-α2sc , MBON-α′1R65B09R11F03MB051BMBON-γ2α′1 , ( MBON-α′2 ) R70B10R19F09MB052BMBON-α2sc , ( MBON-α2p3p ) , ( MBON-α′3m ) , MBON-α′3ap , MBON-α′1R71D08R11F03MB057BMBON-β′1R80G12R53H03MB077BMBON-γ2α′1R25D01R19F09MB077CMBON-γ2α′1R25D01R19F09MB082CMBON-α3 , MBON-α′2R40B08R23C06MB083CMBON-γ3 , MBON-γ3β′1R52G04R94B10MB085CMBON-γ1pedc>α/βR52H01R52B07MB090CMBON-γ2α′1 , ( MBON-α′2 ) R73H08R19F09MB091CMBON-α′2R73H08R20G04MB093CMBON-α3 , ( MBON-α′2 ) R73H08R40B08MB110CMBON-γ3 , MBON-γ3β′1R20A02R94B10MB112CMBON-γ1pedc>α/βR93D10R13F04MB210BMBON-γ5β′2a , MBON-β′2mp , ( MBON-β′2mp_bilateral ) R15B01R27G01MB242AMBON-calyxR64F07R57C10MB262BMBON-γ1pedc>α/βR52B07R52H01MB298BMBON-γ4>γ1γ2R53C03R24E12MB310CMBON-α1R52G04R17C11MB399BMBON-β2β′2aR21D02R22C12MB433B ( MBON-γ4>γ1γ2 ) , MBON-β1>αR30E08R11C07MB434BMBON-γ4>γ1γ2 , MBON-β1>αR30E08R53C10MB438BPPL1-γ1pedc , ( PPL1-α2α′2 ) R30E11R22B12MB542BMBON-α2p3p , ( MBON-α′3m ) , MBON-α′1R65B09R51D04MB543BMBON-α′3m , ( MBON-α′3ap ) , MBON-α′1R65B09R81E11MB549CMBON-α2sc , MBON-α′3apR71D08R49C12MB622BMBON-calyxR64F07R64F07For each of 31 driver lines used in this study , the MBON cell types in which expression is seen as well as the enhancer fragments used for the activation domain ( p65ADZp ) and DNA binding domain ( ZpGAL4DBD ) hemi-driver constructs are given . All ZpGAL4DBD constructs were inserted in attP2 . The insertion sites of the p65ADZp constructs are indicated by the letter at the end of the driver name as follows: A , su ( Hw ) attP8; B , attP40; C , VK00027 . The cell types shown in brackets indicate expression in those cells was weak , stochastic or was only observed with a subset of UAS-reporters . To combine the expression patterns observed in two split-GAL4 lines , flies were generated by standard genetic crosses that contained the two DNA-binding ( DBD ) halves and the two activation-domain ( AD ) halves found in the parent split-GAL4 lines; the AD and DBD components of all split-GAL4 lines are given in Table 1 of the accompanying manuscript ( Aso et al . , 2014 ) . In general , these lines contained more off-target cells than the parent split-GAL4 lines , due to the interactions of the AD and DBD combinations not present in the parent lines . To identify lines for behavioral experiments , we directly assessed the expression patterns; the majority of combinations produced useful reagents ( Figure 4—figure supplement 1 shows the expression patterns of those combinations used in this work ) . The following constructs were used for activating or silencing neuronal function: 5XUAS-CsChrimson-mVenus ( attP18 ) , 10XUAS-CsChrimson-mVenus ( attP18 ) , 20XUAS-CsChrimson-mVenus ( attP18 ) ( Klapoetke et al . , 2014 ) ; pJFRC124-20XUAS-IVS-dTrpA1 ( attP18 ) ; 10XUAS-dTrpA1 ( attP16 ) ( Hamada et al . , 2008 ) ; pJFRC100-20XUAS-TTS-Shibirets1-p10 ( VK00005 ) ( Pfeiffer et al . , 2012 ) ; UAS-Shi x1 was generated in Thomas Preat's lab by segregating one of the multiple insertions found in the lines described by ( Kitamoto , 2001 ) . To compare the expression levels driven by split-GAL4 drivers in specific cell types , 3–7 days post-eclosion female brains were dissected , antibody-stained , mounted and imaged at 20× under identical conditions ( see the accompanying manuscript for details ) ( Aso et al . , 2014 ) . The relative expression levels in individual cell types are presented as a 0–5 unit gray scale based on the intensity of the signals in the dendrites obtained for each cell type in each split-GAL4 line . The signal intensity depends on the morphology of individual cell types as well as how many cells of the same cell type innervate the same compartment; thus comparing intensities across cell types is less accurate than the comparisons between lines for the same cell type . Since the purpose of obtaining these data was to estimate the expression levels of the UAS-effectors used to manipulate cell function , and expression levels are known to vary with genomic insertion site ( Pfeiffer et al . , 2010 ) , we sought to collect expression data using UAS-indicator lines inserted into the same genomic location as the effectors . We believe that this practice addresses a potential weakness in many prior studies where expression patterns have been determined with an indicator construct inserted in one chromosomal site , while perturbing function with an effector residing at another site . The lack of precise correlation between the expression pattern of the indicator and effector introduces significant uncertainty . The best practice would be to directly measure the expression of the effector protein itself , something we were able to do for the red-shifted channel rhodopsin CsChrimson-mVenus ( Klapoetke et al . , 2014 ) by staining for mVenus . The next best approach is to have the indicator of expression and the effector inserted at the same chromosomal location , which we were able to achieve for all cases except the assays of the 2 hr aversive and appetitive odor memory which used a weaker UAS-Shibire effector based on a P-element insertion . We used the following UAS-indicators for the matrices shown in Figures: Figure 2B , 20XUAS-CsChrimson-mVenus ( attP18 ) reared at 22°C; Figures 7B , 9B , 10B , 11B , pJFRC2-10XUAS-IVS-mCD8::GFP ( VK00005 ) reared at 18°C [similar expression was observed with pJFRC225-5xUAS-IVS-myr::smGFP-FLAG ( VK00005 ) reared at 25°C]; Figure 12B , pJFRC200-10xUAS-IVS-myr::smGFP-HA ( attP18 ) reared at 22°C . Full confocal stacks of these images are available at www . janelia . org/split-gal4 . We used a set of highly specific GAL4 drivers , made using the split-GAL4 intersectional approach ( Aso et al . , 2014 ) . These drivers have much more restricted expression patterns than those previously used , allowing greater certainty in assigning the effects of perturbations to specific cells . Even with these improved GAL4 drivers , there can be significant variation in expression levels between drivers or in different cell types within the expression pattern of a single driver , as well as off-target expression and variations in genetic background . To assign a function to a cell population , we therefore required that the effects of a manipulation be observed using two different GAL4 drivers for that cell population . In cases where we only had one split-GAL4 driver for a cell type , we believe it is only appropriate to interpret an observed phenotype as suggestive , except in cases where we were confirming a previously published result . Finally , we interpret some results as simply raising the possibility of a role for a cell type . For example , where one GAL4 line resulted in a significant effect , but a second line with a very similar expression pattern did not . There were also cases where we saw a consistent tendency in multiple lines , but where none of the individual lines themselves reached statistical significance . Detailed methods for immunohistochemistry and image analysis are described in the accompanying manuscript ( Aso et al . , 2014 ) . For the data in Figure 6—figure supplement 1B , rabbit anti-GABA ( 1:500; A2052 , Sigma-Aldrich , St Louis , MO 63103 ) was used . We asked if using the split-GAL4 lines to activate or inactivate neurons would perturb general innate behaviors , such as locomotion and visual perception that might interfere with our assays of memory , locomotion and sleep . Given the results of these tests , we selected 23 split-GAL4 lines for use in the primary behavioral screening of the MBONs and additional lines to confirm the results of primary screening ( see Table 1 ) . We also performed behavioral assays to verify that the animals carrying the driver line and effector were able to perceive odors , electric shocks and sugar rewards ( see Results ) . Thus , the behavioral phenotypes we observed in these lines are unlikely to result from general defects in innate behaviors . We first screened 27 MBON split-GAL4 driver lines , crossed to a multi-insert UAS-Shibirets1 effector line ( UAS-Shits1 on the third chromosome ) ( Kitamoto , 2001 ) at 34°C . 3–7 days old adult males of each genotype were wet starved for 1–4 hr and then were assayed for 22 parameters of basic locomotion in response to startle , optomotor and phototaxis stimuli using an apparatus ( Fly Behavioral Olympiad , unpublished ) inspired by published assays ( Benzer , 1967; Zhu et al . , 2009 ) . Fourteen split-GAL4 lines showed a difference in one or more parameters from the pBDPGAL4U and these lines were re-screened with pJFRC100-20XUAS-TTS-Shibire-ts1-p10 ( VK00005 ) DL ( Pfeiffer et al . , 2012 ) . Although some lines had phenotypes in some behavioral categories , none of the output lines showed consistent phenotypes across the two Shibire effectors . All of the lines screened were able to appropriately respond to visual stimuli , showed positive phototaxis towards green and UV light , and were able to walk when MBONs were inactivated with Shibire . Only one line , MB549C , showed a significant reduction in walking speed , though subsequent analysis suggests this reduction in speed was due to the genetic background of that line rather than silencing of MBONs . Thus flies can move and orient themselves when MBONs , and any neurons with off-target expression in these split-GAL4 drivers , are inactivated and the small differences from wild-type would not be expected to significantly affect behavior in the assays we performed that used Shibire as an effector . We also assayed the behavior of flies from each split-GAL4 driver line in a high-throughput open-field arena described in ( Kabra et al . , 2013 ) for 15 min during dTrpA1 activation using 10X UAS-dTRPA1 ( attP16 ) ( Hamada et al . , 2008 ) at 30°C . We first tracked the body and wing position of the flies ( Branson et al . , 2009; Kabra et al . , 2013 ) , and then automatically annotated 14 social and locomotor behaviors of flies such as walking , chasing , grooming , etc ( Supplementary file 1 ) ( Kabra et al . , 2013 ) . Although we observed variation of locomotion levels between drivers and their GAL4/+ controls , only one driver , MB052B/dTrpA1 , showed an obvious phenotype . This phenotype of MB052B was limited to male flies and presumably attributed to male specific expression in this line; we therefore did not use males of this driver in experiments involving activation . The choice assay was performed in a 10 cm diameter and 3 mm high circular arena as previously described ( Klapoetke et al . , 2014 ) . Flies expressing CsChrimson were allowed to distribute between two dark quadrants and two quadrants illuminated with 617 nm LEDs ( Red-Orange LUXEON Rebel LED—122 lm; Luxeon Star LEDs , Brantford , Ontario , Canada ) . This wavelength efficiently activates neurons expressing CsChrimson ( Klapoetke et al . , 2014 ) , but was distant enough from the peak absorption spectrum of endogenous rhodopsins that at the light intensity used ( 34 µW/mm2 ) negligible phototaxis of the control genotype was observed . To maintain a constant temperature , the LED board was placed on a heat sink and air ( 150 ml/min ) was exchanged through holes at the center and four corners of the arena in a similar way as in the previously described olfactometer ( Vet et al . , 1983 ) . The four quadrants were separated by 1 mm dividers . The bottom of arena consisted of a 3 mm thick diffuser with an IR absorption film ( YAG , Laser PVC Film; Edmund optics , Barrington , NJ 08007-1380 ) . The intensity of red light decreased from 34 to 3 µW/mm2 over a 10 mm gradient extending from the border between light on and off quadrants , as shown in Figure 5A . Crosses were kept on standard cornmeal food supplemented with retinal ( 0 . 2 mM all-trans-retinal prior to eclosion and then 0 . 4 mM ) at 22°C at 60% relative humidity in the dark . Groups of approximately twenty 4–10 days post-eclosion females were tested at 25°C at 50% relative humidity in a dark chamber . Videography was performed under reflected IR light using a camera ( ROHS 1 . 3 MP B&W Flea3 USB 3 . 0 Camera; POINT GREY , Richmond , BC , Canada ) with an 800-nm long pass filter ( B&W filter; Schneider Optics ) at 30 frames per sec , 1024 × 1024 pixel resolution and analyzed using Fiji ( Schindelin et al . , 2012 ) or Ctrax ( Branson et al . , 2009 ) . Statistical comparisons were performed using Prism ( Graphpad Inc , La Jolla , CA 92037 ) ; Kruskal Wallis One way ANOVA followed by Dunn's post-test for comparison between control and experimental genotype in Figure 2C and Figure 5—figure supplement 1; One way ANOVA followed by Bonferroni's multiple comparison test for Figures 3G and 4; In Figure 5D , p-values for the exit direction were computed using the test of equal proportions from R ( http://stat . ethz . ch/R-manual/R-patched/library/stats/html/prop . test . html ) followed by multiple comparisons with the Dunn-Sidak correction . Only data obtained with 20XUAS-CsChrimson-mVenus ( attP18 ) are shown in this study . Our preliminary results with 5XUAS-CsChrimson-mVenus ( attP18 ) or 10XUAS-CsChrimson-mVenus ( attP18 ) indicate that either too weak or too strong expression may result in a failure to observe a phenotype . Behavioral experiments were performed at 60% humidity in dim red light for training and in complete darkness for test . The odors , 3-octanol ( OCT; Merck ) and 4-methylcyclohexanol ( MCH; Sigma–Aldrich ) were diluted to 1% and 2% , respectively , in paraffin oil ( Sigma–Aldrich ) . Flies were placed in the apparatus and shifted to the restrictive temperature ( 32°C ) from 30 min prior to the commencement of training until the end the experiment . During the 2-hr period between training and testing , trained flies were kept in a vial with moistened filter paper . The trained flies were then allowed to choose between MCH and OCT for 2 min in a modified transparent T-maze . Odors were placed in cups with 30 mm diameter and delivered at a flow rate of 0 . 6 l/min per tube . The distribution of the flies was monitored by videography and the preference index was calculated by taking the mean indices of the last 10 s in the 2-min choice period . Half of the trained groups received reinforcement together with the first presented odor and the other half with the second odor to cancel the effect of the order of reinforcement . For aversive memory , a group of ∼50 flies in a training tube alternately received OCT and MCH for 1 min in a constant air stream; twelve 1 . 5 s 90 V electric shocks spaced over 60 s were paired with one of the odors . In the primary screening using pJFRC100-20XUAS-TTS-Shibire-ts1-p10 ( VK00005 ) as the effector , flies were raised at 25°C . In the secondary screening using the UAS-Shi x1 effector , flies were raised at 18°C . Odor avoidance was measured by asking flies to choose between air and either MCH or OCT at the same concentrations used in the memory assay; these odors are aversive to naïve flies . For shock avoidance , flies were asked to choose between two tubes , both with copper grids , but only one electrically active . For appetitive memory , the conditioning protocol was as described previously ( Liu et al . , 2012 ) . Flies were starved prior to the experiments until ∼10% mortality was reached . For sugar attraction , flies are asked to choose between two tubes , one with plain filter paper and one with sugar-embedded paper . Statistical analyses were performed with Prism5 software ( GraphPad ) . The tested groups that did not violate the assumption of normal distribution ( D'Agostino-Pearson test ) or homogeneity of variance ( Bartlett's test ) were analyzed with parametric statistics: one-sample t-test or one-way analysis of variance followed by planned pairwise multiple comparisons ( Bonferroni ) . The significance level for statistical tests was set to 0 . 05 . As some of the data points in Figure 7B violated the assumption , non-parametric statistics were applied to the dataset ( Kruskal–Wallis test followed by Dunn's multiple test ) . The effector-line pJFRC100-20XUAS-TTS-Shibire-ts1-p10 ( VK00005 ) was used , and all experimental flies were heterozygous ( w+/w- ) or wild-type ( w+/Y ) for white . Flies were sorted by genotype under CO2 anesthesia at least 2 days prior to experiments; each measurement used 30–40 mixed males and females . For appetitive conditioning experiments , 2–4 days post-eclosion flies were starved on moistened filter paper to approximately 20% mortality ( Schnaitmann et al . , 2010 ) ; for aversive conditioning experiments , flies were not starved . Control responses to sugar and shock were measured as described previously ( Schnaitmann et al . , 2010; Vogt et al . , 2014 ) . Appetitive and aversive conditioning paradigms and behavioral tests were as previously described ( Schnaitmann et al . , 2013; Vogt et al . , 2014 ) . Briefly , conditioned stimuli were presented from below using LEDs with peak wavelengths of 452 nm and 520 nm ( Seoul Z-Power RGB LED ) or 456 nm and 520 nm ( H-HP803NB , and H-HP803PG , 3 W Hexagon Power LEDs , Roithner Lasertechnik , Vienna , Austria ) , adjusted to 14 . 1 Cd m−2 s−1 ( blue ) and 70 . 7 Cd m−2 s−1 ( green ) . Each quadrant of the arena was also equipped with an IR-LED ( 850 nm ) to provide background illumination for videography . For appetitive conditioning , filter paper soaked in 2 M sucrose and subsequently dried was presented as reward ( Schnaitmann et al . , 2010 ) . For aversive conditioning , a 1 s electric shock ( AC 60 V ) was applied 12 times in 60 s during CS + presentation using a transparent shock grid made of laser-structured ITO on a glass plate . In both appetitive and aversive conditioning assays , differential training was followed by a binary choice without reinforcement . During the 90 s test period , blue and green light were presented in two diagonally opposite quadrants of the arena and the color choice of flies was recorded from above at 1 frame per second with a CMOS camera ( Firefly MV , Point Grey ) . A preference index for each frame was calculated by subtracting the number of flies on the green quadrants from the number on the blue quadrants , divided by the total number of flies . The difference in average visual stimulus preference between the two groups was used to calculate a performance index . Sugar preference and shock avoidance tests were performed as described previously ( Schnaitmann et al . , 2010; Vogt et al . , 2014 ) . Statistical analyses were performed with Prism5 software ( GraphPad ) . The groups that did not violate the assumption of normal distribution ( Shapiro–Wilk test ) or homogeneity of variance ( Bartlett's test ) were analyzed with parametric statistics: one-way analysis of variance followed by the planned pairwise multiple comparisons ( Bonferroni ) . For data that significantly differed from the normal distribution or did not show homogeneity of variance ( Bartlett's test ) , non-parametric statistics were applied ( Kruskal–Wallis test followed by Dunn's multiple test ) . The significance level of statistical tests was set to 0 . 05 . Only the most conservative statistical result of multiple pairwise comparisons is indicated . For behavioral experiments concerning long-term aversive olfactory memory ( LTM ) , wild type ( Canton S ) or Split-GAL4 female flies were crossed to either pJFRC100-20XUAS-TTS-Shibire-ts1-p10 ( VK00005 ) ( outcrossed to a Canton S genetic background ) or wild type males . Flies were raised on standard medium containing yeast , cornmeal and agar at 18°C and 60% relative humidity under a 12 hr:12 hr light–dark cycle . The day before the experiment , 0–2 days post-eclosion flies were transferred to fresh food vials . Flies were trained with five cycles of aversive conditioning spaced by 15 min inter-trial intervals ( spaced conditioning ) at 25°C . The time course of one cycle of aversive conditioning was as follows: flies were exposed to the first odorant for 1 min while twelve 1 . 5 s , 60-V electric shocks , separated by 3 . 5 s , were delivered; after a 45 s rest , flies were exposed to the second odorant for 1 min . Two odorants , 3-octanol and 4-methyl-cyclohexanol , were used alternatively as the conditioned stimulus . For all assays ( training , memory test and olfactory acuity ) , odorants were diluted in mineral oil at a final concentration of 0 . 36 mM for octanol and 0 . 325 mM for methyl-cyclohexanol , and were delivered by 0 . 4 l/min airflow bubbled through odor-containing bottles . Except during conditioning , flies were kept on food and were maintained at 18°C between training and test . The memory test was performed as described in ( Trannoy et al . , 2011 ) . Flies were allowed to acclimatize to the restrictive ( 32°C ) or permissive ( 25°C ) temperature for 30 min prior to the test . Memory scores are displayed as means ± SEM . In a primary screen , each Split-Gal4 line was tested for aversive LTM ( n = 7-10 , except MB093C , n = 20 ) . The scores obtained for each line were compared by a two-tailed unpaired t-test to the pool of +/UAS-Shits scores ( n > 150 ) . Due to multiple comparisons , a Benjamini-Hochberg procedure ( Benjamini and Hochberg , 1995 ) was applied to control the false discovery rate with a significance level of 0 . 05 . Putative hits after the primary screen were then re-assayed in comparison with +/UAS-Shits and Split-GAL4/+ at both the restrictive and permissive temperatures . In this second set of experiments , the scores from the three genotypes were compared using one-way ANOVA followed by pairwise comparisons by Newman–Keuls posthoc tests . MB052B/Shi showed memory impairment compared to MB052B/+ and +/Shi at the restrictive temperature ( ANOVA , F2 , 33 = 14 . 84 , p < 0 . 0001; ***: p < 0 . 001 by Newman–Keuls pairwise comparison; n ≥ 9 for all genotypes ) but not at the permissive temperature ( ANOVA , F2 , 24 = 0 . 95 , p = 0 . 40 , n ≥ 8 for all genotypes ) . Flies were reared at room temperature ( 22°C–23°C ) under ambient light with no constrained light/dark cycle . Split-GAL4 males were crossed to pJFRC100-20XUAS-TTS-Shibire-ts1-p10 ( VK00005 ) females . A daily control ( pJFRC100 x pBDPGAL4U ) was run alongside all experimental crosses . Split-GAL4/+ crosses were performed at a different time than the original split-GAL4/Shits screen and run alongside UAS-Shits/+ flies . Thermoactivation of Shits1 was carried out at 31°C during both the training and test period . For 24-hr memory , flies were kept at 22°C–23°C under ambient room light between training and test . Odors used were 1:36 ( vol:vol ) ethyl acetate in mineral oil and 1:36 ( vol:vol ) iso-amyl alcohol in mineral oil . Choice tests for groups of 30 flies were performed in a Y-maze ( each arm 2 . 5 cm in length and 1 . 5 cm in diameter ) . Odors were actively streamed individually through the top arms of the Y at 0 . 3 l/min . Vials of flies were placed at the lower Y arm and flies climbed up and chose between opposing arms of the Y into 14 ml culture tubes; one arm contained one of the odors and the other arm contained air streamed through mineral oil . The preference index was calculated by the formula: ( number of flies in odor vial–number of flies in air vial ) / total number of flies . The conditioned preference index was the average of the preference indices in reciprocal trials . Ethanol conditioning was performed essentially as described in ( Kaun et al . , 2011 ) . Groups of 30 males were trained in perforated 14 ml culture vials filled with 1 ml of 1% agar and covered with mesh lids . 96 vials of flies were trained simultaneously in two 30 × 15 × 15 cm training boxes . Training consisted of a 10 min habituation to the training chamber with air , a 10 min presentation of odor 1 ( 1:36 odor:mineral oil actively streamed at 2 l/min ) , then 10 min of odor 2 ( 1:36 odor:mineral oil actively streamed at 2 l/min ) , with 60% ethanol . Air flow was matched for CS+ and CS− experiments , and ethanol was delivered by mixing pure ethanol vapor ( 1 . 5 l/min ) with humidified air ( 1 . 1 l/min ) at a specified ratio ( Wolf et al . , 2002 ) . Reciprocal training was performed to ensure that an inherent preference for either odor did not affect the results . Vials of flies from Group 1 and Group 2 were paired according to placement in the training chamber and tested simultaneously . Flies were tested in the Y-maze described above either 30 min or 24 hr after training . Reciprocal groups were averaged for each n = 1 . Statistical analyses were performed using the statistical software JMP 10 . 0 . 0 ( SAS Institute , Inc . , Cary , NC 27513-2414 ) . Each split-GAL4/UAS-Shits1 cross was run on two separate days and pooled for a total with n = 12/group . Statistical significance for any split-GAL4 line was determined by performing a Wilcoxon test for each split-GAL4/UAS-Shits cross against a pooled control . The pooled control included 12 randomly sampled means from the pBPDG4U/UAS-Shits daily control . A Benjamini-Hochberg False discovery rate ( FDR ) test ( Benjamini and Hochberg , 1995 ) was performed on the p-values for each Wilcoxon comparison . Lines showing p < 0 . 05 following the FDR test were considered statistically significant . GAL4/+ controls were performed for these significant hits , and Kruskal–Wallis comparisons were made comparing each split-GAL4/UAS-shits , split-GAL4/+ and +/UAS-Shits . Lines were considered significant hits if they passed the FDR correction , splitGAL4/+ Kruskal–Wallis test and showed no significantly decreased sensitivity for either odor used in the assay . Split-GAL4 flies were crossed to either 10X UAS-dTrpA1 ( attP16 ) ( Hamada et al . , 2008 ) or pJFRC124-20XUAS-IVS-dTrpA1 ( attP18 ) and maintained at 21–22°C . Virgin female progeny , 3–7 days post-eclosion , ( n = 24–35 ) were placed in 65 mm × 5 mm transparent plastic tubes with standard cornmeal dextrose agar media , placed in a Drosophila Activity Monitoring system ( Trikinetics ) and locomotor activity data were collected in 1-min bins for 7 days . Activity monitors were maintained with a 12 hr:12 hr light–dark cycle at 65% relative humidity . Total 24-hr sleep amounts ( daytime plus nighttime sleep ) were extracted from the locomotor data as described by ( Donelson et al . , 2012 ) ; sleep was defined as 5 min or more of inactivity ( Hendricks et al . , 2000; Shaw et al . , 2000 ) . Sleep profiles were generated representing average ( n = 24–32 ) sleep ( min/30 min ) for day 3 ( baseline ) , days 4 and 5 ( activation ) , and day 6 ( recovery ) . In addition to permissive temperature controls , pBDPGAL4U /dTrpA1 and split-GAL4/+ were used as genotypic controls for hit detection . For all screen hits , waking activity was calculated as the number of beam crossings/min when the fly was awake; consistent with the assays performed in the Flybowl , none of the lines had discernable locomotor defects . Statistical comparisons between experimental and control genotypes were performed using Prism ( Graphpad Inc ) by Kruskal Wallis One way ANOVA followed by Dunn's post-test . | An animal's survival depends on its ability to respond appropriately to its environment , approaching stimuli that signal rewards and avoiding any that warn of potential threats . In fruit flies , this behavior requires activity in a region of the brain called the mushroom body , which processes sensory information and uses that information to influence responses to stimuli . Aso et al . recently mapped the mushroom body of the fruit fly in its entirety . This work showed , among other things , that the mushroom body contained 21 different types of output neurons . Building on this work , Aso et al . have started to work out how this circuitry enables flies to learn to associate a stimulus , such as an odor , with an outcome , such as the presence of food . Two complementary techniques—the use of molecular genetics to block neuronal activity , and the use of light to activate neurons ( a technique called optogenetics ) —were employed to study the roles performed by the output neurons in the mushroom body . Results revealed that distinct groups of output cells must be activated for flies to avoid—as opposed to approach—odors . Moreover , the same output neurons are used to avoid both odors and colors that have been associated with punishment . Together , these results indicate that the output cells do not encode the identity of stimuli: rather , they signal whether a stimulus should be approached or avoided . The output cells also regulate the amount of sleep taken by the fly , which is consistent with the mushroom body having a broader role in regulating the fly's internal state . The results of these experiments—combined with new knowledge about the detailed structure of the mushroom body—lay the foundations for new studies that explore associative learning at the level of individual circuits and their component cells . Given that the organization of the mushroom body has much in common with that of the mammalian brain , these studies should provide insights into the fundamental principles that underpin learning and memory in other species , including humans . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2014 | Mushroom body output neurons encode valence and guide memory-based action selection in Drosophila |
Phylogenetic relationships among extinct hominoids ( apes and humans ) are controversial due to pervasive homoplasy and the incompleteness of the fossil record . The bony labyrinth might contribute to this debate , as it displays strong phylogenetic signal among other mammals . However , the potential of the vestibular apparatus for phylogenetic reconstruction among fossil apes remains understudied . Here we test and quantify the phylogenetic signal embedded in the vestibular morphology of extant anthropoids ( monkeys , apes and humans ) and two extinct apes ( Oreopithecus and Australopithecus ) as captured by a deformation-based 3D geometric morphometric analysis . We also reconstruct the ancestral morphology of various hominoid clades based on phylogenetically-informed maximum likelihood methods . Besides revealing strong phylogenetic signal in the vestibule and enabling the proposal of potential synapomorphies for various hominoid clades , our results confirm the relevance of vestibular morphology for addressing the controversial phylogenetic relationships of fossil apes .
Catarrhine primates ( Old World anthropoids ) include two extant subclades: cercopithecoids ( Old World monkeys ) and hominoids ( apes and humans ) . Based on molecular ( e . g . , Springer et al . , 2012 ) and paleontological ( e . g . , Harrison , 2013; Stevens et al . , 2013 ) data , both groups diverged during the late Oligocene ( ≥25 Ma ) , but experienced very different evolutionary histories . Hominoids first radiated in the early Miocene of Africa ( Harrison , 2010; Begun , 2013; Begun , 2015 ) and subsequently diversified into Eurasia during the middle and late Miocene ( Alba , 2012; Begun , 2015 ) . Their diversity and geographic distribution ( humans excluded ) was much greater during the Miocene than at present , being currently restricted to a few genera in southeastern Asia and Africa . In contrast , extant cercopithecoid lineages started to diversify later and experienced a major radiation during the late Miocene ( Jablonski and Frost , 2010 ) , being currently much more diverse and widely distributed than hominoids in both Africa and Asia . The decimated current diversity of hominoids , coupled with the fragmentary nature of their fossil record , abundant homoplasy ( e . g . , Larson , 1998 ) , and the lack of known fossil hylobatids prior to the latest Miocene ( Harrison , 2016 ) make it difficult to confidently infer the phylogenetic relationships of extinct hominoids and thus reliably infer the morphotype of the last common ancestor ( LCA ) of various hominoid subclades . This is required not only to adequately understand the evolutionary history of the group as a whole , but also to reconstruct the LCA of chimpanzees and humans , from which the earliest hominins evolved during the late Miocene . The contribution of Miocene apes to debates about hominoid evolution is thus diminished by the numerous controversies about the phylogenetic position of the former . For example , putative stem hominoids from the early Miocene of Africa ( Begun , 2013; Begun , 2015 ) , such as the proconsulid Ekembo , lack most of the synapomorphies of crown hominoids , such that some authors still contend that they might represent stem catarrhines instead ( Harrison , 2010; Harrison , 2013 ) . Even more uncertain is the position of dendropithecids ( e . g . , Micropithecus ) and other small-bodied catarrhines from the early to middle Miocene of Africa , which are generally interpreted as stem catarrhines ( Harrison , 2010; Harrison , 2013 ) but might include stem hominoids as well ( Alba et al . , 2015; Begun , 2015 ) . The same controversy applies to the European middle to late Miocene Pliobates , recovered as a stem hominoid more derived than proconsulids ( Alba et al . , 2015 ) or alternatively as a member of the stem catarrhine pliopithecoid radiation ( Nengo et al . , 2017 ) . Regarding undoubted hominoids , the distinction between stem and crown taxa is by no means less controversial , in part due to the virtual lack of fossil hylobatids since their origin in the early Miocene until the latest Miocene ( Harrison , 2016 ) . This is best exemplified by the late Miocene Oreopithecus from Italy , variously considered a hominid ( Moyà Solà and Köhler , 1997; Harrison and Rook , 1997 ) or a stem hominoid ( Nengo et al . , 2017 ) . Finally , the phylogenetic placement of Miocene apes from Eurasia is also controversial . For example , most Asian forms have been classically considered pongines ( Begun , 2013 ) , but most recently the late Miocene Lufengpithecus from China has been reinterpreted as a stem hominid ( Kelley and Gao , 2012; Begun , 2015 ) . Even more controversial is the position of the middle to late Miocene European dryopithecines , interpreted as either stem hominids ( Alba , 2012; Alba et al . , 2015 ) or hominines ( Begun , 2013; Begun , 2015 ) , and further controversies apply when trying to decipher the phylogenetic relationships among various members of this extinct group . Deciding among phylogenetic hypotheses for Miocene apes has consequences for our current understanding of hominin origins , from calibrating molecular data to estimate their divergence time to the reconstruction of the ancestral locomotor repertoire from which the earliest bipeds arose . For example , the recently described dryopithecine Danuvius from Germany has been used to argue that bipedal and suspensory adaptations characterized the last common ancestor of crown hominids ( Böhme et al . , 2019 ) . However , without a phylogenetic analysis supporting a more basal branching of Danuvius compared with the older dryopithecine Pierolapithecus ( Moyà-Solà et al . , 2004 ) , which was orthograde but lacked adaptations to both bipedalism and suspension ( Alba et al . , 2010; Alba , 2012 ) , the implications for the ancestral condition of the group remain moot . The recent recovery of enamel proteome sequences from the early Pleistocene ( 1 . 9 Ma ) fossil pongine Gigantopithecus ( Welker et al . , 2019 ) offers some hope that molecular data will become available for Miocene apes sometime in the future . In the meantime , to better resolve the phylogeny of Miocene hominoids , the search for morphological features not very prone to homoplasy is crucial . Anatomical structures that develop early during development and are not remodeled thereafter ( such as the enamel-dentine junction of teeth ) represent the best candidates ( Corruccini , 1987; Skinner et al . , 2008 ) . In this regard , the inner ear is also a very promising anatomical area ( Spoor and Zonneveld , 1998; Spoor et al . , 2003 ) , even if thus far its phylogenetic implications have been mostly explored for fossil hominins only ( Quam et al . , 2016; Conde-Valverde et al . , 2018; Ponce de León et al . , 2018; Beaudet , 2019a; Beaudet et al . , 2019b ) , while its application to fossil apes has been mainly devoted to locomotor inferences ( David et al . , 2010; Malinzak et al . , 2012; Rook et al . , 2004; Ryan et al . , 2012 ) . Housed in the highly mineralized petrosal bone , which is frequently preserved in the fossil record , the inner ear is composed of a series of endolymph-filled membranes ( encased in the corresponding bony labyrinths ) , namely the cochlea or cochlear duct ( involved in hearing ) and the vestibular apparatus ( devoted to balance and vision stability ) . The vestibule consists of three ( anterior , posterior , and lateral ) semicircular canals ( SCs ) and two macular organs ( utricle and saccule ) . The approximately orthogonal SCs sense angular accelerations and decelerations of the head , while the maculae perceive linear accelerations and thus provide gravity reference ( Spoor and Zonneveld , 1998; Rabbitt et al . , 2004; Johnson Chacko et al . , 2018; Cheung and Ercoline , 2018 ) . Differences in the relative size and morphology of the SCs have been correlated with locomotor agility ( Spoor et al . , 1994; Spoor et al . , 2007; Walker et al . , 2008; Silcox et al . , 2009; Ryan et al . , 2012; Perier et al . , 2016 ) and positional behavior ( Le Maître et al . , 2017 ) , albeit not without criticism ( Rae et al . , 2016; Benson et al . , 2017; Coutier et al . , 2017 ) . On the other hand , the bony labyrinth morphology has been considered of great importance for phylogenetic reconstruction in various mammals including hominins ( Quam et al . , 2016; Conde-Valverde et al . , 2018; Ponce de León et al . , 2018; Beaudet , 2019a; Beaudet et al . , 2019b ) and nonhuman primates ( Lebrun et al . , 2010; Lebrun et al . , 2012; Grohé et al . , 2016; Mennecart et al . , 2017; Costeur et al . , 2018; Schwab et al . , 2019 ) . While previous research in hominoids has yielded encouraging results for phylogenetic reconstruction ( Spoor and Zonneveld , 1998; Spoor et al . , 2003; Rook et al . , 2004; Gunz et al . , 2012 ) , according to some authors phylogeny may not be a major component of ape vestibular morphology ( Le Maître et al . , 2017 ) . Determining if and to what extent inner ear anatomy reflects phylogeny among extant hominoids is central for assessing the potential of this anatomical area to more confidently resolve the controversial phylogenetic relationships of fossil apes . To provide insight into this question , we use phylogenetically-informed statistical analyses to test the significance and quantify the amount of phylogenetic signal of vestibular shape captured by three-dimensional geometric morphometrics ( 3DGM ) in living hominoids and a broader sample of extant anthropoids . To capture vestibular shape , we mostly rely on a landmark-free , deformation-based 3DGM approach that takes the whole surface into account ( Glaunés and Joshi , 2006; Durrleman et al . , 2012b; Durrleman et al . , 2012a ) and enables integrating the spatial trajectory of the semicircular canals with their thickness and volumetric variations , the latter two being more difficult to assess based on mainstream landmark-based 3DGM . Since our results reveal the presence of strong phylogenetic signal in the vestibular morphology , we also employ maximum likelihood methods ( Felsenstein , 1988; Schluter et al . , 1997 ) to reconstruct the ancestral vestibular morphology for the LCA of main hominoid subclades ( crown hominoids , hominids , and hominines ) , with the aim to identify phylogenetically informative characters that can be used in formal cladistic analysis . To test the reliability and illustrate the usefulness of our approach from a phylogenetic viewpoint , we also include two extinct hominoid taxa: the early hominin Australopithecus and the aforementioned controversial late Miocene ape Oreopithecus . The well-known phylogenetic placement of Australopithecus as the sister-taxon of humans predicts that its vestibular morphology will be somewhat derived towards the modern human condition . On the other hand , our analysis will enable testing the competing phylogenetic hypotheses for Oreopithecus ( stem hominoid vs . hominid ) , thereby illustrating the potential of our method for clarifying the controversial affinities of extinct apes .
In spite of a similar spatial configuration of the three SCs , their trajectory , stoutness and relative proportions are quite variable among anthropoids ( Figure 1 ) . A bgPCA performed among major anthropoid groups ( platyrrhines , cercopithecoids , hylobatids , and hominids ) enables their accurate distinction ( Figure 2 , Figure 2—figure supplement 1a , Figure 2—source data 1 ) , as shown by classification results ( more than 95% individuals correctly classified after cross-validation; Table 1 ) . In particular , bgPC1 discriminates hylobatids from hominids and all the remaining taxa ( Figure 2 , Figure 2—figure supplement 1c ) . A landmark-based analysis applied to the same sample yields very similar results except for hylobatids , due to the reasons explained in the next section . Shape variation in the analyzed sample accounts for a strong phylogenetic signal ( Kmult = 1 . 248 , p<0 . 001 ) , and this also holds for the first three bgPCs separately ( see below ) . When analyzed individually , we recover a similarly strong phylogenetic signal ( λ = 1 and K = 1 . 15; Table 2 ) for bgPC1 ( 68 . 8% of total variance ) . bgPC1 captures differences in thickness and cross-section of the SCs , and is also driven by the development of the macular organs relative to the canals . Great apes fall on positive values ( Figure 2 , Figure 2—figure supplement 1a ) due to their stout and flattened SCs combined with an extensive vertical compression of the anterior canal , a more anterosuperior insertion of the lateral canal into the vestibule , and a greater volume of the vestibular recesses relative to the canals ( Figures 3a–e , 4a–e and 5a–e ) . Hylobatids , as well as colobine and papionin cercopithecoids , showing slender and elongated canals but maintaining well developed ampullae , largely overlap on negative values , while cercopithecins and platyrrhines display intermediate values due to their slightly inflated SCs ( Figures 2 , 3f–h , 4f–h and 5f–h ) . bgPC2 ( 19 . 6% of total variance ) also bears strong phylogenetic signal ( λ = 0 . 91; K = 1 . 51; Table 2 ) , with variance accumulating among rather than within clades ( as indicated by K > 1 ) . This axis separates platyrrhines—especially Ateles ( Figure 1i ) —from other anthropoids due to the more reduced lateral canal in the former , which is inversely proportional to anterior canal development and vertical elongation ( Figure 2a ) . In contrast , Gorilla ( Figures 1a , 3a , 4a and 5a ) occupies the opposite end of the distribution due to its large lateral canal and reduced anterior one , the latter being also vertically compressed , whereas the remaining hominoids show intermediate values along bgPC2 . bgPC3 ( 11 . 6% of variance ) , which is driven by both trajectory and relative size of the SCs ( Figure 2b ) , still displays a strong and significant phylogenetic signal ( λ = 1 and K = 1 . 16; Table 2 ) . Hylobatids ( Figures 1f–h , 3f–h , 4f–h and 5f–h ) display the highest scores for bgPC3 , as a result of the right to acute angle formed by the apex of the common crus ( CC ) , the latter being also shorter , a more anterosuperiorly projecting and long anterior canal , an obtuse angle between the planes identified by the anterior and lateral canals , and a more developed lateral canal relative to the posterior one , which is also posteriorly oriented . Most of the taxa fall within moderate positive and moderate negative values , with some cercopithecoids ( Cercopithecus , Macaca , Papio and Nasalis , among others ) and Cebus occupying the negative end of the distribution . African apes , cercopithecines , and Nasalis occupy an intermediate position , being characterized by a well-developed and anteriorly inclined posterior canal , an obtuse to right angle of the CC apex , and an obtuse angle between the vertical canals , combined with a larger lateral canal . Orangutans fall on moderately positive values and are distinguished from other great apes and humans by an anterosuperiorly projecting anterior canal ( less than in hylobatids ) . When fossil specimens are plotted a posteriori onto the tangent space identified by extant taxa , Oreopithecus ( BAC 208; Figures 1n and 6a ) falls on moderately positive scores for bgPC1 ( where the distributions of hominids , cercopithecins , and platyrrhines overlap ) , while the two Australopithecus individuals ( StW 573 and StW 578; Figures 1o , p and 6b–c ) fall within the range of living great apes and humans ( Figure 2 , Figure 2—figure supplement 1a ) . This is due to the volumetric proportions of their SCs and the possession of voluminous vestibular recesses ( although the latter character is less pronounced in Oreopithecus ) . In BAC 208 , SC volume is greater on the lateral canal , while the two Australopithecus specimens possess stouter vertical canals . When the relative size of the SCs is taken into account , StW 573 shows more evenly developed canals than StW 578 and BAC 208 , which both display a smaller lateral canal . Their position along the bgPC2 axis reflects these differences , with StW 573 falling on positive scores ( close to the mean value for hominids and cercopithecoids; Figure 2a , Figure 2—figure supplement 1b ) and the other two specimens occupying moderately negative values ( within the range of extant catarrhines and approaching that of platyrrhines; Figure 2a , Figure 2—figure supplement 1b ) . This is caused by the comparatively smaller lateral canal and by the large vertical canal in StW 578 and in BAC 208 . Due to the acute angles between the planes identified by the anterior and lateral canals and that between the planes of the posterior and anterior canals , Oreopithecus falls at the negative end of the extant anthropoid distribution for the bgPC3 ( Figure 2b , Figure 2—figure supplement 1c ) . On the other hand , the two Australopithecus specimens occupy more intermediate values because of the possession of a right angle between the planes of the aforementioned canals . When the bgPCs are considered at the same time ( Figure 2 ) , the australopith specimens fall well within the great ape and human range . This is further supported by their posterior probabilities of group membership based on the proximity of fossil specimen scores to groups centroids , with StW 573 and 578 being classified as hominids with p=0 . 678 and p=0 . 190 , respectively ( Table 3 ) . StW 573 falls close to Pan and Homo , whereas StW 578 occupies an intermediate position between humans and orangutans due to its stouter volumetric proportions . When the posterior probabilities are computed using the centroids of the hominoid genera , StW 573 is classified as Pan as first option ( p=0 . 368 ) and as Homo as second ( p=0 . 264 ) , while StW 578 is more clearly classified as Homo with p=0 . 727 ( Table 4 ) . These results suggest that both australopith specimens show vestibular similarities with extant humans , but that StW 573 display a more plesiomorphic ( chimpanzee-like ) morphology . In turn , Oreopithecus shows a mosaic of vestibular features ( Pan-like volumetric proportions , small lateral canal , and acute angles between the anterior and both the posterior and the lateral canals ) that does not match the condition of any extant taxon ( Figure 2 ) . The posterior probabilities indicate closest similarities with cercopithecoids , followed by hominids and platyrrhines , although in all instances it falls outside the variability of the extant members of these groups ( p<0 . 05; Table 3 ) . When comparisons are restricted to hominoid genera , Oreopithecus appears more similar to humans than to any ape genus , but again with a posterior probability that indicates significant differences with all of them ( p<0 . 05; Table 4 ) . The multivariate regression between shape ( deformation fields ) and size ( log-transformed volume of the vestibule ) shows a significant correlation ( i . e . , allometry ) at p<0 . 001 , but nevertheless explains only a limited portion of the variance ( R2 = 0 . 192 ) . Bivariate regressions of the bgPCs against log-transformed cube root of vestibular volume reveal a significant correlation only for bgPC1 ( R2 = 0 . 635 , p<0 . 001 , Table 2 ) . A visual inspection of the scatter of points ( Figure 7a ) suggests that allometry for bgPC1 is more marked in hominids . This is confirmed when separate regressions are performed for hominids ( R2 = 0 . 480 , p<0 . 001 ) and the rest of the sample ( R2 = 0 . 058 , p<0 . 01 ) , with the former displaying also a higher slope ( Table 5 ) . When phylogeny is considered by means of PGLS regression ( Table 5 ) , the correlation for the whole sample is still significant but explains much less variance ( R2 = 0 . 261 , p<0 . 01 ) , becoming non-significant for hominids and the rest of the sample separately . bgPC3 shows a low yet significant correlation with volume for non-hominids , which becomes non-significant after PGLS correction ( Table 5 ) . Both Australopithecus and Oreopithecus overlap with the hominid scatter of points , with the two australopith specimens falling above the hominid regression line , whereas BAC 208 falls slightly below ( although well above that of non-hominid anthropoids ) . The bivariate regression between the log-transformed cube root of the SC volume and SC length shows in all instances a significant correlation that nevertheless only explains a limited ( ca . 20–30% ) amount of variance ( Table 5 , Figure 7b ) . Isometry cannot be rejected for anthropoids as a whole , but a negatively allometric relationship emerges ( revealing that length increases faster than volume ) when hominids and other taxa are analyzed separately ( Table 5 ) . The latter is confirmed by PGLS regressions for the whole sample and the two groups separately , which further explain a higher proportion of variance ( Table 5 ) , although the hominid regression is not significant with all probability due to small sample size . The bivariate plot ( Figure 7b ) shows an allometric grade shift between hominids and the remaining anthropoid taxa , which is confirmed by ANCOVA results—indicating no significant differences ( F = 0 . 705 , p=0 . 403 ) between the allometric slopes but significantly different intercepts ( F = 263 . 26 , p<0 . 001 ) between the two groups . This indicates that hominids possess more voluminous ( i . e . , stouter ) canals than other anthropoids at equal lengths , with only minimal overlap . All the fossil specimens display hominid-like volumetric proportions ( Figure 7b ) : StW 573 and 578 fall slightly above the hominid regression line , whereas BAC 208 , due to its slenderer semicircular canals , falls below ( although much closer than to the non-hominid regression line ) . Recently , caution has been advised regarding the use of between-group PCA ( bgPCA ) applied to 3D geometric morphometric ( GM ) data , as it could produce spurious grouping when there are fewer groups than variables ( Cardini et al . , 2019 ) . However , the same study also highlighted that the presence of a strong covariance among the variables ( as found in many biological structures ) largely reduces the magnitude of the problem . Interestingly , due to the properties of diffeomorphisms , the set of momenta is expected to be highly correlated , as close momenta tend to covary . Prior to computing the bgPCA , we explored the principal components resulting from the vestibular shape GM analyses to investigate the presence of a preexisting group structure , which was found to be similar to that showed by bgPCA for both extant and fossil taxa ( Figure 8 ) . We used hierarchical clustering analysis ( HCA ) on the deformation fields for assessing the probability of correct classification of individuals according to the groups used in the bgPCA . The confusion matrix resulting from the HCA shows that most individuals are correctly identified in the corresponding groups ( Table 6 ) . Only in the deformation-based analyses hylobatids show a low percentage of classification ( 24% of the individuals ) , mostly due to the great similarity in the volumetric proportions and surface shape of the vestibule between this group and cercopithecoids . Another way to ascertain that the group separation observed in our bgPCA is not the result of any bias is to compare different kind of analyses with a same data set and the same grouping factor . We thus compared our deformation-based results with those obtained from a configuration of 3D semilandmarks commonly used to investigate the vestibular shape ( Gunz et al . , 2012 ) . Both analyses , based on the full primate sample , yielded a similar group separation ( compare Figure 2 with Figure 9 ) , as shown by the components resulting from the bgPCA . These results are also coherent with the biological reality , enabling the discrimination of major anthropoid clades in agreement with their phylogenetic relationships . In the landmark-based approach , hylobatids largely overlap with great apes in both bgPC1 ( occupying an intermediate position ) and bgPC2 ( Figure 9a ) , but can be distinguished from them ( and other anthropoids ) to a large extent based on bgPC3 ( Figure 9b ) . Shape variation in the analyzed sample , as captured by landmark-based 3DGM , accounts for a strong phylogenetic signal ( Kmult = 0 . 973 , p<0 . 001 ) , and this also holds for the first three bgPCs separately ( see below ) . Shape differences along bgPC1 ( 53 . 3% of total variance ) in the landmark-based approach embed a strong phylogenetic signal ( λ = 1 and K = 1 . 26; Table 7 ) . This component correlates with the insertion of the lateral canal on the vestibule , the size and shape of the posterior canal , and the roundness of the SCs . Great apes and humans fall on negative values for the first axis ( Figure 9c ) , as they are characterized by smaller canals compared to the size of the vestibular recesses and less rounded SCs ( particularly the anterior one , which is vertically compressed ) . Hylobatids stand on intermediate scores that largely overlap with the hominid range ( Figure 9 ) due to a combination of long SCs , a vertically compressed anterior canal , and well separated lateral and posterior canals ( since posterior canal is posteriorly displaced and the lateral canal inserts anteriorly in the vestibule ) . In contrast , Old world monkeys tend to be located on positive values of bgPC1 and display a protruding lateral canal that intersects the plane defined by the posterior canal . bgPC2 ( 29 . 3% of variance ) separates platyrrhines—especially Ateles found in the most negative scores ( Figure 1i ) —from other anthropoids due to the more reduced lateral canal in the former , which is inversely proportional to anterior canal development and vertical elongation ( Figure 9d ) . This pattern is shared , although to a lesser extent , by humans ( Figure 1d ) and Theropithecus ( Figure 1j ) , which possess more developed anterior and posterior canals relative to the lateral one . In contrast , Gorilla ( Figure 1a ) occupies the positive end of the distribution due to its large lateral canal and reduced anterior one , the latter being also vertically compressed , whereas the remaining hominoids show intermediate values along bgPC2 . The identified phylogenetic signal displayed by bgPC2 is slightly reduced as compared to bgPC1 , but still significant and very high ( λ = 0 . 9 and K = 1 . 08; ) . The bgPC3 ( 17 . 4% of variance ) is driven by both trajectory and size of the SCs , especially of the posterior one ( Figure 9e ) , and displays less phylogenetic signal than bgPC1 and bgPC2 ( λ = 0 . 93 and K = 0 . 53 , Table 7 ) . Along this axis Homo , Ateles and Theropithecus ( Figure 9b ) occupy the positive end , as due to their large and anteriorly inclined posterior canal that protrudes laterally , an obtuse angle of the CC apex , large posterior canal , and an angle between the anterior and posterior canals that is close to 90° . Great apes and the majority of non-hominoid taxa fall in an intermediate position ( Figure 9b ) , differing from the aforementioned genera by the less obtuse angle in the CC apex and a larger lateral canal . Hylobatids and , to a lesser extent , Trachypithecus show the lowest scores for bgPC3 , as the result of the right to acute angle formed by the apex of the CC , more developed anterior and lateral canals relative to the posterior one , and a posteriorly tilted posterior canal . When plotted onto the tangent space of extant taxa , the two Australopithecus specimens overlap with the range of great apes and humans for all the bgPCs ( Figure 9 , Figure 9—source data 1 ) . They show a vertically compressed anterior canal together with well-separated lateral and posterior canals . The lateral canal is also moderately sinuous and its ampullar portion bends upwards , thus resulting in a negative score for bgPC1 . The two specimens can be distinguished from one another by means of bgPC2 ( Figure 9a ) , with StW 573 falling on positive values and StW 578 occupying negative ones . This is explained by the smaller lateral canal relative to the vertical SCs in StW 578 , which therefore overlaps with extant human variation in both bgPC1 and bgPC2 , whereas StW 573 overlaps instead with chimpanzees and bonobos ( Figure 9a ) . Along bgPC3 , both specimens display intermediate values due to the large posterior canal and for the almost right angle between the planes of the anterior and posterior SCs , overlapping with all extant anthropoids except hylobatids . In turn , along bgPC1 Oreopithecus displays more positive values than hominoids and falls well within the range of non-hominoid anthropoids ( it only slightly overlaps with the positive end of the hominoid distribution; Figure 9 ) due to its more coplanar lateral canal that almost intersects the plane defined by the posterior canal . Furthermore , due to its small lateral canal and fairly short CC , BAC 208 displays an intermediate value for bgPC2 , within the range of extant catarrhines and slightly above the positive end of platyrrhine distribution ( Figure 9a ) . Finally , along bgPC3 Oreopithecus falls on a positive score , differing from hylobatids , as a result of the acute angles found between anterior canal plane and those defined by both the posterior and lateral canals ( Figure 9b ) . Overall , the two 3DGM techniques used in this paper generally yielded similar results except for hylobatids and Oreopithecus ( along bgPC1 alone ) . This is attributable to differences in the underlying methodological assumptions of each method when computing shape variation . In particular , our 3DGM landmark protocol measures the spatial trajectory of SCs based on their midline skeleton ( Gunz et al . , 2012 ) and hence it does not capture differences in volumetric proportions . In contrast , by comparing surfaces as a whole ( Durrleman et al . , 2012b; Durrleman et al . , 2012a ) , the deformation analysis is particularly sensitive to volumetric differences . In addition , the amount of identified phylogenetic signal is very similar for both techniques , affecting the entire variance . Together with the results of the HCA , we confidently show that the separation found between the groups in the bgPCA of this study already exists in the shape data and that is not a spurious effect produced by the bgPCA method itself . The phylomorphospace approach applied to vestibular shape variation in hominoids infers different branch lengths for hominids and hylobatids from the ancestral morphology estimated for crown hominoids , which falls much closer to great apes and humans than to hylobatids for both bgPC1 and bgPC3 ( Figure 10 ) . According to our reconstructions based on the extant taxa ( Figure 10 ) , the crown hominoid LCA vestibular morphology ( Figure 11a ) would be characterized by equally developed and slightly inflated SCs , a fairly vertically compressed anterior canal , and by the slender portion of the lateral canal connecting more anteromedially with the vestibule . The estimated morphology for the LCA of hylobatids ( Figure 11b ) resembles to some extent that of the crown hominoid LCA ( slightly vertically compressed anterior canal and lack of intersection between the lateral and posterior canals ) , combined with more monkey-like features ( markedly slender canals with inflated ampullae ) , and others exclusive to hylobatids ( an obtuse angle between a slightly anteriorly protruding anterior canal and a small posterior canal relative to the others ) . Hylobatid genera are generally less diverging from one other than great apes and humans . Hoolock ( Figures 1f , 3f , 4f and 5f ) apparently displays the most primitive morphology among hylobatids ( with equally developed , rounded , almost orthogonal canals , and less anteriorly protruding anterior canal ) , while Hylobates ( Figures 1h , 3h , 4h and 5h ) and , to a lesser extent , Symphalangus ( Figures 1g , 3g , 4g and 5g ) show the slenderest SCs and an extremely anteriorly protruding anterior canal ( as noted by Le Maître et al . , 2017 ) . The reconstructed morphologies for LCAs of crown hominids ( Figure 11c ) and , to a lesser extent , crown hominines ( Figure 11d ) and the Pan-Homo clade ( Figure 11e ) are not very far from the crown hominoid ancestral condition ( Figure 11a ) for any of the first three bgPCs . The crown hominid LCA is characterized by a lateral insertion of the slender portion of the lateral canal on the vestibule , a moderate medioventral displacement of the posterior canal , and an increased vertical compression of the anterior canal , in combination with thick and bulgy canals and well-developed vestibular recesses . The hominine and Pan-Homo clade LCAs possess stouter canals and are very similar to one another , distinguished only by the size of the lateral and anterior canals . The LCA of the Pan-Homo clade shows a larger and less vertically compressed anterior canal , and a smaller lateral one , which also connects more anteriorly with the vestibule . The hominin ( Australopithecus-Homo ) LCA ( Figure 11f ) is closer to Homo , being characterized by the stoutest volumetric proportions , with larger vertical canals relative to the Pan-Homo clade LCA , yet smaller than those found in humans . Its anterior canal is more vertically compressed than in Homo , rather resembling the morphology of Pan , while the posterior canal is rounded , thus being intermediate between the human ( laterally projecting ) and the chimpanzee ( laterally compressed ) morphology . Among crown hominids , Pan ( Figures 1b–c , 3b–c , 4b–c and 5b–c ) more closely resembles the morphology of the inferred LCAs of crown hominids and hominoids ( Figure 11a , c ) in the moderately inflated and equally developed SCs and in the degree of the vertical compression of the anterior canal . Pongo ( Figures 1e , 3e , 4e and 5e ) occupies the positive end along bgPC1 ( Figure 10 ) due to the possession of relatively small but extremely stout canals ( especially the anterior one and the common crus ) . It also exhibits the most vertically compressed anterior canal and a ‘triangular’ lateral canal ( i . e . , showing straight slender portions of the bony labyrinth close to the ampulla and to the connection with the vestibule , as previously outlined by Spoor and Zonneveld , 1998 ) . Finally , Gorilla and Homo ( Figures 1a , 3d , 4d and 5d ) are derived in opposite directions along bgPC2 ( Figure 10a ) , as the former exhibits increased lateral canal radius with a flattened cross-section , while humans retain quite rounded canals shape and cross-section , and show a reduction of the lateral canal , as opposed to more developed anterior and posterior canals . Australopithecus is closer to humans than to any great ape , due to the possession of large vertical canals and stout volumetric proportions , being closest to the LCA of the Australopithecus-Homo clade and only slightly more derived in the same direction as Pongo due to the stouter canals ( Figure 10 ) . In contrast , Oreopithecus appears more plesiomorphic than humans and extant great apes in volumetric proportions , resembling those found in the LCA of crown hominoids . Nevertheless , Oreopithecus is clearly distinct from hylobatids in both SC stoutness and shape , being most clearly distinguished from gibbons and siamangs based on the acute angles defined the anterior canal with both the posterior and lateral canals ( Figure 10b ) . Oreopithecus is also derived in terms of SC size ( with the vertical canals much larger than the lateral one ) , similarly to humans albeit to a greater extent , and opposite to gorillas , due to the remarkably smaller lateral canal .
Previous research on the morphology of the vestibular apparatus among extant mammals has focused on its relationship to positional behavior ( Spoor et al . , 2007; Perier et al . , 2016; Le Maître et al . , 2017 ) , particularly in order to make locomotor inferences in extinct species ( Spoor et al . , 1994; Walker et al . , 2008; Silcox et al . , 2009; Ryan et al . , 2012 ) . However , the phylogenetic signal embedded in vestibular morphology has not been adequately quantified among hominoids , because previous attempts were either exploratory ( Gunz et al . , 2012 ) or based on a restricted sample ( Le Maître et al . , 2017 ) . Our results indicate that main anthropoid groups can be distinguished based on vestibular shape variation , and that there are important differences not only between hylobatids , great apes , and humans , but also among extant great ape genera . A significant phylogenetic signal is found to affect the entire variance of the anthropoid sample . Thus , the shape of the SCs is overall informative from a phylogenetic viewpoint—as hypothesized for strepsirrhine primates ( Lebrun et al . , 2010 ) and carnivorans ( Schwab et al . , 2019 ) , but in contrast to previous results for hominoids ( Le Maître et al . , 2017 ) and some other mammals ( Grohé et al . , 2016; Costeur et al . , 2018 ) . Based on the analysis of the shape of the vestibular apparatus , we identify several potential hominoid synapomorphies ( Table 8; Figure 12a–g ) , including among others a posteromedially displaced posterior canal and a straight segment between the lateral-most point of the lateral canal and its anteromedially situated insertion on the vestibule . These features result in an anteromedially located lateral canal ( i . e . , the plane defined by the posterior canal is always separated from the trajectory of the lateral canal , even when the latter is well developed , as in Gorilla and Hylobates ) . This would imply an increased sensitivity for angular accelerations occurring along the coronal plane , which has been correlated with orthogrady in extant hominoids ( Le Maître et al . , 2017 ) . The most evident character shared by extant apes and humans ( even if somewhat variable in the latter and in Hoolock ) , and further displayed by the extinct genera analyzed here , is the vertical compression of the anterior canal ( Figure 12a ) , as noted for great apes only in a previous analysis ( Spoor and Zonneveld , 1998 ) . Since the subarcuate fossa arguably constrains the shape of the anterior canal ( Jeffery et al . , 2008 ) , the hominoid morphology might be related to the absence of the fossa in great apes and siamangs ( Moyà-Solà and Köhler , 1993; Gannon et al . , 1988; Spoor and Leakey , 1996 ) . However , the combination of a well-developed fossa and marked vertical compression of the anterior SC found in Hylobates argues against this hypothesis . The latter is further rejected by the cercopithecoid morphology , characterized by a rounded anterior SC even in the largest terrestrial genera ( Papio , Theropithecus , and Mandrillus ) , which unlike other cercopithecoids display a much reduced ( or even absent ) subarcuate fossa ( Gannon et al . , 1988; Spoor and Leakey , 1996 ) . As a result of vertical compression , all hominoids display the anterior canal projected anterosuperiorly to some extent ( less accentuated in Homo and Hoolock ) . Such a projection of the anterior canal in Hylobates has been interpreted as a hylobatid synapomorphy ( Le Maître et al . , 2017 ) . Our results raise doubts about the latter view and indicate instead that the anterior projection of this canal is variable within both hylobatids and hominids , and that the LCA of crown hylobatids might have not shown the patent elongation of the anterior canal that is found in Hylobates and Symphalangus . On the other hand , here we identify two potential synapomorphies for hylobatids ( Table 8; Figure 12h , i ) : an obtuse angle between the planes defined by the anterior and posterior canals , and a posteriorly-inclined posterior canal , which is smaller relative to the anterior and lateral ones . In turn , bgPC1 clearly separates hominids ( great apes and humans ) from cercopithecoids and hylobatids and enables the identification of some potential hominid synapomorphies ( Table 8; Figure 12j , k ) . In particular , hominids differ from other anthropoids , including hylobatids , by derived volumetric proportions of the SCs ( stouter canals relative to their length , or shorter canals relative to their volume ) even when size-scaling considerations are taken into account , as well as by the possession of more extensive vestibular recess for a similar size of the SCs ( Table 8; Figure 12j , k ) , as reflected in bgPC1 . In particular , cercopithecoids and hylobatids completely overlap due to the possession of slender SCs , while hominids as a whole ( even if more markedly orangutans ) differ by their swollen and relatively shorter SCs ( with only few cercopithecins falling within the hominid range ) . This might be related to the fact that hylobatids and cercopithecoids , unlike great apes , are swift moving animals that make fast and large head movements , thus requiring a limited duct sensitivity to avoid overstimulation and a quick response to angular displacement ( Spoor and Zonneveld , 1998 ) . This hypothesis is supported by biophysical models suggesting that the length of the membranous ducts is inversely proportional to their sensitivity and that a larger lumen of the ducts correlates with a reduced steadiness of the response to an external angular stimulus ( i . e . , the abrupt change of the position and/or posture ) ( Muller , 1994; Rabbitt et al . , 2004 ) . Therefore , species with shorter and thicker ducts ( such as hominids ) require more time to perceive and react to sudden head displacements , while being more sensitive to fine movements . Nevertheless , caution must be used when inferring the lumen of the ducts from that of the bony canals , as the amount of the SC cross-section occupied by perilymphatic space is variable depending on the species ( Ramprashad et al . , 1984; Spoor and Zonneveld , 1998 ) . Superimposed on the aforementioned hominoid and hominid synapomorphies ( Table 8; Figure 12 ) , there are also marked differences among hominid genera . Such differences mainly relate the relative size among the SCs ( which varies particularly along bgPC2 and bgPC3 ) , while hylobatid genera are less diverse in this regard . Chimpanzees and bonobos are characterized by equally developed SCs and a moderately short CC . Similarly , orangutans possess evenly proportioned SCs and can be distinguished from Pan by a shorter CC , more inflated canals , and a greater vertical compression of the anterior SC . Gorillas display the largest intrageneric variability in the studied sample , especially with regard to SCs slenderness , coupled with some other distinctive traits ( obtuse angle of the CC apex , and longer lateral SC and CC ) . Humans differ from apes in the enlarged vertical canals , a laterally protruding and inferiorly displaced posterior canal , and a moderately smaller lateral canal . Relatively enlarged vertical canals are also found in Theropithecus . The human morphology has been linked to bipedalism ( Spoor et al . , 1994; Spoor et al . , 2003 ) , as accelerations during bipedal walking mainly occur along the vertical axis , so that the broadly similar morphology of Theropithecus might be related to the bipedal shuffling characteristic of this genus during foraging , causing them spend an extremely large amount of time with an erect trunk posture ( Wrangham , 1980 ) . Irrespective of the functional implications of the variation in vestibular morphology among anthropoids , our analyses show that this variation bears strong phylogenetic signal and , hence , has potential for reconstructing the evolutionary history of the group—particularly hominoids , which in spite of their extant decimated diversity are more variable in this regard ( particularly when size differences among the SCs are considered ) ( Spoor and Zonneveld , 1998; Le Maître et al . , 2017 ) than the taxonomically more diverse cercopithecoids . Although functional demands frequently lead to the independent evolution of similar morphologies ( homoplasy ) , often function is not decoupled from—but superimposed on—phylogeny , with many clades being characterized by synapomorphic features linked to the adaptation for new functions . Therefore , to the extent that vestibular morphology appears to be linked to positional behavior ( Spoor et al . , 1994; Spoor et al . , 2007; Walker et al . , 2008; Silcox et al . , 2009; Ryan et al . , 2012; Le Maître et al . , 2017 ) , the higher variation of vestibular morphology displayed by hominoids compared to cercopithecoids agrees with the more diverse and varied locomotor repertoires of the former . This is because cercopithecoids as a whole are largely pronograde terrestrial quadrupeds that mostly differ in the degree of arboreality vs . terrestriality ( Fleagle , 2013; Gosselin-Ildari , 2013 ) , while crown hominoids are characterized by a derived and versatile orthograde body plan and associated adaptations throughout the body that enable very different and very specialized antipronograde behaviors—vertical climbing ( all apes ) , ricochetal brachiation ( hylobatids ) , arboreal quadrumanous suspension and clambering ( orangutans ) , below-branch arm-swinging as well as semiterrestrial knuckle-walking ( African apes ) , and terrestrial bipedalism ( humans ) ( Hunt , 1991; Thorpe and Crompton , 2006 ) . The fact that , in terms of positional behavior , extant hominoid lineages have more significantly diverged in different directions from their last common ancestor with cercopithecoids explains why extant hominoids more strongly differ in vestibular features—even if the functional link of some vestibular features remains to be better determined . Determining the order in which these features evolved is therefore required to use them for inferring the phylogenetic placement of extinct hominoids . Besides proposing various potential synapomorphies for the hominoid and hominid clades , we further reconstruct the evolution of the vestibular apparatus in this group by estimating ancestral vestibular morphotypes by means of maximum likelihood and a molecular phylogeny . From the LCA of crown hominoids , great apes and humans appear derived in the opposite direction of hylobatids with regard to their volumetric proportions ( stout vs . slender SCs , respectively ) . According to our ancestral state reconstruction for crown hominoids , in this regard hylobatids appear secondarily convergent with cercopithecoids . However , this hypothesis ( and the alternate one , that hylobatids merely reflect more closely the primitive condition for crown catarrhines as a whole ) should be tested by means of adding extinct stem cercopithecoids and fossil hominoids of less controversial affinities than Oreopithecus to the analysis . It is noteworthy that Hoolock , in agreement with the basal position of this genus among extant hylobatids , apparently retains a more primitive morphology than other hominoids for various features . While the characters related to an anteromedial displacement of the lateral SC appear synapomorphic for hominoids , the only incipient vertical compression of the anterior canal SC morphology of Hoolock might be plesiomorphic for hylobatids , in which case the marked vertical compression of this canal would be a synapomorphy of hominids only , with the remaining hylobatid genera having also evolved it in parallel . While various great ape lineages and humans further diverged from one another from the more derived condition of the reconstructed crown hominoid LCA , as noted above gibbons and siamangs might have secondarily converged to some extent with cercopithecoids by evolving slenderer SCs , presumably as a result of similar evolutionary pressures posed by fast-moving types of locomotion . Based on our reconstructed ancestral morphotypes , bonobos and ( to a lesser extent ) chimpanzees would be closer to the LCAs of hominids , hominines , and the Pan-Homo clade , than either humans or the remaining extant great apes ( gorillas and orangutans ) . The latter would have diverged from the hominid LCA in markedly different directions both in terms of SC configuration and stoutness . Our results therefore support the view that not only hominins , but also gorillas and , to a large extent , the orangutan lineage diverged from ancestors with a largely Pan-like vestibular morphology . It would be tempting to interpret this pattern in locomotor terms ( e . g . , by suggesting a semiterrestrial ancestry not only for hominines , but also for crown hominids as a whole ) . However , caution is required as other selection pressures and/or non-adaptive factors could have potentially played an equally , if not more significant , role in determining vestibular shape variation in this group . With regard to the fossil hominoids analyzed here , Australopithecus not only displays the various hominoid synapomorphies mentioned above , but also hominid-like volumetric proportions of the SCs and , as expected in a bipedal hominin , human-like vestibular features such as large anterior and posterior canals . This is in agreement with the large amount of habitual bipedal behaviors inferred for the almost complete skeleton ( StW 573 ) to which one of the analyzed specimens belongs ( Heaton et al . , 2019 ) , as well as previous analyses of the inner ear as a whole ( Beaudet et al . , 2019b ) . Following our ancestral state reconstruction , Australopithecus appears derived in the same direction as humans , although it more closely resembles the morphology of the Australopithecus-Homo clade LCA , which is derived from the reconstructed LCA of the Pan-Homo clade in the opposite direction as chimpanzees and bonobos are . The fact that the two analyzed specimens of Australopithecus are classified as humans with a moderately high probability reflects the fact that their vestibular morphology already approximates the human condition , although maintaining plesiomorphic characters , particularly in the specimen that more closely resembles chimpanzees . On the other hand , their similar classification probability with other extant great ape genera is consistent with a more primitive vestibular morphology . It is noteworthy that , although the two analyzed individuals are very similar to one another ( Beaudet et al . , 2019b ) , they display noticeable differences in vestibular morphology , with StW 578 showing more human-like ( even if stouter , approximating the orangutan morphology ) canals and StW 573 retaining a more African great ape-like morphology . This could be related with diachronic changes within South African Australopithecus or to its previously noted heterogeneity ( Clarke , 2013; Grine , 2013 ) , and might help to discern , coupled with other features , the number of species represented among the current samples . Unlike Australopithecus , Oreopithecus displays a mosaic vestibular morphology that defies a simple phylogenetic interpretation , as it does not fit well among the variation displayed by any extant hominoid genus . This is due to the combination of some hominoid and hominid synapomorphies with more plesiomorphic , cercopithecoid-like , or even platyrrhine-like , features . In particular , the hominid-like volumetric proportions of the Oreopithecus SCs would support the contention that this taxon is a great ape ( Begun et al . , 1997 ) , or even a member of the European dryopithecine radiation , as previously argued by some authors ( Köhler and Moyà-Solà , 1997; Harrison and Rook , 1997 ) . However , other , apparently more plesiomorphic vestibular features , are at odds with such an interpretation , and even with the previous suggestion that Oreopithecus postcranium would be consistent with a giant hylobatid that emphasized cautious climbing ( Sarmiento , 1987 ) . In particular , the vestibular morphology of Oreopithecus does not overlap with that of extant hylobatids in any respect , particularly differing by the acute angle between the anterior and posterior canal planes and by the large posterior canal relative to the others . Oreopithecus also appears more primitive than crown hominoids in the shape of the lateral canal , which is flat ( rather than displaying an upwards bent ampullar portion ) and posteriorly displaced ( especially in the junction between the ampulla and the vestibular recesses ) . In these regard , the Oreopithecus morphology more closely resembles that of cercopithecoids and platyrrhines , respectively , possibly reflecting a plesiomorphic condition that would be more consistent with a stem hominoid status , as recently supported by some other authors ( Nengo et al . , 2017 ) . If our interpretation above is correct , then the stout volumetric proportions of Oreopithecus would be homoplastic with those of great apes and humans , representing an independent acquisition that might be functionally related to the evolution of a slower mode of locomotion—in agreement with previous analysis of the inner ear of this taxon ( Rook et al . , 2004; Ryan et al . , 2012 ) and the possession of an orthograde body plan with adaptations related to cautious vertical climbing and forelimb-dominated suspension ( Sarmiento , 1987; Harrison and Rook , 1997 ) . This is plausible given that hylobatids appear to some extent convergent in this regard to cercopithecoids , due to their agile locomotion . This suggests that volumetric proportions are quite labile in evolutionary terms , so that other features and functional considerations must also be considered when interpreting the vestibular morphology of extinct taxa in phylogenetic terms . Interestingly , Oreopithecus resembles australopiths and humans in the possession of a larger vertical and posterior canals relative to the lateral one , which apparently represents another homoplasy that would lend some support to the controversial claim that bipedalism featured prominently among the locomotor repertoire of this taxon ( Köhler and Moyà-Solà , 1997; Rook et al . , 1999 ) . However , the SCs of Oreopithecus are clearly distinguishable from those present in Homo and Australopithecus regarding volumetric proportions , orientation , and shape . This rules out a hominin-like bipedalism for Oreopithecus—in further agreement with the lack of the lower torso features than in australopiths and humans are functionally linked to committed bipedalism ( Hammond et al . , 2020 ) —but would not be at odds with the possession of more varied orthograde positional behaviors combining climbing with a different type of bipedalism ( more related to a stable bipedal stance and short distance shuffling instead of fast walking or running ) , as previously inferred based on the foot of this taxon ( Köhler and Moyà-Solà , 1997 ) . In conclusion , our study provides new insight into the evolution of the vestibular apparatus in hominoids and confirms the potential of SC shape for investigating further the phylogenetic affinities of fossil apes , which are still controversial due to the inherent limitations of the fossil record . This is not to say that functional considerations must not be taken into account—rather the contrary , as several of the discussed vestibular features are arguably linked with the demands of particular positional behaviors , as noted by previous authors ( Spoor et al . , 2007; Walker et al . , 2008; Silcox et al . , 2009; Ryan et al . , 2012; Perier et al . , 2016; Le Maître et al . , 2017 ) . However , as exemplified by the analysis of the extinct hominin Australopithecus , the various characters identified as potentially synapomorphic for either crown hominoids or hominids offer the prospect of refining the phylogenetic placement of fossil apes for which their stem vs . crown hominoid status is controversial—as these features can be easily scored from CTs of the petrosal bone and incorporated into formal cladistic analyses including information from other anatomical areas . On the other hand , our ancestral state reconstructions rely mainly on living taxa , which is potentially problematic in the case of hominoids , which were much more diverse in the Miocene and appear quite prone to homoplasy , particularly with regard to the locomotor adaptations of the few surviving lineages . Even if the quantification of phylogenetic signal based on the phylogeny of extant taxa indicates that vestibular morphology overall is not significantly affected by homoplasy , the evolutionary history of vestibular morphology presented here on the basis of ancestral morphotypes should be treated with caution as a set of working hypotheses that require further testing based on the information provided by a larger fossil sample . In particular , given the relationship between vestibular morphology and positional behavior , and the fact that the locomotor apparatus of extinct hominoids frequently displays a mosaic of primitive and derived features unknown among the surviving lineages ( e . g . , Moyà-Solà et al . , 2004; Alba , 2012; Alba et al . , 2015; Böhme et al . , 2019 ) , it may be predicted that the vestibular morphology of extinct hominoids will similarly display unique combinations of features . This is illustrated here by the condition of Oreopithecus , which is nevertheless most consistent with that of a stem hominoid somewhat convergent with hominids in terms of locomotion . In any case , our conclusions should be subjected to further scrutiny in the future by means of the inclusion of additional fossil taxa , with emphasis on Miocene hominoids as well as stem cercopithecoids .
The analyzed sample includes microcomputed tomography ( µCT ) scans of 142 dried anthropoid crania belonging to 27 species and 25 genera , including all extant great ape genera and a selection of hylobatids , cercopithecoids , and platyrrhines ( Table 9 and Supplementary file 1 ) . A few specimens are juveniles instead of adults , but this should not affect their vestibular morphology as the bony labyrinth ossifies in early prenatal stages , bounding its shape and size ( Jeffery and Spoor , 2004; Perier et al . , 2016 ) . The hominoid subsample consists of 48 individuals belonging to 8 species and seven genera ( Supplementary file 1 ) . For each specimen , the bony labyrinth was virtually extracted ( from the left side when possible , or otherwise from the right side and mirrored ) by segmenting the µCT image stacks ( voxel size reported in Supplementary file 1 ) . Virtual 3D models were generated using Avizo 9 . 0 . 1 software ( FEI Visualization Sciences Group ) . The fossil sample consists of one left bony labyrinth belonging to the late Miocene stem-hominoid Oreopithecus bambolii ( Rook et al . , 2004 ) and of two right inner ear that have been virtually extracted from the Australopithecus specimens StW 573 and StW 578 from Sterkfontein ( Beaudet et al . , 2019b ) . The vestibular apparatus was separated from the cochlea by cutting right under the oval window and the saccule , and filling the resulting hole with a flat surface in Geomagic Studio 2014 software ( 3D Systems ) . Our 3DGM approach is based on deformation methods ( Durrleman et al . , 2012b; Durrleman et al . , 2012a ) , which do not rely on a priori defined landmarks but consider instead the geometrical correspondences between continuous surfaces , and are particularly convenient for comparing overall shape and complex 3D surface changes ( Durrleman et al . , 2012b; Dumoncel et al . , 2014; Beaudet et al . , 2016a ) . This method relies on the construction of a sample-average surface model ( template ) and its deformation to the investigated surfaces ( Durrleman et al . , 2012b; Durrleman et al . , 2012a; Beaudet et al . , 2016b ) . Unlike in classical landmark-based 3DGM analyses , the surfaces are represented by a set of oriented faces and the comparisons do not assume a point-to-point correspondence between samples ( Durrleman et al . , 2012b ) . Prior to the analysis , the unscaled vestibular surfaces were aligned and scaled using the ‘Align Surface’ module of Avizo 9 . 0 . Then , deformations between surfaces were mathematically modeled as a diffeomorphism ( i . e . , a one-to-one deformation of the 3D space that is smooth , invertible , and with a smooth inverse ) , and a set of momenta ( vectors representing the flow of deformations from the initial position of the control points on the template to the target shape ) were estimated with Deformetrica 3 software . Due to its high-demanding computational power , analyses were run in the CALMIP supercomputing center ( Toulouse , France ) . We inspected interspecific major patterns of shape variation by means of between-group principal components analysis ( bgPCA ) of the deformation-based shape residuals , using major clades ( i . e . , platyrrhines , cercopithecoids , hylobatids and hominids ) as grouping factor . The restricted platyrrhine sample included in this study is aimed to serve as an outgroup for catarrhines , since the description of the vestibular morphology variation among New World monkeys as a whole is beyond the scope of this paper . Each group has been designed to include a large number of individuals ( >>10 ) in order to prevent spurious separation between the groups used in the analysis . We preferred bgPCA over linear discriminant analysis ( LDA ) because the latter produces overexaggerated separation among groups when the number of variables is close to the number of the analyzed individuals ( Mitteroecker and Bookstein , 2011 ) . Taking into account that some authors have recently recommended caution when interpreting bgPCA results , as they might present spurious grouping ( Cardini et al . , 2019 ) , we compared our results with those of a landmark-based 3DGM analysis ran on the same sample and investigated the presence of a preexisting group structure ( See ‘Exploration of a preexisting group structure in the tangent space of the vestibular shape’ section for further details ) . Correlation between SC shape and size ( allometry ) was assessed using multivariate regression of the deformation fields against log-transformed cube root of the entire vestibular volume ( ln Vol , in mm3 ) , as well as bivariate regressions computed for each bgPC against ln Vol . We further inspected the correlation between log-transformed cube root of the SC volume ( ln VolSC , in mm3; including the SCs and the CC cut at their connection with the vestibular recesses ) and log-transformed length ( ln L , in mm; measured along the streamline of the SCs and of the CC ) . All these regressions were performed for the whole anthropoid sample , as well as for hominids and non-hominids separately , as we detected several differences for the former . In the case of the ln VolSC vs . ln L regression , we also checked the homogeneity of slopes and intercepts between hominoids and the rest of the sample via analysis of covariance ( ANCOVA ) . Regressions were performed by means of ordinary least-square linear regression ( OLS ) as well as by taking into account the phylogenetic non-independence between data , that is by fitting our linear models via phylogenetic generalized least squares ( PGLS ) . The data used to compute the regressions are given in Figure 7—source data 1 . The hierarchical clustering analysis ( HCA ) of the deformation fields ( used to identify preexisting group structure embedded in our shape data ) has been computed caret v6 . 0–84 ( Kuhn , 2008 ) and Factominer v1 . 34 ( Le et al . , 2008 ) R packages . The discrimination and the amount of overlap among the groups defined a priori for the bgPCA has been quantified by computing the number of correctly classified individuals with cross-validation using the Morpho v2 . 6 ( Schlager , 2014 ) R package . Finally , we estimated the posterior probabilities of group membership for the fossil specimens based on the Mahalanobis distances between their bgPC scores and group centroids using the ‘typprobClass’ function of Morpho v2 . 6 ( Schlager , 2014 ) R package . These probabilities , which were computed for both bgPCA ( based on main anthropoid groups as well as on extant hominoid genera only ) denote the likelihood of the specimens to belong to each group without assuming that it must belong to one of them ( which is required when comparing the fossils with extant genera ) , so that the sum of probabilities for each specimen does not equal one . Probabilities < 0 . 05 indicate that the specimen falls outside the variability of that particular group . Further statistical analyses were carried out using different R packages in RStudio v . 1 . 1 . 453 for R v . 3 . 5 . 0: ape v5 . 1 ( Paradis , 2012 ) , phytools v0 . 6–60 ( Revell , 2012 ) , Morpho v2 . 6 ( Schlager , 2014 ) , caper v1 . 0 . 2 ( Orme et al . , 2013 ) , and geomorph v3 . 1 . 1 ( Adams et al . , 2019 ) . To assess phylogenetic signal , that is the degree to which related species resemble each other ( Felsenstein , 1985; Harvey and Pagel , 1991 ) , we used a phylogenetic tree ( Figure 13 ) derived from a time-calibrated molecular phylogeny for the extant taxa ( Springer et al . , 2012 ) . Estimated mean divergence dates for the various included extant clades are indicated in Figure 13 ( see Table 1 and S1 in Springer et al . ( 2012 ) for 95% composite credibility intervals ) . The phylogenetic placement of Oreopithecus has been controversial; for our purposes here , we followed Nengo et al . ( 2017 ) in considering this taxon as a stem-hominoid , although other possibilities are discussed in the text . The node of the Oreopithecus-crown hominoid divergence has been placed 1 Ma older than the divergence of crown apes and humans and its tip corresponds to its last occurrence in the fossil record ( 7 . 0–6 . 5 Ma; Rook et al . , 2000 ) . For the South African Australopithecus sp . , we used the published first appearance datum for Australopithecus africanus ( 4 . 02 Ma ) that includes the Jacovec specimens into the taxon ( Wood and Boyle , 2016 ) . We computed Pagel’s λ ( Pagel , 1999 ) , Blomberg’s K ( Blomberg et al . , 2003 ) and Kmult ( Adams , 2014 ) using phytools v . 0 . 6–44 ( Revell , 2012 ) and geomorph v3 . 1 . 1 ( Adams et al . , 2019 ) R packages . These metrics compare the variance in the phylogenetic tree tips relative to that expected under a Brownian motion evolutionary model . Pagel’s λ is a scaling coefficient of the expected correlations between related species on the tree , and varies from 0 ( no correlation due to absence of phylogenetic signal ) to 1 ( covariance proportional to phylogenetic distance , implying maximal phylogenetic signal ) . Blomberg’s K and its multivariate generalization Kmult inform on how precisely the variance-covariance patterns found in the data are matched by the phylogenetic tree and where variance accumulates: K ≈ one implies that the mode of evolution closely resembles that expected under Brownian motion; when K < 1 , close relatives resemble each other less than expected ( variance is accumulated within clades ) , implying an evolutionary pattern that departs from a purely stochastic model ( which could be caused by adaptation uncorrelated with phylogeny , that is homoplasy ) ; finally , K > 1 is found when close relatives are more similar than expected under Brownian motion ( variance is among clades ) , which could indicate stabilizing selection . To quantify major patterns of vestibular shape variation along the branches of the phylogeny we relied on a phylomorphospace approach ( Sidlauskas , 2008 ) , which allows us to intuitively visualize the extent and direction of the inferred shape change by means of branch length and orientation . This method projects the phylogenetic tree ( Figure 13 ) onto the tangent space computed from the bgPCA and estimates the position in the morphospace of the internal nodes ( ancestral morphologies ) via a maximum likelihood ( ML ) method for continuous characters ( Felsenstein , 1988; Schluter et al . , 1997 ) using the ‘fastAnc’ function of phytools v0 . 6–60 R package ( Revell , 2012 ) . Subsequently , the bgPC scores of the ancestral states are rotated and translated from the shape data back into the configuration space for interpolation and 3D visualization using Deformetrica 3 software . | Humans , gorillas , chimpanzees , orangutans and gibbons all belong to a group known as the hominoids . This ‘superfamily’ also includes the immediate ancestors and close relatives of these species , however in many instances the evolutionary relationships between these extinct ape species remain controversial . While DNA can help evolutionary biologists to work out how living species are related to one another , fossils are typically the principle source of information for extinct species . Inferring evolutionary relationships from fossils must be done with caution , but the bony cavity that houses the inner ear – which is involved in balance and hearing and fairly common in the fossil record – has proven useful for tracing the evolution of certain groups of mammals . However , no one had previously looked to see if this structure could give insights into the evolutionary relatedness among living and extinct hominoids . Urciuoli et al . have now used a 3D imaging technique to capture the complex shapes of the inner ear cavities of 27 species of monkeys and apes , including humans and two extinct apes ( Oreopithecus and Australopithecus ) . The results confirmed that the shape of these structures most closely reflected the evolutionary relationships between the species and not , for example , how the animals moved . Urciuoli et al . went on to identify features of these bony chambers that were shared within several hominoid groups , and to estimate what the inner ears of the ancestors of these groups might have looked like . The results for Australopithecus , for example , were consistent with it being most closely related to modern humans than other apes , while those for the enigmatic Oreopithecus supported the view that it was a much older species of ape that converged in some respects with other apes still alive today . The findings highlight the potential of the inner ear for reconstructing the early branches of our family tree . They also offer the prospect of refining the controversial evolutionary relationships within the impressive diversity of extinct ape species . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology"
] | 2020 | The evolution of the vestibular apparatus in apes and humans |
Mammals articulate their jaws using a novel joint between the dentary and squamosal bones . In eutherian mammals , this joint forms in the embryo , supporting feeding and vocalisation from birth . In contrast , marsupials and monotremes exhibit extreme altriciality and are born before the bones of the novel mammalian jaw joint form . These mammals need to rely on other mechanisms to allow them to feed . Here , we show that this vital function is carried out by the earlier developing , cartilaginous incus of the middle ear , abutting the cranial base to form a cranio-mandibular articulation . The nature of this articulation varies between monotremes and marsupials , with juvenile monotremes retaining a double articulation , similar to that of the fossil mammaliaform Morganucodon , while marsupials use a versican-rich matrix to stabilise the jaw against the cranial base . These findings provide novel insight into the evolution of mammals and the changing relationship between the jaw and ear .
In non-mammalian vertebrates , the jaw joint is formed between the quadrate ( or palatoquadrate ) of the upper jaw and the articular part of Meckel’s cartilage , a rod of cartilage that runs through the lower jaw ( Figure 1A ) . This is known as the primary jaw joint . In mammals , this function is carried out by a new joint between the dentary and squamosal bones , known as the temporomandibular joint or TMJ in humans , and is referred to as the secondary jaw joint . In mammals , the bones of the original primary jaw joint have been incorporated into the ear and play a role in hearing ( Anthwal et al . , 2013 ) . In addition to forming a joint with the articular as part of the primary jaw joint , the amniote quadrate also articulates with the cranial base . During the evolutionary transition that gave rise to mammals , the connection between the quadrate and the cranial base simplified ( Luo and Crompton , 1994 ) . The robust quadrate of reptiles moved from being attached to up to five separate skeletal elements , able to bear the mechanical force of feeding , to become the diminutive mammalian incus , suspended by a ligament from a single cranial base bone , the petrosal , in an air-filled cavity allowing sound transmission ( Kemp , 2005; Kielan-Jaworowska et al . , 2004 ) . At the same time , Meckel’s cartilage lost its permanent nature , separating the incus and neighbouring malleus from the rest of the jaw in adults ( Figure 1B; Anthwal et al . , 2017; Urban et al . , 2017 ) . Early mammal-like reptiles had a permanent Meckel’s cartilage and joints between the quadrate and articular ( Q-A ) , and posteriorly between the quadrate and cranial base – similar to extant reptiles ( Figure 1A ) . In mammaliaforms , such as Morganucodon , both a primary Q-A and a secondary dentary squamosal joint were present , in addition to a joint between the quadrate ( incus ) and the paraoccipital process of the petrosal ( Figure 1C ) . This petrosal and incus joint precedes detachment of the middle ear from Meckel’s cartilage in mammal evolution ( Luo and Crompton , 1994 ) . A connection between the future middle ear bones and the cranial base is therefore a feature of fossil mammaliaforms . In extant mammals , the proposed homologue of the paraoccipital process is the crista parotica , which forms as a cartilaginous spur off the petrosal and is derived from neural crest cells , distinct to the rest of the petrosal and otic capsule , which are mesodermally derived ( O'Gorman , 2005; Thompson et al . , 2012 ) . Modern mammals have separated the middle ear from the jaw in adults , and the ossicles ( malleus , incus and stapes ) are now suspended by ligaments from the cranial base to allow free vibration during sound transmission from the ear drum to the inner ear ( Figure 1B ) . Paleontological evidence indicates that the evolution of the definitive mammalian middle ear ( DMME ) occurred at least twice , once in the lineage that gave rise to monotremes and once in the therian ( marsupial and eutherian ) mammals ( Meng et al . , 2018; Rich et al . , 2005 ) , while new developmental data suggests that the two groups of therian mammals may have each independently acquired the DMME ( Urban et al . , 2017 ) . Here , we refer to eutherian mammals rather than placental mammals , as marsupials have a yolk-sac placenta ( Renfree , 2010 ) . Marsupials ( Allin , 1975; Filan , 1991 ) and monotremes ( Griffiths , 1978 ) , exhibit extreme altriciality , greater than is seen in any eutherian ( Werneburg et al . , 2016 ) . This has profound consequences for early feeding as the bones that form the mammalian jaw joint , the dentary and squamosal , have not fully ossified by the time of birth/hatching . The dentary-squamosal joint forms prior to birth in eutherian mammals , and begins to function in the embryo ( Habib et al . , 2007; Jahan et al . , 2014 ) . In the mouse , gestation is approximately 20 days , with breakdown of Meckel’s cartilage , to separate the lower jaw from the ear bones , following during early postnatal stages ( Anthwal et al . , 2013 ) . In contrast , the opossum Monodelphis has a short gestation of just 13 days ( Keyte and Smith , 2008 ) , and is born before development of the dentary-squamosal articulation , which forms between 14 and 20 days after birth ( Filan , 1991; Maier , 1987 ) . Monotremes hatch out of the egg after 10 days post-oviposition ( Griffiths , 1978 ) . The formation of the dentary-squamosal joint in monotremes has recently been followed and shown to form from 10 days after hatching in the platypus ( Anthwal and Tucker , 2020 ) . Breakdown of Meckel’s cartilage in both marsupials and monotremes occurs relatively late postnatally ( Urban et al . , 2017 ) , with a robust Meckel’s still evident in nest young platypuses ( Zeller , 1993 ) . There is , therefore , a significant gap between birth and the advent of a functional mammalian jaw joint in both marsupials and monotremes . The feeding strategies of new-born mammals vary in extant members of each group of mammals . Compared to eutherian mammals , marsupials rely on placental support for a relatively short period of time ( Renfree , 2010 ) and consequently receive the nutrition required for their development via a lengthy and sophisticated lactation ( Tyndale-Biscoe and Janssens , 1988; Tyndale-Biscoe and Renfree , 1987 ) . During their early postnatal life marsupials attach to the mother’s teat and use the comparatively early developed tongue musculature to suck ( Smith , 1994 ) . In the grey short-tailed opossum , Monodelphis domestica , pups are born after 13 days of embryonic development , which is followed by around 14 days permanently attached to the mother’s teat , after which they detach intermittently from the mother but continue to suckle . Weaning occurs around postnatal day 60 ( Keyte and Smith , 2008 ) . In contrast to therian mammals , young monotremes do not obtain milk in quite the same way as therian mammals due to the absence of teats in the mother ( Griffiths , 1978 ) . Instead young monotremes suck up milk vigorously from the flattened but protuberant nipple-like areola on the mother’s abdomen ( Griffiths , 1978 ) . In the case of echidnas , these areolae are within the pouch . Given the lack of a jaw joint at birth , it has been proposed that marsupials and monotremes use the connection between the middle ear bones and cranial base to permit feeding prior to the formation of the articulation between the dentary and squamosal and cavitation of the middle ear ( Crompton and Parker , 1978; Maier , 1987; Sánchez-Villagra et al . , 2002; Zeller , 1993 ) . To investigate this idea further , we have analysed the articulations that link the lower jaw to the cranial base ( cranio-mandibular joints ) in monotremes ( platypus Ornithorhyncus anatinus and short-beaked echidna Tachyglossus aculeatus ) as they develop from hatching , and compare them to a marsupial ( grey short-tailed opossum , Monodelphis domestica ) , and a eutherian ( mouse , Mus musculus ) , with additional comparison to the gecko , guinea pig and bat . We show that in early post-hatching life the monotreme incus and cranial base fuse , and later form an articulation , creating a double cranio-mandibular articulation , similar to the jaw anatomy of fossil mammal-like reptiles . This close association of the incus and cranial base is also observed at embryonic stages in eutherians and is reflected in mouse cell lineage studies . In contrast , opossums at birth utilise a cushion of extra-cellular matrix-rich mesenchyme in between the incus and petrosal to provide an articulation point . Marsupials and monotremes , therefore , have different strategies for coping with an early birth . Our research suggests that the incus retains a transient lower jaw support role across extant mammals but at different stages of pre and postnatal development .
It has been suggested that the joint between the malleus and incus might act as the jaw joint early on in marsupial postnatal development , thereby recapitulating the reptilian function of these bones in mammals ( Müller , 1968; Crompton and Parker , 1978 ) . Alternatively , it has been suggested that the actual articulation point in marsupials is between the incus and the cranial base ( Maier , 1987; Sánchez-Villagra et al . , 2002 ) . Less information is available regarding monotreme development , however , the incus has been described as being in cartilaginous connection with the cranial base during early post hatching development ( Watson , 1916; Zeller , 1993 ) . The development of the malleus and incus , and incus and cranial base , was therefore investigated across the three groups of mammals , with the gecko as an outgroup . In the ocelot gecko ( Paroedura picta ) , the quadrate and articular ( the homologous elements to the incus and malleus respectively in non-mammal amniotes ) form a clear synovial joint in the embryo at mid-gestation ( Figure 2A ) . In mice ( Mus musculus ) , the malleus and incus are initially formed from a single cartilaginous condensation that separates , by the formation of a joint , at Embryonic ( E ) day 15 . 5 ( Amin and Tucker , 2006 ) . At birth , therefore , the incus and malleus are evident as distinct cartilages ( Figure 2B ) . In Monodelphis domestica , the malleus and incus are still connected at birth at the dorsal end by a ridge of cartilage ( Filan , 1991; Figure 2C ) . We observed a similar connection between the malleus and incus in the echidna ( Tachyglossus aculeatus ) just after birth . Like the opossum , the middle ear ossicles were fused dorsally , indicating that they function as a unit ( Figure 2D ) . These findings demonstrate that , like opossums , monotremes do not use the primary jaw joint as the craniomandibular articulation before the development of the dentary-squamosal joint . We therefore investigated the relationship between the incus and the petrosal in the cranial base in mice , opossums , platypus and echidna , comparing the interaction to the developing joint between the quadrate and opisthotic in embryonic geckos . In many reptiles , as shown in the gecko , the quadrate ( incus homologue ) forms a synovial joint with the opisthotic ( also known as the otoccipital ) in the cranial base during embryonic development ( Figure 2E ) . The opisthotic/otoccipital is architecturally equivalent to the petrosal of mammals . In mice , the crus breve ( short process ) of the incus nestled in a fossa created by the crista parotica of the petrosal , but was separated by a region of mesenchymal cells , highlighting the lack of a clear articulation point between the two elements ( Figure 2F ) . The incus at birth , therefore only articulated with the adjacent middle ear bones , the malleus and stapes . Similar to the mouse , the crus breve in neonatal opossums , fitted into a fossa created by the crista parotica , but abutted the petrosal on the inferior aspect of the crista parotica ( Figure 2G ) . The incus and petrosal were therefore positioned much closer than in the mouse . The relationship between the incus and crista parotica in the two monotreme species was significantly different from the therian mammals . In both platypus ( Ornithorhynchus anatinus ) and echidna ( Tachyglossus aculeatus ) , the incus appeared to be fused with the crista parotica at birth ( Figure 2H , I ) , agreeing with Watson , 1916 . The lower jaw , via Meckel’s cartilage , would therefore be physically connected to the upper jaw , via the incus at this timepoint . 3D reconstructions of the incus , malleus and petrosal , showing the relationship of these different elements in the different species is shown in Figure 2—figure supplement 1 . The relatively small size of the incus in both monotremes is striking , as is the extended and tapered crus breve of the incus in the opossum . To investigate the monotreme relationship between the incus and crista parotica further we followed development of these two cartilages from birth to functional use of the dentary-squamosal joint , but before complete cavitation of the middle ear space . Due to the scarcity of available specimens very little is known about monotreme ear and jaw development . In adult platypuses , the incus appears in contact with the crista parotica , forming a fibrous articulation ( Zeller , 1993; Luo and Crompton , 1994 ) . Similarly , in the adult echidna , the incus has been described as tightly attached to the petrosal ( Aitkin and Johnstone , 1972 ) . At 2 days and 6 . 5 days the platypus incus was fused to the crista parotica by immature chondrocytes ( Figure 3A , B ) . Between 10 days and 30 days the connection was difficult to make out , with the two cartilages almost completely integrated together ( Figure 3C , D ) . Strikingly , by 80 days , when the dentary-squamosal joint would have started to become functional , the incus and crista parotica were no longer fused , with the two distinct cartilages abutting each other ( Figure 3E ) . At this stage , in contrast to the other stages investigated , the ear ossicles and petrosal had begun to ossify . However , the regions forming the malleus-incus joint , and the incus-petrosal articulation remained cartilaginous . A cartilaginous articular surface between the incus and petrosal was maintained at 120 days , a period when the young would have started to leave the burrow ( Figure 3F; Holland and Jackson , 2002 ) . A similar move from early fusion , to articulation was observed in the echidna ( Figure 3F–J ) . No evidence of a synovial capsule , however , was identified at any stage . The fusion of the incus and crista parotica coincides with the period when the young would have been feeding from milk , while the move to an articulation was associated with periods when the dentary-squamosal was fully formed and functional . After separation of the incus and petrosal , there was a period where two cranial-mandible articulations were evident in the platypus - between Meckel’s cartilage and the petrosal , via the malleus and incus , and between the dentary and squamosal ( Figure 3—figure supplement 1C ) . The chain of elements linking Meckel’s cartilage to the petrosal in the platypus is shown in Figure 3—figure supplement 1A , B , D at 50 days post-hatching . Middle ear cavitation occurred very late in the monotreme specimens analysed , with only the 120 day platypus showing partial cavitation around the hypotympanum , but this did not extend upwards to where the ossicles are housed . Hearing , thus , must be a very late developing sense in the platypus . Limited expression analysis has been performed in monotremes , with no previous expression data performed in the ear or jaw during development . In order to further understand the change in the relationship between the incus and petrosal , immunohistochemistry staining was carried out in echidna samples 0 and 3 days post hatching . In the fused incus-petrosal region of 0-day-old echidna ( Figure 4A ) , the expression of both a master regulator of cartilage development , Sox9 , and a principal component of cartilage extra cellular matrix , Collagen Type 2 , were continuous between the incus and the crista parotica of the petrosal , as well as between the incus and the malleus ( Figure 4B ) . Since the connection between these elements is lost later in post-hatching development , IF for beta-catenin was carried-out . Nuclear localised beta-catenin is a readout of canonical Wnt signalling , and is known to negatively regulate chondrocytes differentiation and promote joint formation ( Hartmann and Tabin , 2001 ) . Few beta-catenin positive cells were observed within the cartilage of the middle ear and petrosal at 0 days , though beta-catenin was strongly expressed in the neuro-epithelium of the inner ear ( Figure 4C ) . At post-hatching day 3 , the incus and crista parotica were still fused , although the cells joining the two elements resembled fibrocartilage or immature chondrocytes ( Figure 4D ) . Expression of Sox9 was still strong and continuous throughout all elements ( Figure 4E , E’ ) , however Collagen Type 2 expression was weaker in the fusion region ( Figure 4E , E” ) , possibly indicating a change in cartilage type from hyaline cartilage to fibrocartilage . Interestingly nuclear beta-catenin , suggestive of active Wnt signalling , was observed in two stripes , in the chondrocytes between the incus and petrosal , and within the malleus-incus joint , indicating suppression of cartilage fate in these regions ( Figure 4F ) . Upregulation of Wnt signalling between the incus and petrosal therefore , may play a role in formation of a joint between these two , initially fused , structures . While the fusion between the incus and petrosal in echidna and platypus could be explained by the evolutionary distance between monotremes and therian mammals , it has also been suggested that the incus is transiently attached to the cranial base in 7-week-old human fetuses ( Rodríguez-Vázquez et al . , 2018 ) . This suggests that the potential for fusion may be a default state in mammals . In order to examine this , we next undertook fate mapping experiments in the mouse , and investigated the relationship between the incus and petrosal in other eutherian mammals during embryonic development . Sox9 expressing cells were fate mapped by tamoxifen induction at E14 . 5 in Sox9CreERT2; tdTomato mice , which were then collected at P0 ( Figure 5A ) . At this stage Sox9 ( green ) was expressed in the petrosal and incus and suspensory ligaments , overlapping with the red fluorescent Protein ( RFP ) marking the Sox9 lineage cells . In addition , the red Sox9 lineage cells were found in the Sox9 negative mesenchymal cells , in the gap between the petrosal and incus ( Figure 5A ) . A pre-cartilaginous bridge is therefore evident in the mouse between the incus and the crista parotica . Next , expression of Sox9 was investigated at E14 . 5 . The incus , and the crista parotica are both neural crest derived ( O'Gorman , 2005; Thompson et al . , 2012 ) , while the rest of the petrosal is mesodermal . We therefore looked at the expression of Sox9 ( red ) in Mesp1Cre;mTmG mice , where mesoderm-derived tissue can be detected by anti-GFP IF ( Figure 5B ) . Since tissue processing and wax embedding removes endogenous fluorescence , the membrane RFP that is expressed in the non-mesodermal tissue of Mesp1Cre;mTmG mice was not detectable in these slides . Consequently , all red signal was Sox9 immunofluorescence staining . Sox9 protein was expressed continuously between the incus and the petrosal . The incus Sox9 expression domain was continuous with the expression domain of the neural crest -derived crista parotica , which in turn was fused to the mesodermal portion of the petrosal . Since the incus does not fuse with the petrosal in the mouse , despite the expression of Sox9 between the elements , we next looked at the mRNA expression of joint markers Gdf5 and Bapx1 between the incus and petrosal of mice by in situ hybridisation ( Figure 5C–E; Storm and Kingsley , 1999; Francis-West et al . , 1999; Tucker et al . , 2004 ) . Gdf5 was expressed in the mesenchyme between the incus and petrosal , as well as in the malleus-incus joint ( Figure 5D ) . Bapx1 , which specifies both the malleus-incus joint and the quadrate-articular joint ( Tucker et al . , 2004 ) , was not expressed in between the incus and the petrosal ( Figure 5E ) . In the mouse , therefore there is a potential for the incus and crista parotica to fuse but they are prevented from doing so by the upregulation of the joint marker Gdf5 . Very close associations between the incus and crista parotica during development were also observed in other eutherian mammals via PTA stained microCT ( see bat in Figure 5—figure supplement 1 ) , suggesting that interactions between these two elements are observed as a feature prenatally in eutherian mammals , similar to post-hatching monotremes . The function of this prenatal connection between the upper and lower jaw is unclear but may act as a brace to buffer movement during this period . Next we investigated the articulation between the incus and petrosal observed in the developing opossum . It was originally suggested that the marsupial incus forms a joint with the crista parotica ( Maier , 1987 ) , although this was disputed in Monodelphis ( Filan , 1991 ) . Although this latter paper found no evidence of a joint they did show the mesenchyme between the crista parotica and incus as being condensed ( Filan , 1991 ) . We therefore investigated the extra cellular matrix ( ECM ) components of the mesenchyme surrounding the opossum incus in more detail ( Figure 6 ) . It was noted that mesenchyme surrounding the crus breve and superior portion of the body of the incus had a more intense staining with alcian blue compared to those regions around the inferior border of the incus and the other ossicles ( Figure 2C , G ) . This pattern was observed throughout ossicle development ( Figure 6A–C ) . In order to further characterise the differences in the ECM in the different regions of the middle ear mesenchyme , immunohistochemistry for versican was carried out . Versican is a large proteoglycan with side chains of glycosaminoglycans ( GAGs ) , such as hyaluronic acid ( HA ) . Proteoglycan complexes act to attract water , and are held in place by collagen fibres to stiffen the matrix in hyaline cartilage , and act to lubricate articular cartilage ( Wu et al . , 2005 ) . Versican is required during the initial condensation of mesenchyme but is absent from mature cartilage , where aggrecan is expressed ( Kamiya et al . , 2006 ) . Versican expression is maintained in the joint region during limb cartilage development , acting to inhibit maturation of the mesenchyme to form cartilage ( Choocheep et al . , 2010; Snow et al . , 2005 ) . Versican was strongly expressed in the mesenchyme surrounding the short arm of the incus at 5 days , 10 day and 27 days , correlating with the region of strong alcian blue expression ( Figure 6D–F ) . The high level of versican around the crus breve therefore suggests a role for the ECM in providing a buffering function in this region . Cell density of the mesenchyme was measured in regions with strong alcian blue/versican staining and compared against the cell density of regions with low alcian blue/versican staining . Unpaired two-tailed t-test demonstrated that the regions with high alcian blue had a significantly higher ( p=0 . 0152 ) cell density than those regions with lower alcian staining ( Figure 6G ) . Versican is processed by ADAMTS family members for clearing and remodelling ( Nandadasa et al . , 2014 ) . While the full-length form of versican is thought to have a structural role , the cleaved form has an active role in signalling , influencing morphogenesis and tissue remodelling ( Nandadasa et al . , 2014 ) . Interestingly when we analysed the cleaved form of versican , using antibodies against DPEAAE , the expression was largely reciprocal to that of uncleaved versican , with lower levels specifically around the crus breve ( Figure 6—figure supplement 1A ) . This suggests that versican around the incus is protected from cleavage allowing it to maintain its structural role . The lack of cleaved versican around the crus breve , suggests the lack of a signalling role in this region , in agreement with the low level of expression of CD44 , a cell surface receptor and binding partner of versican-hyaluronan complexes . CD44 was not associated with the mesenchyme around the crus breve , but was instead restricted to the perichondrium of the cartilaginous elements and periosteum of the skeletal elements of the ear ( Figure 6—figure supplement 1B ) .
The incus of adult mammals plays a key role in hearing . Our data here suggest that the incus also plays a transient role supporting the lower jaw against the cranial base during both marsupial and monotreme postnatal development . The role of the incus and the points of jaw articulation are summarised in Figure 7A . The incus and petrosal were found to be fused at hatching in both monotremes . During this early fusion period , the puggle would be feeding exclusively on milk and Meckel’s cartilage could therefore act as a flexible elastic strut to facilitate jaw movement ( Zeller , 1993 ) . Interestingly , a potential role of the ear ossicles in jaw support was also observed in eutherians during prenatal development . Fate mapping and gene expression studies in mice indicated that the crus breve of the incus and the crista parotica were formed from a continuous region of Sox9 expressing chondrogenic cells ( Figure 5A , B ) , separated by expression of the joint marker Gdf5 ( Storm and Kingsley , 1999; Figure 5C ) . Furthermore , the incus and cranial base temporarily fuse during the development of the human middle ear region ( Rodríguez-Vázquez et al . , 2018 ) , and abut during bat development ( Figure 5—figure supplement 1 ) . Together these data indicate that the relationship of the incus to the cranial base is not a derived feature of monotremes , and that the common mammal-like reptile ancestors of both monotremes and therian mammals may have formed an articulation between the quadrate/incus and petrosal through fusion of the elements followed by joint formation though Wnt and Gdf5 signalling . The current study indicates that the first pharyngeal arch-derived incus forms a continuous field of chondrocytes with the second arch-derived crista parotica , which in turn is fused with the mesoderm-derived body of the petrosal . The borders between these developmentally distinct populations are , therefore , not always reflected by the mature anatomy . For young monotremes and marsupials , the middle ear must function as part of the mandible postnatally until the dentary-squamosal bones have formed . This is similar , but not identical to the situation in cynodont ancestors of mammals . In these animals , the quadrate/incus articulated with a number of cranial elements , including the quadratojugal , to stabilise the jaw articulation . These connections and many elements like the quadratojugal have been lost in extant mammals in order to free the incus and increase its mobility during sound transmission ( Luo and Crompton , 1994 ) . The mechanical requirements for feeding placed upon the middle ears in monotremes and marsupials during early life have resulted in the fusion of the incus and petrosal in monotremes , and the elongated contact supported by a proteoglycan matrix in marsupials . These adaptations allow for stabilisation of the middle ear before the development of the dentary-squamosal joint and separation of the middle ear from the mandible , but do not compromise the effectiveness of the middle ear in later life . The changing connections between the middle ear ossicles and the cranial base in the different groups are highlighted in Figure 7B . The crus breve of the incus is elongated in the developing opossum compared with other species analysed ( Figure 2—figure supplement 1 ) . In order to feed by suckling in the absence of a dentary-squamosal joint we propose that this anatomy allows for an increased surface contact with the cranial base during postnatal development , which , in combination with the proteoglycan-rich surrounding mesenchyme , acts to stabilise the mandible against the rest of the head . It is noted that many adult marsupials have a relatively elongated crus breve of the incus compared to eutherian species , for example the bare-tailed woolly opossum Caluromys philander , and the grey short-tailed opossum , Monodelphis domestica ( Sánchez-Villagra et al . , 2002 ) . Even when eutherian mammals have a longer crus breve , such as in Talpid moles , the process is thinner and more finger-like compared to that of marsupials ( Segall , 1973; Segall , 1970 ) . This may be a consequence of the developmental requirement for an elongated short process to facilitate feeding before the development of the mature mammalian jaw articulation . In the majority of adult marsupials , including Monodelphis , the incus is suspended from the cranial base by suspensory ligaments , and the crus breve extends into a fossa . One interesting exception is the marsupial mole , in which the crus breve has a connective tissue attachment to a lamella on the petrosal ( Archer , 1976 ) . This results in the middle ear ossicles being affixed to the cranial base , an adaptation to a fossorial niche found in other mammals such as in true moles . In light of the current study , the absence of an incudal fossa in the marsupial mole may be interpreted as a retention of the juvenile petrosal morphology ( paedomorphy ) . In adult non-mammalian amniotes the homologue of the incus - the quadrate - and cranial base are strongly attached by fibrous syndesmoses or cartilaginous synchondroses ( Payne et al . , 2011 ) , and we show that a synovial joint appears to form in geckos during development ( Figure 2 ) . In the neonatal opossum neither type of connection is observed . In neonatal marsupials Sánchez-Villagra and colleagues describe the connection between the incus and petrosal as being an ‘immature syndesmosis’ , which acts as a ‘supportive strut’ during sucking ( Sánchez-Villagra et al . , 2002 ) . In the current study , we demonstrate a specialised condensed mesenchyme surrounds the incus of opossum postnatal juveniles . We show that this condensed mesenchyme is rich in the proteoglycan versican ( Figure 6 ) . In contrast expression studies in human foetuses demonstrate that versican is restricted to the perichondrium of Meckel’s cartilage ( Shibata et al . , 2014; Shibata et al . , 2013 ) , with high hyaluronic acid levels within the joints but not surrounding the incus ( Takanashi et al . , 2013 ) . This concentration of versican around the crus breve therefore may be a feature of Monodelphis , and perhaps marsupials in general . The versican-rich mesenchyme may act to either stabilise the incus by increasing the tension of the surrounding mesenchyme during feeding , ‘lubricate’ the articulation between the incus and cranial base by increasing the hydration of the ECM , or both . In keeping with this role , versican is dynamically expressed at the pubic symphysis during pregnancy in mice ( Rosa et al . , 2012 ) , during which time the mouse pubic symphysis forms a fibrous joint or syndesmosis ( Ortega et al . , 2003 ) . Significantly , there is little cleaved versican ( DPEAAE ) around the crus breve of the incus , suggesting a mechanical , rather than a signalling role ( Figure 6—figure supplement 1 . A ) . Overall it is likely that this mesenchyme is supporting the incus , rather than enabling mobilisation , with the high level of uncleaved versican acting to increase fibroviscocity while also elevating hydration of the ECM . In this way , the mesenchyme around the incus acts as a cushion during the mechanical stress of suckling . Meckel’s cartilage persists to at least 50 days post-hatching in the platypus . At this timepoint , juvenile monotremes have two connections between the lower and upper jaw . The first connection is through the middle ear , which in juveniles remains attached to the mandible and articulates with the cranial base via the incus . The second is the later developing novel mammalian jaw joint . Only much later in the life of the young does it appear that the connection between the middle ear and mandible is lost , and the malleus and incus act as a DMME . The connection of the incus to the cranial base appears to be maintained in the adult echidna and platypus ( Luo and Crompton , 1994; Aitkin and Johnstone , 1972 ) . This would be expected to impact on the movement of the incus , and therefore the efficiency of hearing , reflected in the poor hearing reported for monotremes ( Aitkin and Johnstone , 1972; Gates et al . , 1974 ) . This novel finding of a double cranial articulation in the juvenile has significant implications for the evolution of the middle ear and jaw joint in mammals . Fossil evidence indicates that mammalian ancestors had a persistent connection between the middle ear ossicles and the jaw , as evidenced by the presence of an ossified Meckel’s element , or a dentary groove and post dentary trough , supporting a persistent Meckel’s cartilage ( Luo , 2011; Rich et al . , 2005; Urban et al . , 2017 ) . For these animals , the connection of the middle ear with the jaw took one of two forms , in each case the mammalian secondary jaw joint was present . The first was a more basal mandibular middle ear where the incus and malleus were firmly attached to the cranial base and dentary respectively . More derived fossils had a partial , or transitional mammalian middle ear ( PMME or TMME ) , where the middle ear was medially inflected away from the dentary , presumably allowing for improved vibration , but the malleus was still connected to the jaw , via Meckel’s cartilage ( Luo , 2011 ) . In these fossils with a PMME , little is understood of the rear of the ossicular chain , where the incus meets the petrosal , due to the poor and rare preservation of middle ear ossicles in the fossil record , a consequence of their small size . For example , only recently has a multituberculate with a complete incus been described ( Wang et al . , 2019 ) . Our data suggest that even in these transitional mammals with a PMME , the incus would have still articulated with the cranial base via the crista parotica , at least at some point during the animal’s life history . The DMME appears to have evolved independently in monotremes and therian mammals ( Rich et al . , 2005 ) . Due to the absence of evidence we do not know if the incus articulation in animals with a PMME varied in a lineage specific manner , with the therian lineage resembling juvenile marsupials , and monotremaformes resembling juvenile platypuses and echidna , or if both lineages had similar articulations . The data from transgenic reporter mice ( Figure 5 ) , along with data from humans ( Rodríguez-Vázquez et al . , 2018 ) and non-model therians ( Figure 5—figure supplement 1 ) suggests that the monotreme-type fusion and articulation of the incus with the cranial base may have been common in mammal like-reptiles . As such , the developing monotreme , with a double jaw articulation and a fused or articulated incus and petrosal , provides an exciting model for the study of the developmental basis of mammalian evolution .
Opossum ( Monodelphis domestica ) tissue was collected as previously described ( Anthwal et al . , 2017; Urban et al . , 2017 ) . Archival platypus ( Ornithorhynchus anatinus ) and short-beaked echidna ( Tachyglossus aculeatus ) slides were imaged from the collections at the Cambridge University Museum of Zoology , and the Hill Collection , Museum für Naturkunde , Leibniz Institute for Research on Evolution and Biodiversity , Berlin . Details of samples imaged are in Table 1 . All museum samples have been studied in previously published works ( Green , 1937; Presley and Steel , 1978; Watson , 1916 ) . Stages for platypus are estimated based on Ashwell , 2012 . Staging of echidna H . SP EC5 and H . SP EC4 are estimated by cross-referencing ( Griffiths , 1978; Rismiller and McKelvey , 2003 ) . Post-hatching day 0 to 3 echidna samples were collected by Marilyn Renfree and Stephen Johnston . Wildtype and Mesp1Cre; mTmG were kept at the King’s College London Biological Services Unit . Sox9CreERT2:tdTomato embryos were a gift of Prof Robin Lovell-Badge and Dr Karine Rizzoti at the Francis Crick Institute , London . Phosphotungstic acid ( PTA ) contrasted embryonic Pterobnotus quadridens bat µCT scans were provided by Prof Karen Sears and Dr Alexa Sadier at the University of California Los Angeles . Guinea pig ( Cavia porcellus ) displays samples were collected as previously described ( Anthwal et al . , 2015 ) . Gecko and mouse samples were investigated during embryonic development ( 35 days post oviposition ( dpo ) and E16 . 5 respectively ) . The gestation for geckos is around 60 days , and mice have a gestation of 20–21 days . Much of opossum and echidna development occurs during early post-gestation/hatching life , including formation of the dentary-squamosal joint , and so 4-day-old opossums , and 0- to 3-day-old echidnas were investigated before the onset of this joint . All culling of mouse , opossum , guinea pig and reptile tissue followed Schedule One methods as approved by the UK Home Office and was performed by trained individuals . Use of genetically modified mice was approved by the local GMO committee at King’s , under personal and project licences in accordance with the Animal ( Scientific Procedures ) Act of 1986 , UK . All tissues for histological sectioning were fixed overnight at 4°C in 4% paraformaldehyde ( PFA ) , before being dehydrated through a series of graded ethanol , cleared with Histoclear II , before wax infiltration with paraffin wax at 60°C . Wax-embedded samples were microtome sectioned at 8 µm thickness , then mounted in parallel series on charged slides . For histological examination of bone and cartilage , the slides were then stained with picrosirius red and alcian blue trichrome stain using standard techniques . For immunofluorescence staining slides were rehydrated through a graded series of ethanol to PBS . Heat induced antigen retrieval was carried out by microwaving the samples for 10 min in 0 . 1M Sodium citrate pH6 buffer . Slides were then blocked in 1% Bovine serum albumin , 0 . 1% cold water fish skin gelatine , 0 . 1% triton-X for 1 hr . Sections were then treated over night at 4°C with primary antibodies . The following primary antibodies were used , rabbit anti Sox9 ( Chemicon ) at a dilution of 1/200 , chicken anti GFP ( Abcam ) at a dilution of 1/500 , rat anti RFP ( Chromotek ) at a dilution of 1/200 , Rabbit anti Beta-catenin ( Santa Cruz ) 1/200 , mouse anti type 2 collagen ( DSHB ) at 1/50 , mouse anti CD44 ( DSHB ) at 1/50 , mouse anti Tenascin C ( DSHB ) at 1/40 , mouse anti versican ( DSHB ) at 1/50 , rabbit anti versican V1 ( Abcam ) at 1/400 . Following repeated PBS washes , secondary antibodies were added . For fluorescent labelling the following antibodies were used at 1/300: Alexa568 conjugated Donkey anti-Rabbit , Alexa 488 conjugated Donkey anti-Rabbit , Alexa568 conjugated Donkey anti-Mouse , Alexa568 conjugated Donkey anti-Rat , Alexa488 conjugated Donkey anti-Chicken ( all Invitrogen ) . Secondary antibodies were added in the blocking buffer for 1 hr at room temperature in the dark . The secondary antibody was then washed off with PBS , and the slides mounted with Fluroshield mounting medium containing DAPI ( Abcam ) . Sections were visualised by Leica SP5 confocal microscopy . For Versican and CD44 slides , secondary biotinylated goat anti-mouse antibody ( Dako ) was added to the slides 1/400 in blocking buffer . Slides were then washed in PBS before being treated with ABC-HRP streptavidin kit ( Vector Labs ) , and then revealed with DAB ( Vector Labs ) . Monotreme immunofluorescence staining was carried out in technical replicates due to the rare nature of the samples . Mouse and opossum analysis was carried out in biological triplicates . Radioactively labelled antisense RNA probes were made against mouse Gdf5 and Bapx1 mRNA , and radioactive in situ hybridisations were carried out to detect the expression of these genes in sagittal plain cut sections of wildtype mice , as previously described ( Tucker et al . , 2004 ) . All in situ staining was carried out in biological replicates . Three‐dimensional reconstructions of middle ear and surrounding cranial base cartilages were generated from serial histology images in FIJI ( ImageJ 1 . 47 v ) , using the Trackem2 Plugin ( Schindelin et al . , 2015; Schindelin et al . , 2012 ) . Cell density was counted in a DAPI stained sections of 5-day-old opossums ( n = 3 ) . In FIJI , 20 separate 80 µm2 fields were randomly placed across the mesenchyme surrounding the incus across 5 sections . The total number of nuclei were counted if they were located wholly within the field , or where more than 50% of the nuclei intersected the upper or right hand margin of the field . Next by looking at parallel alcian blue stained sections , the fields were scored as being in proteoglycan-rich or weak regions . Two fields were ambiguous , and so were removed from the analysis . The user did not know the proteoglycan status of the field at the time of counting . Next the mean cell number in each field was calculated in the remaining 8 proteoglycan-rich ( alcian blue stained ) regions and 10 proteoglycan weak ( weak alcian blue stain ) regions , and compared by unpaired two-tailed students t-test in Prism statistical analysis software ( Graphpad ) , with p<0 . 05 . | The defining feature of all mammals is how the jaw works . Fish , reptiles and other animals with backbones have a lower jaw made of many bones fused together , one of which connects to the upper jaw . The lower jaw in mammals , however , is made of a single bone that connects with the upper jaw using a completely unique jaw joint . This new joint emerged as the ancestors of all mammals split from the reptiles around 200 million years ago . The bones that formed the original jaw joint ended up in the middle ear in mammals and switched to a role in hearing . Nowadays , there are three types of mammals: the placentals , marsupials and monotremes ( the egg laying mammals ) . In mice , humans and other placental mammals , the skeleton of the adult jaw joint forms in the embryo before birth . However , marsupials ( such as kangaroos and opossums ) and monotremes ( platypuses and echidnas ) are born at a much earlier embryonic stage , before the adult jaw joint has formed . It is therefore unclear how newborn marsupials and monotremes are able to move their jaws to feed on milk from their mother . Anthwal et al . compared how the jaw develops in mice , opossums , platypuses and echidnas before and after the adult jaw joint becomes functional . The experiments showed that young echidnas , platypuses and opossums use their middle ear bones to articulate the lower jaw with the head before the adult jaw joint forms . In young opossums , the ear bones form a cushion to support the jaw . In juvenile platypuses a double joint is evident , with the ear bones forming a joint at the same time as the newly formed adult jaw joint , similar to the situation observed in fossils of mammal ancestors . The experiments also indicated that mice and other placental mammals may potentially use their ear bones to support the jaw before birth . These findings shed light on why the ear and jaw have such a close connection in mammals . In humans , the ear and jaw bones are still connected by ligaments , explaining why trauma to the jaw joint can cause dislocation of the ear bones . Similarly , defects in the development of the jaw can impact the ear , such as in Treacher Collins Syndrome , where in some cases the jaw joint fails to form and the ear bones appear to try and take this role . Understanding how the ear and jaw evolved will help us understand why they look like they do and why a defect in one can have knock-on consequences for the other . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"evolutionary",
"biology"
] | 2020 | Transient role of the middle ear as a lower jaw support across mammals |
Nutritional regulation by ants emerges from a distributed process: food is collected by a small fraction of workers , stored within the crops of individuals , and spread via local ant-to-ant interactions . The precise individual-level underpinnings of this collective regulation have remained unclear mainly due to difficulties in measuring food within ants’ crops . Here we image fluorescent liquid food in individually tagged Camponotus sanctus ants and track the real-time food flow from foragers to their gradually satiating colonies . We show how the feedback between colony satiation level and food inflow is mediated by individual crop loads; specifically , the crop loads of recipient ants control food flow rates , while those of foragers regulate the frequency of foraging-trips . Interestingly , these effects do not rise from pure physical limitations of crop capacity . Our findings suggest that the emergence of food intake regulation does not require individual foragers to assess the global state of the colony .
Eusocial insects stand out in their ability to achieve collective regulation with no central control . Nutritional management in bees and ants is a compelling example . On the one hand , the colony as a whole displays high levels of collective regulation on the amount of food collected ( Howard and Tschinkel , 1980; Sorensen et al . , 1985; Cassill and Tschinkel , 1999 ) , on its nutritional composition ( Dussutour and Simpson , 2009; Cook et al . , 2010; Bazazi et al . , 2016 ) , and on its dissemination within the colony ( Anderson and Ratnieks , 1999; Sendova-Franks et al . , 2010; Greenwald et al . , 2015 ) . On the other hand , this regulation is achieved by individuals that react to their local environment . Food dissemination often relies on local trophallactic interactions in which liquid food , not fully digested , is regurgitated from the crop of one individual and passed mouth-to-mouth to another ( ( Hölldobler and Wilson , 1990 ) chapter 7 , page 291 and SI of ( Greenwald et al . , 2015 ) ) . Such distributed processes are characterized by intricate interaction networks that include significant random aspects ( Fewell , 2003; Pinter-Wollman et al . , 2011; Mersch et al . , 2013; Sendova-Franks et al . , 2010 ) which may hinder global coordination . How colonies manage to achieve tight nutritional regulation despite the difficulties that are inherent to a distributed process is not fully understood . An essential component in the nutritional regulation of any living system is the adjustment of incoming food rates to the current level of satiation ( Parks , 2012; Simpson and Raubenheimer , 1993; Josens and Roces , 2000 ) . To experimentally approach the principles behind such adjustments it is useful to observe the global process in which food accumulates in the system . Indeed , the dynamics of food accumulation in ant colonies have been a subject of interest for many years ( Wilson and Eisner , 1957; Markin , 1970; Howard and Tschinkel , 1980; Buffin et al . , 2009; Buffin et al . , 2012; Sendova-Franks et al . , 2010; Greenwald et al . , 2015 ) . These studies show that when introduced to a new food source , the levels of food stored within the colony display logistic dynamics . The logistic growth in the amount of accumulated food supports the notion that total food inflow is regulated by the amount of food already stored within the colony . The local origins of this global regulation are still not fully understood . To understand how global food flow regulation emerges from single ant behaviors one should consider the forager ants . These ants , which typically constitute only a small fraction of all workers , are the ones responsible for bringing food into the nest ( Oster and Wilson , 1978; Traniello , 1977 ) . Therefore , any change in the global inflow of food to the colony must be manifested in the rate at which foragers collect and deliver food . Accordingly , colonies can regulate the inflow of food by modulating foraging effort: for example , by varying the number of active foragers through recruitment ( Gordon , 2002 ) . Indeed , many studies on the regulation of foraging have focused on recruitment behavior and have shown that it correlates with the colony’s nutritional state ( Traniello , 1977; Seeley , 1989; Tenczar et al . , 2014; Cassill , 2003 ) . In this work , we explore a less studied aspect of food flow regulation , namely , changes in the behavior of already active foragers . Active foragers engage in repeated trips between the food source and the nest ( Traniello , 1977; Tenczar et al . , 2014 ) , where they use trophallaxis to deliver their food load to multiple recipients ( Seeley , 1989; Gregson et al . , 2003; Huang and Seeley , 2003; Traniello , 1977 ) . The rate at which a forager leaves the nest for her next trip as well as the amount of food that she manages to unload per trip provide potential regulators of the collective foraging effort . These regulators may be tied to the colony’s nutritional state through the experience of returning foragers when they unload in the nest . In this vein , it was shown that honeybee foragers experience longer waiting times between subsequent unloading interactions if the colony is satiated ( Seeley , 1989 ) , and it was suggested that they use this information to adjust their recruitment behavior . Most previous studies of individual forager behavior did not make direct connections between single ant rules and the global dynamics of food accumulation ( Seeley , 1989; Huang and Seeley , 2003; Gregson et al . , 2003 ) . Nonetheless , the observations and interpretations they present are consistent with a simple intuition for the origins of the observed logistic dynamics in the accumulation of food: Initially , when a scout from a hungry colony encounters food she commences a recruitment process in which the number of active foragers increases ( Greene and Gordon , 2007 ) . This positive feedback is followed by a delayed negative feedback that results from the increased difficulty of foragers to locate available recipients as the colony satiates ( Seeley , 1989; Seeley and Tovey , 1994; Buffin et al . , 2009; Sendova-Franks et al . , 2010 ) . A simple prediction follows: if foragers wait to unload their entire crop contents before leaving for their next foraging trip ( Gregson et al . , 2003; Traniello , 1977 ) then the frequency at which a forager exits the nest should gradually decrease as the colony satiates ( Buffin et al . , 2009 ) . Although the above intuition may seem complete , it has only little empirical support . Until recently , microscopic measurements of real-time individual crop loads and food-flows in single interactions were unavailable . As a result , existing explanations for different aspects of the foraging process rely ( either explicitly or implicitly ) on various assumptions . The foragers were assumed to unload their entire crop contents before leaving the nest ( Traniello , 1977; Gregson et al . , 2003; Buffin et al . , 2009 ) and use local experience to assess the colony’s nutritional state ( Seeley , 1989; Seeley and Tovey , 1994; Huang and Seeley , 2003 ) . The recipients were assumed to be either empty or full ( Sendova-Franks et al . , 2010; Seeley , 1989; Seeley and Tovey , 1994 ) and fill upon a single interaction with a forager ( Sendova-Franks et al . , 2010; Seeley , 1989; Seeley and Tovey , 1994 ) . As for the pattern of interactions between foragers and their recipients , it was assumed that in the nest a forager has a constant probability per unit time to interact with potential recipients ( Sendova-Franks et al . , 2010; Seeley , 1989; Seeley and Tovey , 1994 ) , that there is a formation of queues of returning foragers and available receivers ( Seeley , 1989 ) , and that interaction patterns are random ( Seeley and Tovey , 1994; Buffin et al . , 2009; Sendova-Franks et al . , 2010 ) . Relying on individual-level assumptions may be deceiving since multiple sets of microscopic rules can lead to similar macroscopic outcomes . For example , the slowing down of foragers’ unloading rates may stem from reduced rates of trophallactic interactions but can also be the result of smaller amounts of food transferred per interaction . Both will affect the global outcome similarly . To uniquely identify the micro-scale mechanisms of food inflow regulation and examine the assumptions outlined above , we tracked fluorescently-labeled food in crops of individually tagged ants ( Greenwald et al . , 2015 ) . This technology allowed for a non-intrusive study of the dynamics of food accumulation in ant colonies with a spatial resolution of single-ant crop loads and a temporal resolution sufficient to capture single trophallactic events . We thus present the missing experimental data on the crop contents of encountered ants , the amount of food transferred per interaction , the dynamics of forager unloading at different satiety states of the colony , and the amount of food in the foragers’ crops when they exit the nest . In the following sections we use these highly resolved measurements to quantitatively link the microscopic and macroscopic scales of food accumulation dynamics in ant colonies . We demonstrate how the global dynamics and the regulation of individual foraging effort rely on individual crop loads . Specifically , we delineate how individual crop loads affect a forager’s unloading rate as well as her decision to exit the nest for the next food collection trip . Our findings suggest a distributed regulation mechanism which does not require individual foragers to assess global , colony-scale variables .
Food accumulation dynamics were studied by introducing starved colonies of Camponotus sanctus ants to fluorescently-labeled food . Food was supplied ad-libitum to isolate the effects of colony satiation from the effect of resource availability on the inflow of food . As the colony replenished , we followed the traffic and storage of the food within the crops of individual ants using real-time fluorescent imaging ( Figure 1a , Video 1 and Video 2 , for details see Materials and methods ) . We found that the total amount of food in the colony gradually accumulated until saturation ( Figure 1b , ( Buffin et al . , 2009 ) ) . The level at which food saturated was defined post-hoc as the colony’s intake volume target . We define the ‘colony state’ at time t , denoted F ( t ) , as the total amount of food in the colony at time t divided by the colony’s intake target . The ‘colony state’ is thus a normalized measure of the colony’s satiety level , starting from F=0 when the colony is starved and gradually approaching F=1 as the colony approaches its target ( Figure 1b , black line ) . To enable tracing of the food flow process on the single-ant level , all ants were individually tagged ( Figure 1a , and see Methods , Experimental setup ) . As could be expected ( Gordon , 1989; Tenczar et al . , 2014 ) , this labeling showed that a few consistent foragers were accountable for the transfer of food from the source to the ants in the nest ( Figure 1b ) . This allowed us to study the dynamics of food accumulation by expressing colony state as the sum of the contributions of individual foragers: ( 1 ) F ( t ) =∑i=1Nfi ( t ) where fi ( t ) is the portion of the colony state contributed by forager i by time t , and N is the number of foragers . Collective food inflow ( dFdt ) , as well as the flow of food through each individual forager ( dfidt ) were derived by differentiating the measured colony state and the individual contributions with respect to time ( Figure 1c , Figure 1—figure supplement 2 , and Materials and methods , Data Analysis ) . This revealed that flows of food through each individual forager declined with increasing colony satiation state , F , and were roughly proportional to the available space in the colony , 1-F: ( 2 ) ∀i , dfidt≈m ( 1-F ( t ) ) where m is a constant ( Figures 1c and Figure 1—figure supplement 2 ) . This linear relationship holds for each forager , regardless of when she began foraging and is thus incompatible with feed-forward control in which a forager slows down as a function of her own history . Rather , it supports a mechanism by which the colony state feeds back on the food transfer rate of each individual forager . Breaking down the total inflow of food into individual forager contributions and using Equation 2 , we obtain: ( 3 ) dFdt=∑i=1n ( t ) dfidt≈n ( t ) ⋅m ( 1-F ( t ) ) where n ( t ) is the number of foragers that have begun foraging by time t . This formulation provides simple intuition for the non-monotonicity of food flow as apparent in Figure 1c . Specifically , an initial rise in collective inflow occurred when the number of foragers grew at a rate that overcame the rate of individual flow decay . Once the number of active foragers stabilized total flow rates declined linearly with colony state . Equation 3 describes a feedback process in which the rate of change in the colony state depends on the colony state itself , and more specifically - on 1-F , the space left to fill until the colony reaches its target . This is a direct consequence of individual foragers that deliver food at slower rates as the colony fills ( Equation 2 ) . However , the satiation state of the colony is a global factor that is , most likely , not directly available to individual ants . In the next section , we demonstrate how the observed feedback emerges from pairwise trophallactic interactions . The average food flow through a single forager ( dfidt ) can be estimated by the product of two macroscopic parameters that may depend on the colony state , F: her average interaction rate , ⟨r ( F ) ⟩ , and the average volume transferred per interaction , ⟨v ( F ) ⟩ . ( 4 ) dfidt≈⟨r ( F ) ⟩⋅⟨v ( F ) ⟩ While both the interaction rate and the interaction volume declined with increasing colony state , the change in the interaction volume was more prominent ( Figure 2a , Figure 2—figure supplements 1 and 2 ) . In fact , interaction volumes were nearly sufficient to account for the inflow dynamics , while the interaction rate introduced a minor second-order correction ( Figure 2b and Figure 2—figure supplement 3 ) . Importantly , interaction rates alone did not suffice to account for the inflow dynamics ( Figure 2b ) . Therefore , we turned to explore the local determinants that affect interaction volumes , under the assumption that interaction volumes are set locally depending on the states of the interacting individuals . The maximal potential volume of any given interaction is constrained by both the donor’s crop load and the available space in the recipient’s crop . To inspect the impact of each of these two local factors , we examined the distribution of all interaction volumes ( v ) from foragers to non-forager recipients for different ranges of crop loads , either of the recipient ( crecipient , Figure 3—figure supplement 1 ) or of the forager ( cforager , Figure 3—figure supplement 2 ) . We found that these distributions all follow an exponential probability density function ( PDF ) of the form: ( 5 ) p ( v|c ) =λce-λcvwhere p ( v|c ) is the conditional PDF of interaction volumes , v , given a crop load c , and c is either cforager or crecipient . We found that while the recipient’s crop load affected the distribution of interaction volumes ( Figure 3a ) , that of the forager had little effect , if any ( Figure 3b ) . Specifically , the distribution of interaction volumes scaled with the space left to fill in the recipient’s crop but was effectively independent of the forager’s crop load ( hence , hereafter we take c=crecipient ) : ( 6 ) λc=λ0C0-c where λ0=7 . 01 and C0=1 . 14 ( Figure 3a ) . C0 may be interpreted as the average crop load target to which recipients aim to ultimately fill ( expected to be close to 1 ) . On average , in an interaction with a forager the recipient receives 1λ0=0 . 14 of the space left to fill in her crop . Consistently , a linear fit relating mean interaction volume to recipient crop load ⟨v ( c ) ⟩=ac+b ( Figure 3—figure supplement 3 ) yields b≈-a≈0 . 13≈1λ0 . This is to be expected if ⟨v ( c ) ⟩=1λ0 ( 1-c ) . Indeed , normalizing interaction volumes by the total amount of available space in the recipient’s crop , v~=vC0-c , we find that all trophallactic volume distributions collapse onto a single exponential function p ( v~ ) =λe-λv~ with 1λ=0 . 12≈1λ0 ( Figure 3c ) . Simply put , the volume of an interaction can be estimated by a random exponentially distributed fraction ( v~ ) of the available space in the recipient’s crop ( C0-c ) . We can now move forward to express mean interaction volume , ⟨v⟩ , in terms of the colony state . Defining p ( v|F ) to be the conditional probability for an interaction of volume v when the colony state is F , the mean interaction volume ( at colony state F ) can be calculated by: ( 7 ) ⟨v ( F ) ⟩=∫vv⋅p ( v|F ) dv In light of our findings that interaction volumes change mainly with respect to the recipient’s crop load , we can decompose p ( v|F ) =∫cp ( v|c ) ⋅p ( c|F ) dc where p ( c|F ) is the probability density that the recipient will have a crop load of size c at colony state F . Equation 7 now becomes: ( 8 ) ⟨v ( F ) ⟩=∫cp ( c|F ) ∫vvp ( v|c ) dvdc The probability ( p ( c|F ) ) changed as the colony satiated and individual ants approached their targets . Figure 3d shows that the ants that interact with a forager reliably represent the satiation level of the colony and that the accuracy of this representation increases as the colony satiates . Altogether , substituting the microscopic interaction rule described by Equations 5 and 6 into the global summation described by Equation 8 demonstrates how the average interaction volume changes in proportion to the empty space in the colony ( 1-F ) : ( 9 ) ⟨v⟩=∫cp ( c|F ) ∫vλ0C0-ce-λ0C0-cvvdvdc=∫cp ( c|F ) C0-cλ0dc=1λ0 ( C0-⟨c ( F ) ⟩ ) =C0λ0 ( 1-F ) where the following identities were used: C0=∑ctargetN , F=∑c∑ctarget , ∫p ( c ) dc=1 and ∫p ( c ) cdc=⟨c⟩ . For each ant , ctarget signifies her crop load at colony satiation . The value of the multiplicative factor C0λ0≈0 . 16 stands in agreement with our experimental measurements ( Figure 2—figure supplement 2 ) . The above analysis demonstrates how the global inflow is determined by interaction volumes that are locally controlled by the recipient’s crop load , which on average represents the colony state . The forager’s crop comes into play in a different aspect of food inflow . Its finite capacity requires foragers to repeatedly leave the nest to reload at the food source in order to supply food to the entire colony . However , leaving the nest encompasses inevitable risks and energetic costs . Therefore , it is interesting to study whether foraging effort is also regulated , and if so , how it is expressed in decisions of individual foragers to leave the nest . To check whether foragers adjust their activity to the changing colony state we examined their behavior with respect to the accumulating food in the colony . The behavior of individual foragers was typically of a cyclic nature , as they alternated between two phases: ( 1 ) an outdoor phase , in which they filled their crops at the food source , and ( 2 ) an indoor phase , in which they distributed their crop contents in trophallaxis to colony members inside the nest ( Figure 4a ) . The frequency of these cycles ( ‘foraging frequency’ ) displayed a tight linear relationship with the available space in the colony ( 1-F ) , demonstrating that foraging effort is matched to the colony’s needs ( Figure 4b and Figure 4—figure supplement 1a , y=0 . 8 10-3+3 . 6 10-3 ( 1-F ) , R2=0 . 98 ) . Additionally , the increase in cycle times was mainly attributed to the prolonged indoor phase of the cycle , rather than the relatively constant outdoor phase ( Figure 4c and Figure 4—figure supplement 1b , Spearman’s correlation test , indoor phase: rs=0 . 77 , p<0 . 001 , outdoor phase: rs=0 . 13 , p=0 . 08 ) . This suggests that foraging frequency was regulated by the colony . To test for a causal effect of colony state on foraging frequency , we elicited an external perturbation on the colony state . Indeed , in experiments where the colony state was actively dropped by introducing new hungry ants after others had reached satiation ( see Methods , Perturbation Experiment ) , foragers’ durations in the nest sharply dropped as well ( Figure 4e and Figure 4—figure supplement 1 ) . This response generated a secondary rise in the amount of food in the colony , relaxing at a new value as durations in the nest gradually lengthened once again ( Figure 4d–e and Figure 4—figure supplement 2 ) . These experiments explicitly decoupled colony state from the time that passed since the initial introduction of food , and thus show that the colony state rather than forager history was the important factor that affects foraging frequency . Overall , these findings portray the following negative feedback process: foragers raise the colony state by bringing in food , while the colony state , in turn , inhibits their foraging frequency ( Figure 4f ) . A possible mechanism for this feedback might be that foragers do not exit the nest for their next foraging trip before they fully unload ( Traniello , 1977; Gregson et al . , 2003; Buffin et al . , 2009 ) . In this case unloading rates directly dictate the foraging frequency . This is consistent with the fact that both unloading rate and the foraging frequency are proportional to the total available space , 1-F . We explore this hypothesized mechanism in the next section . We find that foragers do not leave the nest only after they have fully unloaded ( Figure 5a ) . Nevertheless , the average amount of food in their crops when they exit remains nearly constant over different colony states ( Figure 5a , Spearman’s correlation test , rs=0 . 24 , p=0 . 001 ) . This constant averagesuffices in producing the observed relation between foraging and unloading rates as specified above . To maintain a relatively constant average crop state upon exit despite the declining unloading rates , foragers stayed longer in the nest ( Figure 4c ) and performed more interactions ( Figure 5c , and Figure 5—figure supplement 1a ) . They also actively explored deeper into the nest ( Figure 5d and Figure 5—figure supplement 1b ) . Surprisingly , even though the average crop load with which foragers exit the nest remains constant , this is not because the foragers unload a fixed amount in each visit . Rather , the crop loads with which foragers left the nest were highly variable ( Figure 5b ) . This raises the questions of when foragers decide to exit and how these decisions lead to exits at variable crop contents that , nevertheless , maintain a constant average over different colony states . Do factors other than their own crop load affect their decisions to exit , and more specifically , do the foragers use high-level information regarding the colony state ? To gain insight on the role of individual versus collective information in foragers’ decisions to exit , we were interested in a forager’s probability to exit the nest as a function of her own crop state ( crop ) and the colony state ( colony ) . We first estimated the probability for an individual forager to exit per time unit given her crop load and colony state R ( exit|crop , colony ) ( hereinafter , Rexit ) . We found that this exit rate was strongly dependent on both the forager’s crop state and the colony state ( Figure 6a , Table 1 ) . To understand which information the foragers require to generate this exit pattern we make the simplifying and common assumption that Rexit is an outcome of a Markovian decision process ( Robinson et al . , 2011; Sumpter et al . , 2012 ) , and can be treated as a product of two probabilities: ( 10 ) R ( exit|crop , colony ) =R ( makeadecision ) ⋅P ( decision=exit|crop , colony ) where R ( makeadecision ) is the probability of a forager to make a decision within a time unit ( hereinafter , Rdecide ) , and P ( decision=exit|crop , colony ) is her probability to decide to exit given her crop load and colony state when a decision is made ( hereinafter , P ) . Since the precise timings of an ant’s decisions are beyond our experimental reach , we replaced Rdecide by three assumed decision rates: ( 1 ) a constant decision rate , ( 2 ) a decision rate that is matched to the forager’s interaction rate , and ( 3 ) a decision rate matched to the forager’s unloading rate . Figure 6 shows the corresponding P for each decision rate . For a constant decision rate , P is proportional to Rexit and depends on both the forager’s crop state and the colony state ( Figure 6a , Table 1 ) . For a decision rate that is matched to the forager’s interaction rate ( e . g . a forager considers whether to exit only after an interaction has ended ) , P was calculated by considering only observations at ends of interactions as decision points . In this case , the effect of the colony state on P is present but smaller that the effect of the forager’s crop ( Figure 6b , Table 1 ) . Last , for a decision rate matched to the forager’s unloading rate , P was calculated by considering observations at fixed intervals of the forager’s crop load as decision points . For this case , the effect of the colony state on P approaches zero , such that P varies predominantly with the forager’s internal crop state ( Figure 6c–d , Table 1 ) . Since the probability to decide to exit , P , was effectively independent of the colony state when the rate of decisions , Rdecide , was adjusted to the unloading rate , we learn that the rate of exits Rexit can be decomposed into two functions with a clear separation of variables: ( 11 ) R ( exit|crop , colony ) =U ( colony ) ⋅G ( crop ) where U ( colony ) is linear in the unloading rate ( Equation 2 ) and G ( crop ) is a function of the crop that does not depend on the colony state ( Figure 6d ) . Interpreting this result from the perspective of the individual forager suggests a simple biological mechanism that may underly this separation of variables: In the course of unloading , the forager considers whether to exit or not each time she senses a sufficiently large change in her crop load , and then decides to exit based on her crop load alone . Since the rate at which her crop load changes is mainly affected by the recipients ( Figures 2 and 3 ) , the rate of her decisions is controlled by the colony ( U ( colony ) ∝1-F ) ; once the forager is triggered to make a decision , the decision itself depends on personal information alone ( G ( crop ) , Figure 6d ) .
On the scale of the entire colony , and in agreement with previous studies ( Buffin et al . , 2009; Sendova-Franks et al . , 2010 ) , we find that food accumulation follows logistic dynamics ( Figure 1b ) . Our individual-level measurements confirm that the logistic equation which describes the global dynamics can be interpreted as the product of two intuitive terms ( Equation 3 ) : the number of active foragers times the average unloading rate per forager . As previously speculated ( Buffin et al . , 2009 ) , the initial rise in the global food inflow stems from gradually joining foragers while its subsequent decay is the result of a negative feedback process wherein colony satiation levels work to decrease the unloading rates of individual foragers ( Figure 1c ) . We traced the mechanisms of this large-scale negative feedback to the immediate experience of individual foragers and specifically to the crop-loads of the ants they interact with . When a forager enters the nest she interacts with a ‘representative sample’ of recipient ants , i . e . ants whose crop load is , on average , proportional to the total satiation state of the colony ( Figure 3d ) . Further , in each such interaction the amount of food transferred is random but , on average , proportional to the available space in the recipient’s crop ( Figure 3a , c ) . Together , these findings imply that unloading rates are determined by the colony and directly proportional to the total empty space in the crops of the entire colony . Many sets of local rules could have yielded the same average flow , and thus would have been consistent with similar global dynamics . For example , previous studies have attributed the global negative feedback to the decreasing probability of a forager to encounter an accepting recipient , which delays the time until delivery ( Sendova-Franks et al . , 2010; Seeley , 1989; Seeley and Tovey , 1994; Cassill and Tschinkel , 1999 ) . However , due to experimental limitations these studies relied on the implicit assumption that recipients are satisfied by a single interaction , while in fact recipients may very well be partially satiated ( Huang and Seeley , 2003 ) . Our measurements on the level of single crops show that recipients are typically partially loaded , and the effect of their crop loads on interaction sizes is more dominant than the minor decrease of interaction rates in generating the collective negative feedback . Interestingly , partial crop loads do not affect the interaction volume merely by physical limitation: in most interactions the donor does not deliver her entire crop load , nor does the recipient fill up to her capacity . This finding contradicts the prevalent assumption used by those studies that did take partially loaded recipients into account ( Huang and Seeley , 2003; Gregson et al . , 2003 ) . These studies supposed that the amount of food transferred in an interaction is the maximal possible amount , and partial crop loads result from discrepancies between foragers’ loads and recipients’ capacities . Here we introduce explicit measurements of interaction volumes that reveal that exponentially distributed interaction volumes lead to partially loaded ants . This volume distribution concurs with the global feedback as it is scaled to the available space in the recipient’s crop . Feedback based on interaction volumes that are not set by physical limitations , rather than ‘all-or-none’ interactions , potentially allows individual ants to fine-tune their intake and allow for combinations of several sources towards their desired nutritional target ( Cassill and Tschinkel , 1999 ) . Previous studies typically addressed the global feedback between colony state and the collective foraging effort ( Seeley , 1989; Seeley and Tovey , 1994; Cassill , 2003 ) . Our work complements this by demonstrating how this feedback acts on individual foraging frequencies ( Figure 4 , see also [Tenczar et al . , 2014] and [Rivera et al . , 2016] ) . Furthermore , while previous studies suggested that foragers use local information , such as time delays , interaction rates or number of refused interactions , to infer the colony’s needs ( Seeley , 1989; Seeley and Tovey , 1994; Cassill , 2003; Greene and Gordon , 2007; Gordon et al . , 1993 ) , we propose a mechanism that demonstrates how foragers could adjust their foraging frequency relying on their own crop load alone ( Figure 6c , d ) ( Mayack and Naug , 2013 ) . In brief , foragers could adjust their exit rates to colony needs by modulating their decision rate according to unloading rates , while the decision itself depends on their current crop load alone . Interestingly , foragers usually do not exit completely empty ( Figure 5a , b ) as could be intuitively assumed ( Gregson et al . , 2003; Buffin et al . , 2009 ) . This provides further evidence that , similar to interaction volumes , foraging activity is not regulated by pure physical limitations ( i . e . an empty crop ) . Rather , we have found that foragers exit with a wide range of crop loads . The lack of a well-defined exit threshold entails a potentially wasteful effect in which forager crop loads at exit increase with colony state: The difficulty of unloading at higher colony states means that foragers spend longer times with a relatively full crop . Since there is a probability to exit at any crop load this may lead to an upward drift in the crop loads of exiting foragers . Here we show this drift is minor ( Figure 5a ) and propose different options by which this may be achieved . For example , it could be the case that after each interaction a forager decides whether to exit the nest or , rather , wait for another opportunity to unload . This decision scheme holds an appealing simplicity as it implies that a forager’s decision rate is set externally and not by an internal clock or parameters . On the other hand , it demands that the forager makes complex decisions that integrate the state of her own crop with colony-level information ( Figure 6b ) . Another possibility rids the foragers of the need to use colony level information . In this case , foragers effectively modulate their decisions to exit according to their unloading rates . We suggest a biologically appealing mechanism to achieve this , in which decisions occur at constant crop intervals ( Figure 6c ) . Generally , any mechanism by which the forager’s trigger to leave the nest depends on her unloading rate could yield similar results . Leaving the nest while partially loaded could hold some benefits: foragers may use the food in their crops as provisions to be consumed in their expedition ( Rytter and Shik , 2016 ) , waiting for full unloading in the nest may be time-consuming and limit exploration for other food types , and frequent visits to the food source may ensure its exploitation . This raises the question whether there exists an optimal crop load with which foragers should exit the nest , which could potentially depend on factors such as the abundance and quality of the food source , predation risk , and the demand for food in the nest ( Dornhaus and Chittka , 2004 ) . In light of our findings on both the forager’s decision to exit and the distribution of interaction volumes , we hypothesize that an internal mechanism based on the mechanical tension of the crop’s walls is involved in trophallaxis . Considering the crop as an elastic organ that stretches as it fills , the relative change in the volume of the recipient’s crop may provide a mechanism for the scaling of interaction volumes with available crop space . Additionally , if ants could sense changes in the tension of their crop walls ( Stoffolano and Haselton , 2013 ) , then this would provide an anatomical basis for a model in which foragers adjust decision rates to unloading rates .
Camponotus sanctus are omnivorous ants that are presumed to naturally live in monogynous colonies of tens to hundreds of individuals ( projecting from Camponotus socius , [Tschinkel , 2005] ) , distributed from the near East to Iran and Afghanistan ( Ionescu-Hirsch , 2009 ) . Workers of this species are relatively large ( 0 . 8–1 . 6 cm ) and characterized by translucent gasters , rendering them suitable for both barcode labeling and crop imaging . Our experiments were conducted on lab colonies of 50–100 workers , reared from single queens that were collected during nuptial flights in Neve Shalom and Rehovot , Israel . Table 2 contains further details on each experimental colony . Fluorescent food imaging and 2D barcode identification ( BugTag , Robiotec ) were used to obtain a live visualization of the food flow through colonies of individually tagged ants . See ( Greenwald et al . , 2015 ) for a detailed description of the experimental setup . In short , an artificial nest was placed on a glass platform positioned between two cameras . A camera below the nest filmed through the platform , capturing the fluorescence emitted from the food inside the translucent ants . Meanwhile , a camera above the nest filmed through its infrared shelter , capturing the barcodes on the ants’ thoraxes , allowing identification of single ants inside the nest . Together , footages from both cameras enabled the association between each individual ant and her food load , throughout time and across trophallactic events . The two cameras were synchronously triggered at a fixed frame rate , ( here 0 . 5 Hz . , except for colony B which was recorded at 1 Hz . ) . We chose a temporal resolution that is sufficient to capture events of 2 s since shorter interactions barely involve food exchange ( Greenwald et al . , 2015 ) . Top camera images were used to extract ant identities , coordinates and orientations using the BugTag software ( Robiotec ) . Bottom camera images were used to detect fluorescence with a pixel intensity threshold , using the openCV library in Python . Gasters of fed ants appeared as bright ‘blobs’ and thus passed the image threshold ( for details , see [Greenwald et al . , 2015] ) . In order to associate between the identity of an ant and her appropriate blob , the image from the upper camera was transformed to align with the fluorescent image . Then , for each identified tag , a small area extended from the back of the tag toward the ant’s abdomen was crossed with the thresholded fluorescent image . If a blob intercepted this area , it was assigned to the tag’s identity . Thus , for each experiment a database was obtained , which included for every frame the coordinates , orientation , and measured fluorescence ( in arbitrary units of pixel intensity ) of each identified ant . Even though the fluorescence emitted from an ant’s crop is reasonably indicative of the food volume , it is a noisy measurement mainly due to her highly variable postures . Therefore , assuming that an ant’s crop content remains constant during the intervals between trophallactic events , it is best evaluated as the maximal fluorescence measurement acquired in each such interval ( Greenwald et al . , 2015 ) . In order to precisely consider the relevant intervals for this estimation , the trophallactic interactions were manually identified from the video . Interactions were classified as trophallactic events whenever the mandibles of the participating ants came in contact and the mandibles of at least one of the ants were open . For forager ants , another situation in which their crop loads may change is when they directly feed from the food source . These feedings were also manually identified from the video , as times when a forager’s open mandibles touched the food source . Ultimately , for each ant we obtained a ‘timeline’ , describing at every instance whether she was engaged in trophallaxis ( and if so , with whom ) , whether she was directly feeding from the food source , and the estimated food load in her crop . Figure 4a depicts an example of such individual-level data . Following a food-deprivation period of 3–5 weeks , ant colonies ( queen , workers and brood ) were manually barcoded and introduced to the experimental nest for an acclimatization period of at least 4 hr . The nest consisted of an IR-sheltered chamber ( ~100 cm2 ) , neighboring an open area which served as a yard ( Figure 1a ) . After the acclimatization period , the two cameras synchronously started to record . After 30 min , the fluorescent food ( sucrose [80 g/l] , Rhodamine B [0 . 08 g/l] ) was introduced to the nest yard ad libitum , and the recording proceeded for at least four more hours - a duration sufficient for the colony to reach its desired food volume intake ( Figure 1b and Figure 1 ) . Overall , we analyzed data from three such experiments , that included 12 foragers , who fed from the food source 139 times , and were engaged in 1227 trophallactic interactions . To manipulatively examine the role of the colony’s satiety in the control of food inflow , we characterized the system’s response to a perturbation in the colony’s satiety level . This experiment was conducted as the observation experiment described above , except that it consisted of two phases: Phase 1: The starved colony was segregated between two equally-sized chambers - one with access to the nest yard , and the other blocked behind a removable perspex wall . Thus , when the fluorescent food was introduced to the nest yard , only the ants with access to the yard gradually satiated while the others in the blocked chamber remained hungry . We reasoned that if foragers react to the colony’s satiety through their experience in the nest , they would perceive saturation of the accessible chamber as saturation of the colony , as they could only interact with ants of the accessible chamber . Phase 2: After the first chamber satiated , we introduced the hungry ants of the blocked chamber by removing the wall , effectively dropping the perceived satiety level of the colony at once . Recording then proceeded for at least 90 more minutes , sufficient for the colony to reach secondary satiation ( Figure 4d and Figure 4—figure supplement 2 ) . Segregating the colony into two chambers . In order to avoid artificial biases in the chambers’ populations , ants were initially introduced to the nest without the wall to freely settle within it . Only after a habituation period of at least 4 hr , the wall was gently inserted to divide the ants , that were then left to habituate for at least one more hour before recording started . The blocked chamber included the queen and brood in both perturbation colonies , and the number of ants in the accessible chamber was 33 and 31 in colonies M1 and M2 , respectively . Time of wall removal . Satiation of the first chamber was identified with semi-online approximative image analysis of the videos from the fluorescence camera , by summing the pixel intensities of each frame , which rose as food accumulated . Satiation was determined when this fluorescent signal ceased to rise for at least 1 hr , serving as our cue to remove the wall . Overall , we analyzed data from two perturbation experiments , on colonies of 69 and 95 ants , including 23 and 16 foragers , respectively . All data obtained after crop load estimation was analyzed using Matlab software . Four data files are available with this manuscript . Each experiment consisted of a few individuals who performed consistent foraging cycles between the food source and the nest . Those ants were considered as ‘foragers’ . Some other individuals were occasionally observed at the food source but clearly did not display such foraging cycles . To our purposes they were not considered as foragers . These ants visited the food source no more than four times , while consistent foragers performed an average of 15 . 67 cycles and no less than 8 . The data presented here is from the first return of a forger to the nest from the food source until the end of the experiment . The total accumulated food was calculated as the sum of all interaction volumes between foragers and non-foragers . The volume of an interaction was taken to be positive when food was transferred from the forager and negative when it was transferred to the forager . Each forager’s contribution is the sum of her own interaction volumes . Although food accumulated through discrete local events , we were mostly interested in the average dynamics of food flow , which are convenient to describe in a continuous manner . Therefore , the accumulated food was first smoothed with a moving average with a time window large enough to include several trophallactic events ( 2000 s ) . This window size was chosen by plotting the smoothed data on top of the raw data and assuring that small fluctuations were smoothed while the general shape was maintained . Food inflow was derived by differentiating the smoothed data . Since differentiation is a process highly sensitive to local noise , we differentiated the smoothed accumulated food with a window of 200–500 s , depending on the fluctuations of the experiment . This window size was chosen by verifying that the sum of the obtained inflow is indeed sufficiently close to the raw data of accumulated food . Our experimental method provided us with measurements of food volume in arbitrary units of fluorescent pixel intensity . Due to possible variations in lighting conditions between experiments , the obtained pixel intensities were incomparable . Therefore we used pixel intensities only for analyses perfomed within the same experiment ( Figure 1b–c , Figure 4d–e and Figure 1—figure supplement 1 ) . For all other purposes , food load was estimated in normalized units . In analyses where individual crop loads and interaction volumes were linked to the global dynamics ( Figure 3 ) , absolute loads were important . Therefore , food loads were normalized between experiments , by dividing each measurement by the 90th percentile crop load measurement of its experiment . In analyzing foragers’ responses to their own crop loads ( Figure 6 ) , the relative satiety state of each forager was of interest . Accordingly , food loads were normalized between foragers , by dividing the measurements of each forager by her own maximal measurement . Since exponential distributions could be fit only to ‘positive’ interactions , i . e . where the forager was the donor , when we fit exponential distributions we neglected the negative interactions . Negative interactions constituted 216 out of 962 , and accounted for 12% of the total food flow . The consequence of this approximation is that we effectively lose 12% accuracy in the modeled food flow . Despite this loss of accuracy , the results from this analysis were consistent with parameters obtained otherwise ( without neglecting the negative interactions ) , ensuring that it was indeed sufficient to consider only the positive interactions . | In an ant society , a small group of workers , called foragers , feeds the rest of the colony . Each forager goes out of the nest to find food; any liquid food she collects is stored in her ‘crop’ , a pouch located just upstream of her stomach . When a forager goes back to the nest , she unloads this liquid by mouth-to-mouth contact into the crops of other ants . The foragers need to adjust how often they go on foraging trips based on the amount of food the other ants require at any given time . However , it is still unclear how foragers can assess the changing needs of the colony . For example , it had been assumed that a forager would fully feed the individuals she encounters in the nest and then go for another foraging trip when her crop is empty . Yet , scientists had not managed to track food transfer at the level of the individual insect to confirm if this is the case . Greenwald , Baltiansky and Feinerman have now used laboratory ant colonies and fluorescently labeled food to monitor in real time how food is transferred between individual ants . Contrary to previous hypotheses , when a forager comes back to the nest , she gives small portions of food to many different ants . The insects in the colony are therefore being nourished through these repetitive interactions . As the experiments show , when a forager meets other ants in the nest , the fullness of their crops reliably represents how full the colony is as a whole . Moreover , the portion that the forager gives is , on average , proportional to the space available in the receiver’s crop: the emptier the crop , the more food is given . The amount of food in the crops of the receiving ants therefore controls how much food enters the colony , and the rate at which a forager unloads its crop . A possible mechanism for regulating foraging frequency is that a forager considers whether or not to go on a foraging trip only after she senses a substantial change in the amount of food in her crop . In this case , her decision is based on the fullness of her own crop: the smaller the amount of food left in her pouch , the more likely she is to decide to leave the nest to bring in more food . Because the rate at which the foragers’ crop empties is tied to the amount of food in the receiving ants’ crops , how often the forager goes for food changes with the hunger level of the whole colony , with more trips when the ants are hungrier . These experiments show that the amount of food in the crops of the receiving and foraging ants helps foragers adapt their behavior to the colony’s needs . This mechanism means the insects can achieve a common goal without explicitly knowing it . However , it remains to be explained how exactly the mechanical changes in the fullness of foragers’ crop underpin this decision-making process . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology",
"computational",
"and",
"systems",
"biology"
] | 2018 | Individual crop loads provide local control for collective food intake in ant colonies |
Image-based cell classification has become a common tool to identify phenotypic changes in cell populations . However , this methodology is limited to organisms possessing well-characterized species-specific reagents ( e . g . , antibodies ) that allow cell identification , clustering , and convolutional neural network ( CNN ) training . In the absence of such reagents , the power of image-based classification has remained mostly off-limits to many research organisms . We have developed an image-based classification methodology we named Image3C ( Image-Cytometry Cell Classification ) that does not require species-specific reagents nor pre-existing knowledge about the sample . Image3C combines image-based flow cytometry with an unbiased , high-throughput cell clustering pipeline and CNN integration . Image3C exploits intrinsic cellular features and non-species-specific dyes to perform de novo cell composition analysis and detect changes between different conditions . Therefore , Image3C expands the use of image-based analyses of cell population composition to research organisms in which detailed cellular phenotypes are unknown or for which species-specific reagents are not available .
Single-cell analysis has proven crucial to our understanding of fundamental biological processes such as development , homeostasis , regeneration , aging , and disease ( Goolam et al . , 2016; Kimmel et al . , 2019; Pepe-Mooney et al . , 2019; Philippeos et al . , 2018; Tirosh et al . , 2016 ) . High-throughput analyses of these and other biological processes at single-cell resolution require technologies capable of describing individual cells and subsequently clustering them based on similarities of features like morphology , cell surface protein expression , or transcriptome profile . Recent advances in image-based cell profiling and single-cell RNA sequencing ( scRNA-seq ) allow quantification of differences between cell populations and comparisons of cell type composition between samples ( Caicedo et al . , 2017 ) . Single-cell studies that use traditional research organisms ( e . g . , mouse , rat , or fruit fly ) benefit from the availability of genomic platforms and established antibody libraries . However , the same cannot be said for a growing number of important , yet understudied research organisms lacking such reagents and whose biological interrogation would benefit immensely from single-cell analyses . In these cases , classical histochemical methods are often used to identify and characterize specific cells . Yet , the successful identification and enumeration of biologically meaningful cell types in such studies can be harmed by both the limited number and variety of cellular attributes ( few features or low dynamic range ) available for determination of cell types and by observer bias when using traditional , hand-counting approaches ( e . g . , hemocytometer and Giemsa stain ) ( van der Meer et al . , 2004 ) . These shortcomings , together with the lack of extensive knowledge on cell-specific phenotypes available for training or for a priori assumptions , usually result in the underestimation of the complexity of cellular composition or interactions among cell types within tissues . Automated classification of cells using convolutional neural networks ( CNNs , machine learning [ML] method specialized in image recognition and classification ) has become a promising approach for accurate high-throughput cell analysis that is free from observer bias ( Blasi et al . , 2016; Eulenberg et al . , 2017; Kobayashi et al . , 2017; Lei et al . , 2018; Nassar et al . , 2019; Suzuki et al . , 2019 ) . To date , CNN-based automated clustering and classification techniques require pre-existing knowledge about the organism or cell type of interest ( e . g . , cell-specific morphological traits within an image set ) or the availability of cell-specific reagents ( e . g . , antibodies ) , or genomic sequence ( e . g . , single-cell sequencing ) ( Table 1 shows an overview of the existing methods ) ( Baron et al . , 2019; Blasi et al . , 2016; Cheng et al . , 2021; Eulenberg et al . , 2017; Hennig et al . , 2017; Kobayashi et al . , 2017; Lei et al . , 2018; Nassar et al . , 2019 ) . This means that to make effective use of artificial intelligence ( AI ) approaches for single-cell analysis , one must have information available to train the algorithm or for ML models , which often arises in the form of information gleaned from the use of reagents like antibodies . Research areas that rely on inter-species comparisons or studies on emerging research organisms would benefit from single-cell-based analyses that do not require pre-existing knowledge of cell types ( i . e . , which is required for training a CNN for example ) and/or availability of antibodies or molecular databases . For example , within the interdisciplinary field of eco-immunology , a growing number of researchers are investigating immune system adaptation to different environments by studying immune cell compositions in diverse animals ( Maizels and Nussey , 2013 ) . Given the influences of immune cell composition on the immune system response of an organism ( Kaczorowski et al . , 2017 ) , applying modern single-cell analysis in eco-immunological research would substantially increase our knowledge about the plasticity and conservation of immune responses in a variety of different animals and conditions ( Peuß et al . , 2020 ) . To make sophisticated cellular composition analysis available to any research organism without the need for either pre-existing knowledge about the cell populations or species-specific reagents , we developed Image-Cytometry Cell Classification ( Image3C ) . This method analyzes , visualizes , and quantifies , in a high-throughput and unbiased way , the composition of cell populations by using cell morphological traits and non-species-specific fluorescent probes ( e . g . , nuclear staining or dyes for metabolic states that function well in a variety of organisms ) ( Figure 1 , Table 1 ) . By taking advantage of cell morphology and/or fluorescent dyes related to function or metabolic state , Image3C can analyze single cell suspensions derived from any experimental design , de novo cluster cells present in the sample of interest and compare their abundance between each other or among different assays . Once the de novo clustering based on cell-intrinsic features is obtained , Image3C employs a CNN that uses these clusters as training sets , avoiding in this way user bias or manual classification ( Figure 1 , Table 1 ) . This produces a CNN-based cell classifier ‘machine’ used to quantify subsequently acquired image-based flow cytometry data and compare cellular composition of samples across multiple experiments in a high-throughput manner without the need for repeating time-consuming steps for de novo clustering . The combination of the clustered cell images , the outputs of their functional assays , and the published literature about closely related organisms might allow the identification and description of cell types of interest . In comparison to existing label-free cell clustering methods , Image3C does not require initial antibody staining ( Cheng et al . , 2021; Hennig et al . , 2017; Lippeveld et al . , 2020; Nassar et al . , 2019 ) , pre-existing knowledge of specific cell morphology ( Suzuki et al . , 2019; Yakimov et al . , 2019 ) , and is not limited to a specific cellular phenotype ( Blasi et al . , 2016 ) for a priori identification of certain cell types ( Table 1 ) . This makes Image3C extremely versatile and applicable to virtually any research organism and tissue from which dissociated single cells can be obtained . Parallelly to this de novo clustering approach , Image3C can take advantage of species-specific reagents and prior knowledge to be combined with transcriptomic dataset and provide a new and complimentary layer of information based on cell morphology and function . In sum , Image3C combines modern high-throughput data acquisition by image-based flow cytometry , advanced and unbiased clustering analysis , statistics to compare cellular compositions across different samples , and a CNN classifier component to easily determine changes in cell composition across multiple experiments .
Image3C is an imaging tool developed to study tissue composition at single-cell resolution in research organisms for which antibodies and pre-existing knowledge about cell types are not readily available ( Figure 1 , Table 1 ) . Image3C allows for high-throughput and unbiased analysis in scenarios where manual counting and observer-based cell identification are currently the only options . Image3C includes all the components required for compensating captured images , quantifying multiple features for each event , clustering the events , visualizing and exploring the data , and training and using the CNN for analyzing subsequent samples and integrating multiple experiments ( Figure 1 , Figure 1—figure supplement 1 ) . Once a single cell suspension is prepared from the organism of interest , the cells are stained with a combination of dyes that are expected to function irrespectively of the species used and which have high affinity for specific cellular organelles such as nuclei or molecules associated with metabolic states such as reactive oxygen species ( ROS ) . We validated reagents experimentally by determining that nuclear dyes stain intracellular material matching expected characteristics of nuclear DNA or by activation of cells with drugs to change their metabolic state . The labeled samples are then run on the ImageStreamX Mark II ( Figure 1A ) . ImageStream is a commercially available image-based flow cytometer , whose diffusion in laboratory settings is increasing and that provides highly reproducible images of cells that can be compared across days of acquisitions and experiments . For this approach , no microfluidic devices or custom-made and highly specialized microscopes are required ( Table 1 ) , and , if desired , the users can test the Image3C pipeline also on images acquired at standard microscopes , remembering to carefully control for batch effects . Once images of individual events are collected for each channel of interest , feature values from both morphological and fluorescent data , such as cell size and nuclear size , are extracted from the cell images using IDEAS software ( Amnis Millipore , free for download upon creation of Amnis user account ) ( Figure 1B; see Supplementary files 1 and 2 for feature description ) . Correlation between features is calculated and redundant features are trimmed as well as samples that , among replicates , are outliers ( Figure 2—figure supplements 1 and 2 ) . This prevents clustering artifacts potentially caused by having multiple features providing the same information or including samples that are not representative ( Figure 1B ) . During this step , while the number of features was usually reduced significantly , the correlation between replicates was always high and outliers were rarely observed ( Figure 2—figure supplements 1 and 2 ) . Finally , fluorescence intensity features are transformed to improve homogeneity of variance of distributions and , if used , DNA staining is normalized to remove intensity drift between samples and thus align the 2N and 4N DNA content histogram peaks ( Figure 1B , Figure 2—figure supplement 3 ) . Exported feature quantifications are used for clustering the events . Dimensionality reduction and visualization of clusters is achieved by generating force-directed layout ( FDL ) graphs in the VorteX clustering environment ( Figure 1C ) ( free to install ) ( Samusik et al . , 2016 ) . Cell images for events within each cluster can be visualized using FCS Express Plus together with custom R scripts ( Figure 1D ) . These visualization tools and the cluster feature averages ( i . e . , the mean value of each feature for each cluster ) ( Figure 1E ) allow to explore the images of selected groups of events and the features that differ between cells belonging to separate clusters . If control and treatment samples are included , a statistical analysis using negative binomial regression to compare cell counts per cluster between samples is also available in the Image3C pipeline . This high-throughput and unbiased analysis provides a comprehensive understanding of a cell population composition at higher resolution than what is possible with traditional histological methods . Once this pipeline is run on a first set of samples ( e . g . , homeostatic state ) and the cell clusters are defined for the tissue of interest , the images and the relative clustering IDs can be used to train a CNN classifier in an unbiased way ( Figure 1F ) , including the ability to score frequency of ‘new’ cell types that do not match any of the clusters identified at homeostasis . Therefore , future experiments in the same tissue used for training the CNN classifier can be analyzed directly through the CNN ( Figure 1G ) . This significantly reduces the number of steps and time required to process data collected subsequently . An even greater advantage is represented by the fact that , in the absence of CNN , every time new experimental sets are run it would be necessary to go again through the de novo clustering part of the pipeline ( Figure 1B–E ) and the new set of clusters would need to be cross-annotated to be compared with cell population composition observed in previous experiments . Manually matching clusters between different experimental sets might be a source of errors , mainly if the user is not familiar with the cell types present in the sample and if specific biomarkers or pre-existing knowledge about cell types and morphology are not available . The CNN splits all the cell images in the classes defined during the training step and allows to compare the abundance of cells with same morphology among different samples without the need to cross-annotate clusters ( Figure 1F , G ) . The CNN inclusion in Image3C and the reproducibility of image acquisition through image-based flow cytometry allows use of the clusters defined from one experiment ( e . g . , homeostatic state ) to set up a classifier in an unbiased way for later use as a reference in analyzing the effects of experimental manipulations on these cell populations . We conclude from these results that Image3C can perform de novo high-throughput characterizations of population composition and define specific cell type changes between homeostatic and experimentally perturbed samples across multiple experiments . To test whether Image3C could identify homogeneous and biologically meaningful cell populations , we used the research organism Danio rerio . We obtained cells from adult female zebrafish WKM ( location of the hematopoietic tissue ) in homeostasis condition , stained them , and ran on the ImageStreamX Mark II . We analyzed intrinsic morphological and fluorescent features , such as cell and nuclear size , shape , and darkfield signal ( side scatter [SSC] ) . Feature values were extracted from each cell image and processed through our pipeline ( see Supplementary file 1 for feature description ) . Clustering by the final set of normalized and non-redundant morphological and fluorescent features produced distinct cell populations ( Figure 2A–C , Figure 2—figure supplements 1–3 ) . Image3C can distinguish between the major classes of cells present in zebrafish WKM ( Figure 2; Supplementary files 3 and 4 ) that were described using standard sorting flow cytometry and morphological staining approaches ( Traver et al . , 2003 ) . It is noteworthy that Image3C can clearly identify dead cells and debris ( Figure 2A , B ) allowing to optimize experimental protocols in order to minimize cell death and run the subsequent analysis only on the intact , live cells . Image3C can identify cells with outstanding morphological features , such as neutrophils from other myelomonocytes ( Figure 2B , C ) . Based on zebrafish neutrophil characteristics such as high granularity , high complexity , and low circularity of the nuclei ( Lugo-Villarino et al . , 2010 ) , this type of granulocytes can be easily distinguished . Other types of myelomonocytes , such as monocytes and eosinophils , are here merged in the same cluster since in zebrafish they share many morphological characteristics ( Lugo-Villarino et al . , 2010 ) . Similarly , using only intrinsic morphological features for the clustering , different lymphocytes ( B and T-cells ) and hematopoietic stem cells cannot be separated from each other , but they can be clearly distinguished from the myelomonocytes ( Figure 2A , B ) . Within the lymphocytes/progenitors fraction , we find two clusters ( Dr1 and Dr7 ) that mainly differ in cell diameter ( Figure 2C ) . Whether this morphological difference has a biological implication needs to be studied in future experiments . Image3C also enables the quantification of cell populations ( clusters or CNN classes ) -relative abundance , an important tool for comparing population composition across different treatment groups under different environmental conditions ( Peuß et al . , 2020 ) . Here , we compared our results with previously published data to validate our method . Although a direct comparison with results from classical approaches ( Traver et al . , 2003 ) is not possible since we gated out ( removed analytically ) mature erythrocytes before clustering ( Materials and methods ) , the myelomonocyte to lymphocyte ratio ( M/L ratio = 1 . 59 ) is similar to the one obtained with classic histological approaches ( mean M/L ratio = 1 . 35 ) ( Figure 2D; Traver et al . , 2003 ) . Next , we sought to determine whether Image3C could be used to characterize and quantify biological processes by identifying a tissue of interest and then comparing cellular composition dynamics , functions , and physiological responses of specific cell types across a range of experimental conditions . Our goal was to detect statistically significant changes in cluster relative abundance between control and treated samples to gain a more detailed understanding of cell population dynamics and individual cell function . As proof of concept , we performed a standard phagocytosis assay using WKM tissue from female adult zebrafish . The single cell suspension was incubated with CellTrace Violet labeled Staphylococcus aureus ( CTV-S . aureus ) and with dihydrorhodamine-123 ( DHR ) , a ROS indicator that becomes fluorescent if oxidized to report oxidative bursting following phagocytosis . As controls , we inhibited phagocytosis through cytoskeletal impairment with cytochalasin B ( CCB ) incubation or through incubation with bacteria at lowered temperature by placing culture plates on ice . Events collected on the ImageStreamX Mark II were analyzed with Image3C and clustered in 26 distinct clusters using quantifications of morphological and fluorescent features ( see Supplementary file 2 for feature description ) , including nuclear staining , phagocytized S . aureus , and DHR positivity ( Figure 3A , Figure 3—figure supplement 1 ) . Professional phagocytes are defined by their ability to take up S . aureus ( CTV staining lies within the cell boundary ) and induce a ROS response ( bright DHR signal ) ( Rabinovitch , 1995 ) . In zebrafish , professional phagocytes are mainly granulocytes and monocytic cells and can be discriminated from each other based on morphological differences , such as cell size , granularity , and nuclear shape ( Wittamer et al . , 2011 ) . To compare samples incubated with CTV-S . aureus and the samples where phagocytosis is inhibited ( CTV-S . aureus + CCB and CTV-S . aureus + Ice ) , we used the statistical analysis included in Image3C based on a negative binomial regression model ( Figure 3B , C , Figure 3—figure supplement 2; Supplementary files 5 and 6 ) . Statistical analyses reported clusters with differences in relative abundance between phagocytosis and phagocytosis-inhibited samples . Visualizing these clustered event images ( Supplementary file 7 ) while considering the values and intensities of their morphological and fluorescent features ( Supplementary file 3 ) allowed identification of 4 clusters of professional phagocytes: granulocytes within clusters Dr4_P , Dr12_P , and Dr13_P and monocytic cells in cluster Dr21_P ( Figure 3A , B ) . The morphology of cells in cluster Dr12_P is characteristic of phagocytic neutrophils ( Figure 2B , Figure 3A ) that become adhesive and produce extracellular traps upon recognition of bacterial antigens ( Palić et al . , 2007 ) . Overall , the relative abundance of professional phagocytes is 5–10% ( Figure 3C ) , which is in line with previous studies that estimated the number of professional phagocytes in WKM tissue of adult zebrafish using classical morphological approaches ( Wittamer et al . , 2011 ) . It is also noteworthy that in line with other studies ( Page et al . , 2013 ) we did not observe a cluster of lymphocytes ( e . g . , B-cells ) that actively phagocytize CTV-S . aureus bacteria ( Figure 2; Supplementary file 7 ) . Compared to the classical morphological approaches , Image3C allows to analyze thousands of events in a high-throughput and unbiased fashion , allowing the study of rare cell morphologies and increasing results confidence and reproducibility . These results show that Image3C can successfully analyze biological processes since we were able to recapitulate the presence , cell type , and frequency of professional phagocytes in adult zebrafish WKM organ . A new aspect that Image3C highlighted is that CCB selectively affects cell viability based on cell identity , introducing artifacts and cell damage , actions not specific to inhibition of phagocytosis ( Figure 3B ) . All mature erythrocyte-containing clusters had a significantly higher cell count in the CTV-S . aureus samples compared to the CTV-S . aureus + CCB ones ( Figure 3B; Supplementary files 3 and 5 ) . Cluster analysis revealed that erythrocytes were almost absent in samples incubated with CCB ( Supplementary file 3 ) , while there was a significant increase in the relative abundance of clusters containing dead and apoptotic cells ( Figure 3B; Supplementary file 5 ) . Both outcomes are likely due to reduced cell viability of erythrocytes upon CCB incubation . Moreover , we excluded the possibility of higher cell death in the professional phagocytes upon CCB incubation since pseudo-phagocytes ( phagocytes with DHR response but no internalized CTV-S . aureus ) were significantly more abundant in the CTV-S . aureus + CCB sample ( Figure 3B; Supplementary file 5 ) . These results are remarkable since Image3C allowed us to observe a specific effect of CCB on erythrocytes' viability in zebrafish that , as far as we know , was not described before . Image3C analysis also uncovered another important biological observation . When we inhibited phagocytosis by incubating the single cell suspension on ice ( CTV-S . aureus + Ice ) and compared the specificity of inhibition with the CTV-S . aureus + CCB sample ( Figure 3C; Supplementary file 6 ) , we discovered that the inhibition of phagocytosis through low temperature only affects adhesive neutrophils ( cluster Dr12_P ) ( Figure 3C ) . This is suspected to occur as ice inhibits adhesion , while CCB effectively blocks phagocytosis in all professional phagocytes in zebrafish WKM tissue by acting on the cytoskeleton . The use of Image3C allowed us to specifically identify cell types that are sensible to low temperature and those that are not , confirming the existence of different phagocytosis mechanisms and providing additional knowledge about pros and cons of different protocols that can be applied to inhibit phagocytosis based on specific goals and needs . Since we aimed to provide a tool that is widely applicable , we tested Image3C versatility on the apple snail Pomacea canaliculata , an emerging organism for which molecular and cell biological tools have yet to be fully developed . As such , we repeated the same experiments done in zebrafish on the hemolymph of P . canaliculata . For morphological examination of the cellular composition of the hemolymph collected from female adults in homeostasis conditions , we stained the single cell suspensions with Draq5 ( DNA dye ) and ran on the ImageStreamX Mark II . We used Image3C to analyze the images of the events and identified 9 cell clusters ( Figure 4A , Figure 4—figure supplement 1 ) . Two of these clusters comprised cell doublets , debris , and dead cells ( clusters Pc5 and Pc8 ) and the other clusters , based on inspection of cell images , were grouped into two main categories ( Figure 4A; Supplementary file 8 ) . The first category includes small blast-like cells ( cluster Pc4 ) and intermediate cells ( clusters Pc2 and Pc3 ) with high nuclear-cytoplasmic ( N/C ) ratio . These cells morphologically resemble the Group I hemocytes previously described using a classical morphological approach ( Accorsi et al . , 2013 ) . The second category comprised larger cells with lower N/C ratio and abundant membrane protrusions ( clusters Pc1 , Pc6 , Pc7 , and Pc9 ) . Likely , these cells correspond to the previously described Group II hemocytes that include both granular and agranular cells ( Accorsi et al . , 2013 ) . To identify which of these clusters were enriched for granular cells , we looked at the heatmap with feature values for each individual cluster ( Figure 4B; see Supplementary file 1 for feature description ) . Cluster Pc6 had the highest values for the features related to cytoplasm texture and granularity ( i . e . , area granularity , intensity granularity , and signal granularity ) amongst all clusters other than cell doublets ( Figure 4B; Supplementary files 3 and 8 ) . Based on these data , we identified cluster Pc6 as the one containing the granular hemocytes . The clusters obtained by Image3C were not only homogeneous and biologically meaningful but were also consistent with published P . canaliculata hemocyte classification obtained by classical morphological methods ( Accorsi et al . , 2013 ) . Such remarkable consistency was observed both in terms of identified cell morphologies and their relative abundance in the population of circulating hemocytes ( Figure 4C; Supplementary file 8 ) . For example , the relative abundance of the previously reported small blast-like cells is 14 . 0% , a value almost identical to the abundance of the corresponding cluster Pc4 ( 13 . 8% ) . Similarly , the category of larger hemocytes or Group II hemocytes represents 80 . 4% of the circulating cells as measured by traditional morphological methods , while clusters Pc1 , Pc6 , Pc7 , and Pc9 combined represent 72 . 4% of the events analyzed with Image3C ( Figure 4C; Supplementary file 3 ) . A subset of these cells are the granular cells ( cluster Pc6 ) , which correspond to 7 . 7% of all hemocytes by classical histological methods and 8 . 9% by Image3C . The intermediate cells ( clusters Pc2 and Pc3 ) are the least represented in both approaches , with a relative abundance of 5 . 6% and 10 . 6% for the manually and Image3C analyzed events , respectively ( Figure 4C; Supplementary file 3 ) . This difference is likely best explained by the remarkable difference in both the number of cells and the number of features that can be considered for analysis by Image3C . Only a few hundred hemocytes were visually analyzed using traditional histological methods based only on cell diameter and N/C ratio ( Accorsi et al . , 2013 ) . In contrast , the automated pipeline used by Image3C facilitated the analysis of 10 , 000 nucleated events for each sample and considered 25 morphological features for each cell . The significantly higher number of morphological features simultaneously considered by Image3C also explains the higher number of clusters and improved resolution to distinguish cell types compared to the traditional methods . Hence , Image3C not only can properly analyze cells obtained from an emerging research organism generating biologically meaningful and informative clusters but also represents an unprecedented increase in the accuracy of cell type identification over traditional histological methods , while also allowing high-throughput capability . As with zebrafish , we also performed a phagocytosis experiment on hemocytes from P . canaliculata . Our goal was to test if it is possible with an emerging research organism to successfully discover cell phenotypes and functions and obtain information about specific biological processes of interest by using Image3C to compare cell populations among treated and control samples . Here , we setup the phagocytosis assay incubating the cells with CTV-S . aureus and DHR at room temperature . The phagocytosis was inhibited , as control , either adding EDTA ( CTV-S . aureus + EDTA ) or using low temperature by incubating samples on ice ( CTV-S . aureus + Ice ) . Events collected on the ImageStreamX Mark II were analyzed with Image3C and clustered in 20 distinct clusters using quantifications of morphological and fluorescent features ( see Supplementary file 2 for feature description ) , including nuclear staining , phagocytized S . aureus , and DHR positivity ( Figure 5A , Figure 5—figure supplement 1 ) . We compared the phagocytosis-permissive samples ( CTV-S . aureus ) with samples where phagocytosis was inhibited by EDTA incubation or low temperature using the statistical analysis included in Image3C based on a negative binomial regression model ( Figure 5B , C , Figure 5—figure supplement 2; Supplementary files 9 and 10 ) . The clusters with relative abundance significantly higher in the phagocytosis samples ( Figure 5B; Supplementary files 3 and 11 ) and with high intensities of both DHR and bacteria signals ( Figure 5—figure supplements 3 and 4 ) are the two clusters that we identify as enriched with professional phagocyte ( cluster Pc5_P and Pc17_P ) ( Figure 5B , Figure 4—figure supplement 1 , Figure 5—figure supplement 4; Supplementary file 11 ) . The two clusters show a different DHR signal intensity ( ROS response ) from one another upon bacteria exposure ( cluster Pc5_P with high DHR signal , cluster Pc17_P with low DHR signal ) ( Figure 5—figure supplement 3; Supplementary files 3 and 11 ) . Both Pc5_P and Pc17_P relative abundance is significantly higher in the phagocytosis samples compared to the EDTA-treated sample ( Figure 5C; Supplementary file 9 ) , showing that EDTA effectively inhibits phagocytosis for both types of professional phagocytes . In the sample where the phagocytosis was inhibited by low temperature , however , only cluster Pc17_P had a significantly lower relative abundance compared to the phagocytosis sample ( Figure 5C , Figure 3—figure supplement 2; Supplementary file 10 ) . We can conclude that similar to CCB inhibition in the zebrafish phagocytosis experiment , EDTA is a more effective and generalized ( not cell type-specific ) inhibitor of phagocytosis than low temperature . These results show that also in an emerging research organism Image3C allowed discovery of new aspects of this biological process and highlighted differences among professional phagocytes that would have been difficult to detect with other methods . The data analysis with Image3C clearly highlighted that CCB and EDTA , two classical phagocytic inhibitors commonly used in controls for phagocytosis experiments in vertebrates and invertebrates , respectively , result in a drastic change of cell morphology and cell viability . This consequence is not easily detectable by other methods and is therefore often overlooked . In the present work , these changes significantly modified the overall cell cluster number and distribution and indicate that the effects of CCB and EDTA on cell morphology should be taken into consideration in any study of morphological features of cells with phagocytosis properties because artifacts might be significant . When determining differences between control and experimental treatments , Image3C necessarily combines images and data from all samples and then clusters the cells . This must be taken into consideration for experimental planning . Experiments meant to analyze cell composition and morphological diversity in one biological domain ( e . g . , homeostasis condition ) ( Figure 2 , Figure 4 ) should be carried out separately from those in other domains that are likely to introduce changes in the cell population composition or cell morphologies representing a confounding factor for the de novo clustering in homeostasis condition . Image3C clustering works best when used , at the same time , only on samples belonging to a single experimental domain , such as homeostasis or the phagocytosis assay . An issue that emerges when analyzing different experimental sets independently is the increase of time requirement for analytical steps , the likelihood of introducing errors , and the need to repeatedly annotate the clusters in the FDL graph obtained from each experimental set . This last element is required for comparing cell type behaviors among multiple experiments and provide a global understanding of their functions and response to treatments ( i . e . , cluster #1 from one analytical run cannot be expected to match cell morphologies with cluster #1 from another run since there is a stochastic element to the process ) . This last point is probably the most challenging since mistakes can easily be introduced based on user biases or lack of sufficient pre-existing knowledge about cell morphologies or of cell biomarkers that would allow a confident cross-annotation between multiple FDL graphs . In addition , we observed that the number of clusters drastically increases when including treatments that influence cell morphological properties of the cell . As an example , while we detected 9 unique clusters in naïve hemolymph samples , we detected 20 clusters in the phagocytosis experiment ( Figure 3A , Figure 4A ) . This is in part due to the fact that professional phagocytes change their morphology upon detection of pathogens ( Palić et al . , 2007 ) , thus creating new clusters . Similarly , the complexity of the clustering is also increased by treatments , such as CCB and EDTA incubations , that are necessary to ensure identification of professional phagocytes , but have a strong impact on the morphology of the cells making the clustering and annotation steps more challenging and prone to mistakes since treated samples contain aberrant populations not found at homeostasis ( Figure 5A; Supplementary file 11 ) . To provide an alternative for streamlining the analysis of multiple experimental sets upon initial de novo clustering and cell type identification in homeostasis samples , we included in Image3C the possibility to use these initial images and their cluster IDs to train a CNN without manually classifying the images ( Figure 1 ) . This trained classifier can then be used to assign the cell images subsequently collected from additional experimental sets to one of the clusters defined in the homeostasis condition in a high-throughput way . In this way , it will be possible to determine the behavior of a specific cell type through multiple experimental sets without re-clustering whenever new data is acquired . A crucial element that allows this approach is also represented by the ImageStreamX Mark II system that provides highly reproducible and comparable images of cells coming from different experiments and acquired at different days , introducing much less variability than standard light or electron microscopy . For our pipeline , then , a CNN ( LeCun et al . , 1989 ) based on the architecture of DenseNet ( Huang et al . , 2017 ) was deployed to ( 1 ) use , as training set , images and clusters obtained from a first group of samples ( e . g . , homeostasis conditions , naïve cells , or WT samples ) analyzed in an unbiased way by de novo clustering and ( 2 ) assign new cell images acquired through ImageStreamX Mark II system to their respective classes . As proof of concept , we used the clusters identified for P . canaliculata hemocytes in homeostasis condition with the first part of the pipeline ( Figure 4A ) for training and setting up the CNN classifier . This approach would define the classes based on the unbiased de novo clustering of thousands of cells with no need for formal annotation or previous knowledge about cell types and tissue composition . To prepare a dataset for training the classifier , we first combined clusters that strongly overlapped with one another in terms of morphological characteristics ( e . g . , doublets and dead cells ) to increase accuracy of the classifier ( Figure 6A ) . We used 80% of the cells obtained in the original P . canaliculata dataset together with their cluster IDs to train the classifier through over 25 , 000 iterations . After each iteration , we tested the training with 10% of the original dataset and determined the relative accuracy by scoring numbers of cells whose cluster ID assigned by the classifier matched the original cluster ID ( Figure 6B , C ) . The remaining 10% of the original dataset was used to calculate the precision of the trained classifier . Clusters with higher support numbers obtained higher precision scores . The weighted average precision score ( f1-score , precision average score across clusters controlling for support numbers ) of 0 . 75 is relatively high considering the complexity of the phenotype ( brightfield [BF] , darkfield , and Draq5 images ) ( Figure 6D ) and comparable to other studies using ML for cell classification ( Blasi et al . , 2016 ) . The true probability match for each individual cell ( probability for each cell that the class assigned by the classifier would match the original cluster ID ) demonstrated that lower true probability matches occurred where clusters strongly overlapped or where cell phenotypes are intermediate between clusters , providing an additional layer of information about our dataset ( Figure 6D ) . To test the efficiency of this pipeline , we extracted all the images belonging to the two clusters identified in the phagocytosis assay as cluster-containing phagocytes and determined to which naïve cell type they correspond using the CNN classifier and only the BF , SSC , and Draq5 channels ( i . e . , DHR and labeled bacteria signals were not used ) . We found that 59 . 4 , 6 . 2 , and 9 . 2% of the phagocytes belonged to cluster Pc1_CT , Pc6_CT , and Pc7_CT , respectively ( Figure 6E ) , where CT stands for classifier training . These results confirmed a previously published result that used classical morphological staining and manual annotation to conclude that the hemocytes able to phagocytize were primarily Group II hemocytes ( Accorsi et al . , 2013 ) . Only 8% of the phagocytes were clustered in the Group I hemocytes , here represented by clusters Pc2_CT , Pc3_CT , and Pc4_CT , while the remaining 17 . 2% were assigned by the CNN to the cluster Pc5_CT , constituted by doublets and dead cells ( Figure 6E ) . This result can be explained by the fact that in vitro phagocytosis triggers microaggregate formation ( hemocyte-hemocyte adhesion ) in invertebrate hemocytes that resembles the nodule formation observed in vivo ( Walters , 1970 ) . It is important to observe how this analysis allowed us to assign phagocytes to cell types using the annotation already performed in Figure 4A ( de novo clustering of hemocytes in homeostasis condition ) without the need to reannotate the FDL obtained during the phagocytosis assay ( Figure 5A ) . To test the adaptability of the trained CNN to new datasets , we collected hemocytes from male apple snail specimens , stained the cells with Draq5 , and recorded BF , SSC , and nuclei images from 10 , 000 cells on the ImageStreamX Mark II as previously described . We extracted the images of the cells and used the trained CNN classifier to determine the relative abundance of hemocytes collected from male snails in the seven classes of the classifier ( Figure 6F ) . First , we visually compared the female hemocytes clustered by the de novo clustering with the male hemocytes that were run on the ImageStreamX Mark II and were assigned to a class by the classifier ( Figure 6F ) . This comparison shows that the female and male hemocytes belonging to the same cluster are morphologically extremely similar and different from the hemocytes assigned to other clusters ( Figure 6F ) . This demonstrates that the CNN classifier can be trained with a first group of samples and then it can successfully analyze new datasets acquired later on . The comparison between female and male hemocyte compositions revealed that the clusters significantly different in terms of relative abundance are Pc1_CT and Pc7_CT ( Group II agranular large hemocytes ) and Pc6_CT ( Group II granular large hemocytes ) ( Figure 6F ) . Significantly , prior studies detected no differences between females and males hemocytes composition through manual classification and counting using a classical morphological approach ( Accorsi et al . , 2013 ) . The reduced user bias and high-throughput analysis presented here , in contrast , allowed us to determine that one of the two subpopulations of agranular large hemocytes was significantly more abundant in the female animals ( Pc1_CT: 53 and 38% in females and males , respectively ) while the other agranular ( Pc7_CT ) as well as the granular large hemocytes ( Pc6_CT ) was significantly more abundant in the male animals ( Pc7_CT: 14 and 20% in females and males , respectively; Pc6_CT: 6 and 10% in females and males , respectively ) ( Figure 6F ) . While the biological significance of this observation is not going to be further investigated in this paper , the discovery highlights the power of Image3C analysis compared to traditional methods for determining and quantifying the composition of cell populations . These experiments demonstrate that Image3C , in combination with the presented convolutional classifier , can analyze large experimental datasets and identify significances with small effect sizes . Importantly , Image3C analysis is independent of observer biases and does not require prior knowledge about expected tissue composition or the expected effect of treatment on cell morphology . We have developed a powerful new method to analyze at single-cell resolution the composition of any cell population obtained from research organisms for which species-specific reagents ( such as fluorescently tagged antibodies ) , biomarkers , single-cell atlases , or a high-quality genome for a scRNA-seq approach are not available . We demonstrated that Image3C can identify different cell populations based on morphology and/or function through de novo clustering and highlight important changes in cell type abundance and cell population composition caused by experimental or natural perturbation ( sex , treatment , experimental protocol ) . Image3C does not require , at any point , prior knowledge about the tissue composition or cell type-specific markers , although , if available , they can be included and used . Furthermore , in combination with the CNN classifier trained on these clusters , we demonstrate that Image3C is capable of bias-free and high-throughput analysis of large experimental datasets making it possible to compare a specific cell type behavior or population composition across multiple experiments . Image3C is extremely versatile and can be applied to any tissue or cell population of interest and is adaptable to a variety of experimental designs . Although Image3C was developed in response to the need of analyzing cell composition of tissues in emerging research organisms , the Image3C tool could be potentially used also to add to transcriptomic dataset an additional and complementary layer of information based on cell morphology . Given the recent advancement in image-based flow cytometry that enables image capturing together with cell sorting ( Nitta et al . , 2018 ) , a scRNA-seq approach in combination with the Image3C pipeline would enable simultaneous analysis of both the morphological/phenotypic and genetic properties of a cell population at single-cell resolution .
Twelve-month-old , wild type , female , adult zebrafish were euthanized with cold 500 mg/L MS-222 solution for 5 min . WKM was dissected as previously described ( Traver et al . , 2003 ) and transferred to 40 µm cell strainer with 1 mL of L-15 media containing 10% water , 10 mM HEPES , and 20 U/mL Heparin ( L-90 ) . Cells were gently forced through the cell strainer with the plunger of a 3 mL disposable syringe . The strainer was washed once with 1 mL of L-90 and the resulting single cell suspension was centrifuged at 500 rcf at 4 °C for 5 min . The supernatant was discarded , and the cells were resuspended in 1 mL of L-90 containing 5% fetal calf serum ( FCS ) , 4 mM L-glutamine , and 10 , 000 U of both penicillin and streptomycin ( L-90 media ) . The cells were counted in a 1:20 dilution on the EC-800 flow cytometer ( Sony ) using scatter properties . Specimens of the apple snail P . canaliculata ( Mollusca , Gastropoda , Ampullariidae ) were maintained and bred in captivity , in a water recirculation system filled with artificial freshwater ( 2 . 7 mM CaCl2 , 0 . 8 mM MgSO4 , 1 . 8 mM NaHCO3 , 1:5000 Remineralize Balanced Minerals in Liquid Form [Brightwell Aquatics] ) . The snails were fed twice a week and kept in a 10:14 light:dark cycle . Wild-type adult snails , 7–9 months old and with a shell size of 45–60 mm , were starved for 5 days before the hemolymph collection ( Accorsi et al . , 2013 ) . If not differently specified , female snails were used for the experiments . The withdrawal was performed by applying a pressure on the operculum and dropping the hemolymph directly into an ice-cold tube . The hemolymph collected from different animals was not pooled together . The hemolymph was immediately diluted 1:4 in Bge medium + 10% fetal bovine serum ( FBS ) and then centrifuged at 500 rcf for 5 min . The pellet of cells was resuspended in 100 µL of Bge medium + 10% FBS . The Bge medium ( also known as Biomphalaria glabrata embryonic cell line medium ) is constituted by 22% ( v/v ) Schneider’s Drosophila Medium , 4 . 5 g/L lactalbumin hydrolysate , 1 . 3 g/L galactose , 0 . 02 g/L gentamycin in MilliQ water , pH 7 . 0 . WKM cells from zebrafish were isolated as described before and plated at 4 × 105 cells/well in a 96-well plate in 200 µL of L-90 media and incubated for 3 h at room temperature . Cells were stained with 5 µM Draq5 ( Thermo Fisher Scientific ) for 10 min and subsequently run on the ImageStreamX Mark II ( Amnis Millipore Sigma ) , where 10 , 000 nucleated and focused events were recorded for each sample ( n = 8 ) . Erythrocytes were outgated to enrich for immune-relevant cells and prevent overclustering in the subsequent analysis . The latter is due to the fact that fish erythrocytes are nucleated and their biconcave shape results in different morphological feature values only depending on their orientation during image acquisition . The P . canaliculata hemocytes were stained with 5 µM Draq5 for 10 min , moved to ice , and subsequently run on the ImageStreamX Mark II , where 10 , 000 nucleated and focused events were imaged for each sample ( n = 5 ) . Staphylococcus aureus ( Thermo Fisher Scientific ) were resuspended in PBS at the final concentration of 100 mg/mL and incubated with 5 µM CTV ( Thermo Fisher Scientific ) for 20 min . Labeled bacteria were centrifuged and resuspended in PBS for three times to remove unbound dye and then stored at −20 °C as single-use aliquots . Cells , obtained from fish WKM or snail hemolymph and in a single cell suspension , were plated in a 96-well plate at a concentration of 4 × 105 cells/well in 200 µL of medium and incubated with 2 × 107 CTV-coupled S . aureus/well for 3 h at room temperature . As control , the phagocytosis was inhibited incubating the cells + CTV-S . aureus mix either on ice ( for both species ) or with 0 . 08 mg/mL CCB for zebrafish cells or with 30 mM EDTA and 10 mM HEPES for apple snail cells ( Cueto et al . , 2015; Li et al . , 2006 ) . After 2 h and 30 min , we added 5 µM DHR ( Thermo Fisher Scientific ) to the cell suspension to stain cells positive for ROS production . To control for this treatment with DHR , we incubated one aliquot of cells with 10 ng/mL phorbol 12-myristate 13-acetate ( PMA ) to induce ROS production . At 2 h and 50 min since the beginning of incubation with CTV-S . aureus , all the samples were stained with 5 µM Draq5 for 10 min . After 3 h incubation with bacteria , cells were moved and stored on ice and subsequently run on the ImageStreamX Mark II , where 10 , 000 nucleated and focused events were imaged for each sample ( at least n = 4 snail and n = 6 fish ) at a speed of 1 , 000 images/s . Following cell preparation , data were acquired from each sample on the ImageStreamX Mark II at 60× magnification , slow/sensitive flow speed ( 1 , 000 images/s ) , using 633 , 488 , and 405 nm laser excitation . BF was acquired on channels 1 and 9 , DHR ( 488 nm excitation ) on channel 2 , CTV-S . aureus ( 405 nm excitation ) on channel 7 , Draq5 ( 633 nm excitation ) on channel 11 , and SSC was acquired on channel 6 . Single-color controls were also acquired for each fluorescent channel to allow for fluorescence spillover correction . An interactive map representing the pipeline , the software used , the format of the exported files , and an approximation of time required for running the individual steps is provided in Figure 1—figure supplement 1 . Raw images from the ImageStreamX Mark II system ( RIF files , a type of modified 16-bit TIFF file ) were compensated ( spillover and other corrections applied ) , background was subtracted , and features were calculated using IDEAS 6 . 2 software ( Amnis Millipore , free for download once an Amnis user account is created ) . The resulting compensated image files ( CIF files ) were used to quantify features for all cells and samples . Supplementary files 1 and 2 report the list of features used for each organism and for each experiment and their description . These per-object feature matrices ( DAF files ) were then exported from IDEAS into FCS files . Exported FCS files were processed in R ( R Development Core Team , 2014 ) . In order to trim redundant features that contribute noise but little new information , Spearman’s correlation values for each pair of features were calculated using all events of a representative sample and one of the features of the pair was trimmed when correlation between the two was ≥0 . 85 ( Figure 2—figure supplement 1; Caicedo et al . , 2017 ) . The Spearman’s correlation of the mean values of remaining features per each sample was then used to identify outliers among sample replicates . Samples with correlation of mean feature values below 0 . 85 with the set were discarded ( Figure 2—figure supplement 2 , Figure 3—figure supplement 1 , Figure 4—figure supplement 1 , Figure 5—figure supplement 1 ) , although in general the replicates were consistent . Also , while morphological features did not require any transformation or normalization , fluorescence intensity features were transformed using the estimateLogicle ( ) and transform ( ) functions from the R flowCore package ( Ellis et al . , 2018; Hahne et al . , 2009 ) to improve homoscedasticity ( homogeneity of variance ) of distributions . DNA intensity features were also normalized to align all 2N and 4N peak positions and remove intensity drift between samples ( Figure 2—figure supplement 3 ) using the gaussNorm ( ) function from flowStats package ( Hahne et al . , 2018 ) . The processed data was exported from R ( R Development Core Team , 2014 ) using writeflowSet ( ) function in flowCore package ( Ellis et al . , 2018; Hahne et al . , 2009 ) as CSV or FCS files , depending on downstream needs for the file output . These processed data files were then imported into the VorteX clustering environment for X-Shift k-nearest-neighbor clustering ( free to install ) ( Samusik et al . , 2016 ) . X-Shift was selected as a clustering method for Image3C based on a previously published analysis and comparison of clustering methods ( Weber and Robinson , 2016 ) . From that work , we determined that X-shift represents an optimal trade-off: identifying low-frequency populations , accurately identifying ‘true’ clusters ( i . e . , F1 scores ) , not requiring for a priori knowledge of the number of clusters ( populations ) and having reasonable runtimes ( due to hardware CPU requirements ) . We also did an early comparison of X-shift with K-means ( data not shown ) and determined that K-means was insufficient for our purposes as known cell populations were not well represented by clusters , and we did not want to specify the number of expected clusters into the method’s input parameters since this would not be known in experimental use . During the import into VorteX , all features were scaled to 1 SD to equalize the contribution of features towards clustering . Clustering was performed in VorteX testing a range of k values ( typically from 5 to 150 ) , choosing a final k value using the ‘find elbow point for cluster number’ function in VorteX and confirming visually that over- or underclustering did not occur . FDL graphs of a subset of cells obtained from each set of samples were also generated in VorteX , and cell coordinates in the resultant 2D space were exported along with graphML representation of the FDL graph . Finally , tabular data ( CSV files ) was exported from VorteX including a master table of every cell event with its cluster assignment and original sample ID , as well as a table of the average feature values for each cluster and counts of cells per cluster and per sample . Clustering results were further analyzed and plotted in R ( R Development Core Team , 2014 ) by merging all cell events and feature values with cluster assignments and X/Y coordinates for FDL graph . Using this merged data and the graphML file exported from VorteX , new FDL graphs were created for each treatment condition using the igraph package ( Csardi and Nepusz , 2006 ) in R ( R Development Core Team , 2014 ) . Statistical analysis of differences in cell counts per cluster by condition was performed using negative binomial regression of cell counts per cluster , plots of statistic results and other results were generated , and CSV files containing cell ID , sample ID , feature values , and X/Y coordinates in FDL graph were exported for each sample . The subsequent use of FCS Express Plus version 6 ( DeNovo software , free alternative are mentioned later in the text ) allowed visualization of cell images using DAF/CIF files by cluster and customized subsets of the FDL graphs . DAF files were opened in FCS Express Plus software , and the ‘R add parameters’ transformation feature with a custom script was used to merge the clustering data saved in the CSV files generated above with both DAF and CIF files ( feature values and image sets , respectively ) . FCS Express Plus was utilized at this stage of work because it is the only platform currently available that works with Amnis DAF and CIF files while also running transformation processes driven by R scripting . ImageJ Bio-Formats allows reading images from DAF and CIF files , but we got pixels with a value much higher than expected , probably due to a bug that has not been fixed yet . This allowed to visualize image galleries of cells within each cluster and gate by features of interest on 2D plots ( more traditional flow cytometry analysis ) for exploring the clustering results and identifying clusters and populations of interest . FCS Express Plus is a proprietary software , but similar results can be obtained with IDEAS software where a text file with cluster IDs for each image event can be imported and the cluster information can be matched to the event images . The full complement of R packages used includes flowCore ( Ellis et al . , 2018; Hahne et al . , 2009 ) , flowStats ( Hahne et al . , 2018 ) , igraph ( Csardi and Nepusz , 2006 ) , ggcyto ( Jiang , 2015 ) , ggridges ( Wilke , 2018 ) , ggplot2 ( Wickham , 2016 ) , stringr ( Wickham , 2010 ) , hmisc ( Harrell and Dupont , 2019 ) , and caret ( Kuhn , 2008 ) . Figure 1—figure supplement 1 can be used as an interactive map of Image3C pipeline , where , upon clicking on the different portions of the pipeline , the users will be automatically directed to the corresponding sections of our GitHub repository . The GitHub repository at https://github . com/stowersinstitute/LIBPB-1390-Image3C reports a detailed description of all these processing steps and includes sample scripts , workflow files , and example datasets and tutorials . Once the clusters were defined with the previously described de novo clustering analysis , we used a CNN ( LeCun et al . , 1989 ) based on the architecture of DenseNet ( Huang et al . , 2017 ) for image classification . Out of all the events captured with the ImageStreamX Mark II system , we selected only single nucleated objects applying gates on area vs . aspect ratio plot and Draq5 intensity plot to achieve this selection , respectively . For these objects , we exported 16-bit TIFF images ( one channel per fluorescence/BF image ‘color’ ) using IDEAS 6 . 2 software . Because images from the ImageStreamX Mark II system have nonuniform sizes , each image was cropped or padded to 32 × 32 pixels using NumPy indexing ( van der Walt et al . , 2011 ) in a Python script . The CNN consists of three dense blocks that transition from three-channel image input of 32 × 32 × 3 to a final size of 4 × 4 × 87 with 87 feature maps . A dense block includes three convolution layers , each followed by leaky ReLU activation . The last step of the block is a strided convolution used to down-sample the width and height of the feature maps by a factor of 2 . The final layer of the CNN flattens the 4 × 4 × 87 array into a 1D vector of length 1392 and is fully connected to the output layer , that is , a 1D vector with a length of the number of classes for prediction . The CNN used softmax cross-entropy for the loss function with L2 regularization , and the Adam optimizer ( Kingma and Ba , 2014 ) was used to minimize the loss function . The CNN was implemented in Python using the TensorFlow platform ( Abadi et al . , 2015 ) and the SciPy ecosystem ( Oliphant , 2006; Oliphant , 2007; Pedregosa et al . , 2011 ) . The CNN was used to train a classifier using over 35 , 000 images of P . canaliculata hemolymph cell types in homeostasis condition acquired with the ImageStreamX Mark II system . The event images were randomly split in 80% for training , 10% for testing during training , and 10% for final validation . The first 80% of the images were used together with their cluster IDs obtained by the de novo clustering to train the classifier . The learning rate for the Adam optimizer was set to 0 . 0006 with a decaying learning rate starting at 0 . 001 and decreasing by 1% each step . The training proceeded for 25 , 000 iterations with a size of 256 randomly selected images for each iteration . After each iteration , 10% of the cells of the original P . canaliculata dataset was used to test the classifier . The loss and accuracy of the CNN were recorded after every 100 iterations to monitor the performance . The CNN loss was defined by the softmax of the cross-entropy ( Dahal , 2017 ) between the final output and the one-hot-encoded image labels . To avoid the CNN memorizing the training set , L2 regularization was applied to the weights . The training and test sets follow the same accuracy and loss trends over all iterations , indicating the training set is not memorized and can generalize to predict the test set . The finally trained classifier was tested on the remaining 10% of images that were completely new for the CNN . The trained model was saved for future use , so new images can be inferred by the network to predict cell types . The inference is very fast because only one forward pass is made through the network and no backpropagation occurs . The result of the inference is a vector with length equal to the number of cell type classes . Each element of the vector will be the probability of the cell belonging to the corresponding class; the sum of the vector must be 1 . Inferring a complete experiment will provide a probability vector for each image; the list of probability vectors can be saved as CSV file . For new images , the inference results will need to be examined to ensure the predictions are reliable . A large majority of the probability vectors should have a maximum greater than 0 . 5 , and a subset of the images should be visually inspected to verify proper class assignment . The interactive map of Image3C pipeline ( Figure 1—figure supplement 1 ) includes also the training and the use of the CNN . The GitHub repository at https://github . com/stowersinstitute/LIBPB-1390-Image3C reports a detailed description of all these processing steps , includes sample scripts , workflow files , and example datasets and tutorials and can be easily accessed by clicking on the right side of the Image3C interactive map ( Figure 1—figure supplement 1 ) . Negative binomial regression was performed on tables of cell counts per cluster and per sample and plots were generated using R ( R Development Core Team , 2014 ) with the edgeR package ( Robinson et al . , 2010 ) , which was developed for RNA-seq analysis , but includes generally applicable and user-friendly wrappers for regression , modeling analysis , and plotting of results . For the comparison of cellular hemolymph composition between females and males of P . canaliculata , a one-way ANOVA with subsequent FDR ( Benjamini–Hochberg , correction for multiple testing ) was used . | Cells are the building blocks of all living organisms . They come in many types , each with a different role . Understanding the composition of cells , i . e . , how many cells and which types of cells are present inside an organ can indicate what that organ does . It can also reveal how that organ changes under different conditions , like during an infection or treatment . The most powerful methods for studying cells work well for species researchers already know a lot about , such as mice , zebrafish or humans , but not for less studied animals . To change this Accorsi , Box , Peuß et al . created a new tool called Image3C to be used for studying the composition of cells in less researched organisms . Instead of using reagents that only work for specific species , the tool uses molecules that work across many species , like dyes that stain the cell nucleus . A cell-sorting machine , known as a flow cytometer , connected to a microscope then takes pictures of hundreds of stained cells each second and Image3C groups them based on their appearance , without the need for any prior knowledge about the cell types . Accorsi et al . then tested Image3C on immune system cells of zebrafish , a well-studied animal , and apple snails , an under-studied animal . For both species , the tool was able to sort cells into groups representing different parts of the immune system . Image3C speeds up the grouping process and reduces the need for user intervention and time . This lowers the risk of bias compared to manual counting of cells . It can sort cells even when the types of cells in an organism are unknown and even when specialized reagents for an organism do not exist . This means that it could characterise the cell make-up of new tissues coming from organisms never studied before . Access to this uncharted world of cells stands to reveal previously inaccessible clues about how organs behave and evolve and allow researchers to investigate the impact of environmental changes on these cells . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"tools",
"and",
"resources"
] | 2021 | Image3C, a multimodal image-based and label-independent integrative method for single-cell analysis |
To evolve and to be maintained , seasonal migration , despite its risks , has to yield fitness benefits compared with year-round residency . Empirical data supporting this prediction have remained elusive in the bird literature . To test fitness related benefits of migration , we studied a partial migratory population of European blackbirds ( Turdus merula ) over 7 years . Using a combination of capture-mark-recapture and radio telemetry , we compared survival probabilities between migrants and residents estimated by multi-event survival models , showing that migrant blackbirds had 16% higher probability to survive the winter compared to residents . A subsequent modelling exercise revealed that residents should have 61 . 25% higher breeding success than migrants , to outweigh the survival costs of residency . Our results support theoretical models that migration should confer survival benefits to evolve , and thus provide empirical evidence to understand the evolution and maintenance of migration .
The adaptive function of migration has often been hypothesized to be a selective advantage to escape adverse situations caused by seasonal fluctuations of food resources or environmental conditions . This seasonality may impose considerable constraints to life , particularly during the winter season . Seasonal migration allows animals to cope with temporal environmental fluctuations by moving between geographically distant habitats ( Fryxell and Sinclair , 1988 ) . Given that much of our planet offers seasonally varying resources , it is not surprising that migration has evolved repeatedly in many taxa ( Chapman et al . , 2011 ) . Theoretical research on the evolution of migration ( Lundberg , 1987; 1988; Taylor and Norris , 2007; Griswold et al . , 2010; Kokko , 2011; Shaw and Levin , 2011; Shaw and Couzin , 2013 ) has yielded a key prediction: migration should offer either survival or breeding benefits compared to residency . In anadromous fish , for example , individuals migrate between freshwater and ocean habitats . Recent comparisons of migrant and resident steelheads ( Oncorhynchus mykiss ) found that female migrants have higher fecundity than females that remain in fresh water streams ( Satterthwaite et al . , 2009; Hodge et al . , 2014 , 2016 ) . Similarly , the noctuid moth ( Autographa gamma ) performs a multi-generational migration which confer substantial reproductive benefits by allowing a lineage to spread to multiple sites ( Chapman et al . , 2012 ) . Regarding survival benefits , individuals of a fresh water fish ( Rutilus rutilus ) , increase their survival during the winter by migrating from lakes to streams to avoid predation risks ( Skov et al . , 2013 ) . In birds , seasonal migration has often been argued to bring about survival benefits , as it allows individuals to avoid inhospitable conditions during the non-breeding season , while the same region can offer abundant resources during the breeding season ( Lack , 1954 ) . Species exhibiting polymorphisms in migratory behavior provide an excellent opportunity to test predictions of fitness components . In partially migratory species , some individuals migrate while others remain as year-round residents , thereby allowing for between-group comparisons within the same population . Theory predicts that if residency enhances breeding success in territorial birds , then there should be a corresponding benefit to migrants; higher survival over non-breeding season is a clear , but empirically understudied , possibility ( Lundberg , 1987; Kokko , 2011 ) . Despite the extensive research done on bird migration , there is limited empirical evidence regarding its fitness benefits , as data on fitness-related variation in migratory strategies are logistically difficult to collect in the field . Despite logistical challenges , studies on European robins ( Erithacus rubecula ) and American dippers ( Cinclus mexicanus ) report that migrants have lower survival and reproductive success than residents ( Adriaensen and Dhondt , 1990; Gillis et al . , 2008; Green et al . , 2015 ) . Further , a recent study comparing fitness measures of resident and migrant cormorants ( Phalacrocorax aristotelis ) reported higher reproductive success in residents compared to migrants ( Grist et al . , 2017 ) . We studied a partially migratory population of European blackbirds ( Turdus merula ) ( Figure 1 ) to test whether migration confers survival benefits during the winter . The migrants of our population overwinter , on average , 800 km west-southwest from the breeding grounds ( Fudickar and Partecke , 2012 ) ( Figure 2a and b ) and the majority of migrants are females ( Fudickar et al . , 2013 ) . We fitted multi-event survival models using presence-absence data obtained by capture-mark-recapture and radio-telemetry of 192 resident and 70 migrant free-living blackbirds over the course of seven years . These models account for variation in re-encounter probabilities in relation to space , time and behaviour of the birds . We compared the survival probabilities between residents and migrants during two different seasons: summer ( mean start date: March 2 ± 14 . 5 days - mean end date: November 2 ± 7 . 4 days ) and winter ( mean start date: November 3 ± 7 . 4 days - mean end date: March 1 ± 14 . 5 days ) . Based on theoretical models of partial migration in birds ( Kokko , 2011 ) , which assume that residency offers reproductive benefits ( access to better breeding territories ) and that migration should confer survival benefits for at least some individuals if the winter conditions at the breeding ground are harsh , we predicted that migrants should have higher survival probabilities during the winter period , whereas summer survival might not differ between migrants and residents .
We found that winter mortality is an important determinant of lifespan , as blackbirds had lower probability to survive the winter ( Φ = 0 . 60; 95% CI = 0 . 55–0 . 66 ) than the summer season ( Φ = 0 . 89; 95% CI = 0 . 82–0 . 94 ) ( Table 1 , model 3 ) despite the shorter duration of the former season . This result strongly supports the hypothesis that migration confers survival benefits compared with residency as an alternative strategy . There was no difference between juveniles and adults in survival probability within a season . Juveniles ( Φ = 0 . 89; 95% CI = 0 . 80–0 . 94; model 4 Table 1 ) have similar probability to survive the summer as adults ( Φ = 0 . 90; 95% CI = 0 . 83–0 . 94; model 4 Table 1 ) . During winter , juveniles also have a comparable probability ( Φ = 0 . 59; 95% CI = 0 . 49–0 . 68; model 4 Table 1 ) to survive as adults ( Φ = 0 . 61; 95% CI = 0 . 55–0 . 67; model 4 Table 1 ) . In line with our predictions , migratory European blackbirds had higher winter survival rates than resident blackbirds . The best model ( model 1 , Table 1 ) estimated markedly higher winter survival for migrants ( Φ = 0 . 73; 95% confidence intervals ( CI ) = 0 . 62–0 . 81 , Figure 3 ) than for residents ( Φ = 0 . 57; 95% CI = 0 . 50–0 . 63 , Figure 3 ) , taking into account the lower detection probability for migrants ( P=0 . 19; 95% CI = 0 . 13–0 . 26 , Figure 3 ) compared to residents ( P=0 . 74; 95% CI = 0 . 69–0 . 78 ) . Our second model , which included sex and had modest support ( model 2 , delta AICc = 0 . 95 , Table 1 ) , predicted that migrants have higher winter survival probability than residents , which was also predicted by model 1 . Sex differences were not substantial ( during summer: male residents Φ = 0 . 90; 95% CI = 0 . 83–0 . 95; female residents Φ = 0 . 89; 95% CI = 0 . 89–0 . 94; male migrants Φ = 0 . 95; 95% CI = 0 . 89–0 . 98 , female migrants Φ = 0 . 94; 95% CI = 0 . 87–0 . 97; during winter: male residents Φ = 0 . 58; 95% CI = 0 . 51–0 . 65; female residents Φ = 0 . 53; 95% CI = 0 . 44–0 . 62; male migrants Φ = 0 . 75; 95% CI = 0 . 63–0 . 84 , and female migrants Φ = 0 . 71; 95% CI = 0 . 60–0 . 81; detection probability was lower for migrants ( P=0 . 19; 95% CI = 0 . 13–0 . 26 ) than for residents ( P=0 . 74; 95% CI = 0 . 69–0 . 78 ) ) . It is reassuring that both models 1 and 2 agree on the importance of residency vs . migration in winter , while we refrain from making strong statements regarding the effect of sex , given that Burnham and Anderson , 2002 advise against considering inferior models competitive in cases like our model 2 ( delta AIC within about 0–2 units of the best model , the difference being caused by one parameter added to the best model and the log-likelihood essentially unchanged ) . Our findings support the theoretical predictions that migration yields survival benefits during the winter . In addition , our results provide an explanation for the maintenance of the migrant phenotype in the partially migratory population of European blackbirds that we studied . Residency is predicted to provide reproductive benefits given that year-round occupancy provides , for example , advantages in establishing breeding territories ( Kokko , 2011 ) . The two phenotypes can persist as evolutionary stable strategies ( ESS ) due to frequency dependent selection if the overall fitness of migrants and residents is equal ( Lundberg , 1987 ) . Given the lack of data on the reproductive performance of migrants and residents in our present study , we estimated how much the reproductive performance of residents should be to compensate the survival benefits of migration . If we assume migrant and resident winter survival to be 0 . 73 and 0 . 57 respectively , and summer survival of 0 . 89 for both strategies , then we can estimate the expected number of reproductive attempts for a migrant as 0 . 73 + 0 . 73×0 . 891-0 . 73×0 . 89 = 2 . 58 , and 0 . 57 + 0 . 57×0 . 891-0 . 57×0 . 89 = 1 . 60 for residents . Therefore , the expected lifetime number of reproductive seasons is 61 . 25% higher for migrants than for residents due to higher survival of the former . This calculation assumes that the first breeding season requires one overwintering to be completed successfully , while all other events require an additional surviving sequence of summer followed by winter before a new breeding event can happen . The format for this assumption is s1s2 / ( 1– s1s2 ) which is the solution for the series s1s2 + ( s1s2 ) 2+ ( s1s2 ) 3+… , ( s1 corresponds to winter survival probability and s2 to summer survival probability ) , each subsequent term requiring one sequential survival event through one summer and one winter season . We conclude from this calculation that the reproductive performance of residents would have to be 61 . 25% higher than in migrants to achieve equal fitness of the alternative strategies . Such benefits could come about from prior residency effects ( either occupying a better territory or avoiding floater status ) , combined with a longer time spent in the breeding habitat which can make multiple nesting or re-nesting ( in case of failure ) more likely . Considering that blackbirds are a multi-brood species ( 2–3 broods a year ) , it could be possible that residents gain a 61 . 25% higher breeding success compared to migrants . Future studies need to confirm these calculations . If resident breeding success is higher than 61 . 25% , then the fitness of migrants will be lower than the fitness of residents and migration would be a conditional strategy operating under frequency-dependent selection . For conditional migration strategies , some intrinsic phenotypic characteristics ( sex , age , dominance ) result in a need to adopt a strategy that might yield overall lower fitness than what residents on average achieve , but it is the better choice to optimize individual fitness ( Lundberg , 1987 ) . To distinguish between these two alternatives , data of reproductive success for this species are needed ( note that comparisons within existing studies , such as Grist et al . , 2017 on cormorants , do not incorporate all the processes we have envisaged above ) . It is also plausible that year-to-year variation of winter environmental conditions at the breeding grounds play a role shaping the incidence of migration versus residency over time . For instance , during a harsh and long winter , the survival of residents might be lower compared to a mild and short winter . If fewer residents survive an unusually harsh winter and establish breeding territories during the subsequent breeding season , many high-quality territories would remain vacant for migrants to take advantage of after arrival in the spring . Furthermore , if residents that do survive harsh winters begin the breeding season in poor condition , then physically dominant migrants could successfully take-over prime territories from residents . Under this scenario , the prior residency effect would not be acting at full strength ( Drent et al . , 2003; Jahn et al . , 2010; Kokko , 2011 ) and migrants would gain breeding benefits . We found no evidence for sex differences in survival ( though some ambiguity remains , as a moderately supported model two includes sex as an explanatory variable — note that the best model does not ) . This raises the question: why are females more likely to migrate than males in the study population ( Fudickar et al . , 2013 ) ? We can think of two potential reasons for this observation: either there is differential survival , or differential breeding success for each sex . Regarding survival , one line of thinking is to argue that residency is more dangerous for females than for males , because overwintering blackbirds form foraging flocks and an individual’s access to food is related to its position within the flock hierarchy ( Lundberg and Schwabl , 1983 ) . Within these flocks , females are subordinate to males ( Lundberg and Schwabl , 1983; Lundberg , 1985 ) . Therefore , females would be predicted to suffer higher mortality if they remain as residents during winter , when food is limited , than if they migrate . However , our data do not align perfectly with this interpretation: if overwinter survival during residency was a strong factor driving sex differences in migratory strategy , we ought to have seen lower winter survival in resident females than in males , but this was not the case . The other possible explanation relies on differential breeding success between sexes . It is conceivable that resident males enjoy priority access to prime territories as soon as the breeding season starts . However , it should always be remembered that females , too , may benefit from better territories , thus an early presence may be beneficial for them as well ( Creighton , 2001; Kokko et al . , 2006; Kokko , 2011; Snow , 1956 ) . It would be important to understand exactly how territory acquisition differs between males and females , especially because earlier data from the same geographical area have shown that reproductive success of migrant and resident blackbirds is sex-dependent ( Schwabl , 1983 ) such that male residents have higher reproductive success than male migrants , while female residents and migrants have similar reproductive success . Understanding the mechanisms of territory acquisition could help explain why fewer males migrate: if frequency-dependency penalizes late-arriving males whereas late breeding females are not severely penalized , then the same magnitude of survival differences will favor a larger migratory population within females than within males . In our study , we excluded 11 birds that migrated during the winter and 11 that switched strategies between years , as we considered these sample sizes to be too small for detailed inferences . Departures during the winter usually occurred during periods of cold temperatures and snow accumulation ( Fudickar et al . , 2013 ) . Extreme weather conditions and low food availability might trigger these movements during winter . Future , more extensive studies could conceivably determine lifetime fitness of these strategies . By examining the fitness benefits conferred by migration , our study is able to provide strong support for the hypothesis that migration confers winter survival benefits . Our methodology , which allows comparisons between classes that differ greatly in detectability , can hopefully also shed light on systems where benefits and risks of migration are shared by all individuals of a population , many of which are threatened by risks along their migratory flyways ( Wilcove and Wikelski , 2008 ) . Further understanding of how , where and why migratory animals die will illuminate the path to direct conservation efforts to protect migratory species .
A total of 469 blackbirds were captured and tagged in a mixed deciduous/coniferous forest in southern Germany ( N 47° 47’ , E 9° 2’ ) during spring and summer over seven consecutive years ( 2009–2016 ) . Sex and age were determined using plumage differences ( Svensson , 1992 ) . Juvenile birds were sexed using DNA-based sex identification ( Griffiths et al . , 1998 ) . To this end , a blood sample ( 50 µl ) was collected and stored in Queen’s Lysis buffer ( Seutin et al . , 1991 ) . Each bird was equipped with a radio transmitter in combination with ( i ) a light-level geolocator ( Mk 10S , and Mk 12S ≤ 1 . 2 g; British Antarctic Survey , Cambridge , UK ) during 2009–2011 , or ( ii ) light-level geolocator ( Intigeo-P65 ≤1 . 2 g Migrate Technology , Cambridge , UK ) during 2012–2013 or ( iii ) a Pinpoint GPS logger ( ≤2 g; Biotrack Ltd , Dorset , UK ) during 2014 . Birds tagged during 2015 , however , were equipped just with a radio transmitter ( mean weight ±SD: 1 . 94 g ± 0 . 12 ) . Radio transmitters were provided in 2009–2012 and 2014–2015 by Sparrow Systems , Fisher , IL , USA and in 2014 by The Swiss Ornithological Institute , Sempach , Switzerland . The total weight of the devices carried by the birds was ( mean ±SD ) 3 . 9 g ± 0 . 19 in 2009–2011; 3 . 3 g ± 0 . 20 in 2012; 4 . 15 g ± 0 . 11 in 2013; 4 . 13 g ± 0 . 11 in 2014 . The total weight of the tracking devices was less than 5% of the body weight of the birds in each year of the study . The life span of the battery was at least 12 months . The tags were attached by means of leg-loop harnesses . We collected presence–absence data at regular intervals through a manual and/or an automated radio telemetry system . Manual radio tracking was carried out twice per week using a handheld three element Yagi antenna ( AF Antronics , Inc . , Urbana , IL , USA ) and AR 8200 MKIII handheld receiver ( AOR U . S . A . , Inc . , Torrance , CA , USA ) or a handheld H antenna ( Andreas Wagener Telemetry Systems , Köln , Germany ) and a Yaesu VR 500 handheld receiver ( Vertex Standard USA , Cypress , CA , USA ) . The automated radio telemetry system consisted of 4 to 6 stationary automated receivers ( ARU ) ( Sparrow Systems , Fisher , IL , USA ) deployed at the study site . Each receiver was connected to an H antenna ( ATS , Isanti , MN , USA ) and was able to search for up to 16 different radio frequencies every 60 s . The migratory strategy of each bird was assigned based on the presence–absence data . Birds were classified as migrants if they departed at night ( determined by ARUs ) from the study site during the autumn ( September-November ) . All migrants departed between 19 September and 12 November ( mean departure date: 16 October ) . Migrants arrived during spring between 17 February and 25 March ( mean arrival date: 14 March ) . Birds were classified as residents if they remained present and alive at the study site at least until 31 November of each year . Individuals that had left the study area were searched using a Cessna airplane fitted with two H antennas and two Biotrack receivers ( Lotek Wireless Inc . , Newmarket , ON , Canada ) and a vehicle carrying a telescopic mast ( 6 m height ) and three-element Yagi antenna ( Vargarda Radio , Vargarda , Sweden ) . Out of 469 birds , 158 were excluded because their migratory strategy could not be determined before 31 November due to various reasons ( technical failure of the tracking devices , dispersal from the core study area or mortality ) . In 9 out of the 158 excluded birds , we found a radio tag with a broken harness and in 16 cases we found the tag but no signs of predation nor malfunction were evident . We concluded that 27 birds were predated ( predation signs e . g . the carcass and/or feathers were found near the radio transmitter ) . In 106 cases , we do not know the fate of the birds . Forty-nine of the 106 birds with unknown fates were juveniles . In blackbirds , as in many other altricial bird species , post-fledgling mortality is high ( Naef-Daenzer and Grüebler , 2016 ) and fledglings can disperse several kilometres ( Paradis et al . , 1998 ) . Eleven birds that departed from the breeding grounds during the winter and 11 birds that switched strategies across years were excluded from the analysis . Finally , we excluded 27 juveniles from the analysis because we could not determine the sex due to poor quality of the blood sample . Conversely 262 birds were classified during the autumn and were included in the survival analysis . Out of 262 , 192 were classified as residents ( 69 females: 52 adults and 17 juveniles; 123 males: 96 adults and 27 juveniles ) and 70 birds were classified as migrants ( 45 females: 28 adults and 17 juveniles; 25 males: 17 adults and 8 juveniles ) . To estimate seasonal survival probabilities , each calendar year was divided into two operationally defined ‘seasons’: summer and winter . Summer was defined as the period of time between the date of the first spring arrival of a migrant bird and the date of the last departure in the fall ( mean start date: March 2 ± 14 . 5 days , mean end date November 2 ± 7 . 4 days ) . To define the start of the first summer season in 2009 , the date of the very first capture ( April 23rd ) was used . Winter , in turn , corresponded to the period of time between the date of the last departure in the fall and the date of first arrival the subsequent spring ( mean start date: November 3 ± 7 . 4 days , mean end date: March 1 ± 14 . 5 days ) . Based on the presence–absence data , we generated a matrix of 15 columns , each corresponding to one respective season ( summer 2009 , followed by winter 2009–2010 , followed by summer 2010 , etc . ) and 262 rows ( one for each individual ) . Additional columns containing the covariates sex ( males and females ) , age at capture ( juveniles and adults ) and migratory status ( migrants and residents ) were added to the matrix . We implemented multi-event models using the software E-SURGE 1 . 9 . 0 ( Choquet et al . , 2009 ) . These models belong to the family of hidden Markov models . They assume that the individuals in a population independently transition between a finite set of N states ( e . g . presence , absence ) through a finite number of sampling occasions . Since the capture regime is imperfect , there is uncertainty in presence or absence of each individual . Multi-event models account for this uncertainty ( Pradel , 2005 ) . They allow a simultaneous estimation of the probability of survival ( Φ ) of a group of individuals and its detection probability ( P ) . Detection probability ( P ) is a decisive parameter because it directly influences the survival estimates and in natural populations often is less than 1 . Failing to account for this parameter can lead to incorrect conclusions in capture mark-recapture analyses ( Gimenez et al . , 2008 ) . We used a model selection procedure to evaluate the performance of 16 candidate models that included the effects of sex , age at capture , migratory strategy and season ( Table 1 ) . Model performance was evaluated using the Akaike Information Criterion corrected for small samples ( AICc ) . Delta AICc ( Δ AICc ) was calculated and the models ranked based on this value . After recapture , Mk 10S and Mk 12S geolocators were processed in the following way: Raw data were corrected for clock drift using Bastrak ( British Antarctic Survey ) . After visually inspecting light values of Mk 10S geolocators , a light level threshold of 16 was identified . In 2010 , we reduced the stalk length of Mk 12 s geolocators , resulting in interference from feathers with light censors . Due to interference at sunrise and sunset , we found that a light threshold of 2 was most reliable for Mk 12 s geolocators . Individual sun elevation angles were calculated using all dates that an individual was known to be on the breeding grounds . Transitions were calculated using TransEdit2 ( British Antarctic Survey ) and anomalous transitions were rejected . Latitude and longitude were calculated using Locator ( British Antarctic Survey ) following Tøttrup et al . ( 2012 ) . Intigeo-P65 geolocators were processed in the following way: Transitions were calculated using IntiProc v . 1 . 01 ( Migrate Technology Ltd ) . A threshold of 16 was used and anomalous transitions were discarded . Transition data were imported and analyzed in R using GeoLight Package ( Lisovski and Hahn , 2012 ) . The ‘in-habitat calibration’ was used to calculate individual sun elevation angles . Locations ( latitude and longitude ) were estimated using the function ‘Coord’ of the GeoLight Package . | Winter is one of the most challenging seasons for many animals . Cold temperatures , bad weather , short days , long nights and a shortage of food can impose a deadly threat . To avoid these inhospitable conditions , some animals migrate to warmer climes during the winter . These animals include many songbirds , which return to the same habitat in the following spring because it offers abundant resources that are thought to help them to breed more successfully . Yet , migration itself can be risky , and there is little empirical data on the survival benefits of migration in songbirds . Zúñiga et al . tested whether songbirds that migrate are actually more likely to survive the winter than those that do not migrate . The study focused on a population of European blackbirds over a period of seven years . Some of these birds migrated from the breeding grounds in Germany to their wintering sites in southern Europe , whereas others remained all year at the breeding grounds . Zúñiga et al . found that migrant blackbirds were 16% more likely to survive the winter than the residents . Yet during the summer , there was no difference in survival between the two groups . This raised the question , if migration confers survival benefits , why do some birds do not migrate at all ? Theory predicts that those birds that do not migrate should have some reproductive benefit instead . This makes sense given that birds which remain at the breeding grounds would have access to prime breeding sites which are limited . Using mathematical modelling , Zúñiga et al . estimated how much of reproductive benefits the residents would need to outweigh their greater risk of not surviving the winter . The model predicted that residents should have at least 61 . 25% higher breeding success than migrants . The results provide empirical evidence to help scientist understand how migration evolves and becomes maintained in animal population . Future studies are now needed to confirm the estimated breeding success of both groups . Also , because many songbirds are threatened by human activity during migration and at their overwintering sites , future studies to understand how , where and why migratory songbirds die will be important to direct the conservation efforts to protect migratory species . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"ecology",
"short",
"report"
] | 2017 | Migration confers winter survival benefits in a partially migratory songbird |
Bovines have evolved a subset of antibodies with ultra-long heavy chain complementarity determining regions that harbour cysteine-rich knob domains . To produce high-affinity peptides , we previously isolated autonomous 3–6 kDa knob domains from bovine antibodies . Here , we show that binding of four knob domain peptides elicits a range of effects on the clinically validated drug target complement C5 . Allosteric mechanisms predominated , with one peptide selectively inhibiting C5 cleavage by the alternative pathway C5 convertase , revealing a targetable mechanistic difference between the classical and alternative pathway C5 convertases . Taking a hybrid biophysical approach , we present C5-knob domain co-crystal structures and , by solution methods , observed allosteric effects propagating >50 Å from the binding sites . This study expands the therapeutic scope of C5 , presents new inhibitors , and introduces knob domains as new , low molecular weight antibody fragments , with therapeutic potential .
By the end of 2019 , over 60 peptide drugs have received regulatory approval , with an estimated 400 more in active development globally ( Lau and Dunn , 2018; Lee et al . , 2019 ) . As a potential route to discover therapeutic peptides , we previously reported a method for deriving peptides from the ultra-long heavy chain complementarity determining region 3 ( ul-CDRH3 ) , which are unique to a subset of bovine antibodies ( Macpherson et al . , 2020 ) . We have shown that knob domains , a cysteine-rich mini-domain common to all ul-CDRH3 , can bind antigen autonomously when removed from the antibody scaffold ( Macpherson et al . , 2020 ) . This allows peptide affinity maturation to be performed in vivo , harnessing the cow’s immune system to produce peptides with complex stabilising networks of disulphide bonds . For the discovery of knob domain peptides , immunisation of cattle is followed by cell sorting of B-cells using fluorescently labelled antigen . A library of antigen-specific CDRH3 sequences is created by performing a reverse transcription polymerase chain reaction ( RT PCR ) on the B-cell lysate , followed by a PCR using primers specific to the conserved framework regions which flank CDRH3 ( Macpherson et al . , 2020 ) . Upon sequencing , ul-CDRH3s are immediately evident and the knob domains can be expressed recombinantly as cleavable fusion proteins ( Macpherson et al . , 2020 ) . This method for the discovery of knob domain peptides was established using complement component C5 , and we reported peptides which bound C5 with affinities in the pM–low nM range ( Macpherson et al . , 2020 ) . Herein , we use these novel peptides to probe the structural and functional aspects of C5 activation . C5 is the éminence grise of the complement cascade’s druggable proteins , and the target of effective therapies for diseases with pathogenic complement dysregulation , of which paroxysmal nocturnal haemoglobinuria ( Rother et al . , 2007 ) and atypical haemolytic uraemic syndrome ( Nürnberger et al . , 2009 ) are notable examples . Six monoclonal antibodies targeting C5 have reached , or are entering , clinical trials , closely followed by C5-targeting immune evasion molecules ( Romay-Penabad et al . , 2014 ) , aptamers ( Biesecker et al . , 1999 ) , cyclic peptides ( Ricardo et al . , 2014 ) , interfering RNA ( Borodovsky et al . , 2014 ) , and small molecules ( Jendza et al . , 2019 ) . Currently , C5 inhibitors are being trialled for the treatment of acute respiratory distress syndrome arising from severe acute respiratory syndrome coronavirus 2 ( SARS-CoV-2 ) infection ( Smith et al . , 2020; Wilkinson et al . , 2020; Zelek et al . , 2020 ) and for the neuromuscular disease myasthenia gravis ( Albazli et al . , 2020 ) . C5 is the principal effector of the terminal portion of the complement cascade . At high local C3b concentrations , arising from activation of either or both of the classical ( CP ) and mannose binding lectin ( LP ) pathways , aided by the amplificatory alternative pathway ( AP ) , C5 is cleaved into two moieties with distinct biological functions . Cleavage is performed by two convertases: C4bC2aC3b , formed in response to CP or LP activation ( Takata et al . , 1987 ) ( henceforth the CP C5 convertase ) , and C3bBbC3b , formed in response to AP activation ( DiScipio , 1981 ) ( henceforth the AP C5 convertase ) . Although the constitutive components of the C5 convertases differ , they are thought to be mechanistically identical . Once cleaved , the C5a fragment is the most proinflammatory anaphylatoxin derived from the complement cascade . When signalling through C5aR1 and C5aR2 , C5a is a strong chemoattractant recruiting neutrophils , eosinophils , monocytes , and T lymphocytes to sites of complement activation , whereupon it activates phagocytic cells , prompting degranulation . C5b , meanwhile , interacts with C6 , recruiting C7–C9 to form the terminal C5b-9 complement complex or TCC ( Lachmann and Thompson , 1970 ) . Once inserted into a cell membrane , the TCC is referred to as the membrane attack complex ( MAC ) , a membrane-spanning pore which can lyse sensitive cells ( Götze and Müller-Eberhard , 1970 ) . Aspects of the structural biology of C5 are well understood due to a crystal structure of the apo form ( Fredslund et al . , 2008 ) and a number of co-crystal structures of C5 with various modulators . By virtue of its constitutive role in the terminal pathway , C5 is a recurrent target for immune evasion molecules and structures have been determined of C5 in complex with an inhibitory molecule derived from Staphylococcus aureus , SSL-7 ( Laursen et al . , 2010 ) , as well as several structurally distinct examples from ticks: OmCI ( Jore et al . , 2016 ) , RaCI ( Jore et al . , 2016 ) , and Cirp-T ( Reichhardt et al . , 2020 ) . Additionally , the structures of C5 with the inhibitory monoclonal antibody ( mAb ) eculizumab ( Schatz-Jakobsen et al . , 2016 ) , of C5 with a small molecule inhibitor ( Jendza et al . , 2019 ) , and of C5 with the complement-depleting agent cobra venom factor ( CVF ) ( Laursen et al . , 2011 ) have all been determined . Here , we probe C5 with knob domain peptides and explore the molecular processes which underpin allosteric modulation of this important drug target . This study is the first to investigate the molecular mechanisms and pharmacology of this recently isolated class of peptide .
We have previously shown that antigen-specific , disulphide-rich knob domain peptides derived from bovine antibodies have great potential for therapeutic utility . Using this approach , we obtained four knob domain peptides: K8 , K57 , K92 , and K149 , which we have shown to display tight binding to human C5 . Previously we reported equilibrium dissociation constants of 17 . 8 nM for K8 , 1 . 4 nM for K57 , <0 . 6 nM for K92 , and 15 . 5 nM for K149 ( Macpherson et al . , 2020 ) . For functional characterisation of the peptides , we performed complement assays for CP and AP activation in human serum , assessing C5b neo-epitope formation and C5a release ( schematically presented in Figure 1A , B ) , in combination with orthogonal ELISAs , measuring C3b and C9 deposition . Here , we show that K57 was a potent and fully efficacious inhibitor of C5 activation , preventing release of C5a , and deposition of C5b and C9 . As expected , there was no effect on C3b , which is upstream of C5 ( Figure 1C , D ) . In contrast , K149 was a high-affinity silent binder with no discernible effect on C5a release , formation of C5b neo-epitope or C9 deposition , even at peptide concentrations in excess of 100 × KD ( see Supplementary file 1 Section 1 ) . K8 and K92 exerted more nuanced allosteric effects on C5 ( Figure 1C , D ) . By ELISA , K92 partially prevented C5 activation by the AP , but , intriguingly , no effect was observed in CP assays , suggesting K92 selectively inhibits C5 activation by the AP C5 convertase , but not the CP C5 convertase . Partial antagonists , where the degree of inhibition for the asymptotic concentrations of a dose–response curve ( Emax ) is below 100% , are an impossible mode of pharmacology for orthosteric antagonists ( Klein et al . , 2013 ) , and we therefore propose that K92 operates by a non-steric mechanism . K8 was also demonstrably allosteric , partially inhibiting both the AP and CP in ELISA experiments . For K8 and K92 , no effect on C3b deposition was detected . When tested in CP and AP haemolysis assays ( Figure 1E , F ) , K57 was a potent and fully efficacious inhibitor of complement-mediated cell lysis . Consistent with the ELISA data , K92 was active solely in the AP-driven haemolysis assay , achieving Emax values of 30–40%; while K8 was efficacious in the CP assay but did not show activity in the AP assay below 10 µM , potentially a consequence of the increased serum concentration and stringency of the haemolysis endpoint . To test for cross blocking , arising from overlapping epitopes , or cooperativity between knob domains , we performed a surface plasmon resonance ( SPR ) cross blocking experiment , where , using a Biacore 8K , we saturated a C5-coated sensor chip with two 20 µM injections of knob domain peptide before injecting a different peptide at 20 µM to assess its capacity to bind . This provides a qualitative measure of cross blocking , whereby an increase in response units ( RUs ) indicates ternary complex formation , stoichiometries , or kinetics cannot be reliably derived with concurrent dissociation of both peptides ( Figure 2 ) . Saturation of C5 with K8 , K57 , or K92 did not prevent subsequent binding of the non-functional K149 ( Figure 2D ) , suggesting K149 does not share an epitope with the other ligands , nor does it significantly perturb C5 such that the other binding sites are affected . We detected negative cooperativity between K8 and K92 , whereby saturation of C5 with K8 entirely prevented binding of K92 . When the order of addition was changed and C5 was saturated with K92 , K8 was still able to bind , albeit to a lesser degree ( Figure 2B ) . Saturation of C5 with K8 also entirely eliminated binding of K57 , with a similar order of addition effect , whereby K8 could still partially bind to the C5-K57 complex ( Figure 2A ) . When C5 was saturated with K92 or K57 , only very small amounts of subsequent binding of either peptide were observed by SPR ( Figure 2C ) , suggesting that the epitopes do not overlap but that considerable negative cooperativity exists . To validate the observed K8 and K92 C5-binding modes observed in our crystal structures , we assessed the binding properties for a number of alanine mutants of K8 and K92 . For K8 , R23A and R32A mutants targeted the two salt-bridge interactions with C5 . While for K92 , where there were few electrostatic interactions mediated by side chains , we targeted a hydrogen bond , sustained by H25 , and important hydrophobic interactions with C5 , involving neighbouring aromatics W21 and F26 . While the K92 H25A mutant could not be expressed , the other mutants were tested , alongside unmodified K8 and K92 , in SPR multi-cycle kinetics experiments ( n = 3 ) . For K8 , the R23A resulted in modest twofold decrease in affinity , but R32A was markedly more attenuating , with a 715-fold drop in affinity ( Supplementary file 1 , Table 2 . 6 ) . For K92 , the loss of hydrophobic interactions with C5 in W21A and F26A mutants markedly abridged affinity with a 1209 . 2-fold and 45 . 7-fold drop in affinity , respectively ( Supplementary file 1 , Table 2 . 6 ) . To analyse the interfaces observed in the structures , we performed binding pose metadynamics ( Clark et al . , 2016 ) , an analysis typically employed to computationally evaluate the binding stability of chemical ligands ( Fusani et al . , 2020 ) . This in silico analysis suggested that both K8 and K92’s binding poses were exceptionally stable , with the interface maintaining the key interactions in spite of applied force ( Supplementary file 1 , Tables 2 . 7 and 2 . 8 ) . This , in conjunction with earlier kinetic studies ( Macpherson et al . , 2020 ) , highlights the stability of the interactions made by both knob domains . In the near absence of secondary structure , disulphide bonds appear to act as sources of stability for both peptides . For K92 , both the backbone amide and carbonyl of C23K92 participate in H-bonds with the side chain of S82C5 ( Figure 3B ) . For K8 , an interchain sulphur–π stack between the C27K8-C41K8 disulphide bond and the aromatic of Y1378C5 positions the hydroxyl group of Y1378C5 to make a H-bond with D25K8 ( Figure 3A ) . While for K92 , an intra-chain sulphur–π stack between the C9K92-C23K92 disulphide bond and the aromatic of Y14K92 orientates Y14K92 , such that its hydroxyl group participates in an interchain H-bond with N38C5 . Although antibody-derived , K8 and K92 are structurally unique variable regions . We compared the K8 and K92 knob domains to a non-redundant set of 924 non-identical sequences of paired antibody–protein antigen structures from SAbDab ( Dunbar et al . , 2014 ) . Paratopes were defined as any antibody residues within 4 . 5 Å of the antigen in the structure . The paratopes of K8 and K92 contain 18 and 10 residues , respectively , which are within the typical range of antibody paratope sizes ( Figure 3—figure supplement 4A ) . Given this similarity in size , we searched for structurally and physicochemically similar antibody paratopes from the 924 antibody complexes but no similar paratope sites were found ( Wong et al . , 2020 ) . While the limited examples preclude firm conclusions , this lack of similarity could be due to either the unusual fold of the knob domains or the differences in paratope amino acid composition . In terms of residue usage , one difference in paratope composition that is potentially universal is the presence of cysteine in the knob domains ( Figure 3—figure supplement 4B ) which is uncommon in most antibody paratopes , with the exception of the CDR1-CDR3 disulphides , which have been described in camelid VHH ( Govaert et al . , 2012 ) , and in broadly neutralising antibodies; those which cross react with several strains of a virus , and for which a disulphide bond in CDRH3 has been described in antibodies against HIV-1 ( Hutchinson et al . , 2019 ) and hepatitis C ( Flyak et al . , 2018 ) . Using Arpeggio ( Jubb , 2015 ) to identify inter- ( antigen contacting ) and intra-paratope interactions ( hydrogen bond , polar , ionic , and hydrophobic ) revealed that , on average , antibodies have 16 intra-paratope and 17 inter-paratope interactions; K8 is very close to this , with 15 intra-paratope and 17 inter-paratope interactions , whereas K92 paratope has fewer , with 9 intra-paratope and 10 inter-paratope interactions . A bovine Fab with an ul-CDRH3 was recently crystallised in complex with antigen , in this case a soluble portion of the HIV envelope ( Stanfield et al . , 2020 ) . While the low resolution of the crystal structure hindered analysis , a casual inspection of the paratope suggests that 10 intra-paratope and 10 inter-paratope interactions are sustained by the knob domain , comparable to K92 . A search for structurally homologous proteins , using the DALI protein structure comparison server ( Holm , 2020 ) , did not find any 3D structures similar to K8 or K92 , including the 14 known structures of bovine Fabs with ul-CDRH3 in the PDB . These results highlight the heterogeneity of these structural elements of the bovine immune system which likely arise through selection against a specific antigen/epitope . We next looked at homology with cyclic peptides . A recent review summarised the interactions mediated by cyclic peptides bound to proteins , across 65 co-crystal structures in the PDB ( Malde et al . , 2019 ) . This revealed that cyclic peptides on average sustain eight electrostatic interactions with their protein target , with a range of 1–20 . When we consider K8 , its 19 inter-paratope interactions are comparatively high for a peptide , while the seven inter-paratope interactions of K92 are far more typical ( Figure 3—figure supplement 4C ) . When compared to the binding sites of other C5 modulators ( Figure 4A ) , it can be observed that the epitope for K92 is entirely contained within the binding interface of a previously reported immune evasion molecule , the 23 kDa SSL7 protein from S . aureus ( Figure 4B ) . While the C5-SSL7 structure reveals a shallow binding site involving a series of five H-bonds between SSL7 and a region of β-sheet on the MG5 domain , spanning H511C5-E516C5 ( Laursen et al . , 2010 ) , here we show that K92 is wedged between the MG1 and MG5 domains , inducing a re-orientation of the side chain of H511C5 and forming a backbone H-bond with F510C5 . When comparing K92 and SSL7 , the small changes observed in the binding pose achieve different allosteric effects; SSL7 , either in isolation or in complex with its second ligand IgA , is full , or occasional partial , antagonist of both the AP and CP ( Bestebroer et al . , 2010; Laursen et al . , 2010 ) , while K92 is a selective partial antagonist of the AP . Inspection of the C5a anaphylatoxin domain reveals that the C-terminus of the C5a domain in the C5-K92 complex adopts a helical conformation , which is analogous to the C5-OmCI-RaCI complex , burying the Bb-cleavage site ( R751 ) . In other C5 structures ( including C5-apo and C5-CVF ) , this linker adopts an extended conformation following an unstructured loop and only sparse continuous electron density was observed for the linker extending from MG6 to C5a in the C5-K8 complex , possibly suggesting its R751 scissile bond is more exposed . When the MG5 domains in the C5-K92 complex and the C5-apo structure are superimposed ( Figure 4C ) , a slight twist can be observed in the MG1 domain , caused by the binding of K92 and resulting in a significant rotational movement of the C5 α-chain . A similar conformational change results from the binding of OmCI and RaCI , and to some extent K8 , as these structures are virtually superimposable . CVF can form a highly stable C3/C5 convertase , following incubation with factor D and factor B in the presence of Mg2+ ( Vogel and Müller-Eberhard , 1982 ) , which may offer a surrogate model for C5 convertase ( Laursen et al . , 2011 ) . When superimposing C5-K92 and C5-CVF ( PDB accession code 3PVM ) complexes , C5 does not adopt a similar conformation as when bound by K92 and K8 ( Figure 4—figure supplement 1 ) , potentially indicating both knob domains stabilise a different C5 conformation than binds the C5 convertase . When considering the organisation of the C5 convertases , the C5-CVF crystal structure reveals that CVF and C5 align perfectly to create a mirror image complex , with a conformational change in the C5 convertase site at arginine 751 , potentially placing C5a within range of the catalytic unit of the MG7-associated convertase complex , offering a surrogate model for C5 convertase activation ( Laursen et al . , 2011 ) . We have shown that K92 is not an orthosteric inhibitor of either the CP or the AP convertase , thereby precluding binding of the convertase to a cleft between the MG1–MG5 domains . As the K92 epitope is entirely contained within the SSL7 binding site , this is compatible with the CVF model for C5 activation , with a co-crystal structure of the ternary complex of C5 , CVF , and SSL7 ( PDB accession code 3PRX6 ) , demonstrating that the CVF and SSL7 binding sites do not cross block . Also consistent with the CVF model for C5 activation , binding of K8 to the MG8 domain would not appear to sterically block the catalytic unit . We therefore sought to further explore the apparent conformational changes in our structures . To validate the apparent conformational changes occurring in C5 due to the binding of K8 and K92 as revealed by our crystal structures , we analysed the C5-knob domain complexes by two-solution biophysical techniques – small-angle X-ray scattering ( SAXS ) and hydrogen-deuterium exchange mass spectrometry ( HDX-MS ) . SEC-SAXS , where size exclusion chromatography ( SEC ) immediately precedes the solution X-ray experiment ensuring a monodispersed sample , was performed in concert with SEC multi-angle laser light scattering ( SEC-MALLS ) . Data were collected for C5 and the C5-K8 , C5-K57 , C5-K92 , and C5-K149 complexes ( Figure 5A–C ) . SEC-MALLS confirmed that the increases in molecular weight of the complexes were consistent throughout the elution peaks ( Supplementary file 1 , Table 3 . 1 and Figure 5—figure supplement 1A ) . While SEC-SAXS elution profiles gave stable estimates of the radius of gyration ( RG ) across the tip of the peak , frames ( scattering curves collected during the elution of the sample ) from the descending elution peaks show lower RG values , suggesting the presence of unbound C5 . Frames corresponding to the tip of the peak were averaged and submitted for full SAXS analysis . For the complexes , the scattering curves showed slight increases in both the RG and solute volume ( Supplementary file 1 , Table 3 . 1 ) , with the C5-K8 complex showing the largest change and C5-K149 the smallest change , corresponding with the absence of function and suggesting K149 binds peripherally to a conformation closely resembling C5-apo . For K92 and K57 , the discrepancies observed in the mid s range indicate an overall change in flexibility of C5 upon binding of these peptides , and this tuning of dynamics may contribute to their mechanism . Consistent with earlier observations ( Fredslund et al . , 2008 ) , comparison of the C5-apo experimental data with the theoretical scattering curve revealed discrepancies in the lowest angle range , indicating C5 adopts a more elongated conformation in solution than the crystal structure would suggest ( Figure 5—figure supplement 1B ) . To better approximate C5 in solution , we performed a normal mode model analysis ( NMA ) using SREFLEX ( Panjkovich and Svergun , 2016 ) and found that elongation of the C5 model improved the χ2 from >13 to 1 . 55 . The fit of the C5-K92 complex was also markedly improved by the NMA , whereby elongation and incorporation of the peptide improved the model from an initial χ² of >20 , to 2 . 5 ( with an overall root mean square [RMS] of 3 . 8 in both cases ) . When using the C5-K8 co-crystal structure for fitting of the C5-K8 SAXS data , the absence of the C345c domain was problematic . The generation of a hybrid model where the C345c domain was reinstated initially produced a poor fit ( χ²=75 ) . A restrained rigid body analysis of this model followed by NMA refinement allowed us to significantly reduce the discrepancy to χ²=4 . 1 indicating an overall acceptable fit . The χ² value is still somewhat larger than those observed for the other complexes , which may suggest an increased flexibility around the C345c linker . This result correlates with the absence of clear electron density for the C345c domain in the crystal structure . The latter may be a consequence of K8 inducing additional flexibility to this region , which again could contribute to the efficacy of the peptide . The discrepancies between the crystal structures and the solution scattering data indicate that while permitting elucidation of the molecular interaction of the epitopes , the constraints of the crystal lattice may impede the detection of more subtle , global changes , leading to underestimation of the conformational changes induced by the peptide . To further explore such effects in solution , we used HDX-MS to provide molecular-level information on local protein structure and dynamics . HDX-MS measures the exchange of backbone amide hydrogen to deuterium in the solvent , with the rate of HDX determined by solvent accessibility , protein flexibility , and hydrogen bonding . To interpret the impact of peptide binding on C5 structural dynamics , we performed differential HDX ( ΔHDX ) analysis , comparing C5-knob domain complexes to apo C5 , where shielding of C5 residues through participation in a binding interface will prevent deuteration , while conformational changes may increase or decrease deuterium uptake , in relation to the degree of solvent exposure . For C5-K8 , the sole protected region of C5 corresponded to the epitope on the MG8 domain ( L1380C5-E1387C5 ) , although the interface was not entirely defined ( Figure 6A , see also Figure 6—figure supplement 1 and Supplementary file 1 , Table 3 . 2 ) . Additional conformational changes were observed in the neighbouring C5d domain which becomes more solvent exposed , suggesting K8 is affecting the dynamics of this domain . For the C5-K92 complex , consistent with the crystal structure , there was protection of the C5 residues located in the epitope between the MG1 and MG5 domains ( H70C5-L85C5 ) , shown in Figure 6C . There were also effects distal to the K92 binding site , notably in C5d ( I1169C5-F1227C5 ) and neighbouring CUB domain ( L1303C5-L1346C5 ) , indicating a K92-induced conformational change . Interestingly , the allosteric network can be visualised by changes in solvent exposure which propagate from the K92 epitope through MG2 domain ( L126C5-V145C5 ) and into the C5d and CUB domains . For the C5-K57 complex , the absence of a co-crystal structure meant we had no prior knowledge of the K57 epitope . However , clear protection was observed in the MG5 domain , immediately adjacent to the K92 epitope ( N483C5-L540C5 ) , with sparse areas of increased solvent exposure located in the MG6 ( Q572C5-L590C5 ) , MG8 ( L1379C5-A1388C5 ) , and C5d ( K1048C5-Y1064C5 ) domains ( Figure 6B ) . A single protected peptide was also present in the CUB domain ( G951C5-L967C5 ) , suggesting the K57 epitope may be on either the MG5 or CUB domains . There was little protection or deprotection of proteolytic fragments of the C5a domain in any of the complexes; we therefore propose that the knob domain peptides do not act by inducing conformational changes which shield the scissile arginine bond . Although in the structure of the C5-K92 complex the Bb-cleavage site is more buried compared to that in the C5-K8 complex . Taken in the context of the other changes , notably in the C5d and CUB domains , it is more probable that they affect more global changes in C5 which lower the affinity for C3b or the C5 convertases . The HDX-MS data are in good agreement with our crystallographic data , with the K8 and K92 epitopes defined as clear areas of solvent protection . The conformational change in the C5d domain and significant rotational moment of the C5 α-chain , which were evident upon alignment of the MG1 domain of apo C5 with the C5-K8 and C5-K92 co-crystal structures , also appears to manifest in solution in response to binding of the knob domains . To further home in on the K57 binding site , we measured binding to C5b and C5b-6 in an SPR single-cycle kinetics experiment ( Supplementary file 1 , Table 4 . 1 ) . Upon cleavage of C5a , the remaining domains of the α-chain undergo a substantial conformational change , mediated by rearrangement of the MG8 , CUB , and C5d domains ( Hadders et al . , 2012; Aleshin et al . , 2012 ) . The resulting C5b subunit is metastable and prone to aggregation and decay , which leaves it unable to bind C6 or form the MAC . By SPR , K8 did not bind C5b but could C5b-6 . However , K57 , K92 , and K149 all bound C5b and C5b-6 ( Supplementary file 1 Table 4 . 2 ) . For C5b , this was within twofold of their previously published affinities for C5 ( 17 . 8 nM for K8 , 1 . 4 nM for K57 , and <0 . 6 nM for K92; Macpherson et al . , 2020 ) , except for K149 , which displayed threefold higher affinity for both C5b and C5b-6 than previously reported for C5 ( 15 . 5 nM; Macpherson et al . , 2020 ) . K92 did exhibit lower affinity for C5b-6 complex than C5 , binding the complex at 6 . 7 nM , relative to <0 . 6 nM for C5 alone . As the CUB domain is significantly altered in C5b , this increases the likelihood that , of the two protected regions identified by HDX-MS , the K57 epitope is on the MG5 domain .
Functional characterisation at the level of individual complement pathways identified K57 as a novel C5 inhibitor , which is a fully efficacious inhibitor of the terminal pathway in response to both CP and AP activation , and a potential therapeutic candidate for complement-mediated disorders , such as paroxysmal nocturnal haemoglobinuria and atypical haemolytic uraemic syndrome . Additionally , the discovery of K149 as a ‘silent binder’ of C5 may be of considerable value as a non-inhibitory reagent for the detection of native C5 ( Figure 1 ) . K8 and K92 both displayed allosteric inhibitory activity against C5 . K92 achieved selective inhibition of the AP through a non-competitive mechanism ( Figure 1 ) . To our knowledge , this is the first reported example of complement pathway-specific inhibition through C5 and the first experimental evidence reported for mechanistic differences between the AP and CP C5 convertases . This suggests an expanded therapeutic scope for C5 , whereby tuning of the conformational ensemble or dynamics with allosteric compounds can bias activation to leave certain complement pathways intact . Complete inhibition of the terminal pathway has been shown to increase the susceptibility of eculizumab patients to Neisseria meningitidis infections ( McNamara et al . , 2017 ) . Selective inhibition of C5-cleavage by the AP C5-convertase , and not the CP C5-convertase , may partially preserve serum bactericidal activity , thereby lowering the risk of meningococcal disease . Structural analyses , utilising X-ray crystallography , revealed the unique topologies of knob domain peptides K8 and K92 and their distinctive binding modes in C5 . Due to the apparent structural homology of knob domains with certain venomous peptides , of which conotoxins and spider venoms are examples , it has been proposed that the knob domains of ul-CDRH3 might be similarly predisposed to target the concave epitopes of ion channels . Likewise , structural homology with defensin peptides has garnered hypotheses regarding an improved ability to bind viral capsid coats . Indeed , bovine antibodies with ul-CDRH3 have been raised against the viral capsid of HIV with exceptional efficiency , given the challenging nature of the antigen ( Sok et al . , 2017; Stanfield et al . , 2020 ) . However , the study presented here shows that , in the case of C5 , concave epitopes are not the knob domain’s sole preserve . Notably , the MG8 domain epitope of K8 offers a planar pharmacophore and , while the K92 epitope is more undulating , casual inspection of the C5 structure reveals numerous deeper cavities available ( Figure 3 ) . This may mean that knob domains can be raised to inhibit flat surfaces involved in protein–protein interactions , which might not offer binding sites for orthosteric small molecules . We note that the structural architecture of the knob domains varies for the epitope . Their immune derivation means that , unlike cysteine-rich peptides derived from other natural sources , such as venoms , the bovine immune system can be used to define specificity for any antigen . Comparative structural analysis suggests that knob domain paratopes are differentiated from conventional antibodies and the structures of known cyclic peptides , offering a different binding architecture to other small antibody fragments , such as the camelid VHH . While firm conclusions are hampered by limited examples , the number of interactions does not seem dissimilar from cyclic peptides or mAbs , both of which have been successfully applied to tackling high-affinity protein–protein interactions . Our structures demonstrate that the importance of the network of disulphide bonds goes beyond a stabilising role . An apparent paucity of secondary structure would suggest that while stabilisation of the domain is indeed critical , disulphide bonds also participate in sulphur–π interactions to sustain intra- and inter-chain interactions . The structures of knob domains bound to their target antigen demonstrate both the diversity and versatility afforded to the bovine immune repertoire by these sequences . Structural alignment of the C5-K92 co-crystal structure with the apo C5 structure ( Figure 4 ) , using the MG5 domain , revealed a rotational movement of the MG1 causing the α-chain to adopt a twisted conformation accompanied by a rotational movement in the C5d domain , in response to knob domain binding . Comparison of the C5-K8 structure with the apo C5 structure revealed a similar conformation but with less movement in the C5d domain . The helical C5d domain is the target of two immune evasion molecules which have evolved in ticks , OmCI and RaCI , both of which inhibit C5 by crosslinking C5d to neighbouring domains ( Jore et al . , 2016 ) . Additionally , it has been shown that polyclonal antibodies raised against C5d inhibit binding of C5 to C3b ( DiScipio , 1981 ) . The binding site of OmCI is contained within the CUB and C5d domains , with only a single , non-bonded interaction to the C345c domain visible in the crystal structure ( Jore et al . , 2016 ) , which appears mediated by crystal contacts . Interestingly , the C5-OmCI-RaCI crystal structure ( PDB accession code 5HCC ) reveals similar conformational changes in C5d , relative to the apo C5 structure . We therefore propose that rearrangement of C5d can lower the affinity , or preclude the interaction , of C5 for the convertases and that this may be a common inhibitory mechanism for OmCI , RaCI , K8 , and K92 . Should K92 and K8 inhibit C5 by modulating the C5d domain , in the case of K92 , this occurs at a range of over 50 Å . Such remote effects are not unprecedented; allosteric structural changes can be propagated at over 150 Å in response to drug binding ( Haselbach et al . , 2017 ) . Subsequent solution biophysics methods substantiate our crystallographic observations . HDX-MS analysis revealed areas of solvent protection changes in the MG8 domain , resulting from the binding of K8 , and in the MG1 and MG5 domains of the C5-K92 complex ( Figure 6 ) , corresponding to their respective epitopes , as identified with X-ray crystallography ( Figure 3 ) and confirmed by site-directed mutagenesis analysis . Changes in solvent exposure were also observed in the α-chain for C5-K92 , providing a route to visualise the allosteric network . As similar conformational or dynamic changes occur both in solution and in the crystal structure , this suggests that the effects are ligand induced and are not the result of crystal packing interactions . SAXS analysis also suggests that K8 and K92 increase the flexibility of C5 and effects on dynamics may be a contributing factor in realising efficacy ( Figure 5 ) . For K57 , which had an Emax of 100% for both pathways and was not demonstrably allosteric , HDX-MS and biacore experiments with the metastable C5b suggested a putative epitope on the MG5 domain . This could support an orthosteric mechanism of action as CVF , which can form a stable C5 convertase , contacts the MG5 domain in the C5-CVF co-crystal structure ( PDB accession code 3PVM ) . However , by SAXS , similar changes in conformation and/or dynamics to the C5-K8 and C5-K92 complexes were apparent and an allosteric network , including changes in the C5d domain , was observed in the C5-K92 complex by HDX-MS . These observations could support an allosteric mechanism for K57 . Importantly , we also saw a high degree of negative cooperativity between the different knob domains by SPR ( Figure 2 ) , suggesting that all the functional knob domains perturb the conformational state or dynamics of C5 . K57 also showed cooperativity with other functional knob domains , suggesting providing further evidence that it stabilises a conformation of C5 that is less energetically favourable for binding of the other ligands . Our observations with K92 suggest that further work may be required to elucidate the mechanism of action of another binder of the MG1 and MG5 domains , SSL7 . Given that SSL7 can be a partial inhibitor ( Laursen et al . , 2010 ) , even with co-binding of IgA , this precludes a steric mechanism and invites biophysical studies in solution . Additionally , another tick-derived inhibitor , Cirp-T , was also recently reported as predominantly binding to the MG4 domain , with an orthosteric mechanism of action attributed . However , we note that published data only showed an Emax of <90% in AP-driven assays ( Reichhardt et al . , 2020 ) , indicating that it is an allosteric C5 inhibitor for the AP , and potentially also the CP , C5 convertase , which may merit further investigation . In conclusion , we introduce knob domains as a new peptide modality , with unexplored therapeutic potential for the modulation of proteins and protein–protein interactions . This study is the first application of knob domain peptides and reveals an unexpectedly high incidence of allosteric modulators of complement C5 , expanding its scope for complement-targeted therapies and providing important mechanistic tools for the study of C5 convertases . Knob domains can offer a range of advantages over the current macromolecular C5 inhibitors , including their use as peptide therapeutics , while grafting knob domains into the CDRH3 of well-characterised Fabs or using Fc tags could provide routes to extend half-life in vivo by attenuating renal clearance .
Human C5 was affinity purified using an E141A , H164A OmCI column ( Macpherson et al . , 2018 ) . Briefly , human serum ( TCS Biosciences , Botolph Claydon , UK ) was diluted 1:1 ( v/v ) with phosphate buffered saline ( PBS ) , 20 mM ethylenediaminetetraacetic acid ( EDTA ) , and applied to a 5 mL Hi-Trap NHS column ( GE Healthcare , Amersham UK ) , which contained 20 mg of E141A H164A OmCI protein , at a rate of 1 mL/min . The column was washed with 5× column volumes ( CV ) of PBS , C5 was then eluted using 2 M MgCl2 and immediately dialysed into PBS . C5b was prepared from human C5 by incubating C5 with CVF , factor B and factor D , at a 1:10 molar ratio , as previously described ( Jore et al . , 2016 ) . C5a was removed using a spin column with 30 kDa cut-off ( Thermo Fisher Scientific , Horsham , UK ) . Knob domain peptides were expressed fused to the CDRH3 of the PGT-121 Fab , as previously described ( Macpherson et al . , 2020 ) . Plasmid DNA for each construct was amplified using QIAGEN Plasmid Plus Giga Kits . Expi293F cultures were transfected with Expifectamine 293 Transfection kits ( Invitrogen , Renfrew , UK ) as per the manufacturer’s instructions . The cells were cultured for 4 days and supernatants harvested by centrifugation at 4000 rpm for 1 hr . Harvested supernatants were applied to a Hi-Trap Nickel excel columns ( GE Healthcare , Amersham , UK ) using an Akta pure ( GE Healthcare , Amersham , UK ) . Cell supernatants were loaded at 2 . 5 mL/min , followed by a wash of 7× CV of PBS , 0 . 5 M NaCl . A second wash with 7× CV of buffer A ( 0 . 5 M NaCl , 0 . 02 M Imidazole , PBS pH 7 . 3 ) was performed , and samples were eluted by isocratic elution with 10× CV of buffer B ( 0 . 5 M NaCl , 0 . 25 M ) Imidazole , PBS ( pH 7 . 3 ) . Post elution , the protein-containing fractions were pooled and buffer exchanged into PBS using dialysis cassettes ( Thermo Fisher Scientific , Horsham , UK ) . For isolation of the knob domain peptide , PGT-121 Fab-knob peptide fusion proteins were incubated with tobacco etch virus protease , at a ratio of 100:1 ( w/w ) , for a minimum of 2 hr at room temperature . Peptides were purified using a Waters UV-directed FractionLynx system with a Waters XBridge Protein BEH C4 OBD Prep Column ( 300 Å , 5 µm , 19 × 100 mm , Waters Corp . , Milford , MA ) . An aqueous solvent of water , 0 . 1% trifluoroacetic acid ( TFA ) , and an organic solvent of 100% MeCN was used . The column was run at 20 mL/min at 40°C with a gradient of 5–50% organic solvent , over 11 min . Fractions containing knob peptide were pooled and lyophilised using a Labconco Freezone freeze drier . For the C3 and C9 ELISAs , microtiter plates ( MaxiSorp; Nunc ) were incubated overnight at 4°C with 50 µL of a solution of in 75 mM sodium carbonate ( pH 9 . 6 ) containing either 2 . 5 µg/mL aggregated human IgG ( Sigma-Aldrich , Gillingham , UK ) for CP or 20 µg/mL zymosan ( Sigma-Aldrich , Gillingham , UK ) for AP . As a negative control , wells were coated with 1% ( w/v ) bovine serum albumin ( BSA ) /PBS . Microtiter plates were washed four times with 250 µL of wash buffer ( 50 mM Tris-HCl ) , 150 mM NaCl and 0 . 1% Tween 20 ( pH 8 ) between each step of the procedure . Wells were blocked using 250 µL of 1% ( w/v ) BSA/PBS for 2 hr at room temperature . Normal human serum was diluted in either gelatin veronal buffer with calcium and magnesium ( GVB++: 0 . 1% gelatin , 5 mM Veronal buffer , 145 mM NaCl , 0 . 025% NaN3 , 0 . 15 mM calcium chloride , 1 mM magnesium chloride , pH 7 . 3; for CP ) or Mg-ethylene glycol tetraacetic acid ( Mg-EGTA ) ( 2 . 5 mM veronal buffer [pH 7 . 3] containing 70 mM of NaCl , 140 mM of glucose , 0 . 1% gelatin , 7 mM of MgCl2 , and 10 mM of EGTA; for AP ) . Serum was used at a concentration of 1% in CP or 5% in AP and was mixed with serially diluted concentrations of peptides ( 16 µM – 15 . 6 nM ) in GVB++ or Mg-EGTA buffer , and preincubated on ice for 30 min . Peptide–serum solutions were then incubated in the wells of microtiter plates for 35 min for CP assays ( both C3b and C9 detection ) or 35 min for AP ( C3b ) or 60 min for AP ( C9 ) , at 37°C . Complement activation was assessed through detection of deposited complement activation factors using specific antibodies against C3b ( rat anti-human C3d HM2198 , Hycult , Uden , The Netherlands ) and C9 ( goat anti-human C9 , A226 , Complement Technologies Tyler , TX ) at a 1:1000 dilution . Bound primary antibodies were detected with horse rdischHRP-conjugated goat anti-rat ( ab97057 , Abcam , Cambridge , UK ) or rabbit anti-goat ( P0449 , Agilent Dako , Santa Clara , CA ) secondary antibodies at a 1:1000 dilution . Bound HRP-conjugated antibodies were detected using TMB One solution ( Eco-TEK – manufactured by Bio-TEK , Winooski , VT ) with absorbance measured at 450 nm . For the C5b ELISA , assays were run using the CP and AP Complement functional ELISA kits ( Svar Life Science , Malmö , Sweden ) . For sample preparation , serum was diluted as per the respective protocol for the CP and AP assays . Serial dilutions of peptides were prepared and allowed to incubate with serum for 15 min at room temperature prior to plating . For the C5a ELISA , assays were run using the Complement C5a Human ELISA Kit ( Invitrogen , Renfrew , UK ) . For sample preparation , at the end of the 37°C incubation of the serum/peptide samples on the C5b ELISA plate , 50 µL of the diluted , activated serum was transferred to a C5a ELISA plate containing 50 µL/well of assay buffer . All subsequent experimental steps were performed as described in the protocol . GVB++ or Mg EGTA buffers , which had been supplemented with 2 . 5% glucose ( w/v ) , were used for the CP and AP assays , respectively . For the AP , 150 µL of rabbit erythrocytes ( TCS Biosciences , Botolph Claydon , UK ) were washed twice , by addition of 1 mL of buffer and centrifugation at 800 ×g for 1 min , and finally resuspended in 500 µL of buffer . For the CP , 150 µL sheep erythrocytes ( TCS Biosciences , Botolph Claydon , UK ) were washed twice with 1 mL of buffer and sensitised with a 1/1000 dilution of rabbit anti-sheep red blood cell stroma antibody ( S1389 , Sigma Aldrich , Gillingham , UK ) . After a 30°C/30 min incubation , with shaking , the cells were rewashed and resuspended with 500 µL of buffer . Serial dilutions of peptide were prepared in the respective buffers and normal human serum was added at 1% for the CP and 4 . 5% for the AP ( corresponding to CH50 of the serum ) . Also , 90 µL of peptide–serum mixtures were plated into a V-bottom 96-well microtiter plate ( Corning ) and 10 µL of erythrocytes were added . Plates were incubated for 30 min at 37°C , with shaking . Finally , 50 µL of buffer was added , the plates centrifuged at 800 ×g , and 80 µL of supernatant was transferred to an ELISA plate ( Nunc ) and absorbance measured at 405 nm . 6 . 1 mg/mL C5 ( 20 mM Tris-HCl , 75 mM NaCl , pH 7 . 35 ) was mixed at a 1:1 molar ratio with either the K8 or K92 peptides . Crystallisation trials were initiated by the vapor diffusion method at 18°C with a 1:1 mixture of mother liquor ( v/v ) . C5-K8 crystals were grown in a mother liquor of 0 . 1 M N- ( 2-acetamido ) iminodiacetic A , 14% ethanol ( v/v ) , pH 6 . 0 . For C5-K92 crystals , the mother liquor was 0 . 1 M bicine/Trizma ( pH 8 . 5 ) , 10% ( w/v ) polyethylene glycol 8000 , 20% ( v/v ) ethylene glycol , 30 mM sodium fluoride , 30 mM sodium bromide , and 30 mM sodium iodide ( Gorrec , 2009 ) . Prior to flash freezing in liquid nitrogen , C5-K8 crystals were cryoprotected in mother liquor with 30% 2-methyl-2 , 4-pentanediol ( v/v ) . C5-K92 crystals were frozen without additional cryoprotection . Data were collected at the Diamond Light Source ( Harwell , UK ) , on beamline I03 , at a wavelength of 0 . 9762 Å . The C5-K8 structure was solved using the automated molecular replacement pipeline Balbes ( Long et al . , 2008 ) using the apo C5 structure ( PDB accession code 3CU7 ) , minus the C345c domain . The C5-K8 complex crystallised in space group P212121 with one molecule in the asymmetric unit . A backbone model of the K8 peptide was produced using ARP-wARP ( Langer et al . , 2008 ) which informed manual model building in Coot ( Emsley et al . , 2010 ) , within the CCP4 suite ( Winn et al . , 2011 ) . The model was subjected to multiple rounds of refinement in Refmac ( Murshudov et al . , 1997 ) and Phenix ( Adams et al . , 2010 ) . The overall geometry in the final structure of the C5-K8 complex is good , with 97 . 2% of residues in favoured regions of the Ramachandran plot and no outliers . The C5-K92 complex crystallised in space group C2 with one molecule in the asymmetric unit . C5 was solved by molecular replacement with Phaser ( McCoy et al . , 2007 ) using the C5-OmCI-RaCI structure ( PDB accession code 5HCC ) , with OmCI and RaCI removed . Manual building of the K92 peptide in Coot was greatly informed by mass spectroscopy disulphide mapping experiments . The model was subjected to multiple rounds of manual rebuilding in Coot and refinement in Phenix ( Adams et al . , 2010 ) . The overall geometry in the final structure of the C5-K92 complex is good , with 95 . 2% of residues in favoured regions of the Ramachandran plot and no outliers . Structure factors and coordinates for both C5-knob domain peptide complexes have been deposited in the PDB ( PDB accession codes: 7AD6 ( C5-K92 complex ) and 7AD7 ( C5-K8 complex ) ) . Crystal trials were also performed with the C5-K57 and C5-K149 complexes , but the resulting crystals diffracted poorly . A 250 µL K92 peptide at 1 mg/mL was alkylated with addition of 18 µL of 2-Iodoacetamide ( Thermo Fisher Scientific , Horsham , UK ) at room temperature for 30 min . Overnight dialysis into assay buffer ( 7 . 5 mM Tris-HCl , 1 . 5 mM CaCl2 , pH 7 . 9 ) was performed using 2 kDa slide-a-lyzer cassettes ( Thermo Fisher Scientific , Horsham , UK ) . Chymotrypsin ( sequencing grade , Roche Applied Sciences ) was reconstituted to 1 µg/µL in assay buffer , and 5 µL of reconstituted enzyme was added to 80 µL of sample . Once mixed , the sample was incubated at 37°C for 1 . 5 hr before being quenched with 5 µL of 1% TFA . Samples were diluted 1 in 10 and 5 µL was loaded onto the analytical column . Liquid chromatography electrospray ionisation mass spectrometry was acquired using an Ultimate 3000 UHPLC system ( Thermo Fisher Scientific , Horsham , UK ) coupled with a Q-Exactive Plus Orbitrap ( Thermo Fisher Scientific , Horsham , UK ) . Separations were performed using gradient elution ( A: 0 . 1% formic acid; B: 0 . 1% formic acid in acetonitrile ) on an Acquity UPLC BEH C18 column ( 130 Å , 1 . 7 µm , 2 . 1 × 150 mm; Waters Corp . , Milford , MA ) with the column temperature maintained at 40°C . The following analytical gradient at 0 . 2 mL/min was used: 1% B was held for 2 min , 1–36% B over 28 min , 36–50% over 5 min , and 50–99% B over 0 . 5 min . There were sequential wash steps with changes in gradient of 99%–1% B over 0 . 5 min ( at a higher flow rate of 0 . 5 mL/min ) before equilibration at 1% B for 6 . 5 min ( at the original 0 . 2 mL/min ) . A full MS/dd-MS2 ( Top 5 ) scan was run in positive mode . Full MS: scan range was 200–2000 m/z with 70 , 000 resolution ( at 200 m/z ) and a 3 × 106 AGC target ( the maximum target capacity of the C-trap ) , 100 ms maximum Injection time . The dd-MS2: 2 . 0 m/z isolation window , CID fragmentation ( NCE 28 ) with fixed first mass of 140 . 0 m/z , with a 17 , 500 resolution ( at 200 m/z ) , 1 × 105 AGC target , 200 ms maximum injection time . The source conditions of the MS were capillary voltage , 3 kV; S-lens RF level , 50; sheath gas flow rate , 25; auxiliary gas flow rate , 10; auxiliary gas heater temperature , 150°C; and the MS inlet capillary was maintained at 320°C . Data were acquired using XcaliburTM 4 . 0 software ( Thermo Fisher Scientific , Horsham , UK ) , and raw files were analysed by peptide mapping analysis using Biopharma Finder 2 . 0 software ( Thermo Fisher Scientific , Horsham , UK ) by performing a disulphide bond search with a chymotrypsin ( medium specificity ) digest against the K92 peptide sequence . Assignments and integrations from Biopharma Finder were filtered to include only peptides identified as containing a single disulphide bond and with an experimental mass within |5| ppm of the theoretical mass . Intensities for all peptides containing the same cysteines pairing were summed and percentages were obtained from the summed against total intensities . Kinetics were measured using a Biacore 8K ( GE Healthcare , Amersham , UK ) with a CM5 chip , which was prepared as follows: 1-ethyl-3- ( −3-dimethylaminopropyl ) carbodiimide hydrochloride ( EDC ) /N-hydroxysuccinimide ( NHS ) was mixed at 1:1 ratio ( flow rate , 10 µL/min; contact time , 30 s ) , and human C5 at 1 µg/mL in pH 4 . 5 sodium acetate buffer was injected over flow cell one only ( flow rate , 10 µL/min; contact time , 60 s ) . Final immobilisation levels in the range of 2000–3000 RUs were obtained to yield theoretical Rmax values of ~50–60 RU . Serial dilutions of K8 and K92 knob domains , and various mutants , were prepared in HBS-EP ( 0 . 01 M HEPES pH 7 . 4 , 0 . 15 M NaCl , 3 mM EDTA , 0 . 005% v/v Surfactant P20 ) buffer and injected ( flow rate , 30 µL/min; contact time , 240 s; dissociation time , 6000 s ) . After each injection , the surface was regenerated with two sequential injections of 2 M MgCl2 ( flow rate , 30 µL/min; contact time , 30 s ) . Binding to the reference surface was subtracted , and the data were fitted to a single-site binding model using Biacore evaluation software . Simulation structures were prepared using Schrodinger’s Maestro Protein preparation wizard . The molecular dynamics runs were performed using the Schrodinger’s default implementation of the binding pose metadynamics with the peptide chain considered in place of a ligand . Additional RMSD calculations for the peptide internal structure assessment in the last 20% of the dynamics were performed relative to the starting structures . Data was collected at the EMBL P12 beam line ( PETRA III , DESY Hamburg , Germany; Blanchet et al . , 2015 ) . Data was collected with inline SEC mode using the Agilent 1260 Infinity II Bio-inert LC . Also , 50 μL of complement component C5 at 31 . 6 μM ( 5 . 96 mg/mL ) was injected onto a Superdex 200 Increase 5/150 column ( GE Healthcare , Amersham , UK ) at a flow rate of 0 . 35 mL/min . The mobile phase comprised 20 mM Tris pH 7 . 35 , 75 mM NaCl , and 3% glycerol . The column elute was directly streamed to the SAXS capillary cell , and throughout the 15-min run , 900 frames of 1 s exposure were collected . After data reduction and radial averaging , the program CHROMIXS ( Panjkovich and Svergun , 2018 ) was employed . Around 100 statistically similar buffer frames were selected and used for background subtraction of the sample frames from the chromatographic peak . This results in the final I ( s ) vs s scattering profiles , where s = 4πsinθ/λ , 2θ is the scattering angle , and λ = 1 . 24 Å . The scattering data in the momentum transfer range 0 . 05 < s < 0 . 32 nm−1 were collected with a PILATUS 6M pixel detector at a distance of 3 . 1 m from the sample . ATSAS 2 . 8 ( Franke et al . , 2017 ) was employed for further data analysis and modelling . The program PRIMUS ( Konarev et al . , 2003 ) was used to perform Guinier analysis ( lnI ( s ) versus s2 ) from which the radius of gyration , RG , was determined . Distance probability functions , p ( r ) , were calculated using the inverse Fourier transformation method implemented in GNOM ( Svergun , 1992 ) that provided the maximum particle dimension , Dmax . The concentration-independent molecular weight estimate , MWVC , is based on the volume of correlation ( Rambo and Tainer , 2013 ) . The values are reported in Supplementary file 1 , Table 3 . 1 . Theoretical scattering profiles were computed from X-ray coordinates using Crysol ( Svergun et al . , 1995 ) , and SREFLEX ( Panjkovich and Svergun , 2016 ) was used to refine the models . For this , the program partitions the structure into pseudo-domains and hierarchically employs NMA to find the domain rearrangements minimising the discrepancy χ2 between the SAXS curve computed from the refined model and the experimental data . Because of the absence of electron density for the C345c domain in the C5-K8 complex structure , we included a round of restrained rigid body refinement followed by NMA to obtain an improved fit . On the same day , MALLS data were collected with a separate SEC run under the same experimental conditions ( set-up , buffer , run parameters , etc . ) . For this , a Wyatt Technologies miniDAWN TREOS MALLS detector coupled to an OptiLab T-Rex differential refractometer for protein concentration determination ( dn/dc was taken as 0 . 185 ) was used . The MALLS system was calibrated relative to the scattering from toluene . The MWMALLS distribution of species eluting from the SEC column were determined with the Wyatt ASTRA7 software package . The experimental SAXS data and the models derived from them were deposited to the Small Angle Scattering Biological Data Bank ( SASBDB accession number SASDJA6 ) . A 6 µM of C5 was incubated with 10 µM of peptide ( K8 , K92 , or K57 ) to achieve complex during deuterium exchange conditions . Then , 4 μL of C5 or the C5-peptide complex were diluted into 57 μL of 10 mM phosphate in H2O ( pH 7 . 0 ) or into 10 mM phosphate in D2O ( pD 7 . 0 ) at 25°C . The deuterated samples were then incubated for 0 . 5 , 2 , 15 , and 60 min at 25°C . After the reaction , all samples were quenched by mixing at 1:1 ( v/v ) with a quench buffer ( 4 M guanidine hydrochloride , 250 mM Tris ( 2-carboxyethyl ) phosphine hydrochloride , 100 mM phosphate ) at 1°C . The final mixed solution was pH 2 . 5 . The mixture was then immediately injected into the nanoAcquity HDX module ( Waters Corp . , Milford , MA ) for peptic digest using an enzymatic online digestion column ( Waters Corp . , Milford , MA ) in 0 . 2% formic acid in water at 20°C and with a flow rate of 100 μL/min . All deuterated time points and undeuterated controls were carried out in triplicate with blanks run between each data point . Peptide fragments were then trapped using an Acquity BEH C18 1 . 7 μM VANGUARD chilled pre-column for 3 min . Peptides were eluted into a chilled Acquity UPLC BEH C18 1 . 7 μm 1 . 0 × 100 using the following gradient: 0 min , 5% B; 6 min , 35% B; 7 min , 40% B; 8 min , 95% B , 11 min , 5% B; 12 min , 95% B; 13 min , 5% B; 14 min , 9 5% B; and 15 min , 5% B ( A: 0 . 2% HCOOH in H2O; B: 0 . 2% HCOOH ) in acetonitrile . The trap and UPLC columns were both maintained at 0°C . Peptide fragments were ionised by positive electrospray into a Synapt G2-Si mass spectrometer ( Waters Corp . , Milford , MA ) . Data acquisition was run in ToF-only mode over an m/z range of 50–2000 Th using an MSE method ( low collision energy , 4 V; high collision energy: ramp from 18 V to 40 V ) . Glu-1-Fibrinopeptide B peptide was used for internal lock mass correction . To avoid significant peptide carry-over between runs , the on-line Enzymate pepsin column ( Waters Corp . , Milford , MA ) was washed three times with pepsin wash ( 0 . 8% formic acid , 1 . 5 M Gu-HCl , 4% MeOH ) and a blank run was performed between each sample run . MSE data from undeuterated samples of C5 were used for sequence identification using the Waters Protein Lynx Global Server 2 . 5 . 1 ( PLGS ) . Ion accounting files for the three control samples were combined into a peptide list imported into DynamX v3 . 0 software ( Waters Corp . , Milford , MA ) . The output peptides were subjected to further filtering in DynamX . Filtering parameters used were minimum and maximum peptide sequence length of 4 and 25 , respectively , minimum intensity of 1000 , minimum MS/MS products of 2 , minimum products per amino acid of 0 . 2 , and a maximum MH + error threshold of 10 ppm . DynamX was used to quantify the isotopic envelopes resulting from deuterium uptake for each peptide at each time point . Furthermore , all the spectra were examined and checked visually to ensure correct assignment of m/z peaks and only peptides with a high signal to noise ratios were used for HDX-MS analysis . Following manual filtration in DynamX , confidence intervals for differential HDX-MS ( ΔHDX ) measurements of individual time point were calculated using Deuteros ( Lau et al . , 2019 ) software . Only peptides which satisfied a ΔHDX confidence interval of 98% were considered significant . The ΔHDX was then plotted onto the C5 structure in Pymol . On a Biacore 8K ( GE Healthcare , Amersham , UK ) , human C5b was immobilised on a CM5 chip ( GE Healthcare , Amersham , UK ) . Flow cells were activated using a standard immobilisation protocol: EDC/NHS was mixed at 1:1 ratio ( flow rate , 10 µL/min; contact time , 30 s ) . C5b , at 1 µg/mL , or C5b-6 ( Complement Technologies , Tyler , TX ) , at 2 µg/mL , in pH 4 . 5 sodium acetate buffer , were immobilised on flow cell two only ( flow rate , 10 µL/min; contact time , 420 s ) . Finally , ethanolamine was applied to both flow cells ( flow rate , 10 µL/min; contact time , 420 s ) . A final immobilisation level of 500–700 RUs was obtained for C5b and 1000–1150 RUs were obtained for C5b-6 . Single-cycle kinetics were measured using a seven-point , threefold serial dilution ( spanning a range of 1 µM to 1 . 4 nM ) in HBS-EP buffer ( GE Healthcare , Amersham , UK ) . A high flow rate of 40 µL/min was used , with a contact time of 300 s and a dissociation time of 2700 s . Binding to the reference surface was subtracted , and the data were fitted to a single-site binding model using Biacore evaluation software . All sensorgrams were inspected for evidence of mass transport limitation using the flow rate-independent component of the mass transfer constant ( tc ) . On a Biacore 8K ( GE Healthcare , Amersham , UK ) , human C5 was amine coupled to a CM5 chip using the same protocol as for C5b . A final immobilisation level of approximately 1000–2000 RUs was obtained . For cross blocking , the surface was saturated with two sequential injections of a 20 µM knob domain solution in HBS-EP buffer ( GE Healthcare , Amersham , UK ) using a flow rate of 30 µL/min and contact time of 300 s . This was immediately followed with an injection of a second knob domain peptide , again at 20 µM in HBS-EP , with a flow rate of 30 µL/min and a contact time of 270 s , and the dissociation phase was measured for 600 s . Binding to the reference surface was subtracted , and sensorgrams were plotted in GraphPad Prism ( GraphPad Software , San Diego , California USA , www . graphpad . com ) . | Antibodies are proteins produced by the immune system that can selectively bind to other molecules and modify their behaviour . Cows are highly equipped at fighting-off disease-causing microbes due to the unique shape of some of their antibodies . Unlike other jawed vertebrates , cows’ antibodies contain an ultra-long loop region that contains a ‘knob domain’ which sticks out from the rest of the antibody . Recent research has shown that when detached , the knob domain behaves like an antibody fragment , and can independently bind to a range of different proteins . Antibody fragments are commonly developed in the laboratory to target proteins associated with certain diseases , such as arthritis and cancer . But it was unclear whether the knob domains from cows’ antibodies could also have therapeutic potential . To investigate this , Macpherson et al . studied how knob domains attach to complement C5 , a protein in the inflammatory pathway which is a drug target for various diseases , including severe COVID-19 . The experiments identified various knob domains that bind to complement C5 and inhibits its activity by altering its structure or movement . Further tests studying the structure of these interactions , led to the discovery of a common mechanism by which inhibitors can modify the behaviour of this inflammatory protein . Complement C5 is involved in numerous molecular pathways in the immune system , which means many of the drugs developed to inhibit its activity can also leave patients vulnerable to infection . However , one of the knob domains identified by Macpherson et al . was found to reduce the activity of complement C5 in some pathways , whilst leaving other pathways intact . This could potentially reduce the risk of bacterial infections which sometimes arise following treatment with these types of inhibitors . These findings highlight a new approach for developing drug inhibitors for complement C5 . Furthermore , the ability of knob domains to bind to multiple sites of complement C5 suggests that this fragment could be used to target proteins associated with other diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics",
"immunology",
"and",
"inflammation"
] | 2021 | The allosteric modulation of complement C5 by knob domain peptides |
PD-1 blockade therapy has revolutionized cancer treatments . However , a substantial population of patients is unresponsive . To rescue unresponsive patients , the mechanism of unresponsiveness to PD-1 blockade therapy must be elucidated . Using a ‘bilateral tumor model’ where responsive and unresponsive tumors were inoculated into different sides of the mouse belly , we demonstrated that unresponsive tumors can be categorized into two groups: with and without systemic immunosuppressive property ( SIP ) . The SIP-positive tumors released uncharacterized , non-proteinaceous small molecules that inhibited mitochondrial activation and T cell proliferation . By contrast , the SIP-negative B16 tumor escaped from immunity by losing MHC class I expression . Unresponsiveness of SIP-positive tumors was partially overcome by improving the mitochondrial function with a mitochondrial activator; this was not successful for B16 , which employs immune ignorance . These results demonstrated that the ‘bilateral tumor model’ was useful for stratifying tumors to investigate the mechanism of unresponsiveness and develop a strategy for proper combination therapy .
Cancer immunotherapy using immune checkpoint blockade , particularly antibodies against programmed cell death receptor 1 ( PD-1 ) or its ligand ( PD-L1 ) , has made a revolution in cancer treatments as this treatment has durable response even to terminal stage cancers and lesser side-effects compared to the conventional cancer treatments ( Brahmer et al . , 2010; Couzin-Frankel , 2013; Hodi et al . , 2010; Mahoney et al . , 2015; Topalian et al . , 2015 ) . The success of clinical trials for the PD-1/PD-L1 axis blockade led the FDA to approve antibodies for PD-1 ( e . g . nivolumab , pembrolizumab ) or PD-L1 ( e . g . Atezolizumab , Avelumab , Durvalumab ) for different types of human cancers including metastatic non-small cell lung carcinoma ( NSCLC ) , squamous cell lung cancer , renal cell carcinoma , hodgkin's lymphoma , head and neck squamous cell carcinoma , and recently , for microsatellite instability-high ( MSI-H ) or mismatch repair deficient ( dMMR ) cancers that include many late-stage cancers ( Chowdhury et al . , 2018a ) . Despite the impressive success rate of PD-1 blockade therapy , a significant fraction of patients is unresponsive . To further improve its efficacy , we must ( i ) identify biomarker ( s ) that predict the responsiveness/unresponsiveness and ( ii ) develop improved strategy including the combination therapy . For these improvements , we need to understand the mechanism of unresponsiveness to PD-1 blockade therapy . Most studies on biomarkers and resistance mechanisms have focused only on the tumor’s intrinsic properties ( Cristescu et al . , 2018; Ribas , 2015; Rieth and Subramanian , 2018; Wellenstein and de Visser , 2018; Zou et al . , 2016 ) . We need to elucidate the mechanism for unresponsiveness related to immune effector T cells to understand the complicated interaction between cancer and immunity . Several studies have worked on the unresponsive mechanism from the immunity side in different models . In one such model , the ‘Cold and Hot tumor hypothesis’ , tumors can be roughly classified as ‘immunologically hot ( inflamed ) ’ with an abundance of tumor-infiltrating lymphocytes ( TILs ) and ‘immunologically cold ( noninflamed ) ’ with an absence of a sufficient population of pre-existing immune cells ( Bonaventura et al . , 2019; van der Woude et al . , 2017 ) . In addition , some groups claim that clinical failures in many patients could be due to an imbalance between T-cell reinvigoration and tumor burden ( Borcoman et al . , 2018; Huang et al . , 2017 ) . CD8+ T cells , the major immune effector cells for attacking tumors , are subject to negative regulation by multiple mechanisms in tumor-bearing hosts . Some of the well-known negative regulatory cells and soluble factors include myeloid-derived suppressor cells ( MDSC ) , innate lymphoid cells ( ILC ) , tumor-associated macrophages ( TAM ) , regulatory CD4+ T cells ( Tregs ) , regulatory B cells ( Bregs ) , transforming growth factor β ( TGF-β ) , interleukin-10 ( IL-10 ) , adenosine , granulocyte-macrophage colony-stimulating factor ( GM-CSF ) , prostaglandin E2 ( PGE2 ) , and L-Kynurenine ( Artis and Spits , 2015; DeNardo and Ruffell , 2019; Facciabene et al . , 2012; Sarvaria et al . , 2017; Tauriello et al . , 2018 ) . Lack of MHC class I and neo-antigen on tumor cells also cause unresponsiveness because T cells cannot recognize the tumor ( Garrido et al . , 2016; McGranahan et al . , 2017; Rodríguez , 2017 ) . The tumor microenvironment , influenced by the above mechanisms , allows tumor cells to escape from immune attack ( DeNardo and Ruffell , 2019; Russo and Protti , 2017 ) . Due to this complexity of tumor and immunity interactions , it is difficult to determine which tumor employs which immune escape mechanism . Energy metabolism mediated by mitochondrial activity regulates the fate of T cells . It has been reported that mitochondria play an important role in antigen-specific T cell activation through signaling of mitochondrial-derived reactive oxygen species ( ROS ) ( Mallilankaraman , 2018; Murphy and Siegel , 2013; Sena et al . , 2013 ) . We recently reported that mitochondria are activated in tumor-reactive CTLs during PD-1 blockade therapy in MC38 tumor-bearing hosts ( Chamoto et al . , 2017 ) . Boosting fatty acid oxidation with a metabolic modulator enhanced the PD-1 blockade effect ( Chowdhury et al . , 2018b ) . Therefore , attenuation of the mitochondrial activity of T cells by tumor-mediated factors could be an immune escape mechanism . In this study , we developed a novel approach using a ‘bilateral tumor model’ and studied the immunosuppressive nature of unresponsive tumors to PD-1 blockade therapy . This model allowed us to categorize unresponsive tumors into two: those which have immune ignorance properties at tumor local sites and the others which have systemic immunosuppressive properties ( SIP ) . SIP is mediated by small molecules to downregulate mitochondrial function directly and to inhibit T cell proliferation . Boosting the mitochondrial activity by the addition of bezafibrate , a pan-PPAR agonist , partially improved the efficacy of the PD-1 blockade against unresponsive tumors with SIP but not for tumors with immune ignorance at the local site .
We first determined which tumor was responsive and unresponsive using an anti-PD-L1 monoclonal antibody ( mAb ) to block PD-1 signal ( PD-1 blockade ) or the Pdcd1-/- mouse model ( Figure 1—figure supplement 1 ) . As summarized in ‘Table 1’ , GL261 , MC38 , and MethA were characterized as responsive tumors while LLC , B16 , Pan02 , and CT26 were characterized as unresponsive tumors . Since CD8+ cytotoxic T lymphocytes ( CTLs ) are the main effector cells during PD-1 blockade therapy , we examined the difference in the host's immune responses to a responsive tumor and an unresponsive tumor according to the schedule shown in Figure 1A . We found both the total lymphocytes and the effector memory CD8+ T cells ( defined as CD62Llow CD44high , P3 ) in draining lymph nodes ( DLNs ) significantly increased in the group of responsive tumors , but did not change in unresponsive tumor-bearing hosts after PD-1 blockade ( Figure 1B and C ) . Further , total CD44+ T cells which include both central memory ( CD62Lhigh CD44high , P2 ) and effector memory ( CD62Llow CD44high , P3 ) were also larger after the PD-1 blockade therapy over ctrl IgG treated group in the hosts with responsive tumor ( Figure 1—figure supplement 2 ) . The frequency of CD8+ tumor-infiltrating lymphocytes ( TILs ) also increased after PD-1 blockade in the group of responsive tumor-bearing hosts , but not in unresponsive tumor-bearing hosts ( Figure 1D ) . The expression of T-bet and IFN- γ , which reflect the activity of Th1-type cytotoxic activity , increased after PD-1 blockade treatment in the group bearing responsive tumors , but did not in the unresponsive tumor-bearing group ( Figure 1E and F; Sullivan et al . , 2003 ) . Similar results were obtained in mice on another genetic background ( BALB/c ) ( Figure 1—figure supplement 3 ) . Taken together , anti-tumor immune responses increased only in hosts with responsive tumors but not in hosts with unresponsive tumors . We and others have previously reported that mitochondrial activation in CD8+ T cells is a marker of CTLs activation ( Buck et al . , 2016; Chamoto et al . , 2017 ) . Thus , to determine whether there was an association between the responsiveness to PD-1 blockade therapy and mitochondrial activation in T cells , we measured several markers of mitochondrial activation using the Seahorse Analyzer ( Figure 2—figure supplement 1A ) . We found that DLN CD8+ T cells from responsive ( MC38 and GL261 ) tumor-bearing hosts had significantly higher basal respiration , maximal respiration , spare respiratory capacity ( SRC ) , and ATP turnover by PD-1 blockade , which was not observed in unresponsive ( B16 and LLC ) tumor-bearing hosts ( Figure 2A ) . Similar results were obtained in mice on the BALB/c background ( Figure 2—figure supplement 1B ) . Besides , mitochondrial superoxide production ( MitoSox ) and Cellular ROS ( CellRos ) in CD8+ TIL were increased by PD-1 blockade therapy only in responsive tumor-bearing mice ( Figure 2B and C ) . Together , increased activity in CD8+ T cells by PD-1 blockade in responsive tumor-bearing mice parallels with their activation status of mitochondria . To investigate the mechanism of the systemic immune suppression of unresponsive tumors , we next employed a ‘bilateral tumor inoculation model’ where unresponsive and responsive tumors were inoculated on different sides of the host ( Figure 3A ) . This model facilitates disclosing how much humoral factors derived from unresponsive tumors would contribute to the growth of responsive tumors in the other side . As shown in Figure 3B , we found that when unresponsive tumors ( LLC or Pan02 ) were present on the left side , the growth inhibition of the responsive MC38 on the right by the PD-1 blockade therapy was inefficient . However , when the unresponsive B16 was on the left , the responsive MC38 or GL261 were rejected by PD-1 blockade as efficiently as the case in which no tumor was on the left side ( Figure 3B and Figure 3—figure supplement 1A ) . The sizes of the left unresponsive tumor in the same experiment were not inhibited by the PD-1 blockade therapy ( Figure 3—figure supplement 1B ) . Therefore , we speculated that the unresponsive LLC and Pan02 tumors may have released immune suppressive factors , while the unresponsive B16 did not . Following the same experimental design , we performed the bilateral tumor experiment in mice on another background ( BALB/c ) and identified that CT26 is an unresponsive tumor with SIP ( Figure 3C and Figure 3—figure supplement 1C ) . Taken together , we classified unresponsive tumors into two groups: those with or without SIP ( Table 2 ) . Since we observed mitochondrial activation in CD8+ T cells as a parameter of responsiveness ( Figure 2 ) , we used the bilateral tumor model to investigate how immunosuppressive factors released from unresponsive tumors ( on the left side ) inhibited the immune response against responsive tumors ( on the right side ) from the aspect of mitochondrial activation ( Figure 4A ) . As shown in Figure 4B , the absolute number of lymphocytes in the DLN on the side with MC38 was increased by PD-1 blockade in mice with the SIP-negative B16 on the other side , but not when the SIP-positive LLC was on the other side . Accordingly , mitochondrial ROS production , mass , OCR and ATP turnover in DLN CD8+ T cells were also enhanced by PD-1 blockade on the MC38 side in the presence of B16 on the other side , but not the case when SIP-positive LLC was inoculated on the other side ( Figure 4C and D ) . In contrast , the PD-1 blockade treatment did not change the mitochondrial activation status in the unresponsive tumor sides ( B16 and LLC ) ( Figure 4E and F ) . In summary , while both LLC and B16 were unresponsive , only the LLC systemically inhibited the mitochondrial activation of CTLs during the PD-1 blockade therapy . We suspected that unresponsive tumors without SIP may not be recognized by the acquired immunity . We compared tumor growth between wild type and immune-compromised mice ( Rag2-/- ) . As shown in Figure 5A , the growth of responsive tumors ( MC38 , GL261 , and MethA ) was significantly restricted in wild type compared with Rag2-/- mice . In contrast , unresponsive tumors were more or less insensitive to acquired immunity ( Figure 5B ) . Note that some unresponsive tumors with SIP ( LLC and CT26 ) were sensed to a small extent by acquired immunity while unresponsive tumors without SIP ( B16 ) were completely ignored ( Table 2 ) . This complete ignorance could be attributed to deficiencies in the ‘T cell - tumor cell interaction’ probably due to less neoantigen and/or lack of MHC class I expression . Indeed , we found that B16 does not express MHC class I even after stimulation with IFN-γ , but others do ( Figure 5C and D ) . In other words , B16 acquired unresponsiveness by employing local immunological ignorance instead of SIP . These data indicate that one of the mechanisms of unresponsiveness in tumors without SIP is lack of MHC class I expression , and suggest that elimination of the suppressive factor would facilitate the enhancement of PD-1 blockade therapeutic efficacy only in unresponsive tumors with SIP . To examine whether immune suppressive factors are released from unresponsive tumors , naïve CD8+ T cells were stimulated with anti- ( CD3+CD28 ) mAb-coated beads in the presence of supernatants collected from responsive and unresponsive tumor cell cultures ( Figure 6A ) . Proliferation assays ( thymidine incorporation and Ki67 detection assays ) demonstrated that T cell proliferation was significantly inhibited in the presence of supernatants from LLC or CT26 , but not in the presence of supernatants from B16 , GL261 or MethA ( Figure 6B and Figure 6—figure supplement 1A and B ) . The suppressive effects of soluble factors from the LLC supernatant was further evidenced by the restoration of T cell proliferation when the supernatant was diluted ( Figure 6—figure supplement 1C ) . It is of note that the SIP factor production is not only specific to mouse cell lines , but also to human cell lines ( Figure 6—figure supplement 2 ) . In addition , different parameters of mitochondrial activation such as cellular ROS and mitochondrial potential were significantly inhibited by the LLC supernatant compared with the B16 and GL261 supernatants ( Figure 6C ) . The OCR and the extracellular acidification rate ( ECAR ) , a parameter for glycolytic function , were severely reduced in CD8+ T cells cultured for 48 hrs in the presence of LLC supernatants compared with those from B16 and GL261 ( Figure 6D and E ) . Similar suppressive activities were observed by supernatants from BALB/c background tumor CT26 ( Figure 6—figure supplement 1D ) . To clarify whether this mitochondrial suppression is direct or bystander , we examined mitochondrial activation parameters within 2 hrs of coculture with the supernatant . As shown in Figure 6F , mitochondrial activation parameters were inhibited in the presence of LLC supernatants immediately , indicating that SIP factors highly likely inhibit mitochondrial activity directly , but not cellular transcriptional activity . Indeed , the transcriptional levels of PGC-1α/β , a master regulator of mitochondrial activation , did not change within 2 hrs ( Figure 6—figure supplement 3A ) . The SIP factor inhibited B cell mitochondria as well within 2 hrs , showing this suppressive effect is more general ( Figure 6—figure supplement 3B ) . These results indicate that the immunosuppressive factors released from SIP-positive tumors directly and generally inhibit the mitochondrial function . Further , to understand the molecular properties of suppressive factors , we performed heat-inactivation to denature protein components and used a dextran-coated charcoal ( DCC ) treatment to adsorb small molecules in the culture supernatants . As shown in Figure 6G and Figure 6—figure supplement 1E , heat-inactivation of LLC and CT26 culture supernatants did not abolish their suppressive activity , whereas removing low molecular weight compounds using the DCC treatment eliminated their suppressive activity , suggesting that the suppressive factor ( s ) may be comprised of non-proteinaceous small molecules . We further fractionated the supernatant into ‘Fraction A ( <3 KDa ) ’ and ‘Fraction B ( 3 ~ 50 KDa ) ’ and found that ‘Fraction A’ had almost the same inhibition potential as the total culture supernatants ( Figure 6H and Figure 6—figure supplement 1F ) . Again , removing small molecules from ‘Fraction A’ using the DCC treatment restored the proliferation of CD8+ T cells . We further tested whether previously reported small molecules could be candidates of the SIP factor such as adenosine , Prostaglandin E2 ( PGE2 ) and kynurenine , the transcriptional levels of key enzymes to produce them were examined . However , there was no relationship between the suppressive property and the expression levels of enzymes including CD39 , CD73 , COX-2 , mPGES1 and IDO1 ( Figure 6—figure supplement 4 ) , suggesting the low possibility of known factors . Since SIP reduced the mitochondrial activity , we examined whether mitochondria activation drug combination can reverse the immune suppression by SIP-positive tumors . As bezafibrate activates mitochondria and synergizes with PD-1 blockade therapy , we first tested whether bezafibrate can reverse the suppression of mitochondrial function and proliferation caused by suppressive factors from the LLC culture supernatants in vitro ( Chowdhury et al . , 2018b ) . Mitochondrial function of naïve CD8+ T cells was regained significantly when bezafibrate was used along with culture supernatant in vitro ( Figure 7A ) . Encouraged with these in vitro results , we performed PD-1 blockade combinatorial therapy with bezafibrate for LLC tumor-bearing hosts ( Figure 7B ) . We found that the tumor-killing effect by the PD-1 blockade was enhanced and mouse survival was increased in the combination therapy ( Figure 7C ) . Of note is the fact that the combinatorial treatment could not rescue the B16 tumor-bearing hosts ( Figure 7C ) . We observed similar results in tumors on the BALB/c background . The survival of SIP-positive CT26 tumor-bearing hosts was improved with the combinatorial therapy with bezafibrate ( Figure 7—figure supplement 1 ) . In summary , the SIP effects of unresponsive tumors were partially rescued by a mitochondrial activation chemical , bezafibrate in vitro and in vivo .
One of the biggest issues in PD-1 blockade cancer immunotherapy is how to reduce the rate of unresponsiveness . Although there are many unresponsive mechanisms , cancers employ at least two strategies to escape from the immune attack: local or systemic immune suppression . Some reports have suggested ‘hot tumors’ and ‘cold tumors’ to distinguish responsive and unresponsive tumors based on the level of immune cell infiltration in the tumor mass ( van der Woude et al . , 2017 ) . However , it is difficult to explain the molecular mechanisms of unresponsiveness by this definition because it explains the results of immune responses in local tumor areas , but not the induction phase of immune escape . In this paper , we employed the bilateral tumor inoculation model , which can distinguish local immune ignorance from systemic immune suppression , and categorized unresponsive tumors into two groups , with or without SIP . Small molecule ( s ) with less than 3 kDa size which is released from SIP-positive tumors appear to attenuate mitochondria-mediated energy metabolism in T cells . We rule out the known factors such as suppressive cytokines , adenosine , Prostaglandin E2 ( PGE2 ) and kynurenine . Tumor cells show dysregulated cellular metabolism and the metabolic products often induce immune suppression ( DeBerardinis , 2008; Munn and Mellor , 2013; Vazquez et al . , 2016 ) . Although it has been reported that methyl-nicotinamide ( MNA ) , which is converted by nicotinamide N-methyl-transferase ( NNMT ) , acts as an immune suppressive factor ( Gebicki et al . , 2003 ) , this compound showed no suppression at physiological levels ( data not shown ) . Other metabolites could be candidates , which are derived from the tumor’s metabolic activity . For successful PD-1 blockade therapy , the ‘tumor-immunity cycle’ needs to operate smoothly ( Chen and Mellman , 2013; Pio et al . , 2019 ) . Hindrance in the pathway at any step of antigen recognition , activation , recruitment and killing at the tumor site , DLN or bloodstream would lead to the unresponsive state ( Mushtaq et al . , 2018 ) . DLN is generally considered as a place where naïve T cells are primed to effector T cells . Our bilateral tumor model analysis suggests that LLC systemically inhibits T cell priming at DLN of responsive tumor sides via suppressive factors , but B16 does not . However , it seems to contradict that T cells in DLN on the side of B16 were not activated in spite of the deficiency of SIP . This observation suggests that tumor recognition by the local tumor area is critical to trigger T cell priming in DLN and to establish a successful tumor-immunity cycle . Therefore , tumors lacking MHC take advantage of the ignorance or escape mechanism not only in the local tumor area but also in DLN . Given that LLC expresses MHC and is sensitive to the acquired immunity to some extent , it is reasonable that LLC but not B16 is susceptible to the combination therapy . Mitochondrial activation is essential for the full activation of T cells . In our in vitro assay system for mitochondrial activities , we stimulated naïve CD8+ T cells by anti- ( CD3+CD28 ) mAb beads because CD28 in addition to CD3 signal is necessary for robust mitochondrial activation during the proliferation ( Klein Geltink et al . , 2017 ) . Although our OCR data suggest that the suppressive factors downregulate the mitochondrial activity , ECAR also severely inhibited . Therefore , the suppressive factors may inhibit glycolysis , resulting in the attenuation of subsequent OXPHOS reactions . This hypothesis agrees with the fact that T cells rely on glycolysis more than OXPHOS when they differentiate from naïve to effector T cells ( Menk et al . , 2018 ) . Another possible mechanism for suppression of mitochondrial function by the suppressive factors is inhibition of the downstream signals of CD3 and/or CD28 because these two signals are necessary for the upregulation of glycolysis and OXPHOS in T cells . In this work , we applied bezafibrate to unresponsive LLC or CT26 tumors . We found this combination therapy partially restored the PD-1 blockade effect per the in vitro assays where bezafibrate partially removed the mitochondrial inhibition by the suppressive factors in the supernatant . This partial effect suggests that under the situation of ‘brake’ induced by the suppressive factors , the ‘acceleration’ by PGC-1α/PPAR activation would not fully work . To obtain the maximum benefit , we need to define the suppressive factors and remove the ‘brake’ . Our data suggest the possibility of unknown small molecules for suppressive factors . The purification of this small molecule by bio-assays will enable us to identify its structure by mass spectrometry . Once we know such a compound , we may be able to find the enzymes responsible for the synthesis of this product and target them for combinatorial treatment .
C57BL/6N and BALB/c inbred mice were purchased from ‘The Charles River Laboratories , Japan ( Kanagawa , Japan ) ’ . Pdcd1-/- and Rag2-/- inbred mice lines were maintained under specific pathogen-free conditions at the Institute of Laboratory Animals , Graduate School of Medicine , Kyoto University . Female , 6–8 weeks-old mice were used in all the experiments . Cell lines were cultured in RPMI or DMEM medium ( Gibco , Grand Island , NY , USA; catalog #11875–093 and 11995–065 respectively ) with 10% ( v/v ) heat-inactivated fetal bovine serum and 1% ( v/v ) penicillin-streptomycin mixed solution ( Nacalai Tesque , Kyoto , Japan , 26253–84 ) as per the instructions recommended by the ATCC . Cell lines were free of mycoplasma contamination . Cell cultures were maintained at 37°C with 5% CO2 in a humidified incubator . Details of different murine cell lines used in the experiment e . g . source of cell lines , background , and origin of cancer , etc . are mentioned in Table 1 . The tumor cell lines MethA and GL261 were passaged in vivo once before use in experiments . Tumor cells were intradermally ( i . d . ) injected into the right flank of mice ( day 0 ) . Monotherapy with the anti-PD-L1 antibody was started when the tumor size reached 50–60 mm3 ( around day 5 ) . Mice were intraperitoneally ( i . p . ) injected with 80 μg of anti-PD-L1 mAb ( clone 1-111A . 4 ) ; mAb injection was repeated every fifth day . For untreated mice , an isotype control for the anti-PD-L1 mAb ( Rat IgG2a , κ ) was injected . Tumor sizes were measured every alternate day using a digimatic caliper ( Mitutoyo Europe GmbH , Germany ) and tumor volume was calculated using the formula for a typical ellipsoid [π × ( length ×breadth × height ) /6] . First , unresponsive tumor cells were i . d . - injected into the left flank of mice ( day 0 ) . When the size of the unresponsive tumor was around 60–70 mm3 ( around day 6–7 ) , responsive tumor cells were i . d . - injected into the right flank . Two-three days after the responsive tumor injection ( around day 9–10 ) , anti-PD-L1 antibody was injected following a monotherapy treatment model ( for the dose of antibody and interval between two injections ) . Tumor sizes of responsive and unresponsive tumors were measured every alternate day and tumor volume was calculated according to the formula mentioned earlier . Bezafibrate ( Santa Cruz Biotechnology , Dallas , TX , USA ) was used at the dose of 5 mg/kg for in vivo combination therapy . Bezafibrate was freshly prepared , immediately before use , in DMSO . Dissolved bezafibrate was diluted in PBS and 200 μL was i . p . -injected per mouse . Bezafibrate was added at the concentration of 5 μM for in vitro assays throughout this work wherever it is used unless specified . For combination therapy experiments , the therapy started when the tumor size was 60–70 mm3 . Mice were i . p . - injected with 40 μg of anti-PD-L1 mAb ( clone 1-111A . 4 ) ; the mAb injection was repeated every sixth day . Mice were i . p . -injected with bezafibrate at 5 mg/kg dose every third day . For control groups , an isotype control for the anti-PD-L1 mAb ( Rat IgG2a , κ ) and DMSO vehicle for bezafibrate were injected . All groups were subjected to the same dose of DMSO . Tumor measurement was performed as stated above . To isolate naïve CD8+ T cells from C57BL/6N inbred wild-type mice , the spleen and three LNs ( axillary , brachial , and inguinal LNs ) from both the right and left sides were harvested . The spleen was minced , treated with ACK lysis buffer ( 0 . 15 M NH4Cl + 1 . 0 mM KHCO3 + 0 . 1 mM Na2-EDTA ) for 2 min to lyse the erythrocytes , and mixed with pooled and minced LN cells . Naïve ( CD62Lhigh CD44low ) CD8+ T cells were then purified from total pooled lymphocytes according to the manufacturer’s instructions ( Miltenyi Biotec , 130-096-543 ) . For in vitro analysis , naïve CD8+ T cells were stimulated with anti-CD3 and CD28 mAb-coated dynabeads ( Thermo Fisher Scientific , Gibco , Catalog# 11452D ) . We seeded 0 . 5 million cells/well in 6-well plates in 4 mL total volume of respective media as recommended by the ATCC . After 48 hrs of incubation , we harvested the culture supernatant , centrifuged at 10 , 000 x g for 15 min at 22°C , collected the supernatant , and kept it at −80°C for storage . We added culture supernatant one-fourth of the total volume in the well ( 96-well round-bottom plate ) throughout the in vitro assays with naïve CD8+ T cells in this work , unless specified . Thymidine solution diluted in spleen RPMI ( Basal RPMI media with 10% FCS , 1% Penicillin-Streptomycin , 50 μM 2-Mercapto ethanol , L-Glutamine , Na-pyruvate , NEAA ) was added to cells and incubated for 4 hrs at 37°C in a humidified incubator with 5% CO2 . After incubation , cells were transferred to a 96-well filter plate followed by the addition of scintillation buffer . Thymidine uptake was measured on a Microbeta2 microplate counter ( PerkinElmer , # 2450–0120 ) machine . To inactivate the protein component , culture supernatant was boiled for 10 min at 95°C followed by centrifugation at 10 , 000 x g for 30 min . The supernatant was collected and stored at −80°C for storage . To remove small molecules , the supernatant was treated with DCC , which removes small molecules ( e . g . nucleotides , vitamins , lipids ) from the sample by adsorbing them on the surface . To remove small molecules , 12 mg DCC ( for 500 μL supernatant ) was added and incubated for 20 min at 25°C , followed by centrifugation at 10 , 000 x g for 30 min . After centrifugation , the supernatant that was free from small molecules was collected . Cultures supernatants were fractionated into different fractions using amicon ultra-centrifugal filters ( Merck Millipore Ltd . , Ireland ) with cut-off sizes of 3 KDa and 50 KDa . Supernatants were added to 3 KDa filter and centrifuged at 12 , 000 x g for 30 min at 4°C . The filtered supernatant was collected and further fractionated using a higher cut-off filter ( 50 KDa ) in a similar way . For draining lymph node ( DLN ) analysis , axillary , brachial , and inguinal LNs ( one of each ) were harvested from the tumor-bearing side ( left or right flank ) of mice . All LNs were minced and pooled . Average LN cell numbers ( total pooled LN cells/3 ) were used as absolute cell numbers . For tumor-infiltrating lymphocyte ( TIL ) analysis , tumor tissue was harvested and cut into 1–2 mm pieces with scissors followed by digestion with collagenase type IV ( Worthington Biochemical Corporation , Lakewood , NJ , Catalog # LS004188 ) using a gentle MACS Dissociator ( Miltenyi Biotec ) . The numbers of TILs per mg of tumor tissue were used as the absolute numbers . The following monoclonal antibodies ( mAbs ) were used to detect the respective antigens during FACS staining: CD8 ( 53–6 . 7 ) , CD62L ( MEL-14 ) , CD44 ( IM7 ) , CD45 . 2 ( 104 ) , T-bet ( 4B10 ) , IFN-γ ( XMG-1 . 2 ) from BioLegend ( San Diego , CA , USA ) ; and Ki67 ( SolA15 ) from eBioscience ( San Diego , CA , USA ) . All flow cytometry experiments were performed on a FACS Canto II ( BD Biosciences , Franklin Lakes , NJ , USA ) , and analyzed using the FlowJo software ( FLOWJO , LLC , Ashland , OR , USA ) . Mitochondrial mass , membrane potential , mitochondrial superoxide , and cellular ROS were determined by MitoTracker Green , MitoTracker Deep Red , MitoSOX Red , and CellROX Green reagents , respectively ( all from Life Technologies , Carlsbad , CA , USA ) . The cells were washed twice with D-PBS buffer followed by the addition of dye solution with final concentrations of 0 . 125 , 0 . 125 , 5 . 0 , and 0 . 625 μM , respectively , in RPMI media and incubated at 37°C in a 5% CO2 humidified incubator for 30 min . After incubation , cells were washed twice with D-PBS buffer followed by surface staining . For intranuclear staining , cells were fixed and permeabilized using the Foxp3 staining kit ( Thermo Fisher Scientific , Catalog # 00-5523-00 ) following the manufacturer’s instructions . After fixation and permeabilization , cells were incubated with the respective antibody for 15 min at 4°C in the dark , followed by washing with FACS buffer ( PBS , 0 . 5–1% BSA or 5–10% FBS , 0 . 1% NaN3 sodium azide ) . Homogenized tumor mass cells from in vivo treated experimental mice were incubated for 4 hrs at 37°C in a 5% CO2 humidified incubator . After incubation , Brefeldin A and Monensin ( eBioscience , Invitrogen , Carlsbad , CA , USA; catalog # 4506–51 and 4505–51 respectively ) were added at the concentration of 5 μg/mL and 2 μM as per the manufacturer’s instructions and incubated for further 2 hrs . Following a total of six hours of incubation , cells were washed once with D-PBS and further stained for surface proteins , if any . Cells were then fixed with 1 . 5% paraformaldehyde solution ( incubated for 15 min at 4°C ) and washed twice with FACS buffer . Cells were then treated with 0 . 5% Triton-X-100 in PBS and incubated for 15 min at 4°C to permeabilize the cells . Monoclonal antibodies to IFN- γ were added ( the concentration was pre-optimized ) and incubated for 15 min at 4°C followed by washing with FACS buffer . We isolated RNA from the experimental groups with the RNeasy mini kit ( QIAGEN , Hilden , Germany ) and synthesized cDNA by reverse transcription ( Invitrogen ) . The primers used to perform quantitative reverse transcription PCR ( qRT-PCR ) are listed here . The primers pairs used were FP: TACCACCCCATCTGGTCATT , RP: GGACGTTTTGTTTGGTTGGT for CD39; FP: CAAATCCCACACAACCACTG , RP: TGCTCACTTGGTCACAGGAC for CD73; FP: CAAGGGAGTCTGGAACATTG , RP: ACCCAGGTCCTCGCTTATGA for COX2; FP: ATGAGTACACGAAGCCGAGG , RP: CCAGTATTACAGGAGTGACCCAG for mPGES1; FP: CACTGAGCACGGACGGACTGAGA , RP: TCCAATGCTTTCAGGTCTTGACGC for IDO1; FP: CGGAAATCATATCCAACCAG , RP: TGAGGACCGCTAGCAAGTTTG for PGC-1α; FP: GGTGTTCGGTGAGATTGTAGAG , RP: GTGATAAAACCGTGCTTCTGG for PGC-1β; and FP: TATTGGCAACGAGCGGTTCC , RP: GGCATAGAGGTCTTTACGGATGT for β-actin . β-actin was used as loading control . The oxygen consumption rate ( OCR ) and extracellular acidification rate ( ECAR ) of treated cells were measured using an XFe96 Extracellular Flux analyzer ( Seahorse Biosciences , North Billerica , MA , USA ) . One day before the experiment , first the XFe96 plate was coated with CellTak solution as per the manufacturer’s recommendation . On the day of the experiment , all chemicals ( e . g . Oligomycin , FCCP , and Rotenone/Antimycin A ) were prepared in OCR media as per the manufacturer’s recommendation and the machine was calibrated using the calibrant buffer in the calibrant plate prior to the experiment . 400 thousand cells per well were seeded in the precoated XFe96 plate and the OCR/ECAR was measured . Different parameters from the OCR graph were calculated . ATP turnover was defined as follows: ( last rate measurement before oligomycin ) - ( minimum rate measurement after oligomycin injection ) . Maximal respiration was defined as follows: ( maximum rate measurement after FCCP ) - ( non-mitochondrial respiration ) . Spare respiratory capacity ( SRC ) was calculated by subtracting basal respiration from maximal respiration . We measured the ECAR value in the same well , which contained an optimal glucose level so the basal ECAR ( or glycolysis ) value is the reading we obtained immediately before oligomycin injection . We prepared the assay medium as described in the XF cell Mito Stress Test Kit ( Kit 103015–100 ) . The glucose concentration in this medium is 10 mM . In the classical glycolytic assay procedure ( glucose-free media ) the final concentration of glucose added to the port was 10 mM while measuring flux . The basal ECAR value in this classical method is calculated by subtracting the last rate measurement before the glucose injection from the maximum rate measurement before the oligomycin injection , which gives essentially the same value if calculated by our method . Glycolytic capacity was defined as the rate measured after the oligomycin injection . The glycolytic reserve was defined as follows: ( glycolytic capacity ) – ( basal ECAR value ) . Statistical analysis was performed using Prism 6 ( GraphPad Software , La Jolla , CA , USA ) . One-way ANOVA analysis followed by Sidak's multiple comparison test was utilized to analyze three or more variables . To compare two groups , student’s t-test was used . All statistical tests were two-sided assuming parametric data , and a p-value of <0 . 05 was considered significant . The variations of data were evaluated as the means ± standard error of the mean ( SEM ) . Five or more samples were thought to be appropriate for the sample size estimate in this study . Samples and animals were randomly chosen from the pool and treated . No blinding test was used for the treatment of samples and animals . Mice were maintained under specific pathogen-free conditions at the Institute of Laboratory Animals , Graduate School of Medicine , Kyoto University under the direction of the Institutional Review Board . | Immunotherapy is a fast-emerging treatment area that turns the body’s own immune system against cancer . One powerful group of treatments are the PD-1 blockers . PD-1 is an inducible protein that is sometimes found on healthy immune cells called T cells and normally acts to stop T cells mistakenly attacking healthy cells . However , it can also prevent T cells attacking cancer . This happens when cancer cells make a protein called PD-1 ligand , which interacts with PD-1 to switch off nearby T cells . Antibodies that block PD-1 or PD-1 ligand can reactivate T cells , allowing them to destroy the cancer , but this PD-1 blocking therapy currently works in less than half of all patients who receive the treatment . To mount a successful defense against cancer , a T cell needs to be able to perform two key tasks: recognize cancer cells and prepare to attack . T cells are alerted to the presence of the disease by MHC class I proteins on the surface of cancer cells holding up small fragments of molecules that are tell-tale sign that the cell is cancerous . To prepare to attack , a T cell depends on its mitochondria – the powerhouses of the cell – to send a cascade of signals inside the T cell that help it to activate and multiply . It is possible that cancer cells escape PD-1 blocking treatments by interfering with either one of these two tasks . They may either hide their MHC class I proteins to become invisible to passing T cells – a phenomenon known as “local immune ignorance”; or they may release long-range molecules to stop T cells preparing to attack – “systemic immune suppression” . To explore these options further , Kumar , Chamoto et al . developed a new tumor model in mice . Each mouse had two tumors , one that responded to PD-1 blocking treatment and one that did not . The idea was that , if the unresponsive tumor was simply hiding from passing T cells , its presence should not affect the other tumor . But , if it was releasing molecules to block T-cell activation , the other tumor could become unresponsive to PD-1 blocking treatment too . Kumar , Chamoto et al . examined different types of unresponsive tumor in this model system and found that they fell into two groups . The first group simply hid themselves from nearby T cells , while the second group released molecules to dampen all T cells . The identity of these molecules is unknown , but further experiments suggested that they likely work by blocking the mitochondria in T cells . In mice with these tumors , drugs that boosted mitochondria activity made anti-PD-1 treatment more effective . If the findings in this mouse model parallel those in humans , it could open a new research area for immunotherapy . The next step is for researchers need to identify the molecule responsible for systemic immune suppression . This could help to make PD-1 blocking treatments more effective in people who do not currently respond . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"immunology",
"and",
"inflammation"
] | 2020 | Tumors attenuating the mitochondrial activity in T cells escape from PD-1 blockade therapy |
Variability within isogenic T cell populations yields heterogeneous ‘local’ signaling responses to shared antigenic stimuli , but responding clones may communicate ‘global’ antigen load through paracrine messengers , such as cytokines . Such coordination of individual cell responses within multicellular populations is critical for accurate collective reactions to shared environmental cues . However , cytokine production may saturate as a function of antigen input , or be dominated by the precursor frequency of antigen-specific T cells . Surprisingly , we found that T cells scale their collective output of IL-2 to total antigen input over a large dynamic range , independently of population size . Through experimental quantitation and computational modeling , we demonstrate that this scaling is enforced by an inhibitory cross-talk between antigen and IL-2 signaling , and a nonlinear acceleration of IL-2 secretion per cell . Our study reveals how time-integration of these regulatory loops within individual cell signaling generates scaled collective responses and can be leveraged for immune monitoring .
Understanding how collective biological function emerges from individual cell signaling remains challenging: rapid , binary decisions at the individual cell level ( e . g . , signal transduction and gene activation ) must be bridged to graded , longer-term outcomes at the population level ( e . g . , proliferation , survival , and differentiation ) . This issue is particularly acute for the adaptive immune systems of metazoans , in which multicellular lymphocyte responses scale to the strength of pathogenic challenge across many spatial and temporal ranges . Investigations of information processing in mammalian systems have shown that regulations such as feedback ( Becker et al . , 2010 ) or population averaging ( Cheong et al . , 2011 ) are required to generate a large dynamic range of biological response from the limited resolution of proximal cell signaling . Previous studies have demonstrated that T cell population responses to antigens vary widely with the quality and quantity of pathogenic stimuli several days after immunization ( Zehn et al . , 2009; van Heijst et al . , 2009 ) . Indeed , T cell receptors ( TCRs ) are highly sensitive to differences in antigenic signal strength within the first minutes and hours following contact with peptide-MHC ( pMHC ) complexes ( Lanzavecchia and Sallusto , 2001 ) . However , phenotypic variability within clonal populations of T cells results in heterogeneous sensitivity to shared antigen stimuli ( Feinerman et al . , 2008 ) . Therefore , isogenic T cells in the same antigenic environment can display completely different signaling responses . Furthermore , early antigen discrimination is all-or-none ( Das et al . , 2009 ) , with limited dynamic ranges in both antigen input sensing and response output ( Altan-Bonnet and Germain , 2005; Tkach and Altan-Bonnet , 2012; Huang et al . , 2013 ) , but long-term signaling can widen the gradation of functional responses . For example , a 100-fold difference in bacterial infection load is poorly resolved by a 1 . 5-fold change in the number of activated clones , yet ultimately results in a 20-fold shift in the magnitude of T cell expansion , notably as a result of sustained antigen signaling ( van Heijst et al . , 2009 ) . How the noisy , bimodal decisions of single cells are coordinated to reflect global cognate antigen load over several days of response is currently unclear . Paracrine cytokines that are secreted upon activation present an elegant solution for scaling population-level lymphocyte responses by translating individual antigen stimulation into a ‘public good’ on a longer time scale . Yet collective cytokine accumulation may be highly dependent on population size , as anticipated by quorum sensing systems ( Nealson , 1977; Dai et al . , 2012 ) . Surprisingly , the antigen-scaling of in vivo clonal effector responses is mostly unaffected by large variability in the initial number of responding T cells ( Smith et al . , 2000; Jenkins and Moon , 2012 ) . Thus , lymphocyte populations require mechanisms to create a wide dynamic antigen response range ( Becker et al . , 2010; Tay et al . , 2010; Cheong et al . , 2011; Waysbort et al . , 2013 ) , independently of initial population size ( Smith et al . , 2000; Quiel et al . , 2011 ) . It has been qualitatively proposed that competition for antigen may normalize for clonal density by limiting the duration of antigen signaling—and consequently , outcomes such as proliferation—within larger populations ( Smith et al . , 2000; Tkach and Altan-Bonnet , 2012 ) . However , the quantitative molecular mechanisms required to compensate for hundreds-fold differences in population size within physiological timescales remain unknown . In bridging the molecular , cellular , and population-level scales that regulate immune function , there is much to be gained from quantitative and theoretical approaches ( O’Garra et al . , 2011 ) . Biological studies frequently apply genetic tools to dissect the nodes of regulatory networks . However , the ability to quantitatively track the route from molecular perturbation to functional phenotype requires an integrative and dynamic framework . Experimentally validated computational models have the capacity to generate quantitative predictions and establish the minimal requirements for the emergence of biological phenotypes . Moreover , experimental characterization and predictive modeling of the quantitative , dynamic relationships between system components can reveal regulatory architecture without genetic perturbation ( Yosef et al . , 2013 ) . Such quantitative approaches have been successfully applied in varied studies of the immune system , from lymphocyte signaling ( Chakraborty et al . , 2009; Das et al . , 2009 ) to receptor repertoire generation and thymic development ( Weinstein et al . , 2009; Kosmrlj et al . , 2010; Mora et al . , 2010; Georgiou et al . , 2014 ) , competition for cytokines ( Busse et al . , 2010; Feinerman et al . , 2010 ) , lymphocyte differentiation ( Schulz et al . , 2009; Francois et al . , 2013 ) and host-pathogen interactions ( Ciupe et al . , 2007; Althaus and De Boer , 2008 ) . Ultimately , quantitative analysis of immune responses can facilitate discovery in settings where gene modification is impractical , as in primary human cells , or where the direct and indirect effects of genetic alterations mask subtle or unanticipated interactions ( Sontag et al . , 2004 ) . Although models always fail to capture the full complexity of immune responses , we posit that model building allows thorough , iterative interrogation of the sufficiency of molecular steps to account for large-scale functional properties ( Kemp et al . , 2007; Janes et al . , 2008 ) . Therefore , computational models are ideal for directly testing the emergence of collective responses from signaling within individual lymphocytes . One candidate mediator of lymphocyte cooperation is interleukin-2 ( IL-2 ) , a paracrine cytokine produced and shared by activated T cells ( Smith , 1988 ) . Since IL-2 is secreted early after antigen challenge yet quantitatively tunes late decisions such as the magnitude of expansion and differentiation program of T cells ( Williams et al . , 2006; Bachmann et al . , 2007; Pipkin et al . , 2010; Liao et al . , 2011; McNally et al . , 2011; Boyman and Sprent , 2012 ) , its accumulation may also link disparate time scales of cellular activation . Studies of IL-2 at single , early timepoints have reported that IL-2 scaling is limited in dynamic range , reflecting only the number of digitally activated T cells ( Podtschaske et al . , 2007; Huang et al . , 2013 ) . However , IL-2 production and consumption are modulated by several known feedbacks downstream of IL-2 signaling ( Smith , 1988; Long and Adler , 2006; Villarino et al . , 2007; Boyman and Sprent , 2012; Yamane and Paul , 2012 ) , which could alter the dynamic range ( Nevozhay et al . , 2009; Becker et al . , 2010 ) and T cell number-dependency of IL-2 output over time . Here , we quantitatively characterize the cue-signal-response module of antigen-driven IL-2 secretion , and find that the empirical scaling of IL-2 accumulation challenges current understanding of this cytokine’s production . Simulations of the known regulatory elements of the IL-2 pathway ( Feinerman et al . , 2010 ) predict a low saturating threshold and a strong population size dependence for IL-2 output . However , we demonstrate that IL-2 accumulation by T cells scales as a power law with antigen quantity , independently of population size , providing a shared quantitative readout of the global antigen load . Through experimental and computational probing , we uncovered two critical regulatory elements—a cross-talk interaction and a non-linear feedback—whose inclusion in the model captured the dynamics and scaling of collective IL-2 accumulation , and allowed for accurate prediction of the IL-2 pathway in a polyclonal system . Our study demonstrates how integration of feedbacks over long timescales enables variably sized populations of cells to respond proportionally to a large range of stimuli . Furthermore , these feedbacks carry information about the initiating TCR signal . Indeed , the observed cross-talk between TCR and IL-2 receptor signaling can be used to estimate the degree of antigen signaling experienced by activated T cells in response to un-calibrated stimuli such as explanted tumor tissue .
We measured various input/output relationships for activation of different numbers of primary 5C . C7 TCR transgenic T cells responding to antigen presenting cells pulsed with varied doses of K5 antigen in vitro . In single timepoint snapshots , our observations aligned with previous work ( Altan-Bonnet and Germain , 2005; Podtschaske et al . , 2007; Huang et al . , 2013 ) , showing that T cells respond in a quantal manner to graded doses of antigen across various readouts ( ERK phosphorylation , IL-2Rα or IL-2 expression , and cell cycle entrance—Figure 1A ) . Increasing the stimulating dose of antigen for 100 , 000 T cells did amplify the frequency of activation , but not in proportion to the shift in input stimulus: several readouts were saturated for higher antigen doses , and at best , a 1000-fold increase in antigen translated into 50-fold gain in the IL-2 production response ( Figure 1B ) . IL-2 accumulation at 12 hr reflected this limited antigen-scaling in the frequency of IL-2 producers; furthermore , scaling at 12 hr was greatly influenced by the numbers of T cells in the system , with smaller populations demonstrating poorer antigen resolution ( Figure 1C ) . However , probing further the IL-2 dynamics over days post-activation , we found that IL-2 accumulated rapidly and non-linearly with time , then exponentially decreased ( Figure 2A–B , D ) , despite limited variation in the number of cells ( Figure 2C ) . For each condition , we characterized these dynamics by their apex , [IL-2]max , a quantity which was proportional to the T cell population’s total [IL-2] accumulation over time ( Figure 2E ) . Strikingly , [IL-2]max scaled as a power law that could distinguish more than three orders of magnitude of input antigen dose ( Figure 2F ) , despite the aforementioned saturated dynamic range of early T cell responses . Furthermore , [IL-2]max was essentially independent of the number of antigen-specific T cells in the culture ( NT cells ) ( Figure 2G ) . Given the same stimulus , 1000 T cells accumulated equal or greater amounts of IL-2 as 100-fold larger populations , in less than twice as much time . This could not be accounted for by differential changes in cell number ( through proliferation or death ) , which were at most two-fold by the time parity was gained in IL-2 accumulation ( Figure 2C ) . From 118 conditions over six independent experiments , we derived a simple empirical scaling law that summarizes the population size-independence and large scalability of IL-2 accumulation ( Figure 2H ) : ( 1 ) [IL−2]maxexperiment∝ ( NT cells ) −0 . 10 ( ±0 . 02 ) ×[Antigen]+0 . 76 ( ±0 . 05 ) . 10 . 7554/eLife . 01944 . 003Figure 1 . Limited dynamic range of T lymphocyte activation snapshots at the individual cell level . Varied numbers ( NTcell ) of 5C . C7 TCR Transgenic Rag2−/− T cells were cultured in duplicate in 200 µl of complete medium with 5 . 105 I-Ek-expressing B10 . A Cd3e−/−splenocytes pulsed with varied concentrations of K5 peptide ( [Antigen] ) . ( A and B ) Cells were harvested at varied timepoints and analyzed by flow cytometry for phosphorylation of ERK , upregulation of IL-2Rα , production of IL-2 via a Miltenyi IL-2 secretion assay , or dilution of Cell Trace Violet ( CTV ) upon cell proliferation . These measurements demonstrate ( A ) the bimodality of T cell activation as well as ( B ) the limited dynamics range of response for varied doses of antigens . ( C ) Supernatants were also collected at 12 hr and [IL-2] was measured by ELISA . DOI: http://dx . doi . org/10 . 7554/eLife . 01944 . 00310 . 7554/eLife . 01944 . 004Figure 2 . Dynamics and scaling of IL-2 production and consumption by T cells in vitro . For T cell cultures described in Figure 1 , ( A and B ) supernatants were collected at different timepoints ( typically every 4 hr ) , and [IL-2] was measured by ELISA . For each condition , we recorded the maximal concentration of accumulated [IL-2] ( [IL-2]max , filled symbol ) . Data are represented as mean ± SEM . ( C ) Number of live T cells in cultures as a function of time . For a given quantity of stimulating antigen ( indicated by the symbol ) , the filled point marks the time at which the cytokine accumulation of smaller populations ( 104 and 103 T cells ) is equal to or surpasses the larger population’s [IL-2 ]max ( 105 cells ) . ( D ) Nonlinear accumulation and consumption of IL-2 for a culture of 104 T cells activated with 5 . 105 B10 . A Cd3e−/− splenocytes pre-pulsed with 250 nM of K5 antigen illustrated in linear scale . ( E ) Correlation of [IL-2]max with the total accumulated [IL-2] over time for 118 different conditions ( varied doses of antigen and varied numbers of T cells ) over six experiments . ( F ) [IL-2]max scales almost linearly with [Antigen] over a large dynamic range . ( G ) [IL-2]max is practically independent of NTcell . These data were compiled from independent experiments in which T cells were stimulated with either 1 , 2 , or 2 . 5 * 10−6/10−7/10−8/10−9 M antigen . Thus , results in ( G ) are grouped according to order of magnitude of antigen dose . ( H ) Scaling law for experimentally determined [IL-2]max as a function of [Antigen] and NTcell . The grey plane is fitted for the PLSR result . DOI: http://dx . doi . org/10 . 7554/eLife . 01944 . 004 This result challenged prior understanding of the extensively studied IL-2 pathway ( Figure 3A ) . It has been established that T cells bimodally secrete IL-2 ( Podtschaske et al . , 2007 ) and express the α chain of the IL-2 receptor ( IL-2Rα ) ( Sheldon et al . , 1993 ) following TCR engagement . IL-2 production is subsequently switched off after loss of TCR signaling ( Huppa et al . , 2003 ) and/or gain of IL-2 response via phosphorylation of the transcription factor STAT5 into pSTAT5 ( Long and Adler , 2006; Villarino et al . , 2007; Feinerman et al . , 2010; Waysbort et al . , 2013 ) , which also mediates further upregulation of IL-2Rα ( Smith and Cantrell , 1985; Waysbort et al . , 2013 ) . Implementation of this classical model ( Figure 3B–E ) predicted that T cells would accumulate IL-2 commensurably with the size of their population . Furthermore , thiss model of IL-2 production predicted a weak antigen dependency that saturated near the canonical IL-2 signaling threshold of 10 pM ( Smith , 1988 ) ( Figure 3E ) , revealing a discrepancy between the established pathway regulation and our experimental results . IL-2 receptor exposure to only 10 pM of IL-2 is biophysically sufficient to trigger ligand binding ( Smith , 1988 ) , and subsequent STAT phosphorylation sharply inhibits IL-2 synthesis ( Long and Adler , 2006; Villarino et al . , 2007; Waysbort et al . , 2013 ) . Accordingly , T cells should be incapable of producing much more than 10 pM of cytokine ( Figure 3 ) . Empirically , however , IL-2 accumulation readily exceeded this concentration ( Figure 2 ) . In Figure 3F we show that the two-dimensional fit of the predicted scaling exponents for the classical model ( 2 ) [IL−2]maxpredicted∝ ( NT cells ) +0 . 25 ( ±0 . 005 ) ×[Antigen]+0 . 32 ( ±0 . 008 ) , was also incompatible with our experimental results ( Equation 1 ) . These contradictions prompted further investigation of the IL-2 pathway . 10 . 7554/eLife . 01944 . 005Figure 3 . Shortcomings of the classical model of the IL-2 pathway . ( A ) Sketch of the classical pathway for IL-2 secretion and consumption . ( B ) Cartoon representation of progression through cellular states during production and consumption of IL-2: from naive ( IL-2Rα− , IL-2− ) to activated IL-2 producers ( IL-2Rα+ , IL-2+ and IL-2Rα+ ) and finally to IL-2 consumers ( IL-2Rα++ , IL-2− ) . ( C ) Biochemical model of IL-2 regulation as described in the literature ( ‘classical model’ ) ; parameters in green are derived from experiments . Classical model prediction of IL-2 dynamics for ( D ) 104 T cells stimulated with varied quantities of antigen and ( E ) varied numbers of T cells stimulated with 108 molecules of antigen . ( F ) Two-dimensional dependency for [IL-2]max as a function of [Antigen] and NTcell , as predicted by the above classical model . The red border represents the theoretical 10 pM ceiling of [IL-2] that cells can secrete before switching off IL-2 secretion . DOI: http://dx . doi . org/10 . 7554/eLife . 01944 . 005 To investigate this discrepancy , we systematically probed the dynamics of the IL-2 pathway by measuring IL-2 concentrations via ELISA , as well as cytokine receptor levels ( IL-2Rα and IL-2Rβ ) and response ( pSTAT5 ) by flow cytometry . Visualizing the joint kinetics of cytokine , receptor and phospho-signal among activated cells , we found a striking antigen-dependency in the trajectory of the IL-2 pathway ( Figure 4A , left ) . The pSTAT5 response was proportional to the product of IL-2Rα abundance and accumulated [IL-2] over time leading up to [IL-2]max , but the efficiency of STAT5 phosphorylation lessened with increasing antigen dose ( Figure 4A , right ) . This observation is consistent with previous reports of TCR signaling inhibiting the pSTAT5 response to IL-2 ( Lee et al . , 1999; Yamane et al . , 2005 ) . 10 . 7554/eLife . 01944 . 006Figure 4 . Experimental characterization of the antigen-driven inhibition of IL-2 signaling . ( A ) Dynamics of IL-2 pathway over 150 hr ( arrows indicate progression in time ) for 105 5C . C7 T cells activated in vitro by splenocytes pulsed with varied [K5] antigen ( left ) . STAT5 phosphorylation was measured as the geometric mean fluorescence intensity ( GMFI ) at different times before reaching the maximal [IL-2] concentration ( filled symbol ) , and correlated with the product of [IL-2] and IL-2Rα GMFI for activated cells ( right , representative of more than four experiments ) . ( B ) STAT5 phosphorylation in response to exogenous IL-2 for cells 48 hr post activation with splenocytes pulsed with varied doses of antigen . pSTAT5 is reported as GMFI for all activated IL-2Rα+ T cells . ( C ) Distributions of the abundance of IL-2Rα and IL-2Rβ at 48 hr post activation with splenocytes pulsed with varied doses of antigen . Cell-to-cell variability analysis ( CCVA ) parses these distributions to compare the signaling responses among populations of cells ( bins ) defined by set levels of IL-2Rα and IL-2Rβ ( e . g . , black cross-section across antigen doses ) . ( D ) and ( E ) Cell-to-cell variability analysis , see Experimental Procedures for details . pSTAT5 responses for cultures in ( C ) were parsed according to binned levels of IL-2 receptors . Amplitudes of pSTAT5 for 10 nM ≤ [K5] ≤ 10 µM for each IL-2Rα/IL-2Rβ bin were presented ( D ) as fluorescence intensity ( FI ) or ( E ) as a FI normalized to the pSTAT5 amplitude for [K5] = 10 µM . ( F ) Normalized pSTAT5 amplitude are reported for individual bins of IL-2Rα and IL-2Rβ levels ( top ) or averaged across all IL-2Rα and IL-2Rβ levels ( bottom ) . Error bars are computed as the SEM across all bins . ( G ) Cell-to-cell variability analysis of pSTAT5 response to IL-2 for varied levels of IL-2Rα and IL-2Rβ at different time points . Inset: time dependence of the average pSTAT5 amplitude measured for individual bins of IL-2Rα and IL-2Rβ over time ( n = 3 independent experiments ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01944 . 006 We further characterized the effect of antigen dose on IL-2 signaling by sampling the responses of differentially activated T cells to titrated concentrations of exogenous IL-2 ( Figure 4B ) . 48 hr after antigen activation , cells were collected and stripped of pre-bound cytokine with a low pH buffer , then washed and rested before exposure to serial dilutions of recombinant mouse IL-2 . After 10 min of incubation with IL-2 , cells were fixed , permeabilized and stained for pSTAT5 , IL-2Rα , IL-2Rβ , and γc for flow cytometry analysis . Examining the geometric mean of pSTAT5 in activated ( IL-2Rα+ ) T cells , we found that the quantity of stimulating antigen did not affect the average EC50 , i . e . , sensitivity , of IL-2 response ( Figure 4B ) . Instead , we observed that increasing antigen dose resulted in higher levels of IL-2 receptors at 48 hr ( Figure 4C ) , yet paradoxically dampened the amplitude of STAT5 phosphorylation . Consequently , we postulated that antigen signaling inhibits IL-2 response not at the level of IL-2 receptor engagement , but rather at the level of STAT5 phosphorylation . To disambiguate the antigen/IL-2 signaling cross-talk from concomitant changes in receptor abundance , we performed cell-to-cell variability analysis ( CCVA ) ( Cotari et al . , 2013 ) of IL-2 responsiveness ( Figure 4D–F ) . Using our custom-designed flow cytometry analysis software , ScatterSlice , we parsed the populations of activated T cells into subpopulations ( bins ) of equal IL-2 receptor abundance , and calculated the dose response amplitude of pSTAT5 ( color ) within each bin ( Figure 4D ) . The hindered IL-2 responsiveness of T cells activated with higher antigen doses was easily visualized through these heat maps . We then quantified the pSTAT5-inhibiting effects of TCR signaling by normalizing the pSTAT5 amplitude within each bin to the response of cells that were stimulated with the highest antigen dose ( 10 µM ) , yet expressed equivalent levels of IL-2 receptor α , β ( Figure 4E–F ) and γc chains ( our unpublished data ) . By factoring out the dynamic variation in IL-2 receptor abundance associated with T cell activation , CCVA demonstrated that TCR-driven inhibition of IL-2 signaling scales linearly with antigen dose ( Figure 4F ) . Furthermore , CCVA showed that this inhibitory effect decays throughout the course of the T cell response ( Figure 4G ) , independently of changes in IL-2 receptor levels . Our single cell analyses allowed the deconvolution of downstream signaling events from receptor abundance , and demonstrated the tunability of the inhibitory cross-talk between antigen and pSTAT5 signaling . We hypothesized that this titrated antigen-driven inhibition of IL-2 signaling could delay pSTAT5-mediated shutdown of IL-2 production , especially in strongly activated cells , thus enabling the accumulation of IL-2 beyond the canonical pSTAT5 signaling threshold of 10 pM ( Smith , 1988 ) . We tested this intricate regulation of IL-2 production through signal blocking experiments , using a cytokine-capture assay to identify IL-2-secreting cells ( Figure 5 ) . We demonstrated that persistent TCR signaling is required to sustain IL-2 production ( Huppa et al . , 2003 ) at all times . Administration of an antibody that disrupted TCR-pMHC contact quickly shut down IL-2 production ( Figure 5A–B ) and concomitantly increased pSTAT5 within the population ( Figure 5C–D ) , further suggesting inhibition of IL-2 production by pSTAT5 response , and of pSTAT5 by TCR signaling . Consistent with reports of negative feedback inhibition of IL-2 production by STAT5-mediated IL-2 signaling ( Long and Adler , 2006; Villarino et al . , 2007; Waysbort et al . , 2013 ) , blocking pSTAT5 via a chemical inhibitor of Janus kinase ( JAK ) activity increased the number of IL-2 producing cells ( Figure 5D ) . Strikingly , in contrast to the rapid drop ( τdrop = 0 . 5 ± 0 . 1 hr ) observed in antigen-blocked conditions , dual inhibition of IL-2 and TCR signaling resulted in a slower decline in IL-2 producers ( τdrop = 2 . 2 ± 0 . 1 hr ) ( Figure 5D ) . This demonstrates that pSTAT5 signaling following antigen withdrawal functions as a swifter mechanism to shut down IL-2 secretion compared to the loss of TCR signal alone . 10 . 7554/eLife . 01944 . 007Figure 5 . Coherent feed-forward loop regulation of IL-2 secretion . ( A and B ) 3 × 104 5C . C7 TCR-transgenic Rag2−/− T cells co-cultured with 3 . 5 × 105 APCs pre-pulsed with 500 nM K5 antigen . ( A ) Blocking of cognate pMHC ligand via administration of 20 μg/ml α-I-Ek antibody at varied time points during IL-2 production causes a rapid drop ( detected here within 2 hr ) in the number of IL-2-producing cells . ( B ) Addition of K5 antigen at t = 22 hr to cultures increases the numbers of IL-2-producing cells . ( C ) Phosphorylation of STAT5 is rapidly enhanced upon blocking of cognate pMHC ligand via administration of 20 μg/ml α-I-Ek antibody . ( D ) Cells activated with 1 μM of K5 antigen were treated with a JAK inhibitor or carrier control at 9 hr . Cells of each condition were then treated with antigen-blocking reagent anti-I-Ek or control anti-H2-Db at 30 hr . IL-2 production was measured via cytokine capture assay . All conditions performed in triplicate . ( E ) Sketch of the antigen-driven inhibition of IL-2 signaling ( red ) , which makes IL-2 production contingent on antigen availability . DOI: http://dx . doi . org/10 . 7554/eLife . 01944 . 007 This biochemical network ( Figure 5E ) forms a coherent feed forward loop , in which a signal ( TCR ) and its effect on a target ( inhibition of pSTAT5 ) regulate a common output in the same direction ( promoting IL-2 production ) . More specifically , this is a type IV coherent feed-forward loop , where one arm directly promotes an output , and the other arm represses an output’s inhibitor ( Alon , 2007 ) . Theoretical studies of such motifs have highlighted their signal delay properties and mutual exclusion of target and repressor signals ( Mangan and Alon , 2003 ) , as has been observed for IL-2 and pSTAT5 ( Long and Adler , 2006 ) . Our experiments additionally revealed that this regulatory architecture enforces tight synchronization between output production and sustenance of input cues . Through antigenic control of pSTAT5-mediated repression , IL-2 synthesis is neither terminated before nor extended beyond the loss of antigen signaling . In parallel , we quantified IL-2 secretion for varied T cell population sizes . Using the cytokine-capture assay , we found that greater fractions of smaller populations of T cells maintained IL-2 secretion for longer periods of time ( Figure 6A ) , consistently with in vivo studies ( Sojka et al . , 2004 ) . Paradoxically , though smaller populations of T cells ( 103 ) could accumulate greater concentrations of IL-2 ( [IL-2]max ) than larger populations ( Figure 2B , G ) , they yielded far fewer IL-2-producing cells ( Figure 6B ) . To ‘catch up’ in IL-2 accumulation without converging in numbers of IL-2 producers , T cell population size must adjust cellular rates of IL-2 secretion and/or consumption . 10 . 7554/eLife . 01944 . 008Figure 6 . Count of IL-2 producing cells and rate of IL-2 consumption do not account for scaling law in IL-2 accumulation . We quantitate IL-2 production at the individual cell level for cultures as described in Figure 1 . Percentage ( A ) and counts ( B ) of T cells producing IL-2 as a function of time for varied numbers of 5C . C7 T cells , activated with splenocytes pulsed with 1 µM K5 antigen . ( C ) Negligible IL-2 consumption during IL-2 production phase . Varied numbers of 5C . C7 T cells were activated by splenocytes pulsed with 1 µM K5 antigen and cultured with 10 pM of human IL-2 added 12 hr post initial activation , in triplicate . Left: human IL-2 and Right: mouse IL-2 detected in cultures over time . Graphs are representative of three experiments . ( D ) Depletion of added human IL-2 as a function of IL-2Rα upregulation . 50 , 000 5C . C7 T cells were stimulated with 200 , 000 APCs pulsed with 500 nM K5 antigen during experiment tracking the consumption of 250 or 50 pM human IL-2 added at 6 hr after the start of co-culture . Timepoints were taken every 6 hr between 6 and 96 hr of culture . Percentage ( E ) and counts ( F ) of T cells producing IL-2 as a function of time for 104 T cells activated with splenocytes pulsed with varied doses of K5 antigen . DOI: http://dx . doi . org/10 . 7554/eLife . 01944 . 008 We characterized the effect of population size on IL-2 consumption by adding human IL-2 to large and small populations of T cells 12 hr after antigen activation ( Figure 6C ) . Both mouse and human IL-2 ( hIL-2 ) bind equivalently to the IL-2 receptor but can be measured separately by ELISA ( Deenick et al . , 2003 ) , allowing us to resolve consumption from production . We found that both large and small populations showed similarly limited consumption of hIL-2 before their respective times of maximal mouse IL-2 accumulation ( Figure 6C ) . The onset of hIL-2 consumption correlated with high pSTAT5-driven upregulation of IL-2Rα , which reached its apex several hours after cells amassed maximal mouse IL-2 ( Figure 6D ) . Therefore , the observed parity in IL-2 accumulation between differently sized T cell populations cannot be attributed to differential cytokine consumption during the IL-2 secretion period . In previous studies , single-cell measurements established that T cells are bimodal in secreting IL-2 at early ( <6 hr ) timepoints; stronger antigenic stimuli increases the number of IL-2-producing cells , but not the amount of IL-2 produced per cell ( Podtschaske et al . , 2007; Huang et al . , 2013 ) . Our time series experiments did confirm that greater antigenic stimulus resulted in larger numbers of IL-2 secreting cells over several days ( Figure 6E–F ) . However , the antigen scaling of the number of IL-2 producers was insufficient to account for the observed power law in accumulated IL-2 ( Figure 2F , H ) . We reasoned that if bimodality in IL-2 production indeed sets a constant IL-2 secretion rate per cell ( Podtschaske et al . , 2007 ) , the concentration of cytokine should increase linearly with the cumulative number of secreting cells over time . Surprisingly , we observed the emergence of a nonlinear relationship between these two quantities ( Figure 7A–B ) , demonstrating a time-dependent acceleration in the rate of IL-2 production . 10 . 7554/eLife . 01944 . 009Figure 7 . Experimental characterization of the nonlinear acceleration of IL-2 secretion in activated T cells . ( A and B ) The integral of the number of IL-2 producers over time following activation is compared to [IL-2] ( measured in duplicate ) accumulated ( A ) for 1 µM K5 antigen with different numbers of T cells or ( B ) for 104 T cells exposed to different quantities of antigen . ( C and D ) Fold increase from the average initial rate of IL-2 production per cell as a function of time for varied number of T cells ( C ) or for varied doses of antigens ( D ) . The average rate of IL-2 secretion was estimated to be 7 . 5 molecules per second ( Figure 8—figure supplement 1 ) . ( E ) Sketch of the positive feed-forward loop ( in green ) accounting for the acceleration in IL-2 secretion over sustained periods of antigenic stimulation . DOI: http://dx . doi . org/10 . 7554/eLife . 01944 . 00910 . 7554/eLife . 01944 . 010Figure 7—figure supplement 1 . Criticality of IL-2 boost to achieve antigen-scaling and population-size-independence in [IL-2]max . Top projected values of [IL-2]max if IL-2 secretion is maintained at a constant rate per cell for experimental conditions similar to those described in Figure 7A–B . A constant rate of IL-2 production per cell would yield a linear correlation between the cumulative number of IL-2 producing cells over time and the [IL-2] in the media ( dotted line ) . Points in bold ( on the dotted line ) indicate the projected [IL-2]max if IL-2 secretion were constant per cell . Bottom: discrepancy between observed and projected [IL-2]max scaling with antigen dose and population size . If the IL-2 secretion rate per cell was constant , the dynamic range of [IL-2]max scaling with antigen dose would saturate ( left ) ; additionally , [IL-2]max would scale directly with T cell population size ( right ) . Therefore , a boost in IL-2 production is necessary to achieve the large scalability and population size-independence of [IL-2]max . These results are representative of more than three independent high time-resolution experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 01944 . 010 Despite different stimulating conditions , at early timepoints ( <20 hr ) activated T cells secreted IL-2 at similar rates ( Figure 7C–D , dashed lines & Figure 8—figure supplement 1 ) . However , as smaller populations of T cells sustain longer periods of IL-2 production ( Figure 6A ) , they demonstrated greater time-dependent increases in their rate of IL-2 accumulation per cell ( up to 30-fold ) ( Figure 7C ) . Ultimately , this acceleration allows them to surpass the IL-2 maxima of 100-fold larger populations in a short amount of time ( Figure 2C ) , despite having fewer IL-2 producers ( Figure 6B ) . In parallel , T cells stimulated with low doses of antigen maintained a constant secretion rate for longer periods of time , slowly building their acceleration in cytokine production , while strongly-activated T cells increased their apparent secretion rate sooner ( Figure 7D ) . This time- and antigen-dependent acceleration in the rate of IL-2 secretion ( Figure 7E ) amplified the effect of small differences in the duration and amplitude of T cell signaling ( Supplement to Figure 7 ) . It explained the observed nonlinear kinetics of IL-2 accumulation , and underlied the population size-independence of [IL-2]max and its expanded dynamic range . To understand how these newly found regulations contribute to the emergence of the observed IL-2 scaling laws , we employed biochemically explicit computational modeling of the IL-2 pathway to build an ordinary differential equation model ( see ‘Materials and methods’ ) that captured the experimentally observed progression of molecular states within individual T cells ( Figure 8A ) . Through TCR and subsequent IL-2 signaling , respectively , cells advance from a naïve state to a state of cytokine production , and ultimately to a state of pure cytokine consumption . The amount of antigen per cell regulates IL-2 production through the experimentally characterized inhibitory feed-forward ( Figure 5 ) and stimulatory feedback ( Figure 7 ) loops . The experimentally parameterized theoretical implementation of these regulatory elements is diagrammed in Figure 8B and explained in detail in ‘Materials and methods–Full Model Implementation’ . 10 . 7554/eLife . 01944 . 011Figure 8 . Computational model of IL-2 pathway . ( A ) Cartoon representation of progression through cellular states during production and consumption of IL-2 . Highlighted arrows indicate new regulation uncovered in Figures 5 and 7 . ( B ) Molecular reaction network of IL-2 pathway used to build the mathematical model on the basis of Figure 8A . Detailed description of the model is given in ‘Materials and methods’ . Chemical reactions are represented by solid lines , and dashed lines represent the enzymatic activity of chemical reactions . The parameters associated with the reactions are indicated in the diagram and the values of the parameters are listed in the ‘Materials and methods’ . Experimentally determined/estimated parameters are colored in green . Phenomenologically determined parameters are colored in black . ( C ) Comparison of model-simulated ( top row ) and experimentally observed ( bottom row ) temporal dynamics of IL-2 , IL-2Rα and pSTAT5 for three different numbers of 5C . C7 T cells in 200 μl medium . T cells are co-cultured with 5 . 105 APCs prepulsed with 25 nM of K5 antigen; in the model simulation , the antigen dose is 1 × 108 molecules . Kinetics are representative of six independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 01944 . 01110 . 7554/eLife . 01944 . 012Figure 8—figure supplement 1 . Additional experimental measurements to parametrize the computational model . ( A ) Variation of time to reach [IL-2]max ( τmax ) with number of T cells and antigen dose . Data from six different experiments are plotted together . The grey plane is the best fits to the data , using partial least square regression in MATLAB . The fitting coefficients and standard error of mean for the experimental data are given . We find that the dependency with number of T cells and antigen dose for the time to reach [IL-2]max ( τmax ) is weak . We emphasize here that since the variation in duration for IL-2 production is very small , this cannot be a sufficient mechanism to establish the wide dynamic range of antigen dose scaling and population size-independence of IL-2 accumulation . ( B ) Left: parameterizing the basal rate of IL-2 production per cell . The apparent rate is estimated in molecules per second per cell , as RIL−2 secretion ( T ) =NAvogadroV×[IL−2] ( T ) ∑u=0u=TNIL−2 producers ( u ) Δt , with V being the reaction volume ( V = 2 . 10−4 l ) , [IL-2] the measured concentration of IL-2 ( in Molar ) and Δt the time interval between measurements ( expressed in seconds ) . Distribution of IL-2 secretion rates per cell at 8 hr after the start of co-culture for T cell populations of all sizes ( 105 , 104 , and 103 T cells per well ) stimulated with a range of different antigen quantities ( 1 μM , 100 nM , 10 nM , and 1 nM K5 ) . We estimated the basal rate of IL-2 production to be 7 . 5 molecules per cell per second . Right: parameterizing the rate acceleration for IL-2 production per cell . Maximal acceleration trajectory taken by 103 T cells stimulated with 1 μM K5 . Error bars show standard error of mean of two replicates . Data is representative of four high time resolution experiments . We estimated the maximal boost in IL-2 secretion to be 30-fold over the basal rate of IL-2 secretion , hence 225 molecules per second per cell . ( C ) Parameterizing the upregulation of IL-2Rα , hours after the start of co-culture . Shown: single cell IL-2Rβ distributions for 105 5C . C7 T cells stimulated with 2 . 5 μM K5 antigen at 12 , 24 , 36 , 48 , 78 , and 140 hr . Unstained control is shaded . DOI: http://dx . doi . org/10 . 7554/eLife . 01944 . 012 TCR-mediated inhibition of pSTAT5 signaling is modeled as a reduction in the catalytic ability of the IL2/IL-2R complex ( IL-2Rαβγ•IL-2 ) to induce STAT5 phosphorylation by a factor proportional to the amount of antigen-engaged TCR ( Ag-TCR ) . This mechanism captures the experimental observation that TCR crosstalk modulates the amplitude , but not the EC50 , of IL-2 response ( Figure 4B ) . In modeling the time-dependent acceleration in IL-2 secretion , we followed several lines of evidence that suggested that this feedback depends on antigen signaling . First , this acceleration could be observed despite perturbation of JAK , Phosphoinositide 3-kinase ( PI3K ) and CD28 activity ( our unpublished data ) . Moreover , greater amounts of available antigen and lower numbers of T cells yielded the largest accelerations in IL-2 production ( Figure 7C–D ) ; these conditions are known to increase the length of T cell interactions with antigen presenting cells ( Garcia et al . , 2007 ) . Indeed , recent studies have shown that the duration of antigen priming signals strongly impacts gene expression in T cells ( Tubo et al . , 2013 ) , particularly the upregulation of IL-2 ( Henrickson et al . , 2013 ) . Furthermore , antigen-experienced cells have been shown to exhibit higher rates of IL-2 secretion per cell ( Huang et al . , 2013 ) , possibly through TCR-driven epigenetic modification of the IL-2 locus ( Bruniquel and Schwartz , 2003 ) . We therefore postulated that strength and persistence in TCR signaling determines the extent of acceleration in IL-2 secretion . To model this , we introduced a phenomenological variable , Boost , which upon activation ( Boosta ) increases the rate of IL-2 production per cell . We parameterized Boost’s initial activation by TCR signals to be slow , such that sustained TCR engagement was required to substantially accumulate Boosta . Activated Boost then catalyzes further Boosta , generating a positive feedback that results in the non-linear dynamics of IL-2 secretion . Such phenomenological feedback recapitulates the observed time-dependent acceleration in IL-2 secretion , which is most potent for high quantities of antigen and low numbers of T cells ( Figure 7 ) . Since antigen and secreted IL-2 are shared by the whole T cell population , the number of T cells determines the amount of antigen and cytokine available per cell in the model . Thus , T cell population size regulates the global rate of IL-2 accumulation by setting the number of producers and their antigen availability over time . Additionally , population size controls the global rate of IL-2 depletion by determining the number of consumers , and by dynamically regulating their IL-2 depletion capabilities: the persistent availability of antigen to smaller T cell populations delays pSTAT5-mediated upregulation of IL-2Rα , which postpones the initiation of IL-2 consumption ( Figure 6C–D & 8C ) . While accurately predicting IL-2 consumption will require accounting for cell proliferation and death , which exert stronger effects on longer ( >3 day ) timescales ( Figure 2C ) , our model reproduces the measured dynamics of the IL-2 production pathway for different quantities of antigens and numbers of T cells ( Figures 8C and 9A ) . Most significantly , it recapitulates the scaling law ( Figure 9A bottom ) : ( 3 ) [IL−2]maxmodel∝ ( NT cells ) −0 . 12 ( ±0 . 03 ) × ( NAntigen ) +0 . 82 ( ±0 . 02 ) . 10 . 7554/eLife . 01944 . 013Figure 9 . Mathematical modeling accounts for the scaling law in IL-2 dynamics . ( A ) Simulated dynamics of [IL-2] for different quantities of antigen molecules , NAntigen ( top ) and numbers of T cells , NT cell ( middle ) . [IL-2]max dependency with NAntigen and NT cell ( bottom-filled circles ) can be fitted with Equation 3 ( bottom–grey plane ) . ( B ) Dependence of pSTAT5 response on NAntigen , with ( top ) or without ( middle ) TCR inhibition of pSTAT5 . Abrogating TCR inhibition leads to low saturation in [IL-2]max and spurious scaling with NT cell and NAntigen ( bottom ) . ( C ) Our model recaptures the acceleration of IL-2 secretion as a function of cumulative numbers of IL-2 producing cells ( top ) . Upon removing the boost in IL-2 secretion ( middle ) , low NT cell fail to accumulate comparable [IL-2]max to high NT cell ( bottom ) . ( D ) Model prediction ( top row ) and experimental validation ( bottom row ) of IL-2 accumulation kinetics with JAK inhibitor ( JAKi—solid line ) or without ( DMSO—dashed line ) for different numbers of T cells ( left ) activated with different quantities of K5 peptide ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01944 . 013 We then used our model to examine the relevance of the regulatory mechanisms uncovered in Figures 5 and 7 . In Figure 9B , we tested the importance of the TCR-mediated inhibition of IL-2 signaling . Our complete model accounts for the antigen-dependency of pSTAT5 gain with increasing levels of IL-2Rα and IL-2 ( Figures 4A and 9B , top ) . Removing this inhibitory crosstalk eliminates the antigen dose dependency of pSTAT5 response ( Figure 9B , middle ) . It also results in a decreased antigen-dependency of [IL-2]max that flattens at ∼10 pM ( Figure 9B , bottom ) , the classical threshold for STAT5 phosphorylation ( Wang and Smith , 1987 ) , and thus for termination of IL-2 production ( Long and Adler , 2006 ) . Therefore , TCR inhibition of IL-2 signaling is critical to sustain cytokine secretion beyond 10 pM of IL-2 . We also explored the significance of the nonlinear acceleration in IL-2 production per cell observed in our experiments . The complete model captures the nonlinear correlation of IL-2 with cumulative numbers of IL-2 producing cells ( Figures 7A and 9C , top ) . In contrast , abrogating the acceleration in IL-2 production yields a simple linear correspondence between these variables ( Figure 9C , middle ) , and prevents small populations of T cells from accumulating comparable amounts of IL-2 to those secreted by larger populations . The lack of acceleration also decreases the scaling exponent of [IL-2]max with antigen dose ( Figure 9C , bottom ) , as the antigen-dependence of IL-2 accumulation is limited to differences in the cumulative number of IL-2 producing cells ( Supplement to Figure 7 ) . Therefore , our experimentally determined , quantitative model illustrates the criticality of these new regulatory elements , through which T cells achieve population size-independent power law antigen scaling of IL-2 . We then tested our computational model of IL-2 pathway regulation through in silico and in vitro perturbation of STAT5 signaling . We blocked IL-2 signaling in the model by setting the STAT phosphorylation rate to zero . Our model predicted over ten-fold greater IL-2 accumulation in pSTAT5-inhibited vs unperturbed conditions ( Figure 9D , top ) . Moreover , it forecasted that larger populations of T cells would sustain higher concentrations of IL-2 than smaller populations ( Figure 9D , top left ) . Experimentally treating cells with a JAK inhibitor at time 0 confirmed these predictions , and validated our model’s projections for the dynamics of IL-2 accumulation following JAK blockade ( Figure 9D , bottom ) . These computational and experimental results demonstrate that the empirical scaling of IL-2 accumulation is critically dependent on feedbacks from IL-2 signaling . To further probe the functional significance of our model of IL-2 scaling , we tested numerically and experimentally the joint IL-2 response of two TCR transgenic T cell clones co-cultured at different densities and stimulated with varying concentrations of their respective cognate antigens . The model predicted and experiments confirmed that [IL-2]max for a mixed population of T cell clones is determined by the combined antigen doses , independently of cell numbers ( Figure 10A; Figure 10—figure supplement 1A–C ) . This result demonstrates that IL-2 is a collective measure of global antigenic load with the potential to coordinate polyclonal responses . 10 . 7554/eLife . 01944 . 014Figure 10 . Testing the model of IL-2 regulation through mixed culture of two T cell clones . 5C . C7 and A1 ( M ) TCR transgenic T cells were cultured at varied precursor frequencies ( 5 . 103 and 5 . 104 T cells/well ) with titrated concentrations ( 10−6/10−7/10−8 M ) of cognate antigens ( K5 and HY peptides ) pulsed on separate splenocytes . Graphs show two experiments and are representative of three experiments . ( A ) Model predictions ( top ) and experimental validation ( bottom ) of [IL-2]max for mixed cultures . Color represents total number of T cells ( 5 . 103 + 5 . 103 , 5 . 103 + 5 . 104 , 5 . 104 + 5 . 104 ) . ( B ) Model ( left ) and experimental ( right ) cumulative distributions of the ratio of [IL-2] accumulated by the mixed culture to the sum of [IL-2] accumulated independently by each clone over all conditions . ( C ) Model prediction ( left ) and experimental validation ( right ) that pSTAT5 response to shared IL-2 can resolve the relative activating doses of antigen for 5 . 104 cells of each clone . Marker shape: [Antigen1] = [K5] , marker size: [Antigen2] = [HY] , color: ratio [K5]/[HY] . DOI: http://dx . doi . org/10 . 7554/eLife . 01944 . 01410 . 7554/eLife . 01944 . 015Figure 10—figure supplement 1 . Additional computational predictions and experimental validation for a mixed culture of two T cell clones . ( A ) Model prediction of the temporal dynamics of IL-2 accumulation by a mixture of two different clones of T cells . For given numbers of clone 1 and clone 2 cells in the mixture ( indicated at the top of each box ) , the temporal kinetics of [IL-2] in the 200 µl medium are plotted for combinations of varied doses of antigen 1 ( y-axis ) and antigen 2 ( indicated by the line color ) . Due to the symmetry in the model between clone 1 and clone 2 , we present temporal [IL-2] dynamics with respect to clone 1 only . ( B ) Experiment: temporal dynamics of IL-2 are plotted as in ( A ) for different numbers of T cells from 5C . C7 and A1 ( M ) TCR transgenic mice with varied doses of K5 antigen . ( C ) Experiment: temporal dynamics of IL-2 are plotted as in ( A ) for different numbers of T cells from 5C . C7 and A1 ( M ) TCR transgenic mice with varied doses of HY antigen . ( D ) Scaling for the inhibitory cross-talk between TCR and IL-2 signaling for a mixed culture of 5C . C7 and A1 ( M ) T cells ( see ‘Materials and methods’ for details ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01944 . 015 We compared the IL-2 concentration from mixed cultures to the sum of the IL-2 accumulated independently by each clone by quantifying the ratio of these two values ( Figure 10B ) . For each time point ( colored lines ) , the [IL-2]mix/ ( [IL-2Clone1] + [IL-2Clone2] ) ratios were represented as a cumulative distribution of all 36 culture conditions . Throughout the IL-2 production phase ( t < 50 hr , blue lines ) , the distributions were centered around a ratio of 1 , indicating that IL-2 produced by 2 clones in the same well is approximately equal to the sum of the IL-2 made by the same two populations in separate wells . This result suggests that the IL-2 production phase is dictated by a T cell population’s TCR stimulation , regardless of other nearby immune reactions . At later timepoints ( red lines in Figure 10B ) , the cumulative distribution of ratios of [IL-2] in co-cultures to the sum of the IL-2 from individual clones was drastically shifted to the left , as less IL-2 remained in the co-culture wells due to the increased numbers of IL-2 consumers . This was observed most strikingly when mixing a large number of poorly activated T cells with a small number of strongly activated T cells , as predicted by our model . Furthermore , the model and experiments both demonstrated that each clone’s relative antigen dose could be resolved through the shared cytokine environment via proportional inhibition of the IL-2 pathway ( Figure 10C ) . STAT5 phosphorylation in single cells is determined by the global concentration of IL-2 and cells’ individual IL-2 responsiveness—a convolution of cellular receptor abundance and antigen signaling ( Figure 4A ) . Simple algebra ( ‘Materials and methods–Algebraic relationship between pSTAT5 , IL-2Rα and Antigen for a mixture of two T cell clones ) demonstrates how the cue-signal-response to IL-2 in a mixed population of cells predicts the distribution of stimulating antigens: ( #pSTAT5 ) 2* ( #IL−2Rα ) 1 ( #pSTAT5 ) 1* ( #IL−2Rα ) 2=f ( [Antigen2] ) f ( [Antigen1] ) . Remarkably , this theoretical expression was confirmed experimentally . The relative doses of antigen signals for 5C . C7 and A1 ( M ) T cells , which are of similarly high quality , were detectable through their individual responses to shared IL-2 ( Figure 10C ) . Accordingly , the signaling crosstalk unraveled in Figure 4 allows a mixed population of T cells to perform both a local ( antigen ) and a global ( cytokine ) measurement of collective stimulus through the IL-2 pathway . Furthermore , TCR cross-talk inhibition of the IL-2 pathway provides a graded readout of antigen signaling within activated cells with quantitative resolution across a polyclonal system . We thus sought to apply the potent sensing capacity of this cross-talk toward the quantification of tissue antigenicity . To illustrate how the TCR-IL-2R crosstalk can be used to quantify antigen availability in vivo over a large dynamic range , we probed T cell responses against cells harvested from explanted melanoma tumors and their draining lymph nodes . ( Figure 11A ) . B6 mice were injected in the right flank with 105 B16 melanoma cells , then sacrificed two weeks post-injection . Their tumors and tumor-draining lymph nodes were harvested for co-culture with naïve TRP-1 transgenic T cells , which are specific for the TRP-1 melanoma antigen ( Muranski et al . , 2008 ) . As the tumors varied proportionally in area , weight and cellularity , we asked if the TCR responses of TRP-1 cells would scale with tumor size by re-suspending all tumors in the same volume . In parallel , cells from tumors and tumor-draining lymph nodes were re-suspended at a uniform concentration to probe for differences in antigen presentation per cell . To calibrate this cross-talk assay , spleen and lymph node cells from non tumor-bearing B6 mice were pulsed with titrated amounts of TRP-1 peptide . Throughout the first 50 hr of in vitro activation , we measured the concentration of IL-2 and abundance of IL-2Rα and pSTAT5 for activated TRP-1 T cells in all co-cultures . We then assessed the antigen-induced inhibition of pSTAT5 by calculating the rate of gain in pSTAT5 with increasing IL-2 and IL-2Rα ( SlopeXtalk ) ( Figure 11B , left ) . As in Figure 4A , we found inverse proportionality between SlopeXtalk and titrated concentrations of antigenic peptide , which established a calibration curve to back-calculate fold changes in the antigenic capacity of tumor tissue suspensions ( Figure 11B , right ) . 10 . 7554/eLife . 01944 . 016Figure 11 . Applying antigen-driven inhibition of IL-2 signaling to estimate tumor antigenicity . ( A ) Schematic of experimental design . ( B ) Left: pSTAT5 increase over the first 52 hr of culture as a function of IL-2 and IL-2Rα for titrated TRP1 peptide pulsed on C57BL/6 splenocytes ( Calibration Series , gray ) and tumor samples diluted in equivalent volumes ( Sample Series , green ) . Trajectories were fit with the equation pSTAT5 ( t ) =SlopeXtalk· ( [IL−2] ( t ) ·#IL−2Rα ( t ) ) +Background . Graphs are representative of four experiments . Right: antigen dose as a function of SlopeXtalk , as established by Calibration Series ( black ) . Calibration curve allows estimation of effective antigenicity of tumor samples . Antigenicity of tumors scales with tumor cellularity ( insert ) . ( C ) Back-calculated antigenicity for 105 cells of each mouse’s tumor ( left ) and 106 cells of each mouse’s tumor draining lymph nodes ( right ) . ( D ) Correlation of tumor weight to estimations of antigen presentation per cell for each tissue . DOI: http://dx . doi . org/10 . 7554/eLife . 01944 . 016 We found that the calculated amount of antigen signaling experienced by TRP-1 T cells scaled with tumor cellularity ( Figure 11B , right ) : larger tumors , which yielded more input cells per normalized volume , induced greater suppression of IL-2 signaling , and were thus estimated to have a proportionally greater antigen load ( inset ) . Concurrently , when tumor cells were suspended at the same cellular concentration , the induced SlopeXtalk were very similar , indicating equivalent quantities of presented antigen per cell ( Figure 11C , left ) . Normalizing the antigen calculated for each tissue source by the number of cells plated , we confirmed that both tumor dilution strategies yielded very similar estimates of antigen presentation per cell , which did not depend on tumor size ( Figure 11D , green and red ) . As expected , measurements of SlopeXtalk estimated that tumor-draining lymph node cells presented much less antigen than tumor suspensions ( Figure 11C , right , Figure 11D , blue ) . Strikingly , our measurements showed that the antigen presented per cell in the draining lymph nodes did scale with tumor size ( Figure 11D ) . Our method thus confirms anticipated biological phenomena: tumors with 10 times more cells have 10 times more antigen , there is equal cellular antigenicity between clonal tumors , and there is significantly less antigen per cell in lymph nodes vs tumors . In sum , scalable inhibition of the IL-2 pathway can resolve a wide dynamic range of antigen quantities . This inhibitory cross-talk underlies the [IL-2]max scaling presented in Figure 2 , but can be captured with far fewer timepoints; in fact , a single snapshot measurement of the amplitude of pSTAT5 response to high dose IL-2 can distinguish relative strengths of TCR signaling among activated T cells ( Figure 4 ) . These measurements may help quantify differences in TCR signal strength between conditions resulting in equal binary T cell activation but discrepancies in downstream effector function of activated cells ( van Heijst et al . , 2009 ) , a problem that is particularly acute in the field of tumor immunology ( Joncker et al . , 2006; Engelhardt et al . , 2012 ) .
Our study revealed a scaling law for IL-2 accumulation as a function of antigen dose , and demonstrated that the collective reporting of antigen load through IL-2 production is independent of the number of T cells in mono- and poly-clonal populations . It also quantitatively dissected the regulatory architecture required for this cell–cell communication of antigen input through shared cytokine output . We found that the inhibitory cross-talk between antigen and IL-2 signaling generated a coherent feed-forward loop architecture that ensured direct correspondence between the persistence of antigenic cues and the duration of cytokine production . In parallel , we found a time-dependent acceleration in the rate of IL-2 secretion that allows small T cell populations to compensate for having 20-fold fewer IL-2 producers ( Figure 6B ) in less than twice as much time ( Figure 2B , C ) , and establishes better dynamic range between different doses of antigen . Hence , time integration of IL-2 regulatory loops generates collective-level outputs that reflect global antigen abundance with higher fidelity , functional range and temporal persistence than early activation responses ( Figure 1 ) ( Cheong et al . , 2011 ) . This feedback-controlled , titrated accumulation of IL-2 may help scale and direct the long-term responses of activated cells according their degree of antigenic stimulus . IL-2 signaling critically optimizes lymphocyte differentiation , proliferation , and survival ( Williams et al . , 2006; Bachmann et al . , 2007; Pipkin et al . , 2010; Liao et al . , 2011; McNally et al . , 2011; Boyman and Sprent , 2012 ) . In the case of in vivo CD8+ T cell differentiation , sustained IL-2-driven IL-2Rα expression corresponds to the adoption of short-lived effector , rather than memory , fates ( Kalia et al . , 2010; Obar et al . , 2010; Pipkin et al . , 2010 ) . The tight correlation we find between antigen load and IL-2 accumulation may ensure that large pathogenic challenges are communicated through large IL-2 availability to generate robust effector responses . Indeed , another in vivo study has shown that the expression of IL-2Rα at day four post-infection scales with antigenic affinity , and correlates to greater effector cell expansion and survival ( Zehn et al . , 2009; Kalia et al . , 2010 ) . The higher IL-2Rα levels on OT-I cells exposed to bacteria expressing high affinity OVA vs less potent Q4 peptides suggest that T cell responses to higher affinity epitopes trigger greater accumulation of IL-2 in vivo . Accordingly , we have observed in vitro scaling of [IL-2]max and pSTAT5 inhibition with antigen quality across this altered peptide ligands series ( KT , GV , GA-B , unpublished data ) . Moreover , recent quantitative studies relating CD4+ ( Tubo et al . , 2013 ) and CD8+ ( Henrickson et al . , 2013 ) T cell differentiation to the abundance and persistence of TCR priming signals in vivo have implicated the antigen-scaled production and sensing of IL-2 as possible mediators of divergent cell fate decisions . Finally , graded inhibition of pSTAT5 by TCR signaling may directly influence CD4+ T helper subtype differentiation by blocking time-sensitive cytokine signals ( Yamane and Paul , 2012 ) . This study also introduces the possibility of detecting analog tuning within digitally activated populations by examining cross-talk pathways . Positive and negative feedback loops within proximal TCR signaling generate a sharp digital activation filter that is necessary for efficient discrimination of antigenic ligands ( Altan-Bonnet and Germain , 2005; Das et al . , 2009 ) . As a result , many of the readouts downstream of TCR activation ( i . e . , CD69 , CD25 , ppERK–see Figure 1A ) are all-or-none , and gradations in signal strength are resolved only by counting the number of activated T cells . Such ‘percent activated’ measurements have been fruitfully used to survey tissue antigenicity ( Badovinac and Harty , 2002; Nandi et al . , 2009 ) , but binary single parameter readouts carry some limitations: they are prone to saturation of input detection range , exhibit narrow output dynamic ranges , are subject to shifts in value by cellular death and migration , and provide little information on the functional capacity of activated cells . In contrast , quantifying the degree of TCR-driven inhibition of pSTAT5 signaling inside activated T cells provides an indication of the strength of signaling within the responding population . While these measurements are not immune to saturation , convolution of antigen-dependent STAT5 phosphorylation efficiency with its consequent scaling of IL-2 production ( Figures 4A , 10C , 11 ) increases output resolution of the initiating antigen load . Thus , the regulatory architecture uncovered in this study , which expands the scaling of cytokine production to antigen input , can be similarly exploited to make more informative measurements of antigen signaling ( Figure 11 ) . Finally , our study demonstrated how reductionist , systems biology approaches can quantify the shape , strength and kinetics of immunological regulation when the molecular mechanisms are unknown . By observing shifts in multi-dimensional trajectories of the IL-2 pathway as a function of input numbers of antigens and T cells , we deduced the presence of previously unaccounted for regulatory interactions . Our simplified experimental system allows detailed probing of endogenous pathways with high-resolution time- and dose response- series that strongly constrain our mathematical model , and our quantitative understanding , to the observed biology . Through this bottom-up approach ( O’Garra et al . , 2011 ) , one can dissect the dynamic regulation and scaling laws of cytokine communications that underlie complex in vivo settings .
Experiments used lymph node and spleen cells from 5C . C7 Rag2−/− , A1 ( M ) Rag1−/− or TRP-1 Rag1−/− TCR–transgenic mice and B10 . A Cd3e−/− or C57BL/6 mice aged 2–6 months , cultured in complemented RPMI . B10 . A Cd3e−/− or C57BL/6 splenocytes pulsed overnight with varying concentrations of K5 , HY or TRP-1 peptide were co–cultured with varying numbers of T cells in 200 µl of media in flat–bottomed 96–well plates . The Institutional Animal Care and Use Committee of Memorial Sloan Kettering Cancer Center approved all of the animal experiments . At each timepoint , supernatants were collected and stored at −20°C . Cells were collected and fixed in 1 . 6% cold PFA on ice , permeabilized with 90% MeOH , and stored at −20°C . T cell numbers were quantitatively assessed by flow cytometry: diluted , CTV-labeled samples from each well were stained without any spin steps and run on a BD LSRII for 60 s at a calibrated flow rate , allowing back-calculation of the absolute number of CD45 . 1 5C . C7 T cells . IL-2 producer counts NIL−2+ were obtained by multiplying the above 5C . C7 T cell counts by the frequency of IL-2 producing cells in the same well , measured via the IL–2 Secretion Assay Kit ( Miltenyi Biotech ) . After completion of the time series , levels of ppERK ( Altan-Bonnet and Germain , 2005 ) , pSTAT5 and IL-2 receptor components were measured by FACS , and supernatant [IL-2] was measured by ELISA ( see Feinerman et al . 2010 for details of the method ) . The integral of the number of IL-2 producers over time was calculated as ∑i=0iTNIL−2+ ( i ) * ( ti+1−ti ) , where iT is the index for the last measurement at t = T . Antigen-blocking experiments were performed with anti–MHC class II molecule I–Ek ( clone 14–4–4S ) or anti–MHC class I molecule H2–Db ( clone 20–8–4S ) as a control antibody , at a final concentration of 20 µg/ml . Inhibition of pSTAT5 signaling was performed using Jak inhibitor AZD1480 at a final concentration of 10 µM . Total amount of secreted IL-2 was calculated as the integral of [IL-2] ( t ) over time t , using the method of trapezoidal approximation , normalized over the duration of the experiment and reported in the molar unit M . Tfinal ranges from 80 to 150 hr depending on the experimental conditions . We found a linear correlation between the total amount of secreted IL-2 with [IL-2]max ( χ2=4 . 3 for 118 points and 1 parameter ) :[IL−2]max=0 . 34* ( 1Tfinal∫0Tfinaldt[IL−2] ( t ) ) Cells were stripped of pre-bound cytokine by 2 min incubation in ice–cold pH 4 . 0 0 . 1 M glycine buffer , then washed 2X in RPMI and rested for 10 min at 37°C ( Feinerman et al . , 2010 ) . Cells were added to cytokine titrations and incubated for 10 min at 37°C before fixation , permeabilization and FACS analysis . Single , CD4+ IL–2Rα+ cells were then identified using FlowJo and further analyzed using our custom–designed processing R program ScatterSlice ( Cotari et al . , 2013 ) ( this software can be downloaded at www . Scatterslice . org ) . Each sample’s CD4+ IL–2Rα+ cells were divided into bins according to varied levels of cytokine receptors . Within each bin , a three parameter Hill equation fit the pSTAT5 base , amplitude and EC50 for the IL-2 dose titration . For each stimulation condition , fitted amplitudes for different levels of receptors were then normalized relative to the pSTAT5 amplitude in the corresponding bin within the T cell population stimulated with the highest dose of antigen ( 10 μM K5 ) . For each antigen dose , the mean and standard error of the mean of normalized pSTAT5 amplitude was calculated across all occupied bins of IL–2Rα , β , γ expression . [IL–2]max was fit across a range of doses of antigen ( [Antigen] ) and numbers of T cells ( NT cell ) by partial least squares regression ( PLSR ) for each independent experiment to obtain exponents for [Antigen] and NT cell . The average and standard error of mean of the exponents was calculated from nexperiment = 6 individual fits . Error bars correspond to the 95% confidence interval . We constructed a chemical reaction network ( Figure 8B ) to model IL-2 production and signaling in a population of T cells based on measurements from previous literature and our experiments . Our model comprises 2 global variables: the numbers of free antigen ( Ag ) and IL-2 . These molecules are shared by all cells in the medium . For each cell , there are five independent state variables ( where • represents a complex ) representing the number of molecules of: Ag , IL-2Rα , IL-2Rβγ , IL-2Rαβγ•IL-2 ( where IL-2Rαβγ is the full IL-2 receptor , i . e . , IL-2Rα•IL-2Rαβγ and Activated Boost , and 3 dependent variables ( calculated from the independent variables ) : Ag•TCR , IL-2Rαβγ and Fraction pSTAT5 . The total numbers of TCR and Boost remain constant over time . We observed that the fraction of IL-2 producing cells increases linearly with time: this was modeled phenomenologically by introducing a uniformly distributed random delay ( between 10 and 60 hr ) in the first encounter between T cell and antigen . We implemented a bottom-up approach to simulate different numbers of T cells ( NT cell ) . We simulated ncell number of cells ( typically ncell = 20 ) , and scaled the association/dissociation of Ag to TCR , production of IL-2 , and the association/dissociation of IL-2 to the IL-2Rα by a factor NT cell/ncell to capture the dynamics of NT cell number of T cells in V = 200 µl culture medium . We generated a set of nonlinear ordinary differential equations ( see below ) describing the dynamics of the variables for ncell number of cells . We solved these sets of stiff nonlinear ordinary differential equations using the MATLAB CVODE solver ( Hindmarsh et al . , 2005 ) . See below for details of model . We constructed a chemical reaction network to model IL-2 production and signaling in a population of T cells . Our experimental system tracked how large populations of T cells processed macroscopic numbers of molecules of antigen , IL-2 , IL-2 receptors and signaling molecules . Moreover , the topology of our network does not include strong positive feedbacks ( Vilar et al . , 2002; Artomov et al . , 2010; Francois et al . , 2013 ) , where stochasticity in chemical reactions yielded qualitatively different output . Hence , we relied on a deterministic framework to build an ordinary differential equation model . This model is based on previous literature describing IL-2 signaling and contains the previously determined components ( Feinerman et al . , 2010; Cotari et al . , 2013 ) :T cells constitutively express IL-2Rβ and γ ( CD122 and CD132 ) . Engagement of TCR by antigen leads to activation of T cells . Activated T cells upregulate IL-2Rβ , express the surface receptor IL-2Rα ( CD25 ) and secrete the cytokine IL-2 . The IL-2 receptor chains pre-form a heterotrimeric complex , IL2Rαβγ . Secreted IL-2 accumulates in the extracellular medium and binds to T cells’ IL-2 receptor components to assemble a full tetrameric complex , IL-2/IL2Rαβγ . The full complex phosphorylates STAT5 and is endocytosed and degraded ( allowing IL-2 consumption ) . Phosphorylated STAT5 ( pSTAT5 ) enacts both a negative and positive feedback on its own signaling by shutting down IL-2 production and upregulating production of IL-2Rα , respectively . This classical model of IL-2 cue-signal-response ( Figure 3 ) is insufficient to generate the experimentally observed scaling law ( Equation 1; Figure 2H ) . We appended two new regulatory elements as experimentally characterized in Figures 5 and 7:Our quantitative measurements of the IL-2 signaling pathway demonstrate that TCR signaling inhibits the phosphorylation of STAT5 . This TCR-mediated inhibition of IL-2 signaling occurs in an antigen dose-dependent manner . Single cell measurements of the number of IL-2-producing cells indicate that the rate of IL-2 production per cell accelerates nonlinearly with time . The full reaction network of the model is given in Figure 8B and the nonlinear ordinary differential equations of the model are given in ‘Equations’ . We obtained quantitative parameters from the literature , or parameterized the associated chemical reactions based on our own measurements as detailed below ( ‘Parameters’ ) . As our previous study validated a well-mixed approximation to model our experimental conditions ( Feinerman et al . , 2010 ) , we did not include spatial considerations for the cell-to-cell communications via extracellular IL-2 in these in vitro settings . Our complete model is reductionist by nature and fully parameterized: it enables us to account for our experimental observations ( Figure 9 ) , and to make predictions that we validated experimentally ( Figures 9 and 10 ) . Engagement of antigen to TCR is modeled by a simple chemical equilibrium upon interaction between T cell and antigen-presenting cells . We used the equilibrium constant from recent in situ measurements for I-Ek/MCC antigen and 5C . C7 TCR ( Huang et al . , 2013 ) . Interactions of I-Ek/K5 antigen with 5C . C7 TCR was measured in vitro ( Corse et al . , 2010 ) , and found to be similar to I-Ek interactions with MCC antigen . Hence we tookKTCR·IEk/K5equilibrium=30 , 000 . As previously demonstrated ( Kedl et al . , 2002; Schulz et al . , 2009 ) , the loss of antigen signaling is a critical parameter of the long-term response of T cells . We implemented an exponential temporal loss of antigen from the surfaces of antigen-presenting cells based on engagement with TCR . The typical timescale for such process has been measured in von Essen et al . ( 2004 ) :kantigenconsumption=ln ( 2 ) 3 . 5h−1 . IL-2 receptor chains pre-form heterotrimeric complexes , even in the absence of IL-2 ( Cotari et al . , 2013 ) . As in our previous study ( Feinerman et al . , 2010 ) , we treated IL-2Rβ and IL-2Rγ as a single component IL-2Rβ/γ , which then binds to the IL-2Rα subunit . This association at the membrane is very fast relative to the timescale of protein synthesis; we therefore modeled the pre-association of the IL-2 receptor chains as a steady state process for each time step ( Cotari et al . , 2013 ) . The number of IL-2Rαβγ complexes at each time step is thus calculated as the root of a quadratic equilibrium function of the abundances of IL-2Rα and β/γ and the affinity constant for their binding ( Cotari et al . , 2013 ) . IL-2 then binds to the heterotrimeric IL-2Rαβγ forming the full complex IL-2Rαβγ•IL-2 with a fast association rate ( Wang and Smith , 1987; Smith , 1988; Rickert et al . , 2004 ) . The unoccupied IL-2Rα and IL-2Rβ/γ degrade at very slow rates , with half-lives of 7 hr and 39 hr , respectively ( Duprez et al . , 1988 ) . In contrast , the full complex IL-2Rαβγ•IL-2 is rapidly internalized by T cells , with a half-life of 15 min ( Hemar et al . , 1995 ) . We modeled the phosphorylation of STAT5 as a sigmoid dose response function of the full complex , IL-2Rαβγ•IL-2 . As IL-2Rαβγ•IL-2 formation is limited by the abundance of IL-2Rβ/γ , the relationship between IL-2Rαβγ•IL-2 and STAT5 phosphorylation does not saturate and thus remains linear , as observed in Figures 4A , 10C and 11B . TCR-driven inhibition of IL-2 signaling ( Figure 4 ) was modeled by allowing Ag-TCR to inhibit the catalytic activity of the full complex , IL-2Rαβγ•IL-2 . The strength of the inhibition was parameterized to reproduce ( Figure 9B ) our experimental observations given in Figure 4 . The first test of our model consisted of modeling the effect of Janus kinase ( JAK ) inhibition on the accumulation of IL-2 . This model prediction was made by setting the levels of STAT5 phosphorylation to 0 ( hence abrogating the positive feedback on IL-2Rα and the negative feedback on IL-2 secretion ) . The results of these simulations were compared to the experimental validation in Figure 9D . We then tested our model with a simulation of a mixture of two distinct T cell clones co-cultured at different population densities , and activated by varying doses of their respective cognate antigens . As in the single clone model , we used the ‘bottom up’ approach ( ‘Mathematical model’ ) to model of a mixture of two clones . For each clone we simulate ncell = 20 . We varied ( NTcellClone1/ncell1 ) and ( NTcellClone2/ncell2 ) to simulate various numbers of T cells for clone 1 and clone 2 , respectively . There are three global states: Ag1 ( antigen activating clone 1 ) , Ag2 ( antigen activating clone 2 ) , and [IL-2] , which is shared by all cells from both clones . For each cell , there are five independent state variables representing the number of molecules of: Ag , IL-2Rα , IL-2Rβγ , IL-2Rαβγ•IL-2 , and Activated Boost , and three dependent variables ( calculated from the independent variables ) : Ag•TCR , IL-2Rαβγ and Fraction pSTAT5 . Comparison of the two clone model predictions and their experimental validation are presented in Figure 10 . We modeled Ncells undergoing activation in a volume V . Our previous work ( Feinerman et al . , 2010 ) validated the well-mixed approximation to model IL-2 communications over long timescales . To limit integration times , we modeled ncell individual cells within a volume V·ncells/Ncells . To reproduce the linear increase of the number of activated T cell over time , we assumed that each individual cell ( labeled i ) within this ncell cohort gets activated at a time ti ( typically , ncells=50 ) . These ti represent the time when the ith T cell encounters an antigen-presenting cell . Hence , at a given time t , the number of activated cells nactivated ( t ) is:nactivated ( t ) =∑i=1ncellH ( t , ti ) , where {ti}i=1…ncell are the times of activation of the ncell cells being simulated , and H ( x , y ) is the Heaviside ( thresholding ) function , defined as:H ( x , y ) ={0 :for x<y1 :for x≥y Three variables are described with a steady-state approximation . They are the number of engaged TCR on cell i ( nTCR•pMHCi ) , the number of preformed receptors for IL-2 on cell i ( nIL2Ri ) and the amount of STAT5 phosphorylation within cell i ( PpSTAT5i ) associated with engagement of IL2Ri with IL-2 . In all equations , nX represents the absolute number of molecule X . We compute the amount of complexes between pMHC and TCR , or between the α and β chains of the IL-2 receptor . For a thermodynamic equilibrium between X and Y , X+Y⇌X•Y , the amount of complex C for X•Y is a function nX0 , nY0 and the equilibrium constant Kequilibrium:C ( nX0 , nY0 , Kequilibrium ) =12 ( nX0+nY0+Kequilibrium− ( nX0+nY0+Kequilibrium ) 2−4·nX0·nY0 ) . nTCR•pMHCi ( t ) ={0:for t<tiC ( nTCR0 , npMHC ( t ) /nactivated ( t ) , KTCR•pMHC ) :for t≥tinIL2Ri ( t ) =C ( nIL2Rαi ( t ) , nIL2Rβγi ( t ) , KIL2Rα•IL2Rβ ) PpSTAT5i ( t ) =nIL2•IL2R ( t ) KIL2>pSTAT5+nIL2•IL2R ( t ) ×11+σTCR↘IL2R·nTCR•pMHCi ( t ) The ordinary differential equations for the other variables are:dnpMHC ( t ) dt=−kantigenconsumption ( ∑i=1ncellnTCR•pMHCi ( t ) ) dnIL2Rαi ( t ) dt=kreceptordegradation·[nTCR>IL2Rα ( max ) ·H ( nTCR•pMHCi ( t ) , τreceptorproduction ) +npSTAT5>IL2Rα ( max ) ·PpSTAT5i ( t ) −nIL2Rα ( t ) ]dnIL2Rβi ( t ) dt=kreceptordegradation·[nIL2Rβ ( 0 ) + ( nIL2Rβ ( max ) −nIL2Rβ ( 0 ) ) ·H ( nTCR•pMHCi ( t ) , τreceptorproduction ) ]−kreceptordegradation·nIL2Rβi ( t ) dnIL2•IL2Ri ( t ) dt=kIL2•IL2Rassociation·NcellsncellsVNa[nIL2Ri ( t ) −nIL2•IL2Ri ( t ) ]·nIL2 ( t ) −kIL2•IL2Rendocytosis·nIL2•IL2Ri ( t ) dnBoosti ( t ) dt=[kTCR>Boostproduction·nTCR•pMHCi ( t ) +kBoost>Boostproduction·nBoosti ( t ) ]· ( nBoosttotal−nBoosti ( t ) ) dnIL2 ( t ) dt=Ncellsncells∑i=1ncell[−kIL2>IL2Rassociation·1VNa·nIL2 ( t ) ·[nIL2Ri ( t ) −nIL2•IL2Ri ( t ) ]+ ( kTCR>IL2production+kBoost>IL2production·nBoosti ( t ) nBoosttotal ) ×H ( nTCR•pMHCi ( t ) 1+σpSTAT5↘IL2·PpSTAT5i ( t ) , τIL2production ) ] The initial conditions are:npMHC ( t=0 ) =npMHCtotal·ncellsNcellsnTCR ( t=0 ) =3 . 104nIL2 ( t=0 ) =0nIL2Rαi ( t=0 ) =0nIL2Rβi ( t=0 ) =103nIL2•IL2Ri ( t=0 ) =0nBoosti ( t=0 ) =0 ParameterNotationValueReferenceEquilibrium constant for pMHC-TCR complex formationKTCR•pMHC30 , 000Huppa et al . , 2003Equilibrium constant for IL-2R pre-assemblyKIL2Rα•IL2Rβγ2 , 700Figure S16 in Cotari et al . , 2013Efficiency of TCR inhibition on STAT5 phosphorylationσTCR↘IL2R0 . 01Adjusted to fit Figure 4AEC50 of conversion of full IL-2• IL-2R into pSTAT5KIL2>pSTAT5104Adjusted to fit Figure 4ARate of antigen consumptionkantigenconsumptionln ( 2 ) /3 . 5 h−1von Essen et al . , 2004Threshold number of pMHC–engaged TCR to start activationτpMHC−TCRactivation1Huang et al . , 2013TCR–dependent IL2Rα expression plateaunTCR>IL2Rαmax1000Cotari et al . , 2013Rate of internalization for IL2Rαkreceptordegradationln ( 2 ) /5 h−1Duprez et al . , 1988pSTAT5–dependent plateau for IL2Rα expressionnpSTAT5>IL2Rαmax2 . 106Figure 1D in Cotari et al . , 2013Abundance of IL–2Rβ without TCR activationnIL2Rβ ( 0 ) 1000Figure S8 in Cotari et al . , 2013Maximal abundance of IL2Rβ upon TCR activationnIL2Rβmax10 , 000Figure 8—figure supplement 1CRate of IL-2 binding to full IL-2RkIL2•IL2Rassociation1×1011 h−1Table I in Wang and Smith , 1987Rate of internalization of full complex ( IL-2• IL-2R ) kIL−2•IL−2Rendocytosisln ( 2 ) /0 . 25 h−1Duprez et al . , 1988TCR-dependent secretion rate of IL-2kTCR>IL2production7 . 5×3600 h−1Figure 8—figure supplement 1BpSTAT5–dependent inhibition of IL2 productionσpSTAT5↘IL23 . 105Fit in this studyTotal number of Boost moleculesnBoosttotal105Fit in this studyTCR–dependent activation of IL-2 BoostkTCR>Boostactivation10−3Fit in this studyPositive feedback on the activation of BoostkBoost>Boostactivation3 . 10−1Fit in this studyBoost-dependent secretion rate of IL-2kBoost>IL2production30×kTCR>IL2productionFigure 8—figure supplement 1B As experimentally shown in Figure 10C , the model predicts that the ratio pSTAT52* ( IL−2Rα ) 1pSTAT51* ( IL−2Rα ) 2 is a function of the ratio [Antigen2][Antigen1] . We anticipated that the TCR-driven inhibition of IL-2 signaling that we characterized experimentally ( Figure 4 ) and theoretically ( Figure 9B ) would yield interesting dynamics for the IL-2 pathway in a mixture of T cell clones cultured together . Our quantitative understanding from Figure 4A lead to the following equation for each T cell clone i in the culture#pSTAT5i=[IL−2]* ( #IL−2Rα ) i*f ( [Antigeni] ) , where f is a function and #X represents the number of X . As the concentration of IL-2 is a shared variable for the two co-cultured T cell clones , we can eliminate it by simple algebra:#pSTAT52* ( #IL−2Rα ) 1#pSTAT51* ( #IL−2Rα ) 2=f ( [Antigen2] ) f ( [Antigen1] ) . This lead us to plot in Figure 10C: ( #pSTAT52* ( #IL−2Rα ) 1 ) =F ( [Antigen1] , [Antigen2] ) * ( #pSTAT51* ( #IL−2Rα ) 2 ) , where the correlation coefficient F is defined as:F ( [Antigen1] , [Antigen2] ) =f ( [Antigen2] ) f ( [Antigen1] ) . We fit the antigen dependency for the median #pSTAT55C·C7* ( #IL−2Rα ) A1 ( M ) #pSTAT5A1 ( M ) * ( #IL−2Rα ) 5C·C7 of different time points ( Figure 10—figure supplement 1C ) . This formula fits the experimental data in Figure 10C , with the additional quantification that this prefactor F scales with the quantities of Antigen1 and Antigen2:F ( [Antigen1] , [Antigen2] ) ∝[Antigen1]+0 . 48±0 . 18*[Antigen2]−0 . 49±0 . 18 , where the error bars correspond to the 95% confidence interval in our nonlinear regression parameters . Thus , the study of the mixture of two clones uncovered an additional scaling for the inhibitory cross-talk between TCR and IL-2 signaling in 5C . C7 and A1 ( M ) T cells:#pSTAT52* ( #IL−2Rα ) 1#pSTAT51* ( #IL−2Rα ) 2∝ ( [Antigen2][Antigen1] ) −0 . 49±0 . 18 The quality of the fit is χ2=0 . 8 for N = 9 conditions and 3 parameters . C57BL/6 mice were injected in the right flank with 105 B16 melanoma cells . Two weeks post-injection , samples of tumors and tumor-draining lymph nodes were harvested , prepared as single-cell suspension and irradiated at 3000 rad , then co-cultured with 40 , 000 primary CD4+ TRP-1 transgenic T cells . IL-2 accumulation , IL-2Rα expression and STAT5 phosphorylation among activated TRP-1 cells were measured at intervals of approximately 8 hr for 52 hr . In parallel , the same experiments were performed with lymph node and spleen cells of unchallenged C57BL/6 mice pulsed with titrated amounts of TRP-1 peptide ( Calibration series ) . We fit the IL-2 response as:pSTAT5 ( t ) =SlopeXtalk· ( [IL−2] ( t ) ·#IL−2Rα ( t ) ) +Background , in order to estimate the extent of antigen-driven inhibition of IL-2 signaling , which manifests as a decrease in SlopeXtalk . The antigen-dependency of SlopeXtalk is derived from the data collected in the Calibration Series . This calibration is then applied to back-calculate the antigenicity of the sampled tumors and lymphocytes . | The cells of the immune system face the challenge of removing viruses and other pathogens without endangering healthy tissues . Cells called T cells plays a variety of roles in the immune response: some T cells directly destroy infected cells , some recruit other cells called phagocytes to the site of infection , and some release small proteins called cytokines . These cytokines help cells to communicate with other cells and , therefore , to tailor the overall immune responses to deal with a particular pathogen . It is known that mammals are capable of adjusting the T cell response to match the overall severity of an infection . However , it is not clear how individual T cells coordinate their seemingly binary response—they are either activated when they recognize a pathogen , or they are not activated—into a response at the collective cell level that can be varied continuously over a wide range of values . Here , Tkach et al . show that T cell populations match their production of the cytokine interleukin 2 ( IL-2 ) to the abundance of antigens—molecules released by the pathogen—over an unexpectedly large range of concentrations . Through a combination of experimental and computational analyses , Tkach et al . identified two novel IL-2 feedback loops that help to generate the correct quantity of cytokine , irrespective of the total number of T cells . Furthermore , this model can be used to estimate antigen quantities within diseased tissues . The work of Tkach et al . illustrates the potential of feedback integration in cell signalling and gene regulation as a mechanism to allow cellular populations to respond to environmental stimuli in a graded , collective fashion . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"immunology",
"and",
"inflammation"
] | 2014 | T cells translate individual, quantal activation into collective, analog cytokine responses via time-integrated feedbacks |
Mammalian genomes typically contain hundreds of thousands of endogenous retroviruses ( ERVs ) , derived from ancient retroviral infections . Using this molecular 'fossil' record , we reconstructed the natural history of a specific retrovirus lineage ( ERV-Fc ) that disseminated widely between ~33 and ~15 million years ago , corresponding to the Oligocene and early Miocene epochs . Intercontinental viral spread , numerous instances of interspecies transmission and emergence in hosts representing at least 11 mammalian orders , and a significant role for recombination in diversification of this viral lineage were also revealed . By reconstructing the canonical retroviral genes , we identified patterns of adaptation consistent with selection to maintain essential viral protein functions . Our results demonstrate the unique potential of the ERV fossil record for studying the processes of viral spread and emergence as they play out across macro-evolutionary timescales , such that looking back in time may prove insightful for predicting the long-term consequences of newly emerging viral infections .
Retroviruses ( family Retroviridae ) are abundant in nature and include human immunodeficiency viruses ( HIV-1 and HIV-2 ) , human T-cell leukemia viruses ( HTLV-1 and -2 ) , and the well-studied oncogenic retroviruses of mice and other model organisms , among many others ( Goff , 2007 ) . The hallmark of all retroviruses is reverse transcription of the viral RNA genome to form a DNA provirus , which is inserted at random into host chromosomal DNA . If integration of this viral DNA occurs in the germ line , the resulting insertion is called an endogenous retrovirus ( ERV ) . The inserted sequence ( the ERV ) is replicated as part of the host genome during cell division and can be inherited vertically in a Mendelian fashion . Each ERV integrant is subject to drift and selection and may be lost or , given enough time , become fixed in the population . Over many millions of years , and through repeated rounds of endogenization and copy number expansion , metazoan genomes have become riddled with the remnants of past retroviral infections; in most organisms ( including humans ) , ERVs amount to hundreds of thousands of copies per genome ( Lindblad-Toh et al . , 2005; Lander et al . , 2001; Waterston et al . , 2002 ) . Following endogenization , ERV sequences switch from evolving at the very rapid rate associated with exogenous retroviral replication to a rate approximating the neutral evolutionary rate of the host genome . Thus , ERV sequences embedded in animal genomes serve as long-lasting molecular fossils related to exogenous retroviruses and their ancient , extinct relatives . The group of related ERV elements collectively referred to as ERV-Fc are distantly related to extant gammaretroviruses ( Jern et al . , 2005 ) and form a monophyletic clade with the human ERVs , HERV-H and HERV-W . All ERV-Fc elements possess a simple genome consisting of the canonical gag , pro , pol , and env genes common to all retroviruses but lack additional regulatory or accessory genes associated with complex retroviruses ( Figure 1 ) . The designation ERV-Fc is based on the practice of naming ERV groups after the tRNA complementarity of the viral primer binding sequence ( PBS ) ; in the case of ERV-Fc , the PBS is complementary to a phenylalanine tRNA ( GAA anticodon ) . This viral lineage was first identified and characterized from the genomes of several primate species , including humans , chimpanzees , gorillas , baboons , and multiple New World monkeys ( Bénit et al . , 2003 ) . Estimates of insertion timing suggested independent endogenization in the different primate lineages studied rather than cospeciation after colonization of a common ancestor , and the authors hypothesized that ERV-Fc first infected the common ancestor of all simians and remained actively infectious/mobile for tens of millions of years ( Bénit et al . , 2003 ) . A more recent study described abundant representation of ERV-Fc sequences in the canine genome , and the authors suggested that an ancient cross-species transmission between carnivores and primates could account for the presence of ERV-Fc sequences in the two lineages ( Barrio et al . , 2011 ) . 10 . 7554/eLife . 12704 . 003Figure 1 . Schematic representation of the major features of ERV-Fc proviruses . The region colored in blue indicates gag , brown indicates pol , and yellow represents env coding regions . The gray-colored regions indicate the two long terminal repeat ( LTR ) regions . Vertical lines within these regions indicate where proteolytic cleavage would occur between protein subunits . The identity of these subunits is indicated below the schematic: MA = matrix , CA = capsid , NC = nucleocapsid , PR = protease , RT = reverse transcriptase , IN = integrase , SU = surface , TM = transmembrane , PPT = polypurine tract . The probable location of the viral RNA packaging motif is indicated by ψ . At the termini of the retroviral LTR sequences is shown the canonical TG/CA dinucleotides as well as the 5 nucleotide target site duplications ( TSDs ) flanking the provirus . ERV , endogenous retrovirus . DOI: http://dx . doi . org/10 . 7554/eLife . 12704 . 003 Our goal in the present study was to reconstruct the natural history of a specific exogenous retrovirus lineage , which gave rise to the ERV-Fc group of ERV loci . Because the various mechanisms that influence post-endogenization sequence evolution and copy-number expansion in organismal genomes can erase or alter ERVs in ways that do not accurately reflect the exogenously replicating progenitor virus , we first sought to minimize the effects of post-endogenization evolution . To do this , we first performed an exhaustive search of mammalian genome sequence databases for ERV-Fc loci and then compared the recovered sequences . Next , for each mammalian genome with sufficient ERV-Fc sequence , we reconstructed Gag , Pol , and Env weighted consensus protein sequences representing the exogenous virus that colonized that particular species’ ancestors . Finally , we used these consensus sequences to infer the natural history and evolutionary relationships of the exogenous , ERV-Fc related viruses . In so doing , we uncovered a complex evolutionary history , including a prolonged , ancient global spread of the virus involving multiple instances of cross-species transmission and endogenization , and revealed that recombination played a significant part in the evolution and spread of the ERV-Fc lineage .
Using BLASTn and previously reported ERV-Fc sequences as initial queries , we screened the non-redundant ( nr ) database and 50 mammalian genome sequence databases ranging in completeness from the nearly complete human and mouse genomes to low-coverage genomic scaffolds and unscaffolded trace and contig archives ( Figure 2 and Figure 2—source data 1 ) ( Bénit et al . , 2003 ) . Preliminary amino acid phylogenies of translated consensus sequences generated from the initial BLAST hits were used to confirm or exclude ERV-Fc evolutionary relationships . To extract maximal ERV-Fc sequence information from the genomic databases , an iterative BLAST approach was then undertaken using preliminary hits as query sequences . This approach resulted in the identification of ERV-Fc coding sequences in 28 species , representing every superorder of eutherian mammals except Xenarthra ( Figure 2 ) . No evidence was found for ERV-Fc being present in metatherian mammals . In several cases , a genome possessed evidence of ERV-Fc endogenization , but lacked sufficient sequence information for definitive phylogenetic analysis of Gag/CA , Pol/RT , or Env/TM ( Figure 2—source data 2 ) . These included the genomes of the Chinese hamster and European shrew that harbor gag sequence fragments that branch with ERV-Fc , but are too fragmented to reconstruct complete CA ancestral coding sequences . At the time of sampling , the Chinese hamster and European shrew genomes lacked ERV-Fc pol or env sequences . Similarly , we found that the orangutan genome harbors a single ERV-Fc-associated solo long terminal repeat ( LTR ) element ( Figure 2 and Figure 2—source data 2 ) . 10 . 7554/eLife . 12704 . 004Figure 2 . The genomes of most Eutherian mammals harbor ERV-Fc . A mammalian phylogeny ( adapted from [ ( Bininda-Emonds et al . , 2007] ) including species whose genomes were examined for the presence of ERV-Fc . Species lacking ERV-Fc are depicted in red , while those found to harbor ERV-Fc are depicted in green . Bold font indicates that coding potential in one or more gene regions could be reconstructed , italics indicates that ERV-Fc fragments were identified but coding potential could not be reconstructed; * indicates that only a solo LTR was identified; and †† indicates that a species harbors two genetically distinct ERV-Fc lineages . Background shading indicates higher-order taxonomic relationships: blue = Euarchontoglires , pink = Laurasiatheria , green = Xenarthra , purple = Afrotheria , brown = Metatheria . Envelope icons indicate species in which ERV-Fc env open reading frame ( s ) were identified , and the icons colored green indicate env with homology to HERV-Fc; yellow icons indicate the env had greater similarity to HERV-W . ERV , endogenous retrovirus . HERV , human ERV . DOI: http://dx . doi . org/10 . 7554/eLife . 12704 . 00410 . 7554/eLife . 12704 . 005Figure 2—source data 1 . Genome sequence database summary . List of the genome database builds , fold coverage , and method of sequence acquisition for all species included in this study . DOI: http://dx . doi . org/10 . 7554/eLife . 12704 . 00510 . 7554/eLife . 12704 . 006Figure 2—source data 2 . Overview of recovered ERV-Fc sequences . Summary of ERV-Fc lineages identified , our designations for these lineages , which viral sequences were recovered , and correlation to their RepBase designation ( if applicable ) . Also included is basic taxonomic information about the hosts from which these viral sequences were retrieved . ERV , endogenous retrovirus . DOI: http://dx . doi . org/10 . 7554/eLife . 12704 . 00610 . 7554/eLife . 12704 . 007Figure 2—source data 3 . Sequences of ERV-Fc primer binding sites . ERV , endogenous retrovirusDOI: http://dx . doi . org/10 . 7554/eLife . 12704 . 007 While ERV-Fc was present in the majority of mammalian species examined , its absence from the genomes of several eutherian lineages , such as New World rodents ( degu , chinchilla , guinea pig ) and ruminants ( sheep , cow , water buffalo ) , is inconsistent with a single endogenization event in a common ancestor of all eutherian mammals . Additionally , the genomes of several species , including multiple primate and carnivore species , contained multiple genetically distinct ERV-Fc lineages ( Figure 2 and Figure 2—source data 2 ) . Combined , these findings are consistent with a natural history marked by numerous cross-species transmissions leading to independent episodes of genome colonization in the ancestors of the examined species ( see subsequent section on cross-species transmission ) . Similar to most ancient ERV loci , the viral open reading frame ( ORF ) sequences present in the vast majority of retrieved ERV-Fc elements are disrupted by mutations ( including point-mutations , insertions , and deletions ) , precluding expression of functional viral proteins . However , we found intact ORFs corresponding to the viral env gene in several species ( indicated by an envelope icon in Figure 2 ) . Based on sequence inspection , these are ORFs that potentially retain the capacity to encode retroviral envelope glycoproteins . Species with open env ORFs include aardvark , gray mouse lemur , squirrel monkey , marmoset , baboon , chimpanzee , human , dog , and panda . With one exception , each of these ORFs is unique to the species in which it was identified , indicating independent origins for each . The exception is the previously characterized HERV-Fc1env locus ( Bénit et al . , 2003 ) , which is present in the genomes of all great apes . The env ORF of this locus is open in human , chimpanzee , and bonobo orthologs , whereas mutations have disrupted the ORF in the gorilla ortholog . The primary goal of this study was to use ERV-Fc sequences to gain insight into the nature of the related exogenous viral agents that infected and spread among mammalian hosts . Indeed , by examining our consensus reconstructions in light of the known functions of each viral protein and their roles in replication , we found a number of patterns most consistent with extensive spread and evolution of an exogenous retrovirus ( described in detail in subsequent sections ) . This indicates that our reconstructions reflect the nature of the exogenous agent that left ERV-Fc sequences behind in the germlines of its mammalian hosts . The viral env gene encodes the proteins responsible for binding to the host entry receptor and mediating fusion of the viral and host membranes ( Hunter , 1997 ) . Retroviral Env proteins are expressed as a single large polyprotein which is proteolytically cleaved into the surface ( SU ) and transmembrane ( TM ) subunits by cellular furin-like proteases . The resulting complex is a heterotrimer composed of three SU subunits and three TM subunits . SU is involved in receptor binding , while TM both anchors Env in the membrane and mediates host/viral membrane fusion . As even closely related viruses often utilize different receptors for entry , and because SU is also the major target of antibodies during the host immune response , sequence diversity in SU is usually very high ( Katz and Skalka , 1990 ) . The extracellular portion of retroviral TM proteins creates a series of alpha helices whose orientation shift dramatically during membrane fusion ( Hunter , 1997 ) . Gamma-like retroviral Envs also possess a conserved disulfide motif ( CXnCC ) involved in covalent interactions with SU , and a highly conserved immunosuppressive domain ( Bénit et al . , 2001 ) . Due to these functional constraints , the extracellular region of TM is rather well conserved ( Katz and Skalka , 1990; Bénit et al . , 2001 ) . The diversity within ERV-Fc Env sequences follows a pattern as would be expected based on prior knowledge of Env function; the SU domain displays extremely high sequence diversity , while the N-terminus of TM ( where the alpha helices and the CXnCC and ISD motifs are located ) is well conserved ( Figure 3—figure supplement 2 ) . In order to track the spread and evolution of the virus , we performed phylogenetic analyses using the consensus reconstructions of all three viral precursor proteins from each species . Depending on the viral history , assessing the relationships for all viral proteins can provide either increased confidence in associations between viruses from different species or reveal lineages that have a history of recombination . Initially , we examined the evolutionary history of ERV-Fc Gag . To do so , maximum likelihood ( ML ) phylogenies were generated from viral Gag sequences stripped of p12 , which was omitted due to extremely low amino acid sequence conservation . HERV-H and HERV-W sequences were also included as outgroups based on previous reports that these are the most closely related sequences to ERV-Fc ( Jern et al . , 2005 ) . The analysis revealed that all the ERV-Fc Gag sequences formed a monophyletic branch distinct from HERV-H and HERV-W ( Figure 4A ) . We identified several clades consisting of ERV-Fcs identified in closely related species , including Old World primates , New World primates , lagomorphs , and carnivores . However , the ERV-Fc Gag phylogeny did not recapitulate the known phylogeny of the host species – that is , some clades comprised ERV-Fc from distantly related mammals . Such patterns are suggestive of interspecies transmission ( discussed below ) . 10 . 7554/eLife . 12704 . 011Figure 4 . Phylogenetic relationship between ERV-Fc sequences . Maximum likelihood amino acid trees of ( A ) Gag ( B ) Pol and ( C ) TM generated using the LG substitution matrix . In each panel , HERV-H and HERV-W sequences were included as outgroups . Boostrap confidence values of nodes are depicted by colored spheres . In order to save space , a distance of approximately 0 . 6 was removed from the HERV-W outgroup branch in the Gag phylogeny ( A ) , as indicated by the broken line . DOI: http://dx . doi . org/10 . 7554/eLife . 12704 . 01110 . 7554/eLife . 12704 . 012Figure 4—source data 1 . Full-length ERV-Fc Gag alignment . The phylogeny shown in Figure 4A is based on this alignment excluding the p12 region ( as described in the Materials and methods ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12704 . 01210 . 7554/eLife . 12704 . 013Figure 4—source data 2 . ERV-Fc CA alignment . DOI: http://dx . doi . org/10 . 7554/eLife . 12704 . 01310 . 7554/eLife . 12704 . 014Figure 4—source data 3 . Full-length ERV-Fc Pol alignment . The phylogeny shown in Figure 4B is based on this alignment . DOI: http://dx . doi . org/10 . 7554/eLife . 12704 . 01410 . 7554/eLife . 12704 . 015Figure 4—source data 4 . ERV-Fc RT alignment . DOI: http://dx . doi . org/10 . 7554/eLife . 12704 . 01510 . 7554/eLife . 12704 . 016Figure 4—source data 5 . Full-length ERV-Fc Env sequences , including all recovered open reading frames . DOI: http://dx . doi . org/10 . 7554/eLife . 12704 . 01610 . 7554/eLife . 12704 . 017Figure 4—source data 6 . ERV-Fc TM alignment . The phylogeny shown in Figure 4C is based on this alignment . DOI: http://dx . doi . org/10 . 7554/eLife . 12704 . 01710 . 7554/eLife . 12704 . 018Figure 4—source data 7 . Alignment of ERV-Fc Pol including both inferred and strict consensus sequences . The phylogeny shown in Figure 4—figure supplement 1 is derived from this alignment . DOI: http://dx . doi . org/10 . 7554/eLife . 12704 . 01810 . 7554/eLife . 12704 . 019Figure 4—figure supplement 1 . Inferences made in deriving ERV-Fc consensus sequences do not significantly affect phylogenetic relationships . A ML phylogeny , generated via RAxML , comparing the relationship between our inferred Pol sequences and strict consensus sequences . HERV-H and HERV-W sequences were included as outgroups . Boostrap confidence values of nodes are depicted by colored spheres . HERV , human endogenous retrovirus; ML , maximum likelihood . DOI: http://dx . doi . org/10 . 7554/eLife . 12704 . 019 As retroviruses undergo a high degree of recombination , the evolutionary relatedness of the ERV-Fc sequences identified may differ based on which coding region is examined ( Bénit et al . , 2001; Henzy and Johnson , 2013 ) . To gain a better understanding of the evolutionary relatedness of the identified ERV-Fc sequences , phylogenetic analyses of Pol were also performed ( Figure 4B ) . Similar to what we observed for Gag , the consensus Pol sequences from Old World primates , New World primates , lagomorphs , and carnivores , which comprise distinct clades in the Gag analysis , also clustered together in the Pol phylogeny . In contrast , there were also some noticeable differences in relationships between the Gag and Pol sequences . For instance , the relationship between the carnivore ERV-Fc1 and ERV-Fc2 lineages differed in these phylogenies . In the Gag analysis , the two distinct ERV-Fc lineages ( ERV-Fc1 and -Fc2 ) represented in the ferret genome form a monophyletic clade , and similarly the ERV-Fc1 and -Fc2 lineages from the dog genome are most closely related to each other . In contrast , the Pol phylogeny supports a closer evolutionary relationship between ERV-Fc1 lineages from the dog and ferret genomes and separately the dog and ferret ERV-Fc2 lineages , than between the two lineages from the same species . Such incongruencies in topology are an indication that recombination events involving different viral lineages took place between these two viral regions ( examined further in the discussion that follows ) . Finally , we examined the TM region of Env ( Figure 4C ) . The SU domain of Env was not included as it is known to be one of the most rapidly evolving protein domains of retroviruses ( Bénit et al . , 2001 ) ; indeed , we found that levels of primary sequence identity within ERV-Fc SU were too low for informative phylogenetic analysis ( Figure 4—source data 5 ) . Similar to both Gag and Pol analyses , in the TM analysis ERV-Fc lineages from Old World primates formed a distinct clade within the tree . However , other aspects of the TM phylogeny revealed a history of recombination events involving Env . First , and most strikingly , several TMs formed a monophyletic clade with HERV-W . These include tarsier and a distinct subclade comprised of carnivore TMs ( panda as well as dog and ferret ERV-Fc1 ) . Based on the evolutionary relatedness of the Gag and Pol sequences from these species , it is likely that two recombination events occurred that resulted in these acquisitions of ERV-W Env by ERV-Fc: one leading to the ERV-Fc sequences found in the tarsier genome , and a separate event creating the chimeric virus that spread between several carnivore species and independently endogenized each of them . Second , the mouse and rat TM sequences included here clearly share identity with ERV-Fc; however , these are found in the context of two unrelated betaretrovirus-like genomes . Third , while the Gag and Pol sequences from lagomorphs consistently branched together , the TM sequences from rabbit and pika were most similar to the rat Beta-recombinant and aardvark TM sequences , respectively . Again , such incongruencies argue strongly for a history of recombination involving ERV-Fc coding regions . In summary , the phylogenetic analyses presented here argue for a long viral history of active exogenous replication during which time numerous recombination and interspecies transmission events occurred , resulting in the disparate topologies observed between the different viral protein phylogenies . To further examine the contribution of cross-species transmission events to the distribution of ERV-Fc among mammalian genomes , we performed a tanglegram analysis ( i . e . we compared the viral phylogeny with that of the host species ) . The null hypothesis is that the virus co-speciated with the host ( either as an exogenous virus or as preexisting endogenous elements ) , which would produce host and viral phylogenies with similar topologies . Deviations from the null hypothesis ( co-speciation ) are revealed when lines connecting each virus taxon with that of its host taxon ( the genome in which it was found ) cross one another , indicating instances where cross-species transmission events are likely to have occurred . The results of the comparison between the host phylogeny and ERV-Fc Gag ( which allowed for inclusion of the greatest number of taxa ) are shown in Figure 5A . For this analysis , a supertree was created using ML and Bayesian trees based on both CA or Gag ( stripped of p12 , as described above ) . This approach provided a method for inclusion of ERV-Fc isolates from tenrec , rat , and mouse , for which only a CA sequence could be retrieved ( Figure 2—source data 2 ) , without sacrificing analytical robustness gained by including more residues in the analysis . This analysis revealed numerous incongruencies between the phylogenies of ERV-Fc Gag and their hosts . A similar web of crossing lines was observed when the ERV-Fc Pol phylogeny was used ( Figure 5—figure supplement 1 ) . Furthermore , the vertical distance traversed by the connecting line can provide a proxy for estimating the relative evolutionary relatedness of the species involved in individual cross-species transmission events . Figure 5A provides evidence for a number of cross-species transmission events between species of the same mammalian order ( e . g . human and rhesus ) . Such events might be expected to predominate for several reasons – for example , there are likely to be fewer genetic barriers to viral replication in the new host ( and such barriers should be easier to overcome ) and closely related hosts may be more likely to be sympatric and thus more likely to encounter one another and exchange viruses ( Holmes and Drummond , 2007; Mollentze et al . , 2014; Sharp and Hahn , 2011 ) . However , we also found evidence for cross-species transmission involving hosts belonging to different mammalian orders ( e . g . tenrec , lemur , dolphin ) , which is thought to be much rarer due to the genetic distances involved ( Denner , 2007 ) . Additionally , several mammalian lineages also harbored two lineages of ERV-Fc , including Great Apes , Old World monkeys , New World monkeys , ferrets , and dogs ( Figure 5A and Table S2 ) . In each instance , the molecular data indicate that the two lineages originated from independent cross-species transmissions and genome colonization events in these species . These findings indicate that the distribution of ERV-Fc in the mammalian species included in this study predominantly originated following cross-species transmission events of exogenously replicating viruses . 10 . 7554/eLife . 12704 . 020Figure 5 . ERV-Fc has a multimillion-year history of replication with multiple cross-species transmissions . ( A ) Tanglegram comparison of host ( left ) and ERV-Fc phylogenies ( right ) ; dashed lines match species and the ERV-Fc found within their genome . The host phylogeny was adapted from ( Bininda-Emonds et al . , 2007 ) , while the ERV-Fc phylogeny is a supertree generated using Matrix Representation Parsimony ( MRP ) based on CA and Gag amino acid phylogenies . ( B ) LTR-derived age estimates of ERV-Fc loci derived by applying a neutral evolution rate of 4 . 5×10-9 substitutions per site per year to the nucleotide divergence between the 5’ and 3’ LTRs . Each plotted point represents the age estimate of a single genomic locus . Loci that show clear signatures of gene conversion or recombination have been omitted from this analysis . The average age is indicated by black vertical lines . Dotted lines indicate the approximate boundaries of the Oligocene epoch ( ~33 . 9 to ~23 MYA ) . ERV , endogenous retrovirus . DOI: http://dx . doi . org/10 . 7554/eLife . 12704 . 02010 . 7554/eLife . 12704 . 021Figure 5—figure supplement 1 . Tanglegram comparison of host ( left ) and ERV-Fc phylogenies ( right ) . The host phylogeny was adapted from ( Bininda-Emonds et al . , 2007 ) . The ERV-Fc phylogeny is the ERV-Fc Pol tree shown in Figure 4B . Dashed lines match ERV-Fc lineages and the species from which they were identified . ERV , endogenous retrovirus . DOI: http://dx . doi . org/10 . 7554/eLife . 12704 . 021 These observations in conjunction with other molecular characteristics of the viral lineages led us to estimate a minimum of 26 independent cross-species transmission events resulted in the observed distribution of ERV-Fc among the mammalian species examined . However , this likely underestimates the total contribution of interspecies transmission to the spread of the exogenous virus because transmission may be more common between closely related species ( e . g . due to genetic similarity of the hosts , or sharing of the same or similar range or niche ) . Such jumps between closely related hosts are less likely to be detected using incongruence between virus and host trees and this would be especially true when viral jumps occur close to speciation events . Estimating the time of endogenization can provide a minimal estimate of the age of the corresponding exogenous virus . To produce estimates of when active ERV-Fc endogenization occurred in the mammalian genomes examined , we employed two independent methods . First , where ERV-Fc loci originated in a common ancestor of multiple species this history of vertical transmission was used to assign upper and lower age estimates for these proviral loci . Using established time estimates of speciation ( Hedges et al . , 2006 ) , we could place an upper bound based on the divergence time between species where one harbors a given proviral locus that a second species lacks . The lower bound was similarly determined , but in this case using divergence times between species that share orthologous loci . Second , we used a molecular clock calculation based on the divergence of the 5’ and 3’ LTRs of individual ERV-Fc loci ( Dangel et al . , 1995; Johnson and Coffin , 1999; Martins and Villesen , 2011 ) ( Figure 5B ) . Due to the mechanism underlying viral reverse transcription , the two LTRs of a provirus have identical sequences at the time of insertion , and afterward these sequences acquire mutations in accordance with the neutral evolution rate of the host genome . Therefore , older proviruses would possess LTRs that are more divergent than young loci . Our data revealed that ERV-Fc sequences in several mammalian clades reflected endogenization in a common ancestor with subsequent vertical inheritance by multiple descendant species where they were sampled in this study . Mammalian clades for which this is the case include the Great Apes ( humans , chimpanzee , gorilla , orangutan ) ; Old World monkeys ( grivet , baboon , rhesus , cynomologous ) ; New World monkeys ( marmoset , squirrel monkey ) ; and muridae ( mouse , rat ) . Based on these findings and the reported times of speciation for these clades ( Hedges et al . , 2006 ) , we inferred that ERV-Fc was undergoing genome colonization between 30 and 10 million years ago ( MYA ) . For example , the presence of similar or shared ERV-Fc2 elements among Great Apes must reflect infections that occurred prior to divergence from a common ancestor 15 . 8 MYA ( orangutan/human split ) , but more recent than 19 . 9 MYA ( human/gibbon split ) . Shared ERV-Fc1 elements among Great Apes indicate that this viral lineage endogenized apes between 15 . 8 MYA ( orangutan/human split ) and 8 . 9 MYA ( gorilla/human split ) . Similarly , the New World primate ERV-Fc3-2 found in the genomes of the squirrel monkey and marmoset must be at least 19 . 1 million years old ( marmoset/squirrel monkey split ) , but younger than 43 . 1 million years old ( Platyrrhini/Catarrhini split ) . Also , both the ERV-Fc1 and ERV-Fc2 lineages in Old World primates were found in grivet , baboon , and rhesus , indicating an age of at least 11 . 7 million years , but additional data indicated that these lineages are also present in colobine monkeys [Patel , Senter , Johnson & Diehl , personal communication] , pushing this this date back to 17 . 1 MYA , but no later than 29 . 1 MYA ( Hominoidae/Cercopithecidae split ) . Finally , the ERV-Fcs present in mice and rat genomes pre-date speciation ( 22 . 6 MYA ) but are not found in hamsters ( 43 MYA ) . We also performed molecular clock analysis of ERV-Fc loci based on LTR divergence so that we could estimate the age of endogenization in a larger proportion of species examined in this study . For these analyses , we used a neutral evolution rate of 4 . 5×10-9 substitutions per site per year ( Waterston et al . , 2002 ) . This evolutionary rate is approximately twice that previously estimated for the primate/hominid lineage . However , employing the lower evolutionary rate produced age estimates suggesting insertion of ERV-Fc loci should pre-date major lineage splits where we have examined genome sequencing data and failed to find evidence corroborating such ancient insertional dates ( e . g . Cercopithecidae/Hominoidae and Feliformia/Canidae/Arctoidea ) . Regardless , the data in Figure 5B illustrate two phenomena: 1 ) genome colonization by ERV-Fc in the species examined occurred at many times over the course of many millions of years and 2 ) following colonization , expansion within many lineages ( by reinfection and/or retrotransposition ) often continued for many millions of years . Regarding the first point , the oldest ERV-Fc loci are the ERV-Fc2 elements in the genomes of ferret and canine , which date to 35 . 2 and 32 . 4 MYA , respectively . Nearly as old are ERV-Fc3-2 isolates from the squirrel monkey and marmoset genomes , whose oldest loci date to 31 . 3 and 29 . 3 MYA , respectively . These species diverged from a common ancestor approximately 19 MYA , and they share many orthologous ERV-Fc3-2 loci . Importantly , the LTR-based molecular clock calculations on these ERV-Fc3-2 loci yield age estimates consistent with age estimates based on species divergence times . Several species harbor ERV-Fc isolates whose oldest loci date to around 20 MYA . This group includes rabbit , human ( HERV-Fc2 ) , squirrel monkey ( ERV-Fc3-1 ) , marmoset ( ERV-Fc3-1 ) , and dog ( ERV-Fc1 ) . The ERV-Fc2 from rhesus is likely to fall in this group as well , in spite of the fact that there is a single outlier locus with a calculated age of approximately 30 MYA . We believe that this significantly overestimates the true age of this particular locus , possibly due to the relaxed evolutionary constraints of its position in the centromeric region of the Y chromosome . Nucleotide substitution rates observed on the Y chromosome and near centromeres are significantly higher than for other regions of the genome ( Brown and O'Neill , 2014; Hughes et al . , 2010; Malik , 2009; Xue et al . , 2009 ) . Thus , our data showed that significant cross-species transmission and endogenization by ERV-Fc took place over a span of more than 10 million years . In addition to the long evolutionary period during which ERV-Fc was actively invading mammalian genomes , there is evidence in these data for long periods of post-endogenization amplification of ERV-Fc elements in the majority of host germ lines . For example , for nearly every species’ genome that contained multiple ERV-Fc integrations , the difference in age between the oldest locus and the youngest locus was greater than 10 million years . In addition , the younger loci displayed features known to correlate with expansion by post-insertion amplification mechanisms , including large deletions of the viral genes ( Magiorkinis et al . , 2012 ) , while the older loci tended to retain recognizable gag , pol , and env sequences [68 and data not shown] . The data presented up to this point suggest an interesting history for ERV-Fc within carnivores . The data shown in Figure 4 supports the existence of two distinct lineages ( ERV-Fc2 and ERV-Fc1 ) . ERV-Fc2 sequences encode a canonical ERV-Fc Env ( Figure 4C ) , are found in the cat , dog , and ferret genomes , and appeared in the genomes of the ancestors of dogs and ferrets very early in the history of ERV-Fc ( Figure 5B ) . In contrast , ERV-Fc1 elements encode an Env most similar to ERV-W ( Figure 4C ) , are present in the genomes of giant pandas , dogs , and ferrets , and colonized the ancestors of dogs and ferrets >10 million years after ERV-Fc2 ( Figure 5B ) . Furthermore , we observed an incongruency in the dog and ferret ERV-Fc1/ERV-Fc2 relationship between the Gag and Pol phylogenies ( Figure 4 , panels A and B ) . Incongruency between the Gag and Pol trees could represent either divergent selective pressures acting on these genes , or a history of recombination following the origin of these distinct viral lineages . To distinguish between these possibilities , we generated phylogenies based on nucleotide alignments of gag sequences from individual ERV-Fc loci from the dog and ferret genomes , along with consensus giant panda and feline ERV-Fc nucleotide sequences ( Figure 6 ) . This analysis provided several insights into Carnivora ERV-Fc evolution . First , there is a clear ERV-Fc1 clade comprised of loci from all species known to harbor this recombinant ERV-Fc lineage ( dog , ferret , and giant panda ) , and all canine ERV-Fc1 sequences reside in this clade . However , there are three distinct ERV-Fc1 sublineages present within the dog genome that have distinct relationships with ERV-Fc1 sequences from other carnivores . We interpret this as an indication of at least two , but more probably three , separate cross-species transmission events into dogs . Also found in this ERV-Fc1 clade is a minor population from the ferret genome [ERV-Fc1 ( b ) ] . However , the majority of the ferret ERV-Fc1 gag sequences form a monophyletic clade with gag sequences of the non-chimeric ferret ERV-Fc2 . The observed phylogenetic associations are evidence that a recombination event replaced gag within the ferret lineage . This is supported by the observation that all ERV-Fc1 pol sequences , including those from ferret , form a monophyletic clade ( data not shown ) . Finally , a solitary ferret ERV-Fc2 gag forms a monophyletic clade with the canine ERV-Fc2 sequences . This likely indicates that at least two distinct lineages of ERV-Fc2 jumped from another species into an ancestor of the ferret lineage: one potentially originating in a canine ancestor , and a second coming from an unknown source . 10 . 7554/eLife . 12704 . 022Figure 6 . The evolutionary history of carnivore ERV-Fc1 includes numerous cross-species transmission events and at least one recombination event . Maximum likelihood phylogenetic analysis of carnivore ERV-Fc gag nucleotide sequences: sequences from the dog genome are colored in a shade of red , those from the ferret genome are colored in a shade of blue , and the panda consensus gag is colored in green . The feline ERV-Fc consensus gag sequence has been included as an outgroup and is colored black . For sequences from the dog and ferret genomes , the darker colored taxa are ERV-Fc2 sequences ( defined based on their association with an ERV-Fc envelope sequence ) , while the lighter colored taxa are ERV-Fc1 sequences ( defined by an association with ERV-W envelope sequence ) . Lineages where a large portion of the gag sequence has been replaced with heterologous non-coding sequence is denoted by * in the name . Boostrap confidence values of ancestral nodes are depicted by colored spheres . ERV , endogenous retrovirus . DOI: http://dx . doi . org/10 . 7554/eLife . 12704 . 02210 . 7554/eLife . 12704 . 023Figure 6—source data 1 . Nucleotide alignment of carnivore ERV-Fc gag sequences . The phylogeny shown in Figure 6 is derived from this alignment . DOI: http://dx . doi . org/10 . 7554/eLife . 12704 . 023 Combined , our data support a natural history of ERV-Fc1 such as that depicted in Figure 7 . Briefly , an initial recombination event allowed for the acquisition of an ERV-W envelope by a carnivore ERV-Fc-related virus . Subsequently , multiple cross-species transmission events resulted in colonization of the genomes of the ancestors of dogs , ferrets , and giant pandas . Up to three interspecies transmission events account for the genetic diversity of ERV-Fc1 presently found in the canine genome . Due to the relatively close evolutionary relationship between ERV-Fc1 ( b ) loci found in the ferret genome and one subset of ERV-Fc1 ( b ) in the canine genome , it is possible that this lineage was transmitted directly between the ancestors of dogs and ferrets . However , a similar relationship could also be explained by independent transmissions of similar viruses from a third , unidentified species . Following introduction of ERV-Fc1 ( b ) into mustelids , it appears that additional recombination event ( s ) took place with a pre-existing ERV-Fc2 virus , or viruses , resulting in the acquisition of the ERV-Fc2 gag . It was this double chimeric ERV-Fc1 that most successfully invaded the ferret genome . 10 . 7554/eLife . 12704 . 024Figure 7 . Proposed recombination and transmission sequence involving carnivore ERV-Fc1 . ERV-Fc sequences are depicted in blue , while ERV-W sequences are depicted in orange . See text for a detailed explanation of the arrows . DOI: http://dx . doi . org/10 . 7554/eLife . 12704 . 024
ERV loci can be used to reconstruct the natural history of the ancient , exogenously replicating retroviruses . Previous studies examining retroviral macroevolution via the ERV fossil record have cast an wide net , focusing primarily on the highly conserved RT as a phylogenetic marker and using it to characterize a broad swath of diversity within the Retroviridae family ( Jern et al . , 2005; Hayward et al . , 2013; 2015 ) . However , focusing on RT excludes additional sources of phylogenetic signal available to resolve relationships between closely related taxa , and may overlook the potential role that recombination plays in retroviral evolution . Thus , we sought to examine the deep evolutionary history of a single retrovirus lineage – that which produced the ERV-Fc family of sequences – by collecting and analyzing endogenous retroviral sequence information for all three of the canonical retroviral genes ( gag , pol , and env ) . Doing so allowed us to identify ERV-Fc sequences in 28 of the 50 mammalian genomes examined . Furthermore , we determined that as many as 26 independent cross-species transmission events produced the distribution of identified ERV-Fc elements . This included several species whose genomes appeared to have been independently colonized by two evolutionarily distinct ERV-Fc lineages . These results indicated that the distribution of ERV-Fc among modern mammals is predominately the result of interspecies spread and emergence of the related exogenous forms of the virus . ERV sequences present in the genomes of different species can be related either due to vertical inheritance ( as genomic loci ) or due to independent colonization by an exogenous , infectious virus . The two scenarios differ primarily due to differences in the rate at which exogenously replicating virus sequences and endogenous sequences evolve , as well as any differences in the selective pressures affecting parasitic genomic elements versus those affecting replicating viruses . We found that the patterns of amino acid diversification between ERV-Fc sequences were consistent with selection to maintain functions essential for exogenous viral replication . For example , the critical structural subunits of Gag ( MA and CA ) displayed the least diversity , and within CA , the most conserved residues were in regions involved in essential intrahexameric interactions . In contrast , primary sequence conservation in the non-structural subunits p12 and NC was significantly lower . In spite of this diversity , these regions retained their critical , canonical motifs , instead the number and location of these motifs varied significantly between viral isolates . This is consistent with selection to maintain essential motifs in a system that otherwise lacks structural constraint . The abundance of ERV-Fc sequence information allowed us to explore the evolutionary relationship between , and infer the history of , the individual ERV-Fc lineages uncovered . Our analyses point to a complex life history for the ERV-Fc retroviral lineage . This history began >30 MYA and exogenous replication continued for many millions of years and involved multiple cross-species transmission events . Recent studies have found evidence for cross-species transmissions in examinations of endogenous gammaretroviruses that are similar to extant MLV , and some of the jumps that these viruses made appear to have involved distantly related host-species ( Hayward et al . , 2013; 2015 ) . Taken together , gamma-like retroviruses appear to have had a rich history of cross-species transmissions that contrasts to the life histories of other retroviral genera . For example , exogenous foamy viruses are known to co-speciate with their hosts and the endogenous record suggests that long-term associations between foamy viruses and their hosts are likely to be an ancient feature of this retroviral genus ( Han and Worobey , 2012; Katzourakis et al . , 2009; Switzer et al . , 2005 ) . Furthermore , our analyses revealed that recombination played an important role in the life history of ERV-Fc with instances of acquisition of pol and env sequences from HERV-H or HERV-W-like viruses , as well as evidence that an ERV-Fc-related virus provided env sequence to a betaretrovirus . In this regard , the recombination observed within the carnivore ERV-Fc1 clade of viruses is noteworthy . In this lineage , an ERV-W env gene replaced the ancestral ERV-Fc env; subsequently , this chimeric virus was involved in at least two , and possibly as many as five , cross-species transmission events , giving rise to the endogenous sequences found in the genomes of modern dogs , ferrets , and giant pandas . The Pol and TM regions of the chimeric virus , ERV-Fc1 , form monophyletic clades , clearly indicating a shared ancestry for the viruses in the different species . However , within the dog and separately the ferret genome , the Gag sequences ERV-Fc1 and ERV-Fc2 are more closely related to one another than they are to sequences of the same lineage from the heterologous species . The recombinant ERV-Fc1 lineages were also observed to be younger than the majority of ERV-Fc2 loci in both dog and ferret genomes . Thus , the data revealed a scenario whereby after cross-species transmission the ERV-Fc/ERV-W env chimera acquired the ERV-Fc2 gag present in the genome of its new host species , in this case an ancestor of modern ferrets . Such a scenario would be consistent with the virus acquiring the ability to either interact with positive acting host proteins or avoid host restriction factors , or both . Our analysis suggests that the origins of ERV-Fc date back at least as far as the beginning of the Oligocene epoch ( ~33 . 9 MYA ) . This was a time period of dramatic global change marked by the fusion of the African to the European as well as the Indian to the Asian continental plates ( Briggs , 1995 ) , climatic cooling , development of vast expanses of grasslands , and the emergence of large mammals as the world’s predominate fauna ( Prothero and Berggren , 1992 ) . Continental mergers in the Old World along with a continued Asian-North American connection allowed for significant mammalian migrations throughout the Oligocene . However , we found evidence for ERV-Fc being present in species with little or no known geographic overlap at this early time in the viral life history , including musteloids , canids , Platyrrhini , and Tarsioidae . This makes it difficult to pinpoint a geographic region for the origin of the ERV-Fc viral lineage , as the ancestors of modern species whose genomes harbor ERV-Fc were geographically isolated from each other at the time . The fossil record provides solid evidence that during the Oligocene epoch canids were restricted to North America ( Munthe , 2005 ) , musteloids were present in Asia ( Sato et al . , 2012 ) , and Platyrrhini were likely restricted to South America ( Bond et al . , 2015 ) . The previously widespread distribution of Tarsioidae , which were found in Africa , North America , Europe , and Asia , was contracting to its current geographic isolation in southeast Asia ( Gingerich , 2012 ) . The geographic separation of these host species , coupled with the clear phylogenetic relationships between their viral sequences , provides strong evidence for a rapid global spread of the exogenous forms of ERV-Fc . Based on current phylogeographic knowledge of these early ERV-Fc hosts , and evidence for limited faunal exchange between these continents , we find it unlikely that musteloids , canids , Platyrrhini , or prosimians were solely responsible for this global viral spread . Importantly , the ERV-Fc genomic record in modern mammalian genomes likely represents only a fraction of the total exogenous viral spread: for example , exogenous infections may simply have failed to leave an endogenous footprint in some species , and some unknown proportion of lineages bearing ERV-Fc insertions will have eventually become extinct ( and the corresponding ERV-Fc record lost ) . Thus , it is likely that the ERV-Fc “fossil” record is incomplete , and that either extinct species or species lacking ERV-Fc sequences helped facilitate the worldwide spread of the exogenous virus . Finally , our results indicate that after the birth of ERV-Fc , replication , cross-species transmission , and endogenization continued for approximately another 15 million years . Our data , as well as other published reports ( Bénit et al . , 2003; Barrio et al . , 2011 ) , indicate that active ERV-Fc reinfection may have continued until very recently in some lineages , indicating that at least one ERV locus has retained functional gag and pol coding potential in those species . In ferret and canine , we found evidence that ERV-Fc1 acquired gag sequence from an older , pre-existing endogenous lineage ( ERV-Fc2 ) . Indeed , LTR dating indicated that in both species the oldest ERV-Fc1 locus pre-dates the end of active reinfection of the genome by ERV-Fc2 . Thus , it is plausible that there existed in the genomes of the ancestors of these species at least one functional ERV-Fc2 gag ORF that the newly introduced ERV-Fc1 could have acquired . Observations in laboratory mice as well as in vitro and in vivo experiments provide several well characterized examples of recombination involving ERV sequences giving rise to replication-competent viruses with novel biological properties ( Chong et al . , 1998; Coffin et al . , 1989; Paprotka et al . , 2011; Patience et al . , 1998; Telesnitsky and Goff , 1993 ) . Thus , ERV loci could contribute to adaptive evolution of exogenous viruses by providing a reservoir of novel sequences that can be tapped into by co-packaging and recombination .
Previously published human and baboon ERV-Fc sequences ( Bénit et al . , 2003 ) were used as bait in Basic Local Alignment Search Tool ( BLAST ) queries of 50 mammalian genome-sequencing databases hosted at the National Center for Biotechnology Information ( NCBI ) ( Table S1 ) . These genomes comprise a broad sampling of metatherian and eutherian mammals including representatives of every continent except Antarctica . However , as indicated in Table S1 , these genomes were generated using various approaches that result in ERV sequence information of varying quality . High-coverage Sanger sequencing approaches employing BAC and whole-genome shotgun sequencing utilized to produce genomic sequence for the human and mouse , amongst others , results in maximally reliable assemblage of non-coding sequences . In contrast , genomes assembled using low-coverage Sanger sequencing or only Illumina sequencing data often possess fragmentary non-coding sequence information due to difficulties in accurately assembling short stretches of sequence covering repetitive sequences . To extract maximal sequence information from genomic databases , an iterative BLAST approach was used . Initial BLAST queries utilized CA , RT , and/or TM inputs and were used to perform nt BLAST searches with the following parameters: match/mismatch scores of +1/−1; gap costs of 0 and 2 ( existence , extension ) ; and repeat masking turned off . BLAST hits were extracted along with surrounding sequence information , and these were analyzed using RepeatMasker ( Institute for Systems Biology , Seattle , WA [http://www . repeatmasker . org] ) to identify sequences of ERV origins , and those of ERV identity were in turn used as query sequences with the same scoring parameters as the initial BLAST query . Then , BLAST hits from individual genes were correlated in an attempt to identify complete , or mostly intact , genomes . Approaches for extracting ERV sequences from the genomic databases diverged at this point . In the few instances where the genome sequencing data allowed , ERV LTR sequences were identified , and these were used to extract genomic regions flanked by two LTRs , the vast majority of which represented ERV genomes . Alternatively , if ERV genomes could be identified , they were used as query sequences for a subsequent BLAST interrogation with match/mismatch scores of +2/−3 and gap costs of 5 and 2 ( existence , extension ) . If full-length ERV sequences could not be identified , due to either the specific genomic sequence being of too low quality or the lack of intact ERVs in the genome , full-length gag , pol , and env sequences were used in the final round of BLAST interrogation with the same parameters as used with full-length ERV sequences . When full-length gag , pol , or env sequences could not be obtained , then sequences extracted from the second round of BLAST interrogation ( using initial CA , RT , or TM hits as query sequences ) were utilized . Extracted ERV sequences were aligned , initially using the MUSCLE algorithm ( Edgar , 2004 ) as implemented in Geneious 6 ( Biomatters , Auckland , NZ ) and then further refined by hand . Consensus sequences were generated from these alignments; however , when multiple disparate types of ERV sequence were retrieved , separate consensus sequences were generated for each class . The consensus sequences were then used to infer the proteins encoded by the ERVs . This was done by replacing premature termination codons with the most frequently represented alternate codon and substituting 'R' or 'Y' ambiguities ( transition mutations ) to either cytosine or guanine at locations determining the amino acid encoded ( i . e . NOT third base wobble positions ) . If consensus sequences contained transversion ambiguities , multiple amino acid consensus sequences were generated , differing only at those positions . To generate outgroups for phylogenetic analyses , HERV-H and HERV-W sequences were extracted from the human genome . To do this , Repbase ( Jurka et al . , 2005 ) sequences corresponding to gag , pol , and env of HERV-H and HERV-W ( listed as HERVH and HERV17 in Repbase , respectively ) were used as BLAST queries to interrogate the human genome with match/mismatch scores of +2/−3 and gap costs of 5 and 2 ( existence , extension ) . BLAST hits were extracted and aligned , initially using the MUSCLE algorithm as implemented in Geneious 6 with further refining by hand . Consensus generation proceeded as for ERV-Fc . Multiple alignments were generated from the consensus ERV-Fc , HERV-H , and HERV-W protein sequences . These included CA , RT , and TM subunit alignments for all identified ERV sequences and separately full-length Gag , Pol , and Env alignments for the subset of sequences that possessed full-length sequence information . Sequences can be found in Figure 4—source datas 1–6 . CA alignments comprised approximately 225 amino acids spanning from the N-terminal proteolytic cleavage site to a conserved poly-charged region upstream of the CA/MA proteolytic cleavage site . For phylogenetic analysis , the Gag alignment was trimmed of the p12 region , as sequences in this region possessed low primary sequence homology and varied greatly in length . RT alignments consisted of approximately 216 amino acids that spanned from the N-terminal QΦP that forms portion of the DNA binding domain to 10 amino acids C-terminal of the conserved ΦLG involved in the catalytic function . TM alignments included approximately 130 residues of extracellular sequence from the poly-charged furin cleavage site to a conserved tryptophan adjacent to the poly-hydrophobic putative transmembrane sequence . Phylogenetic trees were constructed using both Bayesian Markov chain Monte Carlo ( MCMC ) and ML algorithms as implemented in MrBayes 3 . 2 . 1 ( Huelsenbeck and Ronquist , 2001 ) and PhyML ( Guindon and Gascuel , 2003 ) , respectively . ML phylogenies were generated for each alignment using combined NNI/SPR ( nearest neighbor interchange/subtree pruning and regrafting ) searching optimizing for topology , branch length , and substitution rate parameters , with the proportion of invariable sites set to 0 . 0 , four substitution rate categories , and an estimated gamma distribution parameter . Support for the ML branching patterns was assessed by performing 200 bootstrap replicates . Separate ML trees were generated for each alignment using the LG ( Le and Gascuel , 2008 ) and RtREV ( Dimmic et al . , 2002 ) amino acid substitution models . Bayesian phylogenies were calculated using the Poisson rate matrix , gamma rate variation with four gamma categories , with unconstrained branch lengths . Two parallel MCMC analyses of 1 , 100 , 000 steps each were performed using four heated chains and a heated chain temperature of 0 . 2 . Sampling of the trees was performed every 200 trees and omitting the first 500 trees ( 100 , 000 steps ) . Effective sample sizes of more than 600 indicated convergence of the MCMC run . In many instances , we included multiple ERV-Fc sequences for individual species . One source of this was due to uncertainties in determining an ancestral ORF sequence , which we resolved by generating multiple consensus sequences each reflecting different possibilities at ambiguous residues ( as described above ) . The other reason for multiple consensus sequences derives from the fact that some species harbored multiple distinct lineages of ERV-Fc , and in these cases , we generated independent consensus sequences for each lineage . Phylogenies were generated including these multiple consensuses and were subsequently collapsed for publication when all formed a monophyletic branch . To assess the influence our method of reconstructing ancestral ORFs had on the generated phylogenetic topologies , 'strict' consensus sequences were generated for all ERV-Fc lineages where we were able to reconstruct a full-length Pol . When necessary , the 'strict' consensus sequence was edited in order to maintain frame . Otherwise , all ambiguities and premature stop codons were maintained . Alignments were produced as per above , and phylogenies were generated via RAxML v7 . 2 . 8 ( Stamatakis , 2006 ) as implemented in Geneious 8 ( Biomatters , Auckland , NZ ) using the LG substitution model and the GAMMA model of rate heterogeneity . Confidence analysis was performed via 100 bootstrap replicates . This set of analyses were performed using RAxML instead of PHYML because inclusion of premature stop residues are forbidden in PHYML . Supertrees were generated using Clann ( ver . 3 . 2 . 3 ) ( Creevey et al . , 2004; Creevey and McInerney , 2005 ) with six total input trees: three based on CA sequence and three based on Gag ( - p12 ) sequence . These CA and Gag inputs include two ML phylogenies , one generated using the RtREV substitution model and the second using LG , and a single Bayesian phylogeny . To reflect the fact that Gag phylogenies are based on a larger dataset of parsimonious information , the input trees were weighted where Gag=2 and CA=1 . Heuristic searches were performed using Matrix Representation Parsimony ( MRP ) with subtrees generated using either the SPR ( sub-tree pruning and re-grafting ) or TBR ( tree bisection and reconnection ) resampling . Minimal differences were observed in phylogenies produced using these resampling methodologies , and phylogenies produced by TBR are presented here . Gag nucleotide alignments were produced of ERV-Fc1 and ERV-Fc2 sequences from carnivores . This included sequences from individual proviral loci from the ferret and dog genomes as well as the ERV-Fc consensus sequences generated from the cat and giant panda genomes . Similar to the amino acid alignment described above , the p12 region was stripped from the alignment as it showed signs of genetic saturation . ML phylogenies were produced in MEGA5 . 2 using the GTR ( general time reversible ) substitution model , pairwise deletion of gapped data ( 'use all sites' option ) , and NNI topology optimization . The Gag supertree and Pol ML phylogeny were further utilized as viral input trees for constructing virus-host tanglegrams . The host phylogeny used in generating these tanglegrams was generated by pruning a previously published pan-mammalian phylogeny ( Bininda-Emonds et al . , 2007 ) . TreeMap 3 was employed to deconvolute the relationship between the viral and host phylogenies and produce the fewest number of incongruencies ( as indicated by crossed lines ) ( Charleston and Robertson , 2002 ) . Gag , Pol , and Env ERV-Fc protein sequences were assessed for amino acid composition . As millennia of exogenous and endogenous replication were likely to have allowed these viruses to explore a nearly infinite evolutionary space , a means for examining the diversity of amino acid properties at a given residue was utilized . This system utilizes the Smith and Smith penalty matrix where identical residues are given a 0 penalty , highly similar residues ( such as Ile and Leu ) are given a penalty of 1 , residues sharing some chemical properties ( such as Leu and Phe ) are given a penalty of 2 , and completely dissimilar residues ( such as Leu and Pro ) are given a penalty of 3 ( Smith and Smith , 1990 ) . Diversity scores at each residue were calculated by summing pairwise Smith and Smith scores between a global ERV-Fc consensus and all ERV-Fc sequences . Global ERV-Fc consensus protein sequences were used to model structures of individual protein domains . Models were generated using the Protein Homology/analogy Recognition Engine V 2 . 0 ( Phyre2 ) server , which uses PSI-Blast , hidden Markov modeling , and profile-profile matching algorithms to thread novel amino sequences onto known protein structures ( Kelley and Sternberg , 2009 ) . For multimeric structures , the Phyre ERV-Fc model was overlayed onto MLV structures . In the case of assembling the CA hexamer , the 1U7K ( Mortuza et al . , 2004 ) MLV structure was utilized as a scaffold . Individual residues were then colored according to their calculated diversity scores . | Viruses have been with us for billions of years , and exist everywhere in nature that life is found . Viruses therefore have had a significant impact on the evolution of all organisms , from bacteria to humans . Unfortunately , viruses do not leave fossils , and so we know very little about how viruses originate and evolve over time . Fortunately , over the course of millions of years , genetic sequences from the viruses accumulate in the DNA genomes of living organisms ( including humans ) . These sequences can serve as molecular “fossils” for exploring the natural history of viruses and their hosts . Diehl et al . have now searched the genomes of 50 modern mammals for “fossil” viral remnants of an ancient group of viruses known as ERV-Fc . This revealed that ERV-Fc viruses infected the ancestors of at least 28 of these mammal species between 15 million and 30 million years ago . The viruses affected a diverse range of hosts , including carnivores , rodents and primates . The distribution of ERV-Fc among different mammals indicates that the viruses spread to every continent except Antarctica and Australia , and that they jumped between species more than 20 times . Diehl et al . also pinpointed patterns of evolutionary change in the genes of the ERV-Fc viruses that reflect how the viruses adapted to different host mammals . As part of this process , the viruses often exchanged genes with each other and with other types of viruses . Such genetic recombination is likely to have played a significant role in the evolutionary success of the ERV-Fc viruses . Mammalian genomes contain hundreds of thousands of ancient viral fossils similar to ERV-Fc . Future work could study these to improve our understanding of when and why new viruses emerge and how long-term contact with viruses affects the evolution of their host organisms . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology",
"microbiology",
"and",
"infectious",
"disease"
] | 2016 | Tracking interspecies transmission and long-term evolution of an ancient retrovirus using the genomes of modern mammals |
Experimental study of the role of disorder in protein function is challenging . It has been proposed that proteins utilize disordered regions in the adaptive recognition of their various binding partners . However apart from a few exceptions , defining the importance of disorder in promiscuous binding interactions has proven to be difficult . In this paper , we have utilized a genetic selection that links protein stability to antibiotic resistance to isolate variants of the newly discovered chaperone Spy that show an up to 7 fold improved chaperone activity against a variety of substrates . These “Super Spy” variants show tighter binding to client proteins and are generally more unstable than is wild type Spy and show increases in apparent flexibility . We establish a good relationship between the degree of their instability and the improvement they show in their chaperone activity . Our results provide evidence for the importance of disorder and flexibility in chaperone function .
Despite years of intense effort , the precise mechanism by which chaperones interact with proteins to enhance their folding is not entirely clear . We reasoned that we might gain insight into this long-standing problem by isolating and characterizing chaperone variants that exhibit improved chaperone activity . A genetic selection that we had developed previously gave us a unique opportunity to pursue these aims . This selection uses a folding biosensor to directly link protein stability to antibiotic resistance . The biosensor consists of an unstable protein inserted into β-lactamase , a selectable marker that encodes penicillin resistance ( Foit et al . , 2009 ) . Stabilization of the unstable protein results in higher levels of antibiotic resistance . We showed that the stabilization could be due to mutations within the unstable protein itself ( Foit et al . , 2009 ) , addition of chemical chaperones to the growth media ( Hailu et al . , 2013 ) , or host variants that stabilize the unstable protein ( Quan et al . , 2011 ) . We isolated host variants that greatly stabilize poorly folded variants of immunity protein 7 ( Im7 ) , increasing their steady-state concentrations in the cell . We found that this stabilization occurs through the induction of a previously uncharacterized chaperone called Spy ( Quan et al . , 2011 ) . We obtained evidence that Spy acts in an ATP-independent manner to help protect bacterial cells from a number of conditions that lead to widespread protein denaturation and aggregation , such as treatment with tannin , ethanol , or butanol ( Quan et al . , 2011 ) . The crystal structure of Spy shows that it forms an unusual cradle shaped dimer ( Figure 1; Quan et al . , 2011; Kwon et al . , 2010 ) . When we attached environmentally sensitive probes to various sites in Spy , including the concave and convex surfaces , nearly all showed substantial changes in fluorescence upon interaction with the client protein casein . These results suggest that client binding may occur over large regions of Spy , that Spy might undergo significant conformational changes upon client binding , or a combination of both ( Quan et al . , 2011 ) . 10 . 7554/eLife . 01584 . 003Figure 1 . Surface presentations of the crystal structure of Spy ( PDB ID: 3O39 ) . The majority of activity-enhancing mutations localize to areas adjacent to hydrophobic patches . ( A ) Surface properties of Spy . Backbone atoms are shown in white , hydrophobic side chain atoms in yellow , and polar and charged side chain atoms in blue . Black dashed lines circle the two predominant hydrophobic patches P1 and P2 . ( B ) Sites accommodating beneficial mutations . Side chain atoms of the residues identified as mutations in the genetic selection are shown in red . Q25 is in the disordered N-terminus , which is not visible in the crystal structure . Q49L , H96L , and Q100L would expand the total hydrophobic area of P1 and P2 . DOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 00310 . 7554/eLife . 01584 . 004Figure 1—figure supplement 1 . Determination of the in vivo specific activity of Spy variants . ( A ) The minimal inhibitory concentration ( MIC ) of strains expressing the folding biosensor as well as the Spy variants measured in the presence of various concentrations of IPTG ( 0 . 01–0 . 5 mM ) . The MIC of penicillin V was measured by plating serially diluted cells ( 10−1 to 10−5 dilutions ) on LB plates containing 0 . 5–7 mg/ml penicillin V as previously described ( Foit et al . , 2009 ) . ( B ) The steady-state levels of different Spy variants increase upon induction with increasing amounts of IPTG ( 0 . 01–0 . 5 mM ) . ( C ) The steady-state levels of the Bla-Im7 L53A I54A biosensor in the presence of different Spy variants are linearly correlated with the relative MIC of the respective strains ( R2 = 0 . 64 ) . The data set shown here was measured when the respective co-expressed Spy variant was induced with 0 . 1 mM IPTG . DOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 00410 . 7554/eLife . 01584 . 005Figure 1—figure supplement 2 . WebLogo representation of a ClustalW ( Thompson et al . , 2002 ) sequence alignment of 29 Spy orthologous sequences . Residues are numbered according to the mature protein region of E . coli Spy . Short dashes indicate residues that are absent from E . coli Spy but are present in other aligned sequences . Red: basic residues K , R , and H; blue: acidic residues D and E; orange: hydrophobic residues including A , V , L , I , F , W , M , and P; black: polar uncharged residues including G , S , T , C , Y , N , and Q . Green boxes indicate the location of our Spy mutations with the original residues shown in green . DOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 005 We decided to investigate the mechanism by which Spy interacts with proteins and learn more about Spy’s properties as a chaperone . We used a genetic selection similar to that used to discover Spy in an attempt to further enhance Spy’s chaperone properties . We have now isolated Spy variants with improved ability to stabilize a poorly folded client protein ( Im7 L53A I54A ) in vivo . These variants also showed improved ability to prevent the aggregation of client proteins in vitro . Many of these Spy variants contain residue substitutions that act to expand a hydrophobic region present on the protein’s concave surface . Crosslinking and hydrogen-deuterium exchange measurements suggest that this hydrophobic region is involved in client protein interaction . Our optimized Spy variants bind the client protein Im7 more tightly than wild type Spy does but are generally less stable suggesting that flexibility is important in the function of Spy as a chaperone .
We expressed a protein stability biosensor in an Escherichia coli strain that co-expresses the gene for the chaperone Spy under the IPTG inducible Trc promoter . The stability biosensor consists of a tripartite fusion that contains the unstable protein Im7 L53A I54A inserted into β-lactamase under the constitutive β-lactamase promoter ( Foit et al . , 2009 ) . This partially unfolded variant of Im7 was chosen because Spy overproduction is known to stabilize it in vivo ( Quan et al . , 2011 ) . Increasing the expression level of the chaperone Spy by increasing IPTG concentrations results in improved penicillin resistance encoded by the β-lactamase-Im7 L53A I54A biosensor ( Figure 1—figure supplement 1A , B , focus on wild-type [WT] traces [black lines] ) . We reasoned that if mutations in Spy increase its specific activity as a chaperone , they should also be capable of enhancing the stability of the biosensor and thereby also enhance antibiotic resistance . Our ability to link protein folding to antibiotic resistance gives us a unique opportunity to select for activity-enhancing mutations in a chaperone . Analysis of the reasons behind the improved chaperone ability of activity enhancing mutants of Spy should inform us about Spy’s catalytic mechanism and perhaps also tell us what makes for a good chaperone . We reasoned that activity-enhancing mutations would be more informative in general than those that decreased function , in part because there are a wider variety of uninteresting reasons that mutations can disrupt function such as those causing chain termination . If we succeeded at all in getting activity enhancing mutations we anticipated obtaining two types of mutations . We might obtain those that acted in a substrate specific manner that improved the action of Spy only against the substrate for which they were selected on , and variants that generally improved the activity of Spy against multiple substrates . If we succeeded in obtaining this latter type of mutations , they should be particularly informative as to what makes a protein an effective chaperone . To obtain activity-enhanced Spy variants , we used an error-prone PCR-based approach ( McCullum et al . , 2010 ) that targeted the mature protein encoding region of the spy gene on pCDFTrc-Spy to create a plasmid library of ∼106 members that contained an average of 1 . 2 nucleotide mutations per spy gene . This variant library was transformed into SQ2041 , a spy null strain of E . coli that contains the stability biosensor ( see strain list in Table 1 ) . We plated the mutant library onto LB plates that contained 0 . 1 mM IPTG ( to induce Spy ) and 4 mg/ml penicillin , the concentration at which a strain co-expressing wild-type Spy and the biosensor ( strain SQ2068 ) fails to grow . Using this selection approach , we isolated 65 Spy variants that , when co-expressed with the biosensor , showed improved antibiotic resistance compared to cells that co-express wild-type Spy . 10 . 7554/eLife . 01584 . 006Table 1 . Strain listDOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 006StrainGenotype or relevant characteristicsSourceSQ765MG1655 ( F¯ λ¯ ilvG¯ rfb-50 rph-1 ) , ΔhsdR ( Quan et al . , 2011 ) SQ2041SQ765 , ΔampC , Δspy , pBR322 bla::GSlinker Im7 L53A I54A ( Foit et al . , 2009 ) This studySQ2068SQ2041 , pCDFTrc-SpyThis studyLW53SQ2041 , pCDFTrc-Spy Q100LThis studyLW54SQ2041 , pCDFTrc-Spy L32PThis studyLW55SQ2041 , pCDFTrc-Spy F115IThis studyLW56SQ2041 , pCDFTrc-Spy Q49LThis studyLW57SQ2041 , pCDFTrc-Spy F115LThis studyLW58SQ2041 , pCDFTrc-Spy H96LThis studyLW59SQ2041 , pCDFTrc-Spy Q25RThis study Remarkably , 48 ( 74% ) of the isolated Spy variants contained a glutamine to leucine mutation at amino acid 100 . For 20 of these variants , this alteration ( Q100L ) was the only mutation present , and strains expressing a Spy Q100L variant emerged from at least four independent mutagenesis and selection experiments . Other single mutations that answered the selection included Q25R , L32P , and F115I . There were also a number of other mutations ( Q49L , H96L , and F115L ) that were found independently 2–3 times in combination with other amino acid substitutions . To verify that these Spy mutations enhance the antibiotic resistance of the co-expressed biosensor when they are present as single mutations , we introduced the individual mutations Q25R , L32P , Q49L , H96L , Q100L , F115I , and F115L into the spy gene on the plasmid pCDFTrc-spy by site-directed mutagenesis and transformed the resulting plasmids into SQ2041 , the spy knockout strain co-expressing the biosensor . All of these strains except the one containing Q49L showed improved penicillin resistance compared to strains expressing wild-type Spy at a wide range of IPTG concentrations ( Figure 1—figure supplement 1A ) ; the relative minimal inhibitory concentrations ( MICs ) were up to twofold higher than the MIC of cells co-expressing wild-type Spy . One simple explanation for the increased penicillin resistance observed in the mutated strains might be increased Spy levels . Such an increase could occur through translational or posttranslational effects such as an increase in Spy stability . To examine these possibilities , we measured the steady-state expression levels of Spy in these strains when induced by different IPTG concentrations . All variants exhibited Spy levels that were within 20% of wild-type except Q49L , which showed a Spy level that was half that of wild-type ( Figure 1—figure supplement 1B ) . These results suggest that the observed increases in MIC ( up to twofold ) for the variant strains are not simply due to increased expression levels of the chaperone . We then measured the specific in vivo activity of our Spy variants by normalizing the maximal MIC values for penicillin V of the variant strains to the amount of Spy variant proteins found in these strains ( Figure 1—figure supplement 1 , ‘Materials and methods’ ) . All strains co-expressing the selected Spy variants ( including Q49L ) had normalized MICs 1 . 4–2 . 2-fold higher than SQ2068 ( Table 2 ) , indicating that the variant Spy proteins have higher specific activity than wild-type Spy . Furthermore , quantitative western blots showed that the increased MICs are linearly correlated with the steady-state levels of the folding biosensor ( Figure 1—figure supplement 1C ) . 10 . 7554/eLife . 01584 . 007Table 2 . Properties of Spy variantsDOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 007Spy variantsMICnormActivity ( aldolase agg . Prev ) Activity ( aldolase refold ) Activity ( α-LA agg . Prev ) kon ( × 105 mol−1 s−1 ) koff ( s−1 ) KD ( µM ) Tm ( °C ) ΔHm ( Kcal mol−1 ) ΔCp ( Kcal K−1 mol−1 ) ΔGNU ( 25°C ) ( Kcal mol−1 ) WT11113 . 98 ± 0 . 110 . 456 ± 0 . 0111 . 15 ± 0 . 02748 . 1 ± 0 . 166 . 6 ± 1 . 50 . 644 . 24 ± 0 . 10Q25R1 . 446 . 90 ± 0 . 641 . 38 ± 0 . 312 . 44 ± 0 . 892 . 29 ± 0 . 130 . 198 ± 0 . 0050 . 87 ± 0 . 03446 . 3 ± 0 . 473 . 7 ± 0 . 80 . 984 . 20 ± 0 . 10L32P1 . 922 . 52 ± 0 . 114 . 85 ± 0 . 572 . 10 ± 0 . 741 . 51 ± 0 . 140 . 030 ± 0 . 0020 . 20 ± 0 . 00331 . 0 ± 0 . 252 . 1 ± 2 . 50 . 710 . 99 ± 0 . 02Q49L1 . 602 . 88 ± 0 . 144 . 25 ± 0 . 661 . 93 ± 0 . 662 . 30 ± 0 . 060 . 176 ± 0 . 0090 . 76 ± 0 . 01852 . 0 ± 0 . 259 . 9 ± 1 . 00 . 684 . 19 ± 0 . 10H96L1 . 622 . 02 ± 0 . 321 . 90 ± 0 . 321 . 64 ± 0 . 492 . 68 ± 0 . 140 . 266 ± 0 . 0101 . 00 ± 0 . 02250 . 1 ± 0 . 156 . 2 ± 3 . 00 . 713 . 66 ± 0 . 23Q100L2 . 191 . 34 ± 0 . 054 . 20 ± 0 . 442 . 12 ± 0 . 751 . 19 ± 0 . 130 . 027 ± 0 . 0020 . 23 ± 0 . 01353 . 8 ± 0 . 628 . 9 ± 1 . 10 . 232 . 24 ± 0 . 11F115L1 . 521 . 98 ± 0 . 324 . 85 ± 0 . 562 . 30 ± 0 . 832 . 73 ± 0 . 010 . 245 ± 0 . 0070 . 90 ± 0 . 02841 . 3 ± 0 . 256 . 6 ± 4 . 90 . 762 . 60 ± 0 . 22F115I1 . 652 . 21 ± 0 . 124 . 33 ± 0 . 492 . 34 ± 0 . 852 . 82 ± 0 . 150 . 328 ± 0 . 0171 . 17 ± 0 . 09741 . 7 ± 0 . 454 . 3 ± 1 . 50 . 982 . 43 ± 0 . 11ΔGNU ( 25°C ) is the free energy of stabilization at 25°C ( NU dictates the transition from folded state to unfolded state ) , ΔHm is the change in enthalpy at Tm which is the melting temperature and ΔCp is the change in heat capacity associated with the unfolding of the Spy variant . agg . prev: aggregation prevention . Fold activity expresses relative to WT . Values after the ± sign are standard errors . MICnorm is measured for cells ( SQ2068 , LW53-59 ) expressing the pBR322 bla::GSlinker Im7 L53A I54A plasmid and various Spy constructs . kon , koff , and KD are kinetic parameters describing the interaction between Im7 L53A I54A and the Spy variants . To test whether the Spy variants’ increased chaperone activity was general or client specific , we purified the variant proteins and tested their chaperone activity in vitro using two standard chaperone clients: reduced denatured α-lactalbumin ( α-LA ) and chemically denatured aldolase . In genetic selections one usually gets what you select for , thus we had anticipated that the variants we obtained would show an improved ability to refold Im7 . These mutations , would at a minimum , likely to be informative about the factors involved in Spy-Im7 interactions . We however considered it unlikely that they would show generally improved chaperone activity for at least two reasons . First , in a wide variety of laboratory evolution experiments where variant enzymes are selected that show improved activity against one substrate , often though not always show decreased activity against other unrelated substrates ( Goldsmith et al . , 2012; Yang et al . , 2013 ) . More specifically , other efforts at improving chaperone activity , though showing some success in generating mutants that were better with the substrates they were selected on , in general showed decreased chaperone activity against other substrates ( Wang et al . , 2002; Aponte et al . , 2010; Schweizer et al . , 2011 ) . For instance Wang et al , through the use of a multistep screening process , succeeded in isolating GroEL variants that enhanced the expression of GFP and circularly permuted versions of GFP 3-8-fold , presumably by enhancing folding of GFP in vivo . However these variants were defective in all other measures of GroEL function tested including ability to support bacterial growth at high temperature , in vivo folding of the GroEL substrate HrcA , and phage lambda and Mu growth ( whose growth dependency on GroEL and GroES historically led to the naming of the GroE genes [Georgopoulos et al . , 1972] ) . These GroEL variants in vitro were also no better than wild type GroEL in enhancing the yield of active GFP . The authors concluded that increased GFP folding of these variants ‘comes at the expense of the ability of GroEL/S to fold its natural substrates’ ( Wang et al . , 2002 ) . The majority of the DnaK variants Aponte et al isolated based on an improved ability to fold an unstable variant of chloramphenicol acetyl transferase in vivo turned out to be inferior to wild type DnaK in refolding luciferase in vitro , though it needs to be mentioned that they did succeed in isolating four variants that showed a slightly improved in vitro refolding yield for luciferase ranging from 1 . 2 to 1 . 9-fold ( Aponte et al . , 2010 ) . Given the apparent difficulty in isolating chaperone or enzyme variants that show generally enhanced activity against a variety of substrates , we were surprised that all 7 Spy variants that we had isolated based on their ability to fold Im7 in vivo were significantly more active in preventing aggregation of both chemically denatured α-lactalbumin and aldolase in in vitro assays . In the aldolase aggregation assay , they were 1 . 3–6 . 9-fold more active than wild type , and in the α-lactalbumin aggregation assay , they were 1 . 6–2 . 4-fold more active ( Table 2 ) . We also tested for the activity of these chaperone variants in their ability to facilitate aldolase refolding . Six of the seven of the variants were found to be more active than is wild type Spy in the range of 1 . 9–4 . 9-fold . The one exception , Q25R , was measured to marginally increase refolding yield ( 1 . 4-fold ) ( See Table 2 ) . Because our Spy variants showed improved chaperone activity towards at least three client proteins ( Im7 , aldolase and α-lactalbumin ) , we called them ‘super-Spy’ variants . We mapped the activity-enhancing mutations identified in the selection onto Spy’s crystal structure and found that many of them were located close to each other . Most of them mapped immediately adjacent to the two predominant hydrophobic patches on the concave surface ( P1 , P2 ) of the cradle interior ( Figure 1A , B ) —the region that we had previously hypothesized might be involved in client binding ( Quan et al . , 2011 ) . The most commonly observed substitutions ( Q100L , which occurred in 74% of the variants , and H96L and the Q49L , which occurred in ∼5% and ∼3% of the variants , respectively ) change polar or charged residues ( glutamine and histidine ) into the hydrophobic amino acid leucine , thereby increasing the area of the hydrophobic region . One proposed mechanism for chaperone function is via blocking hydrophobic regions present on client proteins , thereby preventing their aggregation ( Hartl et al . , 2011 ) . Simple expansion of peptide-binding hydrophobic regions on our Spy variants could thus be one straightforward way to explain their improved chaperone activity . There are some expectations of this simple model: ( 1 ) chaperone variants with a larger or stronger hydrophobic patch will have enhanced affinity for client proteins , and ( 2 ) client proteins are likely to interact with regions on the 3D structure of the chaperone that are adjacent to the sites mutated in our selected Spy variants . Alternatively , the Spy mutations could increase chaperone efficacy in other less direct ways . For example , they could map to sites distant from the active site of the chaperone and exert their beneficial action through allosteric effects . To help distinguish between these possible models and to better understand how Spy interacts with its clients , we decided to map the site ( s ) with which Spy binds the client protein Im7 . To achieve this , we: ( 1 ) examined the effects of Im7 binding on hydrogen-deuterium exchange in Spy , ( 2 ) investigated the proteolytic sensitivity of the chaperone in the presence and absence of Im7 , and ( 3 ) crosslinked Spy to its client . Hydrogen-deuterium exchange at individual peptide bond amides is determined by the protection of amides from solvent , either due to maintenance of secondary and/or tertiary structure or client binding . To ascertain the effects of client protein binding on deuterium exchange , we compared the level of hydrogen-deuterium exchange of free Spy with that of a Spy-Im7 complex using a mass spectrometry approach . We incubated Spy with a 40-amino acid peptide derived from Im7 L53A I54A ( residues 7–45 ) , which binds Spy with a 2 . 6 µM KD ( see ‘Materials and methods’ for details of peptide generation and selection ) . Using 10s exchange times , we obtained evidence that ∼10 Spy amides become more protected upon Spy-Im7 complex formation compared to free Spy ( Figure 2—figure supplement 1A ) . Substantial changes in protection were mapped to the residues located in N- and C-terminal regions of Spy ( Figure 2—figure supplement 1B ) . We were able to localize the improved Spy protection in the presence of client to several specific residues: T5 , H16 , A37 , Q114 , F115 , F119 , and E125 . The protection includes not only the flexible N and C termini ( residues 1–28 and 125–138 ) , which are not present in the crystal structure , but extends into the α1 and α4 helices as well ( Figure 2 ) . Increased protection from hydrogen-deuterium exchange in these regions could possibly indicate involvement of these Spy residues in the interaction with Im7 or could imply the folding of these flexible regions upon client binding , or a combination of both . 10 . 7554/eLife . 01584 . 008Figure 2 . Hydogen-deuterium exchange and limited proteolysis reveal potential hot spots on Spy for Im7 L53A I54A or Im7 7-45 peptide binding . Hydrogen atoms on the backbone amide bond that are protected upon addition of Im7 7-45 are shown as magenta spheres . Peptide bonds of Spy protected from trypsin upon addition of Im7 L53A I54A are shown as yellow and blue spheres , with yellow representing carbon atoms and blue representing nitrogen atoms . DOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 00810 . 7554/eLife . 01584 . 009Figure 2—figure supplement 1 . Hydrogen-deuterium exchange analysis of Spy and the Spy-Im7 7-45 complex by electron capture dissociation fourier transform ion cyclotron resonance mass spectrometry ( ECD-FTICR-MS ) . ( A ) The hydrogen-deuterium exchange pattern for free Spy ( black ) differs from that observed for the Spy-Im77-45 complex ( red ) . The complex shows increased protection against exchange compared to free Spy . Representative c- and z- ion series observed for Spy and Spy-Im7 7-45 indicate that there is a significant difference in exchange within the N- and C-terminal regions . ( B ) Differential deuteration plot of Spy and Spy-Im7 7-45 residues based on mass differences between pairs of corresponding c- ( open circles ) and z- ( closed circles ) fragment ions of Spy and Spy-Im7 7-45 . Most of the protection changes between Spy and Spy-Im7 7-45 can be attributed to residues located in N- and C-terminal regions . Sharp drops ( c-series ) or rises ( z-series ) on the plot indicate Spy residues that form new hydrogen bonding upon complex formation with Im7 . DOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 00910 . 7554/eLife . 01584 . 010Figure 2—figure supplement 2 . Limited proteolysis reveals potential Im7 binding sites in Spy . The susceptibility of Spy's lysine and arginine residues to digestion by trypsin was measured in the presence or absence of Im7 L53A I54A . Protein samples were incubated with trypsin at a constant 100:1 mass ratio ( protein: protease ) . At various times , aliquots were withdrawn , quenched , and analyzed by mass spectrometry . The residue numbers of trypsin-cleavable sites are shown along with their location in Spy's secondary structure ( indicated by the cartoon ) . The bars indicate the time at which cleavage at that residue took place . A missing bar indicates that cleavage was not seen within the 8-min time frame . The unstructured termini are more accessible to trypsin in general , and many sites show apparent increased digestion in the presence of client compared to Spy alone . The interpretation of increased cleavage of these sites is not straightforward . The increased digestion might reflect some real structural rearrangement in Spy upon client binding so that these residues are more exposed to trypsin . Alternatively , they may not be involved in substrate binding and their increased digestion might be an artifact due to a higher total mass of trypsin , which was used to accommodate the addition of client . We were not able to distinguish these two possibilities . On the other hand , the sites showed decreased cleavage despite an increased amount of trypsin are more easily interpreted . Decreased cleavage suggests that they are protected , either by the substrate protein , or by other regions in Spy , which further implies a conformational change upon substrate binding . Cleavage at residues 61 , 84 , 102 , 113 , and 122 ( circled ) is significantly delayed when the client Im7 L53A I54A is present . We conclude that R61 , K84 , K102 , K113 , and R122 are sites but not necessarily the only sites affected by the binding events . DOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 010 As a complementary approach to probe the Im7 binding site in Spy , we used a limited proteolysis assay with a mass spectrometric readout to characterize the exposed and buried regions in Spy before and after Im7 binding as suggested by their accessibility to the protease trypsin . Proteolytic sites in Spy that show altered trypsin susceptibility may either be directly involved in Im7 binding or be near the Im7 binding site . Note that it is also possible that Im7 binding induces a significant conformational change or change in flexibility in Spy that alters the trypsin susceptibility of certain residues . In the absence of Im7 , the flexible N terminus and to a lesser extent the flexible C terminus of Spy are more accessible to trypsin than the structured regions ( Figure 2—figure supplement 2 ) . In the presence of Im7 , a number of sites including R61 , K84 , K102 , K113 , and R122 show significant protection compared to free Spy ( Figure 2 ) . Notably , changes in trypsin susceptibility occur on both the inside and outside of the cradle due to the thinness of the Spy molecule . These susceptibility changes suggest that either client binding occurs over large portions of Spy or that client binding involves major conformational changes or changes in flexibility , or a combination of these factors . Combining these observations with the deuterium exchange results suggests that Im7 peptide binding affects a relative large area on the Spy surface , especially the rim regions ( α4 helix and the N-terminus of the α2’ helix ) and the tips ( N-terminus of the α1 helix , C-terminus of the α3 helix , and the N-terminus of the α4 helix ) of the Spy cradle ( Figure 2 ) . To further map the position of the client-binding site on Spy , we performed crosslinking analysis . Crosslinking provides information about the distances between two cross-linked residues as determined by the length of the spacer in the crosslinking reagent . Identification of the crosslinked sites on a protein complex thus provides spatial information and distance constraints for the two amino acid residues that are crosslinked . We performed crosslinking experiments on Spy and the peptide composed of residues 7-45 from Im7 . Crosslinking was done using our recently developed isotopically-coded collision-induced dissociation CID-cleavable affinity-purifiable amine-reactive 14 Å length crosslinker CyanurBiotinDimercaptoPropionylSuccinimide ( CBDPS-H8/D8 ) ( Petrotchenko et al . , 2011 ) and the newly developed azidobenzoicacidsuccinimide ( ABAS-12C6/13C6 ) , an isotopically-coded photo-reactive 7 Å length crosslinker . Given the relatively long span length of the CBDPS crosslinkers and the flexibility of the crosslinked side chains , we were not surprised to find multiple CBDPS Lys–Lys crosslinks ( Figure 3; Table 3 ) . The most frequently crosslinked residues are K18 and K20 on the unstructured N terminus of Spy . T5 and H16 in this region are also implicated in peptide binding through changes in deuterium protection , suggesting that this flexible N terminus of Spy might be involved in client interaction . 10 . 7554/eLife . 01584 . 011Figure 3 . Crosslinked residues mapped onto the crystal structure of Spy . Spy residues that were found crosslinked to Im7 7-45 peptide are shown in green . Crosslinking with CBDPS-H8/D8 implies a short distance between the N terminus of Im7 and Spy residues; these include Spy K20 , K39 , K47 , K54 , K130 , and K132 . Residues that can be crosslinked with CBDPS also include Spy K18 to Im7 K20 , Spy K20 to Im7 K20 , and Spy K30 to Im7 K43 . Using ABAS , we identified crosslinking between the N terminus of Spy and E21 of Im7 . Zero-length crosslinking using PICUP and EDC reagents identified contacts between Spy Y104 and Im7 Y10 , and Spy K39 and Im7 E12 , respectively . A summary of all identified crosslinks is provided in Table 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 01110 . 7554/eLife . 01584 . 012Figure 3—figure supplement 1 . Distribution of lysine residues on the surface of Spy , lysine residues are copper colored . DOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 01210 . 7554/eLife . 01584 . 013Table 3 . Spy-Im7 crosslinksDOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 013Mass ( Da ) Rt ( min ) Δ ( ppm ) Pr 1SEResSequence*Pr2SEResSequence*CLEnz1853 . 8760916 . 470 . 4Spy-12--SADTTTAAPADAK†PIm7212421KEIEKEABASTr1835 . 7796920 . 770 . 6Spy151818MMHHKGIm7202520LKEIEKENCBDPSPK1447 . 6019522 . 430 . 6Spy161818MHHKGIm7202320LKEIEKCBDPSPK1704 . 7395419 . 180 . 5Spy161818MHHKGIm7202520LKEIEKENCBDPSPK2328 . 0558721 . 481Spy162420MHHKGKFGPHQIm7202520LKEIEKENCBDPSPK1567 . 6817821 . 40−0 . 2Spy171818HHKGIm7202520LKEIEKENCBDPSPK3673 . 6420653 . 63−0 . 4Spy193020KGKFGPHQDMMFKDIm7-20--SISDYTEAEFVQLLKECBDPSTr1781 . 7935734 . 95−0 . 9Spy192420KGKFGPHQIm7192320LLKEIEKCBDPSPK2038 . 9297031 . 40−0 . 1Spy192420KGKFGPHQIm7192520LLKEIEKENCBDPSPK1925 . 8466728 . 32−0 . 6Spy192420KGKFGPHQIm7202520LKEIEKENCBDPSPK1377 . 6108333 . 100 . 3Spy293130MFKDLIm7424543FVKIT-CBDPSPK3793 . 8002556 . 15−0 . 5Spy314339KDLNLTDAQKQQIREIm7-20--SISDYTEAEFVQLLK†ECBDPSTr3266 . 6872345 . 131 . 1Spy314339KDLNLTDAQKQQIREIm7-2012-SISDYTEAEFVQLLKEEDCTr3112 . 4355153 . 530 . 1Spy445047REIMKGQRDIm7-20--SISDYTEAEFVQLLKECBDPSTr3649 . 6933351 . 73−0 . 7Spy516154RDQMKRPPLEERRIm7-20--SISDYTEAEFVQLLKECBDPSTr1838 . 8448831 . 500 . 5Spy546154MKRPPLEERRIm7-8--SISDCBDPSPK2958 . 5342848 . 480 . 1Spy103112104KIYNILTPEQKKIm7-2010-SISDYTEAEFVQLLKEPICUPTr1699 . 7718331 . 88−0 . 3Spy123130130RLTERPAAKGIm7-8--SISDCBDPSPK2822 . 3325455 . 33−0 . 5Spy127132130RPAAKGKMIm7-20--SISDYTEAEFVQLLKECBDPSTr3055 . 3718157 . 27−1 . 6Spy131138132KGKMPATAE-Im7-20--SISDYTEAEFVQLLKECBDPSTrRt: retention time; Δ: mass error for crosslink assignments; Pr: protein; S , E: starting and ending amino acid residues ( sequence numbers ) of the crosslinked peptides , respectively; Res: crosslinked residue ( sequence number ) within corresponding peptide; CL: crosslinking reagent used; Enz: digestion enzyme used . The crosslinked residues are bolded and underlined in the sequences . *The residues shown before and after the sequences are the preceding and following residues of the peptide sequence . They are shown to illustrate digest specificity . †The N-terminal serine was introduced from the Sumo fusion constructs , which were used for Spy or Im7 purification ( Quan et al . , 2011 ) . The symbol ‘−’ is used to indicate this N-terminal serine . The shorter the crosslinking reagent , the more precise and definitive the structural information that crosslinking analysis can provide . To obtain such short-distance constraints for the Spy-Im7 complex , we additionally performed crosslinking using two zero-length crosslinking reactions: tyrosine reactive , Photo-Induced Cross-Linking of Unmodified Proteins ( PICUP ) ( Bitan et al . , 2001 ) and carboxyl/amine reactive 1-ethyl-3- ( 3-dimethylaminopropyl ) carbodiimide ( EDC ) . Using these zero-length reagents , we were able to confidently detect and identify additional short distance Spy-Im7 crosslinks between Y104 of Spy and Y10 of Im7 , and between K39 of Spy and E12 of Im7 ( Table 3; Figure 4 ) . Notably , nearly all crosslinks identified in all four crosslinking reactions occurred on the concave side of Spy , with a few occurring on the rim of the cradle ( Figure 3 ) . Essentially no crosslinks were obtained on the convex side of Spy despite the abundance of lysines on the convex side ( Figure 3—figure supplement 1 ) , providing evidence that peptide binding occurs on the interior concave surface of the Spy homodimer ( Figure 3 ) . The zero-length PICUP crosslinks to Y104 independently suggests that the concave surface is the interface at which Im7 binds . 10 . 7554/eLife . 01584 . 014Figure 4 . Model showing the position of the Im7 7-45 peptide bound on the concave surface of Spy . Mutations that increase the specific activity of Spy are shown in red . Residues on Spy that are crosslinked to Im7 residues by PICUP or EDC are shown in cyan . The enlargement shows the position of residues Q49 , H96 , and Q100 . Mutating these residues to leucine increases the hydrophobic interaction with Y10 , F15 , and I44 on Im7 . DOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 01410 . 7554/eLife . 01584 . 015Figure 4—figure supplement 1 . NMR spectrum of the Im7 7-45 peptide . The average 1HN chemical shift ( 7 . 91 ppm ) is substantially shifted upfield from the expected random coil value for this sequence ( 8 . 14 ppm ) ( Wang and Jardetzky , 2002 ) . This difference verifies that the αhelical nature of Im7 is at least partially maintained in this fragment , suggesting that the αhelical model of Im7 7-45 used in the docking procedure is valid . DOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 01510 . 7554/eLife . 01584 . 016Figure 4—figure supplement 2 . Models of the Spy-Im7 7-45 complex . ( A ) The electrostatic potential of Spy mapped onto its solvent accessible surface . The surface is colored according to its electrostatic potential calculated with the Adaptive Poisson-Boltzmann Solver ( APBS ) plugin in Pymol ( Baker et al . , 2001 ) . Red indicates a more negative potential ( −10 kTe−1 ) and blue a more positive potential ( +10 kTe−1 ) . The Im7 7-45 peptide ( cyan ribbon ) is shown docked into the pocket of Spy . The polar and acidic residues in Im7 as well as the location of the ‘P1’ and ‘P2’ hydrophobic patches of Spy are labeled . The Poisson-Boltzmann ( PB ) calculation was performed using the PARSE force field for the atomic charges and radii that is available with the PDB2PQR server ( http://nbcr-222 . ucsd . edu/pdb2pqr_1 . 8/ ) . The charges and radii were then used as input for the APBS Pymol plugin along with the default parameters ( 0 . 15 M ion concentration , 310 K , a 1 . 4 Å solvent radius , and dielectric constants of 2 . 0 and 78 . 0 for the protein and solvent , respectively ) . A similar result is obtained when directly viewing isosurfaces through the charge distribution grid resulting from the PB calculation . The Im7 7-45 peptide ( cyan ribbon ) is shown docked into the positively charged concave surface of Spy . The polar and acidic residues in Im7 are labeled . ( B ) E14 , Q17 , E21 , E25 and T30 from α1 of Im7 associate with basic residues R50 , R55 , and R61 from Spy monomer A to form an extensive salt bridge and hydrogen-bonding network , with an extra contribution from R122’ from Spy monomer B . ( C ) D31 , D35 , and E39 on Im7 α2 are recognized by nearly the same set of basic residues , including H96’ , R50’ , R55’ , andR61’ from Spy monomer B and R122 from Spy monomer A . ( D ) Residues composing the hydrophobic pocket that Y10 from Im7 is buried into are labeled . The distance between Spy Y104 and Im7 Y10 is shown . ( E ) Hydrophobic interaction at the N- and C-terminal tips of Im7 . ( F ) Electrostatic interaction between Im7 Glu12 and Lys39 and Arg43 of Spy . The distance between Spy Lys39 and Im7 Glu12 is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 016 Using a hierarchical approach that consists of three steps ( docking pose generation , decoy clustering , and structure refinement ) ( see ‘Materials and methods’ ) , we built a tentative , theoretical model of the Spy-Im7 complex by docking the Spy structure ( PDB ID: 3O39 ) with the Im7 7-45 peptide structure ( modeled using I-TASSER [Zhang , 2008; Roy et al . , 2010] ) . Note that the distance constraints we obtained from the crosslinking study were NOT applied during the docking and refinement process . However , of the top 10 models with the lowest energy scores , six fit the zero length crosslinker data ( i . e . , the critical constraints for the two pairs of residues that crosslinked with the zero length crosslinkers: Spy Y104 and Im7 Y10 , Spy K39 and Im7 E12 ) , with atomic contact distances ranging between 2 . 5 and 6 . 5 Å . These models are very similar with average pair-wise RMSDs of about 2 Å . Thus , the docking analysis provides evidence that the crosslinking experiments were sampling an energetically plausible complex . For further analysis , we selected the model with the smallest atomic contact distance for these two pairs of residues found to interact by zero-length crosslinkers . Im7 7-45 was modeled to adopt a two-helix hairpin conformation and to fit into the concave face of the Spy dimer ( Figure 4 ) , consistent with the15N HSQC NMR spectrum of 15N-labeled Im7 7-45 , which suggests that although it is largely unfolded in solution , it is perhaps partially biased towards α-helix formation ( Figure 4—figure supplement 1 ) . Given that this model is strongly constrained by only two pairs of residues that were found to interact using the zero-length crosslinkers , it is best regarded as a very tentative and theoretical model . Never-the-less it is consistent with our experimental data and helps in our interpretation of it . The contacts predicted between Im7 and Spy in this tentative model are moderately extensive , burying a surface of ∼1502 Å2 , which is consistent with medium binding affinity ( KD = 2 . 6 µM ) ( Chen et al . , 2013 ) . The interactions between Im7 and Spy in the model are mediated by a combination of electrostatic and hydrophobic interactions . The two helices of Im7 contain a number of acidic and polar residues that dock into the positive-charged concave surface of Spy ( Figure 4—figure supplement 2A–C ) . The α1 helix of Im7 is surrounded primarily by basic residues from one Spy monomer , whereas α2 of Im7 is symmetrically coordinated by the basic patch from the other Spy monomer . The N-terminal loop of Im7 forms extensive van der Waals contacts with a cluster of hydrophobic residues at the tip of the hydrophobic patch P1 of Spy in our tentative model . The most prominent residue is Im7 Y10 , which is buried into a hydrophobic pocket surrounded by L32 , L34 , I42 , M46 , I103 , Y104 , L107 , and F115 of Spy ( Figure 4—figure supplement 2D ) . Satisfyingly , these residues include those that make up the P1 patch shown in Figure 1 and some of the hydrophobic residues at the tip of P1 ( Quan et al . , 2011 ) . The distance between Im7 Y10 and Spy Y104 is 4 . 5 Å in this model , consistent with the close distance constraint that was defined by the zero-length crosslinking via PICUP . Similarly , Im7 Glu12 and Spy Lys39 form a direct salt bridge , consistent with their interaction via the zero-length crosslinker EDC . This model also helps explain the effects of at least some of our beneficial Spy mutations . Polar-to-apolar Spy mutations ( Q49L , H96L , Q100L ) may enhance the hydrophobic interaction that Spy has with Im7 through the interaction with Y10 , F15 , and I44 of Im7 by expanding and partially fusing the hydrophobic patches P1 and P2 on Spy ( Figures 1 and 4 ) . To further understand why these super-Spy variants enhance the expression and presumably the in vivo stability of Im7 L53A I54A , we measured their interaction affinities and kinetics with Im7 L53A I54A using bio-layer interferometry ( BLI ) . Biotinylated Im7 L53A I54A was immobilized on the streptavidin coated sensor tip . Binding of Spy to Im7 alters the thickness of the molecular layer on the tip surface , which triggers a change in the spectrum signal ( Abdiche et al . , 2008 ) . Thus , we can monitor the binding in real time to measure the association and dissociation rates of the two proteins . In order to get accurate kon and koff rates we found it is vital to use a substrate that is soluble and not bound to the tip as an aggregate . Unfortunately , most chaperone substrates such as aldolase rapidly aggregate when placed under conditions where they are chaperone substrates ( i . e . , at least partially unfolded ) making it very difficult to accurately determine kon and koff rates . Im7 L53A I54A has the very fortunate properties of being not only a clear in vivo substrate of Spy ( indeed Spy was discovered by the ability it has to enhance the yield of folded Im7 L53A I54A ) but also soluble in solution . This nicely behaved , soluble chaperone substrate bound to the tips in a reproducible manner and allowed us to determine the kon and koff rates for the chaperone . All of the Spy variants we tested showed smaller values for both kon and koff compared to wild-type Spy ( Table 2 ) , suggesting that they both bind and release clients more slowly . Overall , the decreases in the koff rates are more dramatic than the decreases in the kon rates ( for instance , Q100L shows a ∼three-fold decrease in kon and a ∼16-fold decrease in koff ) . As a result , all selected Spy variants except F115I show a significantly increased apparent affinity for Im7 L53A I54A , up to 5 . 8 -fold . Note that Q100L , the mutation present in 74% of the variants that answered our selection , has the largest effect on both kon and koff of any of the variants tested . It is thus a good possibility that the increase in the ability of at least some of our Spy variants to stabilize Im7 in vivo is at least in part due to their increased affinity for Im7 . Spy is a very thin molecule , almost entirely lacking a hydrophobic core . This unusually thin nature of Spy is presumably under genetic selection and of functional significance . It also may allow for conformational changes during Spy’s chaperone cycle . Consistent with this hypothesis , major changes in the fluorescence of environmentally sensitive probes attached to various locations in Spy were observed to occur upon client protein binding ( Quan et al . , 2011 ) . One possibility is that our mutations act to alter the flexibility and stability of Spy and , in so doing , alter its chaperone activity . To investigate if the mutations in Spy resulted in changes in the protein’s stability , we performed thermal denaturation experiments to measure the free energy of unfolding . Five of the seven super-Spy variants were significantly less thermodynamically stable than the wild-type protein ( Table 2 ) , with mutant L32P being the most destabilized ( ΔΔGNU = −3 . 25 kcal mol−1 ) . The Q25R and Q49L variants had stabilities that were indistinguishable from wild type . When we plotted thermodynamic stability vs in vivo chaperone activity , we found a significant ( R2 = 0 . 46 ) negative correlation ( Figure 5A ) . Thus the less thermodynamically stable the Spy protein is , the better it tends to function as a chaperone in vivo . There is also a significant correlation ( R2 = 0 . 51 ) between the thermodynamic stability of these Spy variants and their KD of binding Im7 L53A I54A , with the least stable variants showing the tightest binding ( Figure 5B ) . One possibility is that decreased stability results in increased flexibility , which may then result in improved chaperone activity by allowing for more adaptive and tighter binding to client proteins . 10 . 7554/eLife . 01584 . 017Figure 5 . The thermodynamic stability of the Spy variants is inversely correlated with their chaperone activity and their affinity for its client protein . ( A ) The in vivo chaperone activity of Spy variants is expressed as the normalized relative MIC of the strains expressing the variants plus the Bla-Im7 L53A I54A biosensor ( SQ2068 , LW53-59 ) . ( B ) The binding activity of Spy variants towards Im7 L53A I54A is expressed as their dissociation constant ( KD ) to the client; smaller values indicate tighter binding . DOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 01710 . 7554/eLife . 01584 . 018Figure 5—figure supplement 1 . The more thermodynamically stable Spy variants are less flexible . The flexibility of Spy variants is interpreted as the percentage of protected protons in the deuterium exchange analysis measured at 25°C and is inversely correlated with the stability of these Spy variants in the absence of client . This correlation is abolished when Spy variants were bound to the client peptide Im7 7-45 . DOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 01810 . 7554/eLife . 01584 . 019Figure 5—figure supplement 2 . Calculation of relative Spy activity in preventing α–lactalbumin ( α -LA ) aggregation . ( A ) Typical aggregation curves of reduced unfolded α–LA in the absence or presence of wild type Spy added at four different Spy: α–LA ratios . ( B ) 22 Spy concentrations were used in the range of 0–2 Spy: α–LA ratios to produce a series of aggregation curves , from no inhibition to complete inhibition of aggregation . These concentrations were prepared from serial dilution of three starting concentrations . Light scattering endpoints at 300 , 400 , 500 , and 600 s were obtained from the aggregation curves and plotted against Spy: α–LA ratios to make four standard curves . Shown in panel ( B ) is the standard curve generated at 400 S , which can be fit with an exponential equation: y = 0 . 004 + 0 . 108* exp ( −8 . 46 × x ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 01910 . 7554/eLife . 01584 . 020Figure 5—figure supplement 3 . Calculation of relative Spy activity in preventing aldolase aggregation . Six Spy concentrations were used in the range of 0–1 Spy: aldolase ratios to produce a series of aggregation curves from no inhibition to nearly complete inhibition of aggregation . All the ratios were repeated at least three times except for 0 . 1:1 and 1:1 which were done only twice and once respectively . A 0 . 3:1 ratio was chosen to evaluate the relative chaperone activity of the Spy mutants to the wild type since this ratio is locating in the middle range of the standard curve . The relative activity of the mutants was obtained by determining the average slope of the aggregation curves obtained for the mutants and interpolating this number back into the standard curve to calculate the ratio of wt Spy: aldolase that would give with an activity comparable to the mutant . All the mutants are significantly better than the wild type in preventing the aggregation of Aldolase denatured by guanidine . The calculation is diagramed in the figure for the F115I mutant . Each of the mutants were assayed at least three times . The most active mutant Q25R completely suppressed aggregation at a ratio of 0 . 3:1 putting it outside of the range of the standard curve , so for this mutant the assay was repeated at a ratio of 0 . 1:1 . The fitted standard curve equation is y = 29 . 0861x2 − 65 . 0189x + 37 . 9414 and the R2 = 0 . 9967 . DOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 02010 . 7554/eLife . 01584 . 021Figure 5—figure supplement 4 . Calculation of the relative activity of Spy variants in the refolding of denatured aldolase . The change in absorption at 340 nm of the aldolase substrate β-NADH upon hydrolysis by aldolase was used to determine aldolase activity . We measured the ability of Spy to refold aldolase by measuring how much aldolase activity was recovered from a denatured aldolase preparation following 5 min refolding in the presence and absence of Spy . Various concentrations of wild type Spy were first used in order to establish a standard curve . The standard curve was established by fitting all the relative refolding yields of nine different WT Spy ratios ranging from 0:1 to 1 . 75:1 . The curve was fit using the following formula: y = 1 . 27711 ( x + 0 . 02196 ) 0 . 18859 , R² = 0 . 998 . A Spy: aldolase ration of 0 . 25:1 was found to be located in the middle region of the standard curve so this amount of the Spy variants was used in order to establish if the variants were superior or inferior to wild type Spy . All were superior , although Q25R only marginally so . In order to establish activities of the Spy variants their influence on the refolding yields were related to the activity shown with wild type Spy . In the example shown 0 . 25:1 Spy H96L:aldolase performed as well as 0 . 474:1 Spy WT:aldolase , so Spy H96L is calculated to be 1 . 9-fold more active than is Spy WT . DOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 021 As another measure of apparent flexibility , we performed deuterium exchange analysis on these Spy variants in the presence and absence of client . Without the Im7 7-45 peptide , the apparent flexibility of Spy variants as measured by the percentage of protected protons inversely correlates with their thermodynamic stability measured at the same temperature ( 25°C ) ( Figure 5—figure supplement 1 ) . Note that several of the super-Spy chaperones including L32P , F115L , and F115I show either zero or a very small number of protected protons ( 0 , 0 , and 5 , respectively ) , implying a very high degree of disorder in the absence of client proteins . On the other hand , all of our Spy variants show almost the same level of protection when the Im7 7-45 peptide is added , suggesting either that they achieve a similar level of order upon interaction with the client , or that the vast majority of Spy is directly protected via contact with Im7 . These results further imply that binding to the client may induce the folding of the very unstable Spy variants such as L32P .
Although molecular chaperones assist in the folding of a vast number of cellular proteins , including many that are linked to disease states , the mechanism by which they accomplish this feat is still not completely clear ( Horwich et al . , 2009; Kalia et al . , 2010; Hartl et al . , 2011 ) . Defining the binding interface between a chaperone and a client would undoubtedly be very valuable in understanding chaperone action , but these binding sites remain poorly defined . Various regions on the small heat shock proteins , for instance , have been implicated in client binding , including both the N and C termini and sites within the core-alpha crystalline domain ( Jaya et al . , 2009; Basha et al . , 2013 ) . Unfortunately , we only have a few structures of chaperone-client complexes ( Martinez-Hackert and Hendrickson , 2009; Zhu et al . , 1996; Zahn et al . , 2013; Bracher et al . , 2011 ) and importantly , these do not resolve how a chaperone interacts throughout its cycle . To better understand how chaperones bind to proteins , we decided to take a genetic approach with the aim of improving the ability of a chaperone called Spy to protect a poorly folding protein from degradation in vivo . In our selection , we obtained several Spy variants that appear to act as improved chaperones by expanding and partially fusing two hydrophobic patches present on the interior of the cradle-like structure of Spy . By using crosslinking , proteolytic sensitivity , and deuterium protection experiments , we provide additional evidence that implicates these regions in client binding . Our super-Spy variants demonstrate tighter client binding , and we observe a correlation between the in vitro dissociation constants of these variants to the client protein Im7 L53A I54A and their ability to stabilize Im7 L53A I54A in vivo ( Figure 6 ) . This implies that Spy’s binding affinity for Im7 L53A I54A is important in determining its chaperone activity . Tighter binding could act to decrease the steady state concentration of presumably aggregation-prone or protease sensitive Im7 folding intermediates , thus increasing the Im7 level in the cell . Variants of Spy that increased the affinity of clients beyond a certain point may not be obtained by our in vivo selection because they would be expected to fail to release their clients in a timely fashion . Further increases in affinity are expected to be counterproductive . For aldolase aggregation inhibition Q25R , for instance , is the most effective , but it is only very marginally better than wild type Spy in refolding aldolase , perhaps partly because it fails to release aldolase at an optimal rate . Client and variant specific changes in affinity may explain why there appears to be no good relationship between the activity that one specific variant shows with the various specific client proteins tested . 10 . 7554/eLife . 01584 . 022Figure 6 . Kinetic parameters characterizing the interaction of Spy variants with the client protein Im7 L53A I54A . A linear correlation is seen between these in vitro parameters and the in vivo activity of these variants towards the same client ( expressed as normalized MIC ) . Spy variants with better activity in vivo bind and release client slower and have an overall tighter affinity for the client . DOI: http://dx . doi . org/10 . 7554/eLife . 01584 . 022 The genetic selection we performed was specifically designed to enhance the ability of Spy to stabilize a particular mutant of Im7 ( Im7 L53A I54A ) . This selection procedure was expected to result in identifying client-optimized Spy variants specialized for the unstable Im7 L53A I54A but with lower levels of chaperone activity on other clients . A previous attempt to evolve the chaperone GroEL for instance , succeeded in generating highly client specific variants that were defective in the folding of other clients ( Wang et al . , 2002 ) . We were therefore surprised , but pleased , to see that all the Spy variants appeared to function more efficiently than wild-type Spy in preventing aggregation of aldolase and α-lactalbumin in vitro . Based on our biochemical and biophysical evaluation of the super-Spy variants , there appear to be multiple factors that result in the increased chaperone activity . In addition to increasing binding site hydrophobicity , most of our super-Spy variants decrease Spy’s thermodynamic stability and increase the level of disorder that Spy shows in the absence of client proteins . Two of the Spy variants , L32P and F115L , show no protected amide protons whatsoever in the absence of the client , implying extreme levels of flexibility prior to client binding . These data suggest that Spy flexibility is important for chaperone function , presumably by facilitating client binding . The six long coiled-coil tentacles in the jellyfish-like chaperone prefoldin , for instance , are flexible , enabling it to adjust its central cavity and capture various client proteins ( Siegert et al . , 2000 ) . Many other chaperones including GroEL , DnaK , and the small HSPs contain regions of disorder , which have been proposed to aid in client recognition and chaperone function , although the exact roles of these disordered regions are unclear ( reviewed in Bardwell and Jakob ( 2012 ) ) . There are a number of chaperones that are conditionally disordered and that are active as a chaperone in the disordered state . These include HdeA , a chaperone that is activated on the protein level by acidic pH , and Hsp33 , a chaperone that is activated by oxidation ( Bardwell and Jakob , 2012; Foit et al . , 2013 ) . Disorder is also an emerging theme , not only among chaperones but among many hub proteins , which are capable of binding to multiple unrelated targets ( Bardwell and Jakob , 2012 ) . It has been proposed that proteins utilize these disordered regions in the adaptive recognition of their various binding partners ( Oldfield et al . , 2008 ) ; yet , apart from a few exceptions ( Brzovic et al . , 2011 ) , defining the precise roles that disorder plays in these promiscuous binding proteins has proved to be difficult . In general , the majority of mutations in proteins are destabilizing ( Foit et al . , 2009 ) , so if destabilization was the only requirement for enhanced chaperone ability , a wider range of mutations might be expected to destabilize Spy and thus increase its flexibility . If the only requirement is to increase the hydrophobicity of the P1 and P2 region , one might also expect a larger variety of mutants to have answered the genetic selection . The very narrow range of substitutions that answers our selection with 74% of our variants containing the Q100L substitution strongly suggests that these variants may be having more specific effects beyond their effects on surface hydrophobicity and flexibility . Several of the substitutions do not appear to result in a substantial gain in surface hydrophobicity . One alteration , F115L/or I , replaces one hydrophobic residue with another , and two of the substitutions , Q25R and L32P , apparently result in a decrease in hydrophobicity , though not specifically for the patches P1 and P2 . Despite these exceptions , it is striking that the majority of the variants isolated do exhibit changes in flexibility and surface hydrophobicity , which goes at least part way toward explaining their ‘super-Spy’ activity . If the variants we observed do generally enhance Spy’s chaperone activity , then why has evolution not already come up with this solution ? In short , it has . Inspection of an alignment of Spy homologues ( Figure 1—figure supplement 2 ) shows that although evolution has , for some of the residues commonly made the exact same substitutions as have emerged from our selection , and in other cases it has found very similar answers . The F115L substitution is very commonly found in Spy homologue , as is Q49L . The precise substitution that we found , H96L has not been observed in the sequences we compared , but remarkably , the chemically very similar residue M is actually the most common residue found at position 96 . Evolution generally acts to optimize rather than maximize protein activity . One can imagine that some organisms require more or less Spy activity . The optimal level of Spy activity selected for in E . coli is apparently less than seems to be required in other organisms . That other organisms have independently obtained the same ‘Super Spy’ substitutions that we found is testimony of the power of our selection for in vivo stability .
Plasmid pCDFTrc-spy was used as the template for the construction of the Spy mutant library in an error-prone PCR reaction ( McCullum et al . , 2010 ) . In order to introduce random mutations only into the mature protein coding region of Spy , we first introduced a BglI site between the signal sequence and the mature protein coding region of Spy and then amplified only this region using the forward primer containing the BglI site ( 5′CGC GGC CAA CCT GGC CCA TGC C3′ ) and the reverse primer bearing the EcoRV site that is located at the 3′ end of the Spy gene ( 5′GGC CGA TAT CCA ATT GAG ATC TGC CAT ATG GGA TCC TTA3′ ) . The linear fragments were digested with BglI and EcoRV and then re-ligated into the pCDFTrc vector ( digested with BglI and EcoRV as well ) . We used four different concentrations of template DNA ( 0 . 05 , 0 . 5 , 5 , and 50 ng ) to obtain a broad spectrum of mutation frequencies as the lower concentrations of the template are expected to give rise to higher mutation rates . The mutation rates were 1 . 8 , 5 . 0 , 5 . 3 , and 4 . 3 mutations per 1000 nucleotides using 50 , 5 , 0 . 5 , and 0 . 05 ng of template DNA , respectively . Sequencing of 96 random clones ( without selection ) revealed a broad spectrum of transition and transversion mutations with no significant nucleotide bias . We then electroporated the resulting 4 independent ligation reactions into competent cells of strain SQ765 ( genotype: MG1655 , ΔhsdR ) ( Table 1 ) to generate four independent mutant plasmid libraries , each containing ∼106 clones . We then extracted the plasmid from these four independent libraries that had been constructed in SQ765 and transformed the plasmids into the spy and ampC null strain SQ2041 , which contains pBR322bla::GSlinkerIm7 L53A I54A . The transformants were plated onto plates containing 4 mg/ml penicillinV , supplemented by 0 . 1 mM IPTG to induce Spy , and incubated at 37°C overnight . We picked 159 penicillin V resistant colonies and streaked them twice to obtain single colonies . These cells were further tested for penicillin V resistance by spot titration onto plates with increasing amounts of penicillin V ( 3 , 4 , and 5 mg/ml ) supplemented with 0 . 1 mM IPTG , using strain SQ2068 as a negative control . 65 of the 159 strains consistently showed resistance to the three penicillin V concentrations; plasmids were extracted from these 65 strains and sequenced to identify mutations in the mature protein-encoding region of Spy . Activities of the Spy variants were assayed for their ability to prevent the aggregation of two clients: chemically denatured aldolase and reduced denatured α-lactalbumin ( α-LA ) , as well as their ability to assist the refolding of chemically denatured aldolase . The aggregation of bovine α-lactalbumin ( type III from Sigma Aldrich , St . Louis , MO ) was initiated by reducing its disulfide bonds with DTT at 25°C as previously described ( Kulig and Ecroyd , 2012 ) in the presence or absence of Spy . Each reaction contains 50 µM of α-lactalbumin , 0-100 µM of Spy , and 20 mM DTT in a buffer composed of 50 mM sodium phosphate , 100 mM sodium chloride , and 5 mM EDTA , pH 7 . 0 . 100 µl of each reaction solution was incubated in an acrylic UV transparent flat-bottom 96-microwell plate ( Corning , Corning , NY ) sealed with a transparent film ( Denville Scientific , Inc . South Plainfield , NJ ) . The light scattering light due to the aggregation of reduced α-lactalbumin was monitored at 360 nm using a Synergy HT Multi-Mode Microplate Reader ( Biotek , Winooski , VT ) with readings taken every 15 min for 10 hr at 25°C following an initial period of medium speed shaking for 30 s before each reading . Spy variants were added to the aggregation reaction at three concentrations: 5 , 7 . 5 , and 10 µM to make the final Spy: α-lactalbumin ratios equal to 0 . 1 , 0 . 15 , and 0 . 2 , respectively . Light scattering data at four endpoints ( 300 , 400 , 500 , and 600 S ) were used to calculate the relative activities of different Spy variants based on standard curves generated with 22 different Spy: α-lactalbumin ratios . The average of the 12 relative activity values ( three concentrations each calculated according to four standard curves ) of each mutant was then calculated and standard errors ( from the calculation for average ) are given . Typical aggregation curves and standard curves can be found in Figure 5—figure supplement 2 . Quantification was carried out as follows: to measure the relative activity of Spy variants on α–LA , each Spy variants were used at three ratios in duplicate ( Spy: α–LA ratios = 0 . 1 , 0 . 15 , and 0 . 2 ) in the aggregation assay . The standard curve was smooth and could have been usable over a broad range . These specific ratios were chosen because they A ) fell in a steep portion of the standard curve so that the light scattering signal is the most sensitive to the amount of Spy added and because B ) the Spy concentrations at these ratios were high enough to enable pipetting high enough volumes to minimize pipetting errors . The portion of the standard curve we used is indicated by the red box in Figure 5—figure supplement 2 . Light scattering endpoints at 300 , 400 , 500 , and 600 S were taking from each aggregation curve and then back-calculated according to the standard curve at the respective time point to get the effective ratio of Spy: α–LA ( or , ratio equivalent to Spy WT: α–LA ) . The activity of Spy variants relative to wild type is then calculated from dividing these effective ratios by the adding ratios . For example , Q100L added at a Spy: α–LA ratio of 0 . 2 had a light scattering reading of 0 . 009 arbitrary units ( a . u . ) after 400 S of the reaction . According to the standard curve at 400 S , this is equivalent to wild type Spy added at a Spy:α–LA ratio of 0 . 383 . Therefore , Q100L has the activity 1 . 82-fold of wild type based on the calculation from the 400 S standard curve . This calculation is then repeated using the other three standard curves and other light scattering endpoint values taken from the aggregation curves of the same variant added at different Spy:α–LA ratios . This strategy allowed us to generate 12 relative activity values ( in duplicate ) of the same variant , which are then averaged to give the final value . The standard errors we reported reflect the average of these 24 activity measurements . To validate our calculations , we performed the aggregation assay with wild type Spy in exactly the same way as we used for the Spy variants . This gave us a value of 1 . 06 ± 0 . 09 fold activity relative to wild type Spy , suggesting that our approach to compare the activities of the Spy variants to wild type is precise and reproducible . Ammonium sulfate suspension of aldolase from rabbit muscle ( MP Biochemicals , Santa Ana , CA ) was suspended in a 40 mM HEPES , 150 mM NaCl pH7 . 5 buffer that contained 4 mM Beta-mercaptoethanol and then dialyzed into 40 mM HEPES , 150 mM NaCl pH7 . 5 buffer . 100 µM aldolase was then denatured in 40 mM HEPES pH 7 . 5 with 50 mM NaCl , 3 M guanidine , and 2 mM DTT for >2 hr at room temperature . Aggregation of denatured aldolase was initiated by rapidly diluting 6 . 5 µl of the denatured aldolase into 1293 . 5 µl ( a 200-fold dilution ) of 150 mM sodium chloride pH 7 . 5 buffer pre-equilibrated for 10 min at 23°C . The final aldolase concentration reached was 500 nM . The aggregation of aldolase was monitored by measuring light scattering in the absence or presence of Spy variants with a photomultiplier Spectrofluorimeter ( Photon Technology International ( PTI ) , Birmingham , NJ ) and FeliX32 Analysis software ( PTI , inc . ) . with excitation/emission wavelengths and slits of 360 nm and 1 nm respectively . We found that pipetting the aldolase into the cuvette at the same position relative to the stirring bar made for a very reproducible assay . Thus , long gel loading tips ( 83 mm , MultiFlex Pipet Tips from Thermo Fisher Scientific , Waltham , MA ) were used to pipette denatured aldolase into the cuvette through the small sample injection hole on the top of the lid of the PTI photomultiplier Spectrofluorimeter , so that for each injection the end of the tip lay just above the stirring bar . We used different ratios of wild Spy: aldolase to establish a standard curve . For most of the Spy variants we used a ratio of Spy: aldolase of 0 . 3 because this lies in the middle of the standard curve . Our most active variant Q25R , completely suppressed aldolase aggregation at this ratio so it was in addition further diluted and measured at a Spy: aldolase ratio of 0 . 1 . The aggregation rates were measured by calculating the slopes of the aggregation curves ( fitting the data points between 50 s and 600 s of the aggregation reaction ) . Each Spy variant was measured three times . By using the standard curve , we were able to calculate the relative activity of these Spy variants as shown in Figure 5—figure supplement 3 . The error of the variant’s measurements and the error of the standard curve were combined to give a final error . Aldolase used in the refolding assay was prepared by dissolving powdered aldolase ( Cat # 0215985925 , MP Biochemicals , USA ) into 40 mM HEPES , 150 mM NaCl , pH7 . 5 buffer . Aldolase was then denatured in 40 mM HEPES pH 7 . 5 with 50 mM NaCl , 3 M guanidine , and 2 mM DTT at 25 µM overnight at room temperature . To initiate refolding of aldolase , denatured aldolase was diluted 100-fold into the refolding buffer ( 150 mM NaCl , 40 mM HEPES , 5 mM DTT , pH7 . 5 ) to a final concentration of 0 . 25 µM in the absence or presence of different ratios of Spy at 25°C . After 5 min , 14 µl aliquots from the refolding reaction were added into 200 µl assay buffer to test for aldolase activity at 25°C . The assay buffer contained 0 . 15 mM β-nicotinamide adenine dinucleotide reduced disodium salt ( Sigma Aldrich , USA ) , 2 mM Fructose 1 , 6-diphosphate ( Sigma Aldrich , USA ) , 0 . 18 U/ml α-glycerophosphate dehydrogenase/triosephosphate isomerase ( Sigma Aldrich , USA ) , 150 mM NaCl and 40 mM HEPES , pH 7 . 5 . The measurements of absorbance at 340 nm were immediately started using a Synergy HT Multi-Mode Microplate Reader ( Biotek , Winooski , VT ) following a 5-s shaking at medium speed . Data were collected for 5 to 10 min and the absorbance values were plotted against time to obtain a slope , which was then divided by the value corresponding to 100% activity ( i . e . , that of an equivalent concentration of native aldolase ) to obtain the percentage of native aldolase . Using Spy ( wt ) : aldolase ratios of 0 , 0 . 125 , 0 . 25 , 0 . 5 , 0 . 75 , 1 , 1 . 25 , 1 . 5 , and 1 . 75 , we obtained a standard curve shown in Figure 5—figure supplement 4 and found that the ratio of 0 . 25 Spy:Aldolase was located in the middle of the range . Thus we performed refolding assay at a Spy: aldolase ratio of 0 . 25 for all the Spy variants and calculated their relative activities in assisting aldolase refolding according to the standard curve ( see Figure 5—figure supplement 4 legend for details ) . Every Spy mutant was measured at least three independent times and the error of the measurements and the error of fitting the standard curve were combined to give a final error . The steady-state expression levels of the different Spy variants and the β-lactamase-Im7 L53A I54A biosensor in strains LW53-LW59 ( Table 1 ) were quantified by western blotting of whole cell extractions using infrared fluorescence labeled IRDye secondary antibodies . This was done after induction by various amounts of IPTG ( 0 . 01–0 . 5 mM ) for 3 hr . IRDye 680LT goat anti-mouse secondary antibody ( LI-COR Biosciences , Lincoln , NE ) was used to recognize the primary antibodies , which were against either the β-lactamase portion of the biosensor or directed against Spy . As a loading control , we quantified the amount of maltose binding protein ( MBP ) present in the lysate using a IRDye 800CW goat anti-rabbit secondary antibody ( LI-COR Biosciences ) directed against MBP . 0 . 04% maltose was added to the medium to induce MBP expression . MBP is a periplasmic protein that is routinely used as a loading control ( Raivio et al . , 1999 ) ; it has the added advantage that it is not a Spy client ( unpublished results ) . Various dilutions of the whole cell extracts ( A600 = 2 . 5 ) were loaded and their intensities were plotted to quantify the level of Spy and β-lactamase-Im7 L53A I54A biosensor protein in the Spy variants . We performed spot titrations of strains expressing seven different super-Spy variants as well as the β-lactamase-Im7 L53A I54A biosensor onto various concentrations of penicillin V ( 0 through 7000 μg/ml ) and IPTG ( 0 , 0 . 01 , 0 . 1 , 0 . 2 , and 0 . 5 mM ) to quantitatively measure the minimal inhibitory concentration ( MIC ) they exhibited toward the β-lactam antibiotic penicillin V using the methods previously described ( Foit et al . , 2009 ) . To measure the specific in vivo activity of these Spy variants , we then normalized the maximal MIC values of each strain to the levels of Spy expressed in these variants to compensate for possible differences in Spy expression . The maximal MIC values were achieved at 0 . 5 mM IPTG induction for all the variants except for Q49L and WT which reached a maximal MIC starting at 0 . 2 mM . The steady-state protein levels of each Spy variant expressed at various IPTG concentrations were measured using quantitative western blots as described above . The in vivo specific activity was obtained by calculating the ratio between the maximal MICs obtained for the variant divided by the MIC obtained for wild-type Spy when it was expressed to the same level as the variant . These values are reported as normalized MIC values ( MICnorm ) and are a measure of the in vivo specific activity of the Spy variants . Hydrogen-deuterium exchange of Spy and Spy-Im7 complex was performed using top-down ECD-FTICR-MS , as described previously ( Pan et al . , 2009 , 2010; Serpa et al . , 2013 ) . This approach relies on the rapid scrambling-free fragmentation of the intact protein by electron capture dissociation ( ECD ) . The deuteration level of each amino acid residue then can be deduced from the c- and z-ion series produced . Briefly , Spy and Im7 protein stock solutions ( 250-1000 µM ) in 40 mM HEPES , 150 mM NaCl , pH 7 . 5 were diluted with 10 mM ammonium acetate to 25 µM final Spy ( dimer ) and 100 µM Im7 . Spy-Im7 complex was preformed in a 1:4 molar ratio of Spy dimer to Im7 . Spy or Spy-Im7 samples were continuously mixed from separate syringes with D2O in a 1:4 ratio ( 80% D2O final ) via a three-way tee , which was connected to a 100 µm × 21 cm capillary , providing a labeling time of 10 s . The outflow from this capillary was mixed with a quenching solution containing 0 . 4% formic acid in 80% D2O from a third syringe via a second three-way tee , and injected into a Bruker 12 T Apex-Qe hybrid Fourier transform mass spectrometer ( Bruker Daltonics , Billerica , MA ) , equipped with an Apollo II electrospray source . In this analytical format , lower deuteration levels would reflect more protection of the amide protons and formation of hydrogen bonding due to residues involvement in secondary structure elements , interaction , or shielding from the solvent . In-cell ECD fragmentation experiments were performed with an m/z 900-1200 precursor selection range using a cathode filament current of 1 . 2 A and a grid potential of 12 V as previously described ( Serpa et al . , 2013 ) . Approximately 1200 scans were accumulated over the m/z range 250–2600 , corresponding to an acquisition time of approximately 30 min for each ECD spectrum . Deuteration levels of the amino acid residues’ amide groups were determined from centroid masses of the c- and z-ion series . The analysis of the deuteration status of c- and z-ion fragment series thus allowed us to localize the residues that exhibited major differences in the protection between free Spy and Spy that is occupied by its client protein Im7 . Spy-Im7 protein complex was crosslinked with CyanurBiotinDimercaptoPropionylSuccinimide ( CBDPS ) -H8/D8 ( Creative Molecules Inc . , Canada ) . The crosslinked proteins were digested with trypsin and proteinase K and crosslinked peptides were affinity purified with immobilized avidin , then identified by LC-MALDI MS and MS/MS . Zero-length crosslinking brings additional challenges to the mass spectrometric analysis of the crosslinks as the crosslinked peptides formed do not acquire a specific isotopic signature . To facilitate detection and identification of the Spy-Im7 zero-length crosslinks , we applied a variation of the method using 15N metabolically-labeled oligomeric proteins ( Taverner et al . , 2002 ) . We used an equimolar mixture of non-labeled and 15N metabolically-labeled Im7 for crosslinking the Spy-Im7 complex . Following digestion of the crosslinked complex , every Im7 peptide and , most notably , inter-protein Spy-Im7 crosslinks were represented by a pair of light and heavy isotopic forms , derived from non-labeled and 15N-labeled Im7 peptides , respectively . Mass difference between such pairs is then determined by the number of the nitrogen atoms in the Im7 peptides . We have implemented this principle , as an additional selection criterion , into the algorithm for the automatic detection and identification of the inter-protein crosslinks in heteromeric protein complexes . Technical details are described below: Im7 L53A I54A was biotinylated by incubation with EZ-link NHS-biotin ( Thermal Fisher Scientific ) at 1:1 molar ratio ( 100 µM each ) in the assay buffer ( 40 mM HEPES , 150 mM NaCl , pH7 . 5 ) for 30 min at room temperature . Then the reaction was quenched by addition of one tenth volume of 1M Tris pH7 . 5 and the unconjugated biotin was removed by dialyzing in the assay buffer overnight at 4°C . Binding kinetics were performed on the Octet RED system ( Fortebio , Menlo Park , CA ) at 25°C . The streptavidin sensors were pre-wetted in the assay buffer for 15 min before use . Spy variants were serial diluted to 2 , 1 , 0 . 5 , 0 . 25 , and 0 . 125 μM in a 96-well plate in the assay buffer . The binding assay contains the following steps: immobilization of the biotin-conjugated Im7 L53A I54A ( 10 µg/ml ) for 20 min , wash for 5 min , baseline for 1 min , association for 10 min , and dissociation for 15 min . The last three steps were repeated for all the Spy concentrations . Wells with assay buffer only were used as reference wells and were subtracted from the raw data . Curve fitting was performed in Sigmaplot assuming a 1:1 binding model using the following equations for association and dissociation , respectively: y=a∗ ( 1−e− ( kobs∗t ) ) +y0∗t and y=y0+a∗e− ( koff∗t ) , where y0 is the constant to correct for baseline drift . The observed association rate constants ( kobs ) are then plotted against Spy concentrations to obtain the association rate constants ( kon ) according to the following equation: kobs=kon∗[Spy]+koff . The dissociation constant KD is calculated according to: KD=koffkon . Digestion of Spy alone ( 50 µM ) or Spy-Im7 L53A I54A complex ( 50 µM each ) was carried out at 1:100 mass ratio of trypsin to protein in 40 mM HEPES , 100 mM NaCl , pH 7 . 5 at room temperature . At different time points ( 0–8 min ) , aliquots were withdrawn and the digestion was stopped with 10% TFA . Peptides were separated by a reverse phase C18 column ( Zorbax 300SB-C18 , 1 × 50 mm , 3 . 5 µm , Agilent , Santa Clara , CA ) at room temperature and then applied to a Q-TOF dual ESI LC/MS ( Agilent ) for identification . A 15 min linear gradient of 2–80% acetonitrile in 0 . 1% formic acid at a flow rate of 0 . 3 µl/min was used to elute the peptides . Peptide identification was performed using BioConfirm software ( Agilent ) . To identify the minimal length peptide derived from Im7 L53A I54A that mediates the binding to Spy , we applied limited proteolysis to Im7 L53A I54A with trypsin to prepare a series of incompletely digested fragments . After quenching the digestion reaction , the peptide mixture was mixed with the N-terminal strep-tagged Spy and applied to a strep-tactin sepharose column ( IBA , Germany ) . The complex was eluted by 2 . 5 mM desthiobiotin , separated by a reverse phase C18 column ( Zorbax 300SB-C18 , 1 × 50 mm , 3 . 5 µ ) at room temperature , and then applied to a Q-TOF dual ESI LC/MS ( Agilent ) for identification . We obtained fragments 5–43 , 5–70 , 5–73 and 5–76 from Im7 L53A I54A . The shortest fragment ( 5–43 ) constitutes most of the α1 and α2 helices in Im7 . We then screened a few constructs containing extension or deletion at the N or C termini with amino acid sequences derived from the Im7 protein to identify the most stably expressed fragment in vivo . We purified three peptides with the highest expression levels ( 3–45 , 7–45 , and 12–45 ) and characterized their interactions with Spy by isothermal titration calorimetry . Based on their affinities to Spy ( Kd = 2 . 5 , 2 . 3 , 2 . 6 , and 8 µM for the full length Im7 L53A I54A , 3–45 , 7–45 , and 12–45 peptide , respectively ) and their expression levels , we chose the 7–45 fragment as the minimal length peptide that represents the interaction between Spy and Im7 L53A I54A . A 15N-HSQC experiment was performed on an Agilent 600 MHz spectrometer equipped with a triple-resonance cryoprobe . The sample contained 200 mM 15N labeled Im7 7-45 dissolved in 13 mM MES , 13 mM HEPES , 50 mM NaCl pH7 . 0 , supplemented with 10% D2O . The B-factors of residues Phe29-Asn33 in the crystal structure of Spy ( PDB ID: 3O39 ) are very high ( >110 ) , meaning that these residues are possibly flexible . We therefore deleted these residues from the structure before docking analysis because they limit the movable space of the Im7 peptide around the Spy structure . The 3D structure model of Im7 was obtained using the iterative threading assembly algorithm I-TASSER ( Zhang , 2008; Roy et al . , 2010 ) . We also used the ab inito structure modeling algorithm QUARK ( Xu and Zhang , 2012 ) to model the structure of Im7 . Both the QUARK and I-TASSER models of Im7 contain two α-helices , but because I-TASSER modeling has a higher overall confidence score , it was used for docking . A three-step hierarchical approach was used to dock Im7 against Spy: 10 rounds of PSI-BLAST ( Altschul et al . , 1997 ) using the E . coli K12 spy sequence were performed to obtain 503 Spy homologous protein sequences . To separate the sequences of Spy orthologs from the sequences of their homolog , CpxP , the 503 sequences were aligned in SeaView ( Gouy et al . , 2010 ) using the built-in algorithms MUSCLE ( Edgar , 2004 ) and MAFFT ( Katoh et al . , 2002 ) , and then these results were used to construct a phylogenetic tree using MEGA5 ( Tamura et al . , 2011 ) . 252 sequences were assigned to the Spy clade and the rest were assigned to the CpxP clade . The sequences were then aligned with ClustalW ( Thompson et al . , 2002 ) . The thermodynamic parameters characterizing the unfolding of the Spy variants were derived from analysis of the thermal denaturation curves , which were obtained using circular dichroism measurements at 222 nm to follow unfolding . Spy variants were diluted to 6 . 5 μM in 10 mM sodium phosphate buffer ( degassed ) , pH7 . 5 and were denatured at a heating rate of 2°C/min from 10°C to 80°C in a 1 mm path length quartz cuvette . The circular dichroism signal was monitored by a JASCO J-810 CD spectrometer ( JASCO inc . , Easton , MD ) equipped with a Peltier thermoelectric controller . Samples were quickly cooled down to 10°C at the end of the denaturation reaction and each spectrum after renaturation was compared with the original spectrum at 10°C to check the reversibility of the denaturation reactions . All the Spy variants after renaturation gained at least 95% of the initial CD signal , indicating that the denaturation process is reversible for all the variants under the conditions we have used . Data analysis was performed in Sigmaplot according to equation II as described in the supplemental materials in ( Greenfield , 2006 ) to obtain the melting temperature ( Tm ) and the enthalpy change ( ΔHm ) of each denaturation reaction at pH7 . 5 . Each thermal denaturation measurement was performed three times to get the average values of ΔHm and Tm and the standard errors ( Table 2 ) . To find the heat capacity change ( ΔCp ) when the protein unfolds we obtained Tm and ΔHm as a function of pH by measuring the denaturation curves at a wide range of pH from 2 . 0 to 10 . 0 in 20 mM sodium phosphate buffer . We then plotted the ΔHm against Tm to yield ΔCp ( Table 2 ) , which is the slope . Finally , the unfolding free energy change ΔGNU at 25°C is calculated using the modified Gibbs-Helmholtz equation according to Grimsley et al . , ( 2013 ) :ΔG ( T ) =ΔH↓m ( 1−T/T↓m ) −ΔC↓p[ ( T↓m−T ) +Tln ( T/T↓m ) ]where T = 298 . 15K . | Proteins are made from long chains of smaller molecules , called amino acids , that twist and fold into complex three-dimensional shapes . Folding into the correct shape is crucial for a protein to function properly because many proteins work by binding to certain other proteins or molecules , like a key fitting into a lock . Additional proteins called chaperones often help with this folding process , and it has been proposed that chaperones must be particularly flexible in order to cope with the changes in the shape of the different proteins being folded . However , studying this hypothesis directly has proven to be difficult . Now , Quan et al . have tackled this challenge by using a bacterial assay—that they had developed previously—and which links the correct folding of a test protein to cell survival and growth in the presence of an antibiotic . This approach was formerly used to identify a new chaperone called Spy , and Quan et al . have now used it to find variants of this protein that perform as even better chaperones . This assay identified several variants of Spy that could stabilise an unstable test protein even more effectively than the wild-type Spy can . All of these variants were also better than the wild-type Spy at stabilising two other unfolded proteins—and so were dubbed ‘super Spy’ proteins . The mutations in the super Spy variants altered a region on the surface of Spy , which additional experiments revealed was likely to be involved in binding to the partner proteins . Furthermore , prior to binding to these partner proteins , the super Spy variants appear more flexible than the wild-type Spy protein . Quan et al . suggest that this increase in flexibility allows the super Spy variants to bind more tightly to a range of substrates , thus optimising their chaperone function . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2014 | Super Spy variants implicate flexibility in chaperone action |
The endodermis represents the main barrier to extracellular diffusion in plant roots , and it is central to current models of plant nutrient uptake . Despite this , little is known about the genes setting up this endodermal barrier . In this study , we report the identification and characterization of a strong barrier mutant , schengen3 ( sgn3 ) . We observe a surprising ability of the mutant to maintain nutrient homeostasis , but demonstrate a major defect in maintaining sufficient levels of the macronutrient potassium . We show that SGN3/GASSHO1 is a receptor-like kinase that is necessary for localizing CASPARIAN STRIP DOMAIN PROTEINS ( CASPs ) —major players of endodermal differentiation—into an uninterrupted , ring-like domain . SGN3 appears to localize into a broader band , embedding growing CASP microdomains . The discovery of SGN3 strongly advances our ability to interrogate mechanisms of plant nutrient homeostasis and provides a novel actor for localized microdomain formation at the endodermal plasma membrane .
The plant root is a highly selective filter that forages the soil environment for nutrients and water . Its function has been likened to that of an ‘inverted gut’ , displaying a very similar dual role in uptake and protection ( Waisel et al . , 2002 ) . Selectivity of roots is thought to crucially depend on the endodermis , an inner cell layer that surrounds the central vascular strand of the root and represents the main paracellular ( apoplastic ) transport barrier in young roots ( Geldner , 2013 ) . This apoplastic diffusion barrier is set up by the ‘Casparian strips’ , lignin-like , hydrophobic impregnations of the primary cell wall that form a supracellular net-like structure around the central vasculature . In analogy to the tight junction of gut epithelia , this net-like arrangement provides a seal of the extracellular space , while still allowing for nutrient and water transport across outer and inner plasma membrane surfaces . The Casparian strip-bearing endodermis therefore represents an independent development of a polarized epithelium , evolved in the context of multi-cellular organisms with wall-bearing cells . Later during endodermal development , hydrophobic ( cork-like ) suberin lamellae form all around the endodermal surface . This protective suberization should eventually abolish direct transport across the endodermal plasma membrane , forcing symplastic passage through plasmodesmata from outer cell layers . Despite being a central , conserved feature of higher plants , nothing was known until recently about the factors that build and position the Casparian strip . This led to the frustrating situation of not being able to test the many supposed roles of the endodermis and its Casparian strip , because of an absence of specific mutants . Recent papers established that a family of previously undescribed four-transmembrane-span proteins , called CASPs , forms a central , ring-like membrane domain in the endodermis , called the Casparian strip membrane domain ( CSD ) . This domain acts as a lateral diffusion barrier and separates the endodermal plasma membrane into two distinct polar domains ( Roppolo et al . , 2011; Alassimone et al . , 2010 ) . Upon localization , the CASPs show extreme stability and a lack of endocytosis or lateral diffusion . Moreover , CASPs show extensive pair-wise interactions and associate strongly with cell walls . Multiple knock-outs of CASP family members display strongly disorganized formation of Casparian strips , demonstrating their functional importance . It is thought that the CASPs act as scaffolds that spatially organize cell wall biosynthetic enzymes . CASP1 was shown to determine the subcellular localization of a secreted peroxidase , PER64 ( Lee et al . , 2013 ) . In current models , the role of CASPs is to assemble a specific NADPH oxidase with peroxidases , leading to oxidation of mono-lignols and local polymerization of lignin . Lignin formation in vivo must implicate numerous other cell wall proteins and ENHANCED SUBERIN 1 ( ESB1 ) , a dirigent-like protein is a cell wall protein that has also been demonstrated to be crucial for the correct formation of Casparian strips ( Hosmani et al . , 2013 ) . These recent breakthroughs held the promise that we might finally be able to investigate the functional relevance of the Casparian strip through analysis of the different mutants obtained . This however , turned out to be more difficult than expected . Multiple mutants of the CASP family display interrupted Casparian strips in the beginning , but rather quickly deposit more cell wall material in a delocalized fashion , eventually sealing the apoplast . A very similar phenotype occurs in the esb1 mutant ( Hosmani et al . , 2013 ) . Moreover , both mutants display an earlier and stronger production of suberin lamellae ( esb1 was named based on this phenomenon ) . This makes it difficult to use these mutants for assessing the importance of Casparian strips , since the partial defect on Casparian strip formation will always be confounded with the effect of ectopic/enhanced production of suberin in the same cells . In the rbohf mutant , the Casparian strip defect is only partial and restricted to the young part of the root . Moreover , RBOHF has roles in many different aspects of plant development and it would be hard to distinguish what phenotype is directly caused by defects in Casparian strip formation ( Lee et al . , 2013 ) . Here , we report the discovery of a novel Casparian strip mutant , schengen3 ( sgn3 ) . SGN3 encodes a receptor-like kinase with strong expression in the root endodermis . The SGN3 protein accumulates in the plasma membrane in a broad band within which the CSD forms . In its absence , only discontinuous patches of CASPs are observed . SGN3 loss-of-function leads to the strongest defects in Casparian strips known to date , with no indication of a compensatory upregulation of suberin , as seen for other mutants ( Hosmani et al . , 2013 ) . Our analysis reveals a surprising capacity of the mutant to maintain homeostatic control in the absence of the major root diffusion barrier and challenges views according to which the root should lose its ability for selective nutrient uptake , because of a generalized , non-selective bypass of nutrients into the vasculature .
In a forward genetic , GUS-based screen for endodermal barrier mutants ( Lee et al . , 2013 ) ( Alassimone et al . , unpublished ) , we discovered one mutant , sgn3 that displayed a dramatic defect in endodermal barrier formation , as visualized by penetration of the apoplastic tracer propidium iodide ( PI ) into the stele along the entire length of the root . This resembled a phenotype of T-DNA insertion lines in an endodermis-enriched gene that we had analyzed in parallel ( Figure 1A , Figure 1—figure supplement 1A ) ( Birnbaum et al . , 2003; Brady et al . , 2007 ) . Complementation analysis and sequencing confirmed that the causal mutation in sgn3-2 was an early stop codon in the open reading frame of At4g20140 . SGN3 encodes for a leucine-rich-repeat receptor-like kinase ( LRR-RLK ) of subfamily XI that had previously been described as GASSHO1 ( GSO1 ) ( Figure 1B , Figure 1—figure supplement 2 , Figure 1—source data 1 ) ( Tsuwamoto et al . , 2008 ) . In the previous publication , this LRR-RLK was investigated because of embryonic expression of a putative ortholog in Brassica napus , but no single mutant phenotype was found . A phenotype was only observed when gso1 ( sgn3 ) was combined with a mutant of its closest homolog , At5g44700 ( named GSO2 ) . This gso1 gso2 double mutant displayed a protodermal phenotype , characterized by a defective cuticle and fusion of cotyledons ( Tsuwamoto et al . , 2008 ) . True leaves did not differ from wild type . In contrast to SGN3/GSO1 , GSO2 was shown to be expressed only in shoots , but not in roots , explaining the exclusive root phenotype in the sgn3 single mutant that we report here and that had escaped detection earlier on . In order to investigate the cause of this phenotype , we investigated formation of the Casparian strip and found that it is still formed and correctly positioned , but that it is repeatedly interrupted , forming irregularly-sized ‘holes’ of several micrometer in length ( Figure 1C , Figure 1—figure supplement 1B ) . Suberin lamellae formation , by contrast , appeared at a position and in a manner very similar to wild type ( Figure 1D , E , Figure 1—figure supplement 1C , D ) . 10 . 7554/eLife . 03115 . 003Figure 1 . SGN3 receptor-like kinase is important to establish a functional endodermal barrier . ( A ) Lack of endodermal diffusion barrier in sgn3-3 visualized by presence of propidium iodide ( PI ) in stele . ( B ) Diagram of the SGN3 protein showing the different domains , T-DNA insertion lines ( indicated with triangles ) and the ethyl methanesulfonate ( EMS ) -induced mutations ( see also Figure 1—figure supplement 2 ) . ( C ) Surface view of Casparian strip , visualized by autofluorescence after clearing . Note discontinuous lignin deposition in sgn3-3 . Pictures are maximum projections of confocal z-stacks . Arrowheads indicate discontinuities in sgn3 . Spiral-like signal in WT is from deeper-lying xylem vessel . ( D ) Fluorol yellow staining of suberin lamellae deposition in sgn3-3 and WT . Pictures are overlays of transmitted light image ( gray ) with fluorescent signal from suberin dye ( yellow ) . ( E ) Occurrence of suberin deposition along the root is not altered in sgn3-3 . Suberin lamellae deposition was quantified considering three different zones: non-suberized zone , zone of patchy suberization , and zone of continuous suberization ( n = 5 , one representative experiment presented ) . ( F ) Surface view of CSD network visualized with CASP1-GFP expressed under CASP1 promoter showing the net-like structure with discontinuities in sgn3-3 . Projections as in C . ( G ) Absence of a lateral diffusion barrier in sgn3-3 visualized with plasma membrane marker line CASP1::mCherry-SYP122 ( intensity color coded ) . Confocal pictures were taken at the surface of an endodermal cell . Note the mutually exclusive localization with the CSD marker CASP1-GFP ( green ) . Two right images are magnification of the two leftmost images . ( H ) Localization of the outer marker PDR6-Venus expressed under CASP1 promoter is still polar in sgn3-3 . Pictures are median longitudinal sections of endodermal cells . Scale bars: A , C , F , G , H = 20 μm; D = 50 μm . ep , epidermis; co , cortex; en , endodermis; st , stele; LRR , Leucine-rich repeat . DOI: http://dx . doi . org/10 . 7554/eLife . 03115 . 00310 . 7554/eLife . 03115 . 004Figure 1—source data 1 . Detail of SGN3 T-DNA and EMS mutants . Detail of SGN3 mutations . sgn3-1 EMS mutant was lost during the screen and is not available anymore . DOI: http://dx . doi . org/10 . 7554/eLife . 03115 . 00410 . 7554/eLife . 03115 . 005Figure 1—figure supplement 1 . Both Casparian strip domain and Casparian strip but not the suberin are affected in sgn3 . ( A ) Lack of endodermal diffusion barrier in sgn3-3 and sgn3-4 visualized by free diffusion of propidium iodide ( PI ) into stele . ( B ) Surface view of Casparian strips by autofluorescence after clearing shows discontinuous lignin deposition in sgn3-3 and sgn3-4 . Pictures are maximum projections of confocal z-stacks . ( C ) Fluorol yellow staining in WT , sgn3-3 , and sgn3-4 . ( D ) Suberin deposition along the root is not altered in sgn3-3 and sgn3-4 . Suberin lamellae deposition was quantified as in Figure 1E ( n = 5 , one representative experiment shown ) . ( E ) Surface view of CSD network visualized with CASP1-GFP , CASP2-GFP , CASP3-GFP , CASP4-mCherry , and CASP5-GFP under CASP1 promoter showing the discontinuous net-like structures in sgn3-3 . Projections as in B . Scale bars: A , B , E , 20 μm; C , 50 μm . ep , epidermis; co , cortex; en , endodermis; st , stele . DOI: http://dx . doi . org/10 . 7554/eLife . 03115 . 00510 . 7554/eLife . 03115 . 006Figure 1—figure supplement 2 . Diagram of SGN3 genomic DNA with T-DNA and EMS mutants . Schematic of SGN3 genomic region showing the insertions sites of the three T-DNA mutants and the EMS mutants . sgn3-2 was found in a forward genetic , GUS-based screen ( Alassimone et al . , unpublished ) , sgn3-5 to sgn3-18 were found in additional forward genetic screen ( Kalmbach et al . , unpublished ) . sgn3-3 , sgn3-4 , and sgn3-19 are T-DNA insertion mutants in the first exon of SGN3 . DOI: http://dx . doi . org/10 . 7554/eLife . 03115 . 006 We asked whether the Casparian strip cell wall defect is caused by a defect in the formation of the CSD , or whether the function of SGN3 lies downstream of the CASPs , as has been demonstrated for the NADPH oxidase SGN4/RBOHF ( Lee et al . , 2013 ) . In the sgn3 background , CASP1-GFP showed a pattern of interrupted patches that was strikingly similar to that of the Casparian strip itself . This suggests that it is the incorrect localization of the CASPs that causes the Casparian strip defects of the mutant ( Figure 1F ) . Interestingly , in a novel forward genetic screen in which we scored directly for mutants with altered CASP1-GFP localization ( Kalmbach et al . , unpublished ) , we found 14 novel alleles of sgn3 ( Figure 1B , Figure 1—figure supplement 2 , Figure 1—source data 1 ) . This suggests that SGN3 is one of the major non-redundant genes necessary for CASP localization in the Arabidopsis genome . We could show that not only CASP1-GFP , but all four other CASPs ( CASP2-5 ) display identical defects in sgn3 ( Figure 1—figure supplement 1E ) . An interrupted CSD should abolish the lateral diffusion barrier in the plasma membrane of endodermal cells that we had previously described ( Alassimone et al . , 2010 ) . Indeed , a generic plasma membrane protein that is excluded from the CSD in wild type ( Alassimone et al . , 2010 ) ( Figure 1G ) shows a pattern of exclusion in the mutant that represents a perfect negative image to the CASP1-GFP islands ( Figure 1G ) . This illustrates that the lateral diffusion barrier between inner and outer plasma membrane domain is absent in sgn3 mutants , but that the remaining CSD islands are still able to exclude other proteins . To our surprise , this defect did not affect the polar localization of PDR6 , a transporter that localizes to the outer polar domain in the endodermis ( Figure 1H ) . This indicates that a strict polarity in the endodermis can be maintained in the absence of a diffusion barrier , further supporting the notion that polarity of transmembrane proteins in plants might simply be maintained by generally low lateral diffusion , high rates of endocytosis , or both ( Geldner , 2009 ) . We had shown previously that during endodermal differentiation , CASP1-GFP changes from a protein that displays endocytic cycling and lateral diffusion to one that is highly stable and immobile in the CSD ( Roppolo et al . , 2011 ) . In this process , CASP1-GFP localization passes through a ‘string of pearls’ stage , in which individual islands or ‘patches’ of CASP1-GFP are aligned in a band before eventually fusing ( Figure 2A ) . Comparison of 3D-time-lapse observations of endodermal differentiation revealed that sgn3 is not able to progress from this ‘string of pearls’ stage , but only undergoes partial fusion of individual islands , leading to CASP1-GFP patches of heterogenous size ( Figure 2A ) . We further found by fluorescence recovery after photobleaching ( FRAP ) that these CASP1-GFP islands have similar stability and turnover as wild type ( Figure 2B ) . Another feature of the CSD is its very strong adherence to the cell wall , leading to the long-described phenomenon of ‘band plasmolysis’ in endodermal cells ( Krömer , 1903; Behrisch , 1926; Alassimone et al . , 2010 ) . The inability of the CASP1-GFP ring to retract from the Casparian strip upon plasmolysis ( Figure 2—figure supplement 1 ) allow for easy visualization of this phenomenon . This adhesion leads to greatly flattened , often fenestrated protoplasts , that looked very unlike the usual , rounded protoplasts of epidermal cells ( Alassimone et al . , 2010 ) . sgn3 showed the very same inability of CASP1-GFP to retract from the strips , demonstrating that plasma membrane-to-cell wall attachment remains intact in the mutant ( Figure 2—figure supplement 1 ) . We conclude that sgn3 does not cause a general defect in CASP1-GFP trafficking , polymerization or cell wall attachment , but rather a specific defect in the progression of already stable CASP islands towards a contiguous band . 10 . 7554/eLife . 03115 . 007Figure 2 . SGN3 localizes on both sides of the Casparian strip domain and is important for the CASP1-GFP patches to fuse into a contiguous band . ( A ) 3D-confocal time lapse imaging of CASP1::CASP1-GFP in Col-0 and sgn3-3 background reveals problems with progression of CASP1-GFP localization in sgn3-3 . Images show median and surface image of endodermal cells of the same root . Arrowhead indicates CSD , dotted box in sgn3-3 shows immobility of CSD islands . Time in hours . ( B ) Fluorescence recovery after photobleaching ( FRAP ) with pCASP1::CASP1–GFP in WT ( gray ) and sgn3-3 ( white ) . The fifth cell after the onset of CASP1 expression was used . Similar very low recoveries are observed in both genotypes after 15 min . n = 8 independent assays for each series , error bars = s . d . ( C ) SGN3 is expressed before the onset of CASP1 expression in the endodermis . Quantification was done using pSGN3::SGN3-mVenus and pCASP1::CASP1-GFP . The onset of expression corresponds to the first cell in the endodermis file with a clear fluorescent signal not visible in WT . n = 10 . Error bars = s . d . ( D ) Localization of SGN3-mVenus under its own promoter in elongating endodermal cells . ( E ) In differentiated endodermal cells , SGN3-mVenus accumulates in the transversal and anticlinal sides of the plasma membrane , but is depleted from the CSD . Left panel shows a transition from a median ( top ) to a surface ( bottom ) view of an endodermal cell . Right panel shows a close-up of a median view with SGN3 surrounding the CSD . Arrowheads indicate the CSD . ( F ) Maximum projection of a z-stack showing the localization of SGN3 on both sides of the CSD . Note the CSD surrounded by two SGN3-mVenus ‘lines’ ( white arrowheads ) . ( G ) Localization of SGN3-GFP under CASP1 promoter in the ‘string of pearls’ stage . SGN3 surrounds individual CSD patches preceding their fusion ( left panel ) . Close-up showing the CSD patches ( red arrowhead ) surrounded by SGN3-GFP ( right panel ) . ( H ) Schematic of SGN3 putative mode of action . The SGN3 kinase might promote addition of new CASP units to already formed CASP microdomains . Scale bars: A , G ( left panel ) , 10 μm; D , E ( left panel ) , F , 20 μm; E ( right panel ) , G ( right panel ) , 5 μm . Ep , epidermis; co , cortex; en , endodermis; st , stele . DOI: http://dx . doi . org/10 . 7554/eLife . 03115 . 00710 . 7554/eLife . 03115 . 008Figure 2—figure supplement 1 . Plasma membrane-to-cell wall attachment remains intact in sgn3 upon plasmolysis . The CSD highlighted by CASP1-GFP is still attached to the cell wall in sgn3-3 upon plasmolysis . Localization of the generic plasma membrane marker Cit-SYP122 expressed under CASP1 promoter in water and in 0 . 8 M mannitol ( panel 1 and 2 , respectively ) . CASP1-GFP network still attached to the cell wall upon plasmolysis both in WT ( panel 3 ) and sgn3-3 ( panel 4 ) . Pictures were taken 15 endodermal cells after the onset of CASP1 expression . Scale bars: 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03115 . 00810 . 7554/eLife . 03115 . 009Figure 2—figure supplement 2 . SGN3 genomic construct and sgn3 PI phenotype complementation . ( A ) C-terminal mVenus fusion with a 9 . 4-Kb genomic fragment including intron , 5′UTR , and the upstream neighboring gene used to complement sgn3 . ( B ) Complementation of the sgn3-3 PI phenotype in two independent T2 lines containing the pSGN3::SGN3-mVenus construct shown in A ( n ≥ 10 ) , no PI block was observed in sgn3-3 . DOI: http://dx . doi . org/10 . 7554/eLife . 03115 . 00910 . 7554/eLife . 03115 . 010Figure 2—figure supplement 3 . Schematic illustrating quantification of onset of expression along the root . Counting in Figure 2C ( ‘cells after the onset of elongation’ ) was done as described in Alassimone et al . ( 2010 ) . The onset of elongation was determined as the point where an endodermal cell in a median , longitudinal section reached a length more than twice its width . At this point , cells were counted in order to quantitate relative timing of expression between CASP1 and SGN3 , as illustrated on the schematic . DOI: http://dx . doi . org/10 . 7554/eLife . 03115 . 010 We then generated lines of tagged SGN3 variants in order to investigate its expression and subcellular localization . Only a C-terminal mVenus fusion with a 9 . 4-kb genomic fragment containing intron , 5′UTR , and the upstream neighboring gene , provided full complementation of the mutant phenotype , indicating functionality of the protein fusion and of its regulatory sequences ( Figure 2—figure supplement 2A , B ) . In transgenic lines of this construct , expression of SGN3 could be observed in endodermal cells shortly after the onset of elongation , preceding the onset of CASP1 expression ( Figure 2C , Figure 2—figure supplement 3 ) . This is consistent with a function of SGN3 in establishing correct CASP1 localization . Initially , SGN3 could be found on all cell sides , but it then quickly accumulated in the transversal and anticlinal sides of the plasma membrane , yet became excluded from the CSD itself—a hitherto unknown subcellular localization pattern ( Figure 2D–F ) . In 3D-reconstructions this leads to the appearance of a broader ring-like domain that flanks the more restricted , median CSD ( Figure 2F ) . During the critical phase of CASP domain progression from isolated islands towards a fused ring—at which the sgn3 phenotype becomes apparent–SGN3 is seen to surround individual CASP islands on all sides ( Figure 2G ) . Such localization would fit with a role of SGN3 in promoting fusion of growing CASP islands , such that an uninterrupted band can be formed ( Figure 2H ) . In order to establish the functional relationships between SGN3 and other previously characterized genes , we combined sgn3 with casp1 casp3 double and esb1 single mutants . ESB1 is an extracellular protein that localizes to the site of Casparian strip formation and whose absence causes a phenotype resembling that of sgn3 insofar as both the Casparian strip itself and CASP1-GFP are localized in interrupted bands ( Hosmani et al . , 2013 ) . In contrast to sgn3 , however , esb1 mutants additionally display increased , delocalized deposition of additional autofluorescent cell wall material as well as enhanced and earlier deposition of suberin ( Hosmani et al . , 2013 ) . We found ESB1-mCherry in sgn3 to be both present and correctly localized , that is , specifically accumulating at the remaining CASP1-GFP islands ( Figure 3A ) . Thus , ESB1 does not require SGN3 for its accumulation and localization and the Casparian strip discontinuities in esb1 and sgn3 might be caused by independent mechanisms . Consistently , we found that sgn3 esb1 double mutant formed less and/or smaller Casparian strip patches than each single mutant , indicating additivity of the two mutations in this common aspect of their phenotypes ( Figure 3B , C ) . Intriguingly however , sgn3 esb1 double mutants neither display an increased or delocalized deposition of autofluorescent cell wall material , nor an enhanced and earlier formation of suberin . This indicates a full epistasis of sgn3 over esb1 for this phenotype ( Figure 3B , D ) . The same epistatic relationship was found between sgn3 and casp1 casp3 double mutants ( Figure 3B–D ) . The enhanced and ectopic formation of autofluorescence and suberin is thought to be due to a feedback and crosstalk between cell wall components , connecting correct Casparian strip formation to the further deposition of lignin and suberin lamellae formation ( Hosmani et al . , 2013 ) . The epistasis of sgn3 suggests that the perception/signaling that leads to enhanced autofluorescence and suberin formation in esb1 and casp1 casp3 is mediated by—or at least requires—the SGN3 receptor-like kinase . 10 . 7554/eLife . 03115 . 011Figure 3 . Relations between sgn3 and molecular players in Casparian strip formation . ( A ) ESB1-mCherry localizes to the CASP1-GFP microdomains in sgn3 . ESB1-mCherry and CASP1-GFP are expressed under their own promoters . Pictures are maximum projections of confocal z-stacks . ( B ) Surface view of Casparian strip , visualized by autofluorescence after clearing . Top and bottom panels correspond to the early and late stages of endodermal differentiation , 15 and 30 cells after onset of elongation , respectively . When sgn3 is crossed to casp1 casp3 or esb1 , note the additive effect of these mutants concerning the CS patches ( quantified in C ) . Note also the absence of enhance/ectopic deposition of lignin in sgn3 esb1 and sgn3 casp1 casp3 . Pictures are maximum projections as in A . Strong spiral-like signals are from protoxylem cells . ( C ) Quantification of Casparian strip presence shows an additive effect in sgn3 esb1 and sgn3 casp1 casp3 . Casparian strips were quantified as the percentage of cell wall showing autofluorescence along a line of Casparian strip signal of a given length . n = 10 . ( D ) sgn3 is epistatic to esb1 and casp1 casp3 for ectopic suberin deposition . 5-day-old seedlings were stained with Fluorol Yellow . Quantification was done as in Figure 1E . Bars represent the percentage for each zone along the root ( n = 5 , one representative experiment ) . C , Error bars = s . d . , different letters indicate significant differences between genotypes , determined by analysis of variance ( ANOVA ) and Tukey test as post hoc analyses ( p < 0 . 05 ) . Scale bars: A , B , 10 µm . sgn3 is sgn3-3 , esb1 is esb1-1 , casp1 is casp1-1 and casp3 is casp3-1 . DOI: http://dx . doi . org/10 . 7554/eLife . 03115 . 011 Considering the strength of the barrier phenotype of sgn3 and its inability to initiate potentially compensatory suberization , we were surprised to observe a rather mild growth phenotype of sgn3 under certain conditions ( Figure 4A , Figure 4—figure supplement 1A ) . Yet we noticed that sgn3 growth is extremely sensitive to changes in environmental conditions that are non-stressful to wild type ( Figure 4A , Figure 4—figure supplement 1A ) . We ensured that the endodermal barrier phenotype persisted in developed root systems of rosette stage plants both on soil and on hydroponics , thus excluding that the weak phenotype is caused by an eventual repair of the barrier in older plants ( Figure 4—figure supplement 1B ) . A systematic variation of day length , temperature , and light-intensity revealed that growth differences to wild type are more pronounced at higher temperature , as well as under long-day conditions ( Figure 4B , Figure 4—figure supplement 1C ) . In general , sensitivity of sgn3 to growth conditions is such that we observed everything from a near absence of phenotypes to severely dwarfed plants that did not reproduce when grown in different , standard growth chambers that all well supported wild type growth . We wanted to ascertain that the observed growth phenotypes of sgn3 are indeed caused by the Casparian strip defect in the roots and not by some hypothetical , non-redundant function of SGN3 in shoots . We therefore expressed SGN3 under the control of the CASP1 promoter that shows specific expression in differentiating root endodermis , with some expression in the endodermis of the lower hypocotyl and none detectable anywhere else in the plant . These CASP1::SGN3-GFP sgn3 plants could fully rescue the growth phenotype of sgn3 plants ( Figure 4C ) , demonstrating that it is the root endodermal defect of sgn3 that is causative for the overall growth defect of the mutant . 10 . 7554/eLife . 03115 . 012Figure 4 . sgn3 is sensitive to environmental conditions and displays an altered water transport and root pressure . ( A ) Phenotype of 3-week-old WT , sgn3-3 , and sgn3-4 plants grown at 22 or 24°C in long days . Representative pictures are presented . ( B ) Analysis of shoots fresh weight of WT , sgn3-3 , and sgn3-4 plants ( n = 15 ) grown 3 weeks at different temperatures ( 17 , 21 , or 25°C ) . ( C ) Phenotype of 4-week-old WT , sgn3-3 , sgn3-3/pSGN3::SGN3-mVenus ( lines 8 and 46 ) , and sgn3-3/pCASP1::SGN3-GFP ( lines 3 and 4 ) grown at 24°C in long days . Representative pictures are presented . Fresh weight average from n > 8 plants . ( D ) Transpiration of WT , sgn3-3 , and sgn3-4 plants determined as water loss from 3-week-old plants . Error bars = s . d . ( n = 10 ) . ( E ) Root pressure analysis determined as the volume of xylem sap released in 30 min from decapitated WT , sgn3-3 , and sgn3-4 plants grown in short day condition ( n = 7 ) . ( F ) Guttation was collected from WT and sgn3-3 plants grown for 6 weeks in short day conditions ( n = 15 ) . ( G ) Mean hydrostatic hydraulic conductivity of roots ( Lpr-h ) from WT , sgn3-3 , and sgn3-4 plants . Lpr-h was measured during the daytime . Values correspond to means ± SD ( n > 14 ) . B , C , E , F , G . Error bars = s . d . For multiple comparison , different letters indicate significant differences between genotypes , determined by analysis of variance ( ANOVA ) and Tukey test as post hoc analyses; ( B , p < 0 . 01 ) ( E , G , p < 0 . 05 ) . For single comparison in F , stars ( *** ) indicate significant difference determined by Student test ( p < 0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03115 . 01210 . 7554/eLife . 03115 . 013Figure 4—figure supplement 1 . Impact of environmental conditions on growth of sgn3 mutant . Growth conditions considered as ‘standard condition’ are: temperature 21°C , day length 16 hr , and light intensity of 150 μE . Modification of one of those parameters is indicated , the other parameters remaining unchanged . ( A ) Phenotype of 3-week-old WT , sgn3-3 , and sgn3-4 plants grown at 22 , 24 , or 26°C . ( B ) Lack of the apoplastic endodermal diffusion barrier in sgn3-3 visualized with PI is still present when plants were grown in hydroponics or in soil . ( C ) Analysis of shoot fresh weight of WT , sgn3-3 , and sgn3-4 plants ( n = 15 ) grown 3 weeks at different day length ( 8 hr , 16 hr or 24 hr ) , light intensity ( 30 μE , 90 μE or 300 μE ) or in combination with higher temperature ( 25°C ) . Error bars = s . d . ; different letters indicate significant differences between genotypes , determined by analysis of variance ( ANOVA ) and Tukey test as post hoc analyses ( p < 0 . 01 ) . Scale bars: B , 50 μm . ep , epidermis; co , cortex; en , endodermis; st , stele . DOI: http://dx . doi . org/10 . 7554/eLife . 03115 . 013 The numerous , repeated Casparian strip discontinuities in sgn3 must generate an apoplastic bypass for both water and solutes in the mutant root . Especially under conditions of high transpiration , this would be expected to lead to a strong , uncontrolled and possibly detrimental influx of elements into the transpiration stream . We observed that transpiration rate per surface area in sgn3 mutants are very similar to wild type ( Figure 4D ) revealing that there is no compensatory downregulation of transpiration . Another predicted consequence of interrupted Casparian strips is an inability to build up root pressure , a process necessary for vascular transport of nutrients and water in the absence of transpiration and for transport into non-transpiring organs ( Wegner , 2014 ) . Root pressure build-up in the dead xylem vessels is thought to involve an osmotic gradient , generated by living neighbors , that draws water into the vessels . Pressure build-up , however , is thought to crucially depend on the presence of the Casparian strips as an apoplastic diffusion barrier , without which xylem pressure should quickly dissipate into the cortex . We indeed found that root pressure , as estimated by xylem sap exudation from hypocotyls after decapitation , was lower in the sgn3 mutant ( Figure 4E ) . sgn3 also showed a decreased rate of leaf guttation , a process more indirectly dependent on root pressure , but that can be observed less invasively ( Figure 4F ) . As a more direct measure of apoplastic water flow , we inserted excised roots in a pressure chamber and found a significant increase in root hydrostatic hydraulic conductivity ( Lpr-h ) in sgn3 mutants compared to wild type ( Figure 4G ) . The defects of Casparian strip formation may allow an apoplastic bypass for water in pressurized sgn3 roots thereby explaining its increased Lpr-h . Our findings are the first genetic evidence that Casparian strips can be relevant for root pressure buildup and impose a hydraulic resistance; the magnitude of the observed effects nevertheless indicates that an intact Casparian strip is not an absolute requirement and that unmodified walls and/or suberized cells can generate a sufficiently high resistance to radial waterflow so that the positive root pressure can be maintained . sgn3 might be very useful to dissect the mechanisms of root pressure formation and its role in nutrient transport in the future . In order to identify some potential stress and/or compensatory responses in the sgn3 mutant , we generated a general transcriptional profile of sgn3 mutant rosettes , grown under conditions in which wild type and sgn3 were indistinguishable . Only little differences between sgn3 and wild type were observed . Yet , all of the seven more highly expressed genes in sgn3 were found to be also induced by potassium starvation , suggesting a latent , weak potassium deficiency stress under these conditions ( Figure 5A , Figure 5—source data 1 ) . Based on this indication , we tested well-established transcriptional read-outs of potassium deficiency in roots , the potassium influx transport proteins AKT1 and HAK5 , and found that both were upregulated in sgn3 ( Figure 5B , C ) . 10 . 7554/eLife . 03115 . 014Figure 5 . Potassium homeostasis is affected in sgn3 . ( A ) Expression of genes upregulated in 4 week-old sgn3 leaves . The seven genes presented here are the ones whose expression level was significantly increased in sgn3 ( p < 0 . 15 ) . Those genes were investigated in Genevestigator for responses in leaves to nutritional stresses such as potassium starvation , drought , nitrate deficiency , Fe deficiency , S deficiency , P deficiency , and salt stress . Color-code indicates the effect of growth condition to their expression level ( yellow up , gray unchanged , blue down ) . For numerical values see Figure 5—source data 1 . ( B and C ) Quantitative RT-PCR analysis of AKT1 ( B ) and HAK5 ( C ) transcript levels in WT , sgn3-3 , and sgn3-4 roots ( n = 3 ) . Error bars = s . d , different letters indicate significant differences between genotypes , determined by analysis of variance ( ANOVA ) and Tukey test as post hoc analyses ( p < 0 . 05 ) . ( D ) Overview of ionomic analysis performed in sgn3 leaves ( sgn3-3 , sgn3-4 , and sgn3-19 alleles ) in three independent laboratories ( Hokkaido , Lausanne , Aberdeen ) , using 2 growth systems ( hydroponics or soil ) and 2 day-length conditions ( short or long days ) . Elements were determined by ICP-MS ( Hokkaido , Lausanne 05/12 and Aberdeen ) or ion chromatography ( Lausanne 03/11 ) . Color-code indicates significant changes of accumulation in sgn3 mutants compared to WT ( p < 0 . 05; yellow up , gray unchanged , blue down ) . For the numerical values , see Figure 5—source data 2 . ( E and F ) Phenotype of 3-week-old WT , sgn3-3 , and sgn3-4 plants grown in potassium ( K ) deficiency . Plants were watered from germination with a nutritive solution with or without 1 . 5 mM KNO3 ( +K or −K ) . Representative picture are presented . Arrow indicates a chlorotic leave . ( F ) Occurrence of K deficiency phenotype determined as the percentage of plants displaying at least one yellow leaf ( yellow ) vs percentage of plants displaying only green leaves ( green ) . Error bars = s . d . ; data correspond to the mean of two independent experiments with a total of n ≥ 50 plants . DOI: http://dx . doi . org/10 . 7554/eLife . 03115 . 01410 . 7554/eLife . 03115 . 015Figure 5—source data 1 . Transcriptional differences between wild type and sgn3 shoots . Genes affected in sgn3-3 transcriptomic analysis in leaves ( p < 0 . 15 ) and responses of those genes to abiotic stress . Fold change and p-value marked by asterisk were obtained in Genevestigator for K starvation , drought , nitrate starvation , Fe deficiency , S deficiency , P deficiency , and salt ( Hruz et al . , 2008 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03115 . 01510 . 7554/eLife . 03115 . 016Figure 5—source data 2 . Overview of ionomic experiments . Ionomic analysis performed on sgn3 leaves ( sgn3-3 , sgn3-4 , and sgn3-19 T-DNA insertion lines ) in three different labs ( Hokkaido , Lausanne and Aberdeen ) , two growth systems ( hydroponic and soil ) and two growth conditions ( short days and long days ) . Values are presented as mean ± SD . t-tests were performed to determine the significant differences to WT ( corresponding p-values are indicated ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03115 . 01610 . 7554/eLife . 03115 . 017Figure 5—figure supplement 1 . Ionomic comparision of WT and sgn3 grown under low potassium . Ionomic analysis performed on WT , sgn3-3 , and sgn3-4 plants grown in a poor gravel-like substrate and watered from germination with a nutritive solution with potassium ( +K ) or without ( −K ) . The different elements were analyzed by ICP-MS ( Aberdeen ) with a NexION 300D . Each bar represents the mean ± SD of three biological replicates with n ≥ 17 plants . Different letters indicate significant differences between genotype for a given element , determined by analysis of variance ( ANOVA ) and Tukey test as post hoc analysis ( p < 0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03115 . 017 In order to directly measure how sgn3 root barrier defects affect elemental homeostasis in leaves , we undertook a number of elemental profiling experiments using inductively coupled plasma–mass spectrometry ( ICP-MS ) on rosette leaves in different laboratories under different growth conditions . Surprisingly , concentrations of most elements remained essentially unaltered in sgn3 . Despite the dramatic apoplastic bypass in sgn3 roots , the mutants managed to maintain wild type levels of sodium ( Na ) , calcium ( Ca ) , or boron ( B ) , for example . Among the measured transition elements ( Mo , Co , Mn , Zn , Fe , Cu , Zn ) only zinc showed a significant decrease in some measurements ( Figure 5D , Figure 5—source data 2 ) . However , matching the transcriptional profiling results , we invariably found lower levels of potassium ( K ) in the mutant ( ranging from 1 . 4–3 . 0 fold reduction ) . Inversely , levels of magnesium ( Mg ) and cesium ( Cs ) accumulated to higher levels than in wild type , ranging from 1 . 5–2 . 1 and 1 . 3- to 1 . 4-fold increase , respectively ( Figure 5D , Figure 5—source data 2 ) . These findings nicely corroborated the results of our microarrays , but left us wondering about the absence of potassium deficiency symptoms in sgn3 , since the mutant phenotypes under most conditions are simply those of a reduction and delay in growth . We then tested growth of sgn3 plants on a nutrient-poor , gravel-like substrate , which we watered with a nutrient solution with or without potassium . Wild type plants coped well with conditions of low potassium nutrition , with most plants displaying no or only very weak chlorosis on older rosette leaves , indicative of potassium deficiency ( Figure 5E , F ) . Consistently , wild type also maintained higher potassium concentration than sgn3 under these conditions ( Figure 5—figure supplement 1 ) . The sgn3 mutant in contrast , developed severe chlorosis of its rosette leaf margins , characteristic of strong potassium deficiency ( Marschner and Marschner , 2012 ) . sgn3 therefore has an impaired capacity to accumulate sufficient potassium under conditions of low supply of this element .
The deposition of cell wall material with subcellular precision is crucial for the correct function of many plant cell types ( Roppolo and Geldner , 2012 ) . How this localized deposition is achieved is only rarely understood in any mechanistic detail , with the notable exception of wall deposition in metaxylem cells ( Oda and Fukuda , 2012 ) . Localized wall deposition should implicate the formation of microdomains in the plasma membrane that localize cell wall biosynthetic activity . Walls , in turn , signal back to the cell , and this process informs about mechanical stresses , wall polymer breakdown or for regulated expansion growth ( Cheung and Wu , 2011; Lindner et al . , 2012 ) . The precisely localized , ring-like Casparian strip can be used as a model for studying localized wall deposition ( Lee et al . , 2013 ) . The SGN3 LRR-RLK appears to be involved in both the formation of the CSD and the signaling of defective Casparian strips . Lack of SGN3 causes an inability of CASP1-GFP patches to fuse into a contiguous band , without affecting the positioning of the patches along a median ring , nor the stability or functionality of the already established patches . SGN3 localization to the transversal and anticlinal , but not the periclinal membrane domains further increases the complexity of subdomains present at the endodermal plasma membrane . It embeds the forming CASP domain into a larger subdomain within which the narrower CSD is formed . Direct interaction of CASPs and SGN3 could take place early during CASP accumulation at the membrane and occur at the rims of forming CASP patches , possibly promoting patch growth and fusion . This can now be tested in future experiments . A second , currently unrelated role for SGN3 is the induction of increased lignin and suberin formation in other Casparian strip mutants , such as esb1 or casp1 casp3 . We speculate that SGN3 acts as a kinase that reports CSD integrity , signaling an upregulation of lignin , and suberin production . A surveillance of Casparian strip integrity could be of great importance for unimpeded root function and would parallel the surveillance of tight junctions that has been reported for animal epithelia ( Balda and Matter , 2009 ) . We could demonstrate that exclusive expression of SGN3 in the differentiating root endodermis is sufficient to rescue the overall growth defect of sgn3 . This indicates that the endodermal function that we describe here is the only specific , non-redundant activity of this receptor in the plant . The earlier described function of SGN3/GSO1 in the formation of an embryonic cuticle by contrast is fully redundant with that of its homolog GSO2 . The dramatic cotyledonary phenotype of the double mutant is bound to have secondary consequences on seedling growth , making it very difficult to dissect primary from secondary effects . A recent investigation of root meristem defects of the double mutants , for example , could entirely be explained as secondary consequences of the primary defect in embryonic cuticle formation ( Racolta et al . , 2014 ) . It is nevertheless important to speculate how the two functions of endodermal Casparian strip formation and protodermal cuticle formation could be explained by a common molecular activity . By genetic analysis , the GSO1 GSO2 pair has been assigned to a signaling pathway that depends on an endosperm-expressed subtilase ( ABNORMAL LEAF-SHAPE1 , ALE1 ) ( Xing et al . , 2013 ) . In a currently speculative scenario , subtilases might be involved in processing of peptide hormones that could act as ligands for SGN3/GSO1 , promoting formation of a functional cuticle . Yet , the endodermal , lignified Casparian strip and the protodermal , cutin-based embryonic cuticle are two chemically different cell wall modifications that also show distinct modes and sites of deposition within the primary cell wall . These differences make it difficult to perceive a common denominator for SGN3 action in these two cell types . Moreover , our data suggest that a primary function of SGN3/GSO1 is the formation of a continuous band of CASP membrane proteins . No CASP homologs have been implicated in cuticle formation up-to-now , nor are there any indications for the presence of a highly scaffolded membrane protein domain in the epidermis/protodermis that would resemble the CSD . Another SGN3/GSO1 function that we could establish is the promotion of suberin formation in response to a defective Casparian strip . Enhanced suberin formation in esb1 appears to be entirely dependent on SGN3 activity , although the normal developmental progression of suberin formation is not affected in the mutant . Suberin and cutin are chemically closely related polymers whose monomers are formed by sets of homologous enzymes . A common role of SGN3 might be to mediate a cross-talk between cell wall components , possibly regulating suberin production in endodermis and cutin production in epidermis/protodermis . Identification of the SGN3 ligands and downstream kinase substrates will be crucial for an understanding of the common mechanism that might underlie these apparently divergent functions . Regardless of the fact that SGN3 has other roles in the plant , the sgn3 single mutant defect reported here remains exquisitely specific to the endodermis . This is supported by the fact that it is only in the endodermis where SGN3 expression is non-overlapping with its close homolog , GSO2 ( Tsuwamoto et al . , 2008 ) . Moreover , exclusive expression of SGN3 in the endodermis is able to rescue the overall growth defects observed in aerial parts ( this study ) . In addition to its specificity , the sgn3 phenotype turns out to be strong and robust , that is , it displays penetration of tracer along the whole length of a seedling root and the phenotype is stable in the different growth conditions that we tested . Finally , sgn3 does not show any of the enhanced suberin formation that has been observed in other mutants and which confounds interpretation of phenotypes . We therefore propose that the sgn3 mutant is uniquely suited to obtain insights into the specific role of Casparian strips in plants . This mutant might be very useful for investigating not only root nutrient and water uptake , but also for addressing the role of the endodermal barrier in pathogen resistance , hormonal transport , stress , or general growth responses . Our first analysis of the sgn3 mutant has revealed an unexpected robustness of Arabidopsis growth towards a complete lack of its main root diffusion barrier . While still able to grow and complete its life cycle , the mutant nevertheless displays greatly exacerbated growth reductions to non-stressful changes in environmental conditions . This indicates the presence of homeostatic backup systems in the plant that are incompletely compensating for the absence of Casparian strips . Homeostasis of many elements is kept in the sgn3 mutant , and it is intriguing that it is specifically potassium that shows a reproducibly strong decrease in concentration . Potassium ( K ) is the only macronutrient ( N , P , K , S ) that is not integrated into polymers and that remains highly mobile within the plants . It is being used as a counter ion in many transport process and is recycled back to the root along with the phloem sap . At the same time , concentrations of potassium required for plant growth are high and potassium often needs to be concentrated 10–100 fold over the concentration found in the soil ( Marschner and Marschner , 2012; Wang and Wu , 2013 ) . It therefore makes a lot of sense that potassium is an element that cannot be kept at sufficiently high levels in Casparian strip defective plants . We propose that potassium located in the stelar apoplast is continuously lost to the outer cell layers and the soil in the absence of functional Casparian strips . Upregulation of potassium influx carriers in cortex/epidermis , as we have observed in sgn3 roots , would certainly be able to recuperate large parts of the potassium being lost . However , a chronic condition of very low external potassium would lead to a breakdown in this transporter-mediated ‘potassium recycling’ and eventually cause the deficiency phenotypes that we observe . From this scenario , it follows that there should be a higher rate of potassium flux out-of and into the root stele in a sgn3 mutant and that potassium homeostasis in the mutant is extremely dependent on increased activity of potassium transporters . We are currently working on substantiating this ‘potassium recycling’ model by specific interference with and visualization of potassium transporters in the mutant . Further analysis of sgn3 will be instructive for a better understanding of root potassium transport in general and this analysis should also be extended to other macronutrients in the future . What could be the reasons for the lack of effects on many other elements ? It can be expected that similar compensatory responses as seen for potassium exist for other elements . If these elements need to be less concentrated over their external concentration—or if they are less mobile within the plant—the compensatory responses might be sufficient for maintaining homeostasis in the mutant . Interestingly , we found an increase in the levels of the divalent cation magnesium , while concentrations of the divalent calcium , for example , were not affected in sgn3 mutants . This could be explained by a lower concentration gradient between the xylem sap and the external environment and a lower mobility of Calcium within the cell wall space ( Clarkson , 1984; White and Broadley , 2003 ) . Nevertheless , calcium concentration within the symplast is kept extremely low for cellular signaling purposes . Therefore , calcium delivery to aerial tissues is thought to occur either entirely through the apoplast–using the few apoplastic ‘bypasses’ of the endodermis at the root tip or at lateral root emergence sites ( White , 2001 ) —or to employ a very short symplastic route passing exclusively through endodermal cells ( Clarkson , 1984 ) . In both scenarios , it is surprising that the opening of a massive apoplastic bypass in the sgn3 mutant has no effect on calcium homeostasis , while it did affect levels of magnesium . Possibly , magnesium is more mobile in the apoplast than Calcium and undergoes increased apoplastic transport into the transpiration stream in sgn3 ( Thibault and Rinaudo , 1985 ) . Identifying conditions in which calcium homeostasis also breaks down in sgn3 would greatly advance our understanding of the calcium uptake mechanisms in plants . The sgn3 mutant represents a powerful new tool to better understand root uptake and transport mechanism of most other plant nutrients as well . Our initial analysis of the sgn3 mutant challenges the current notions of the role of the Casparian strip as a required barrier for most or all elements . The ability to selectively disrupt this strict endodermal barrier will allow for direct investigations of other mechanisms that maintain and buffer ion homeostasis . Unmodified cell walls , by their properties as ion exchange matrices , in conjunction with vacuolar storage of elements , can play pivotal roles in maintaining homeostasis of many elements . Cytoplasmic magnesium levels , for example , might well remain unaltered in sgn3 , the increased magnesium levels being entirely absorbed by the vacuole . Having taken away the ‘first lock’ of the Casparian strip now puts the focus on the existence of ‘second locks’ that can be studied using the sgn3 mutant as a sensitized background .
Arabidopsis thaliana ecotype Columbia was used for all experiments . For details of knockout mutants , see Table 1 . 10 . 7554/eLife . 03115 . 018Table 1 . Details of knockout mutantsDOI: http://dx . doi . org/10 . 7554/eLife . 03115 . 018Gene numberAccessionMutant numberMutant nameGenotyping primer sequenceReferencesAT4G20140Col-0SALK_043282sgn3-3LP: ATTCTACGAGCCTTCCCATTC RP: CGCAGTGAACACAGTGAGATCPresent workAT4G20140Col-0SALK_064029sgn3-4 or gso1-1LP: CTCGGCTCCCTCGTTAATATC RP: GTTACCTAAACTGGCGGGAAGTsuwamoto et al . ( 2008 ) The Plant JournalAT4G20140Col-0SALK_103965sgn3-19LP: TCCATTATGTGGTTCGAGCTC RP: CTTGTAAACCTTCCCAGAGCCPresent workAT2G28670Col-0n . a . esb1-1n . a . Baxter I et al . ( 2009 ) PLOS Genet , Lahner B et al . ( 2003 ) Nat BiotechnolAT2G36100Col-0SAIL_265_H05casp1-1LP: GCGTTTCAGTACGTCCCTTC RP: CACGTGAGGGAAGTGAGTCTCRoppolo et al . ( 2011 ) NatureAT2G27370Col-0SALK_011092casp3-1LP: GACTCTTCCTTTCTTCACTC RP: GACCAACACAACCGTACGAACRoppolo et al . ( 2011 ) Nature For generation of expression constructs , Gateway Cloning Technology ( Invitrogen , Carlsbad , CA ) or standard molecular biology procedures were used . pESB1::ESB1-mCherry plasmid ( Hosmani et al . , 2013 ) was transformed in sgn3-3 background . pSGN3 ( 5583 bp before ATG ) ::SGN3 gDNA-mVenus was cloned by integrating a 9 . 4-kb genomic fragment including the intron and the 5′UTR into a Basta-resistance pGREENII vector containing mVenus . pCASP1::NLS-GFP-GUS and pCASP4::NLS-GFP-GUS reporters were cloned into a Basta-resistance pGREENII vector and contain 1048 bp and 695 bp promoter fragments , respectively . pCASP1::SGN3 cDNA-GFP was cloned with Gateway using a 1207 promoter fragment followed by SGN3 cDNA without the unique intron and GFP . pCASP1 ( 1207 bp ) ::PDR6 gDNA-Venus-4G ( 4 glycine extension ) was cloned using Gateway in a pB7m34 GW , 3 expression vector ( http://gateway . psb . ugent . be ) . pCASP1 ( 1207 bp ) ::mCherry-SYP122 was constructed with Gateway by cloning the AtSYP122 ( At3g52400 ) cDNA into pH7m42 GW , 3 . pCASP1::CASP2-GFP-4G , pCASP1::CASP3-GFP-4G , pCASP1::CASP4-mCherry and pCASP1::CASP5-GFP-4G were cloned with Gateway using the pB7m34 GW , 3 expression clone . Transgenic plants were generated by introduction of the plant expression constructs into an Agrobacterium tumefaciens strain GV3101 and transformation was done by floral dipping ( Clough and Bent , 1998 ) . For in vitro assays , plants were germinated on 0 . 5 MS ( Murashige and Skoog ) agar plates after 2 days in dark at 4°C . Seedlings were grown vertically in growth chambers at 22°C , under long days ( 16-hr light/8-hr dark ) , 100 μE light , and were used at 5 days after shift to room temperature . For microarray , transpiration , hydraulic conductivity , and ionomic analysis in Lausanne ( May 2012 ) assays , plant were germinated on hydroponic conditions after 2 days in dark at 4°C . Nutrient solution was changed weekly and contained 0 . 5 mM KNO3 , 0 . 25 mM Ca ( NO3 ) 2 , 1 mM KH2PO4 , 1 mM MgSO4 , FeNH4-EDTA 0 . 1 mM , KCl 50 μM , H3BO3 30 μM , MnSO4 5 μM , ZnSO4 1 μM , CuSO4 1 μM , ( NH4 ) 6Mo7O24 0 . 1 μM . Light/dark cycle was 16 hr/8 hr , 23°C/18°C , 100 μE light . For xylem sap and guttation assays , plants were germinated after 2 days in dark at 4°C on soil in short day condition and irrigated with water . Light/dark cycle was 10 hr/14 hr , 22°C/18°C , 100 μE light . For phenotypic assays at different light intensities ( 30 , 90 , 150 , or 300 μE ) , temperatures ( 17 , 21 , or 25°C ) , or day lengths ( 8 , 16 or 24 hr ) , plants were germinated after 2 days in dark at 4°C on soil and irrigated with water . The standard condition was: light/dark cycle 16 hr/8 hr , 21°C/19°C with light intensity of 150 μE . Temperature difference between day and night was always kept at 2°C . For ionomic assays plants were grown in different laboratories as indicated . For the Hokkaido experiments plants were grown hydroponically as described in Takano et al . ( 2001 ) , with slight modifications . Environmental parameters in a growth chamber are as follows: 10-hr/14-hr light/dark cycle , 22°C under fluorescent lamps , 70% humidity . The seeds were sown on rockwool and grown supplied with hydroponic media ( Fujiwara et al . , 1992 ) supplemented with 50 µM Fe-EDTA . After 15 days , the plants were additionally supplied with 3 µM CsCl and 10 µM SrCl2 . The media were changed twice a week . For the Aberdeen experiment plants were grown in short day conditions , in Bulrush multi-purpose compost ( http://www . bulrush . co . uk/retail-range/all-purpose-composts . html ) spiked with various elements as detailed in Lahner et al . ( 2003 ) , and bottom watered using ¼ Hoagland solution containing Fe as 10 μM Fe-HBED as described in Baxter et al . ( 2008 ) . 1–2 adult rosette leaves were harvested 36 days after planting . For the Lausanne experiment in soil ( March 2011 ) , plants were grown in long days 16-hr/8-hr light/dark , 23°C/19°C , and leaves were harvested after 5 weeks . For potassium deficiency assay plants were germinated after 2 days in dark at 4°C on a poor gravel-like substrate ( OIL DRI US-Special ) watered with a ½ Hoagland based solution containing : 1 . 5 mM NH4NO3 or 1 . 5 mM KNO3 ( −K or +K ) . Confocal laser scanning microscopy experiments were performed either on a Leica SP2 , a Zeiss LSM 700 , a Zeiss LSM 710 , or a Zeiss LSM 710 NLO 2-Photon microscope . Excitation and detection windows were set as follows: Leica SP2: GFP 488 nm , 500–600 nm; propidium iodide 488 nm , 600–700 . Zeiss LSM 700: GFP/mVenus 488 nm , 490–555 nm; mCherry/propidium iodide 555 nm , SP 640 . LSM 710: GFP/mVenus 488 nm , 495–554 nm; mCherry/propidium iodide 561 nm , 573–681 nm . Zeiss LSM 710 NLO 2-Photon: GFP/mVenus , 960 nm , 500–550 nm ( NDD ) . For visualization of the apoplastic barrier , seedlings were incubated in the dark for 10 min in a fresh solution of 15 mM ( 10 mg/ml ) Propidium Iodide ( PI ) and rinsed two times in water ( Alassimone et al . , 2010 , Naseer et al . , 2012 ) . For quantification , ‘onset of elongation’ was defined as the point where an endodermal cell in a median , longitudinal section reached a length more than twice its width . From this point , cells in the file were counted until the PI signal was blocked in the endodermal cells . For plants grown in hydroponic and in soil , roots of 13 and 14 days were used , respectively . Casparian strip autofluorescence after clearing was visualized as described ( Alassimone et al . , 2010 , Naseer et al . , 2012 ) . Fluorescence recovery after photobleaching ( FRAP ) was performed with 5-day-old seedlings and imaged with a Leica SP2 confocal microscope as described ( Roppolo et al . , 2011 ) . Plasmolysis was induced by incubating 5-day-old seedlings for 1 hr in 0 . 8 M mannitol and then mounted in the same solution . Plasmolyzed cells were imaged at 15 cells after the onset of CASP1 expression . For time-lapse imaging of CASP1::CASP1-GFP in Col-0 and sgn3-3 , 5-day-old seedlings were into a Lab-Tek II chambered coverglass ( Nunc ) and covered with a small block of agar to prevent drying . Subsequently , slides were mounted and imaged on an upright Zeiss LSM710 NLO confocal microscope . Excitation was provided by a Ti-Sapphire Chameleon II Ultra ( Coherent ) with 960 nm and fluorescence was detected using non-descanned detection ( NDD ) between 500 and 550 nm . Image stacks were taken every 15 min for a total time of 16 hr . Confocal images were analyzed and contrast and brightness were adjusted with the FIJI package ( http://fiji . sc/Fiji ) and Adobe Photoshop CS5 . For the percentage quantification of Casparian strips , autofluorescence after clearing was performed on ten seedlings for each genotype investigated . For each individual seedling , Casparian strips were imaged by confocal microscopy by doing one z-stack and measurements were done using ImageJ . The percentage of Casparian strip was the portion where autofluorescence was visible along a line showing Casparian strip signal . A minimum of 3 . 5 mm of Casparian strip was quantified in each genotype . Suberin lamella was observed in 5-day-old roots after Fluorol Yellow staining as described in Naseer et al . ( 2012 ) . Seedlings were incubated in Fluorol Yellow 088 ( 0 . 01% wt/vol , lactic acid ) at 70°C for 30 min , rinsed with water , and counterstained with aniline blue ( 0 . 5% wt/vol , water ) at RT for 30 min in darkness , washed , mounted on slides with glycerol and observed with epifluorescence microscope . Suberin pattern were observed and counted from the hypocotyl junction to the onset of endodermal cell elongation . Three distinct patterns were considered: continuous suberin lamellae , discontinuous suberin lamellae ( corresponding to the area where suberin lamellae establish ) , and non- suberized ( corresponding to the young part of the root ) . Experiments were repeated at least four times . For ionomic assays plants were grown in different laboratories as indicated . For the Hokkaido and Lausanne experiments ( hydroponic conditions ) , the shoots of plants were harvested , dried in an air incubator at 60°C for more than 60 hr , and the dry weights were measured . The tissues were digested with 3 ml of 61% HNO3 ( for boron determination; Wako Pure Chemicals , Osaka , Japan ) in a tube at 110°C in a DigiPREP apparatus ( SCP Science , Quebec , Canada ) until complete dryness . The residues were dissolved in 10 ml of 2% HNO3 and analyzed for elements by inductively coupled plasma mass spectrometry ( ICP-MS ) ( ELAN , DRC-e; Perkin–Elmer , Waltham , MA , USA ) . For Aberdeen assays , the leaves were cleaned by rinsing with ultrapure water and placed into Pyrex digestion tubes . Samples were dried in an oven at 88°C for 20 hr . After cooling , seven reference samples from each planted block were weighed . The samples together with blank controls were digested with 0 . 90 ml concentrated nitric acid ( Baker Instra-Analyzed; Avantor Performance Materials ) and diluted to 10 . 0 ml with ultrapure water ( 18 . 2 MΩcm ) . The internal standard Indium ( In ) was added to the acid prior to digestion for monitoring technical errors and plasma stability in the ICP-MS instrument . After samples and controls were prepared , elemental analysis was performed with an ICP-MS ( NexION 300D; PerkinElmer ) coupled to Apex desolvation system and SC-4 DX autosampler ( Elemental Scientific Inc . , Omaha , NE , USA ) , monitoring these elements: B , Na , Mg , P , K , Ca , Mn , Fe , Co , Cu , Zn , Sr , and Mo . All samples were normalized to calculated weights , as determined with a heuristic algorithm using the best-measured elements , the weights of the seven weighed samples and the solution concentrations ( Lahner et al . , 2003 ) , detailed at www . ionomicshub . org . For ionic chromatography , for each sample one rosette leaf of a 5-week-old plant was harvested , put in 30 ml Nanopure water and incubated at 80°C for 1 hr . Each sample was then filtered with a 0 . 45-µm non-pyrogenic sterile filter ( Sarstedt , Germany ) . The samples were then diluted to reach a final concentration of 0 . 83 mg fresh weight per 1 ml . Ion concentrations were obtained with an ion chromatography system ( ICS-1100/2100 , Dionex-Thermo Fischer ) . All statistical analyses were done in the R environment ( http://www . R-project . org/ ) . Binary comparisons between wild type and sgn3 were performed using Student's t-test . Multiple comparisons between wild type , sgn3-3 , and sgn3-4 mutants , one-way ANOVA was performed , and Tukey's test was subsequently used as a multiple comparison procedure . The different features were predicted using the following web resources . Signal peptide , Leucine-rich repeats , transmembrane domain , and kinase domain: http://www . uniprot . org/uniprot/C0LGQ5; LRR N-terminal domain: http://pfam . sanger . ac . uk/search/sequence; kinase sub-domains were defined with reference to Hanks and Hunter ( 1995 ) and by aligning to the BRI1 kinase domain ( Vert et al . , 2005 ) . Borders between kinase sub-domains were set manually by comparison of amino acids in the respective border regions for PKA C alpha and BRI1 . In case of conflict , the BRI1 annotation was considered as more meaningful because of its plant origin . | Plant roots forage in the soil for minerals and water , but they must also provide a barrier that stops these nutrients leaking back out of the plant and stops microbes invading and causing disease . The endodermis—an inner layer of cells that surrounds the veins that run along the middle of a root—acts as such a barrier in young roots . Polymers that repel water are deposited between the cells in the roots of almost all vascular plants—which include ferns , conifers , and flowering plants—to form a band around the endodermis called the ‘Casparian strip’ . This strip seals off the young roots and stops water moving through the gaps between plant cells , but still allows minerals , nutrients , and water to be transported through the root cells and into the plant . However , the importance of this structure has yet to be tested due to the lack of mutant plants without a Casparian strip . Pfister et al . now report that deleting the gene that encodes a protein called SCHENGEN3 in the model plant Arabidopsis thaliana causes the Casparian strip to be interrupted by irregularly sized holes . This protein is normally found at high levels in the root endodermis , where it is embedded into the cell membranes . Pfister et al . also showed that without the SCHENGEN3 protein , other proteins called CASPs—that normally mark out a stripe around the root cells where the Casparian strip will form—only accumulated in discontinuous patches . Further experiments revealed that deleting the gene for SCHENGEN3 does not cause general problems in delivering the CASP proteins to the cell membrane; instead , it specifically stops the CASP proteins from forming a single , uninterrupted stripe . Unexpectedly , disrupting the Casparian strip did not appear to hinder many of the functions of a root . The mutant plants could still take up water and nutrients , and the leaves of mutant plants had normal levels of many essential minerals—with the exception of potassium . The level of this mineral was much lower in mutant plants without the SCHENGEN3 protein . Pfister et al . suggest that in plants that lack an intact Casparian strip , potassium is continuously leaked from the root into the soil . These findings reveal that in Arabidopsis , at least , the Casparian strip might not be as important as once thought for helping the plant to take up and accumulate water and nutrients . Further work is now needed to uncover the as yet unknown backup systems that might be able to compensate for the loss of this structure . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology",
"cell",
"biology"
] | 2014 | A receptor-like kinase mutant with absent endodermal diffusion barrier displays selective nutrient homeostasis defects |
Models based in differential expansion of elastic material , axonal constraints , directed growth , or multi-phasic combinations have been proposed to explain brain folding . However , the cellular and physical processes present during folding have not been defined . We used the murine cerebellum to challenge folding models with in vivo data . We show that at folding initiation differential expansion is created by the outer layer of proliferating progenitors expanding faster than the core . However , the stiffness differential , compressive forces , and emergent thickness variations required by elastic material models are not present . We find that folding occurs without an obvious cellular pre-pattern , that the outer layer expansion is uniform and fluid-like , and that the cerebellum is under radial and circumferential constraints . Lastly , we find that a multi-phase model incorporating differential expansion of a fluid outer layer and radial and circumferential constraints approximates the in vivo shape evolution observed during initiation of cerebellar folding .
Recent work to elucidate the mechanics of brain folding has primarily focused on the human cerebral cortex and involved models of directed growth , axonal tension , or differential expansion of elastic materials that generate compressive forces to drive mechanical instabilities leading to folding ( Tallinen et al . , 2014; Ronan et al . , 2014; Bayly et al . , 2013; Xu et al . , 2010; Hohlfeld and Mahadevan , 2011; Bayly et al . , 2014; Lejeune et al . , 2016; Karzbrun et al . , 2018 ) . Current elastic material models are able to create three-dimensional shapes strikingly similar to the final folds seen in the adult human cortex ( Tallinen et al . , 2016 ) . A recent multi-phase model ( Engstrom et al . , 2018 ) that includes elastic and fluid-like layers , differential expansion and radial constraints takes into consideration that multiple factors could lead to folding in the developing brain . However , the cell and tissue level mechanics actually present at the initiation of folding have not been considered or defined , as technological limitations are significant in animals with a folded cerebrum . The murine cerebellum has a simple alignment of 8–10 stereotypical folds along the anterior-posterior axis . Combined with the genetic tools available in mouse this allows for precise developmental interrogation to identify and analyze the in vivo cellular and tissue level behaviors driving growth and folding . The developing cerebellum is distinct from the cerebral cortex , as it has a temporary external granule cell layer ( EGL ) of proliferating granule cell precursors that cover the surface and generate growth primarily in the anterior-posterior ( AP ) direction ( Leto et al . , 2016; Legué et al . , 2015; Legué et al . , 2016 ) . During development a thickening occurs in the EGL at the base of each forming fissure , termed anchoring center ( AC ) ( Sudarov and Joyner , 2007 ) , whereas in the adult cerebellum the inner granule cell layer ( IGL ) , generated by the EGL during the first two weeks of life , is thinnest at the ACs . Previous work on cerebellar folding utilized a tri-layer elastic model incorporating the EGL , the adjacent molecular layer , rich in axons and dendrites , and the IGL ( Lejeune et al . , 2016 ) . However , neither the molecular layer nor the IGL are present when folding is initiated in the embryo . Therefore we argue that a bilayer system consisting of the EGL and underlying core , is a more appropriate approximation for cerebellar folding . Here we show that cerebellar folding emerges from differential expansion between an un-patterned , rapidly expanding EGL and an underlying core . Additionally , we demonstrate that the measured stiffness differential , compressive forces , and the thickness variation in the EGL are all inconsistent with traditional elastic wrinkling models driven by differential growth . Furthermore , we demonstrate that the expansion of the EGL is uniform , and fluid-like , and that the cerebellum is under radial and circumferential constraints when folding initiates . Lastly , we constrain the recent multi-phase model with our in vivo data and find we can capture the temporal shape evolution seen during mouse cerebellum folding initiation . The implications of our findings for human cerebral cortex folding are discussed .
It is well known that differentially expanding bilayer systems can wrinkle to relax building stress ( Richman et al . , 1975; Nelson , 2016; Hannezo et al . , 2012; Shyer et al . , 2013; Wiggs et al . , 1997 ) . We reasoned that in the cerebellum the EGL could behave as a quickly expanding outer layer and its attachment to a more slowly growing core could generate forces that result in a wrinkling-like phenotype . To test whether the cerebellum has differential expansion between the two layers , we measured the expansion of the EGL and the core during the time of initiation of folding from midline sagittal sections ( Figure 1a–d ) . Unlike the cerebral cortex , the unfolded murine cerebellum is a simple cylinder-like structure elongated in the medio-lateral axis ( Figure 1e ) ( Szulc et al . , 2015 ) . All folds in the medial cerebellum ( vermis ) are aligned in the same axis allowing 2-D measurements to estimate expansion in the anterior-posterior axis of the vermis . Therefore the length of the surface of the EGL was used as a measure of the cerebellum surface area and the area of the core as an approximation of cerebellum volume ( Figure 1d ) , and measurements were made each day from embryonic day 16 . 5 ( E16 . 5 ) through postnatal day 0 ( P0 ) . In cross-section the unfolded cerebellum approximates a semicircle , therefore we reasoned that if the cerebellum were to remain unfolded then the ratio of expansion between the length of the EGL and the area of the core should approximate the ratio of the circumference of a semi-circle to its area . Of significance , we found that at E16 . 5 and E17 . 5 the ratios of growth between the EGL and core closely approximated the expansion of a semi-circle . However , at E18 . 5 and P0 the expansion rate of the EGL was greater than the rate of core expansion ( Figure 1f ) . Thus we uncovered that the cerebellum does indeed go through a phase of differential expansion . We next determined whether differential expansion correlates with when folding occurs by calculating a folding index ( the convex curvature of the EGL divided by the length of the EGL ) at each stage ( Mota and Herculano-Houzel , 2015 ) . Indeed , we found that the cerebellum remains unfolded during the initial proportional expansion between the EGL and core and only folds when the differential expansion is initiated ( Figure 1g ) . These results provide quantitative evidence that cerebellar folding involves tissue level mechanical forces arising from differential expansion . Since there is differential expansion between the EGL and the core and as this type of expansion is the driver of elastic bilayer models we tested whether the properties of cerebellar tissue are consistent with the requirements and predictions of such models . Briefly , the initial resulting wrinkling instability defines the distance between folds as the initial sinusoidal undulations increase in amplitude to ultimately turn into lobules . The folding wavelength depends on the thickness of the external layer ( EGL ) and the ratio of the stiffness of the two layers ( EGL/core ) . In particular , for a planar geometry , with the stiffness of the external layer defined as Eo , the stiffness of the core as Ei , and the thickness of the external layer denoted as t , the folding wavelength λ is given by Allen ( 1969 ) λ=2 π t 13EoEi1/3 . If the length of the system is l , then the number of folds is inversely proportional to the thickness of the EGLn=lλ∝ltEiEo1/3 . We explored a standard elastic bilayer model in a circular geometry using the observed ratio of thickness of the EGL to radius of the cerebellum near the onset of shape change ( E16 . 5 ) and invoked a neo-Hookean elastic solid for both layers ( Zhao and Zhao , 2017 ) . The resulting shape change was studied as a function of the ratio of the layer stiffness values ( Figure 2a ) . We found that to produce the observed number of folds ( three in the semi-circular cerebellum and six in the circular model ) at initiation of folding through wrinkling based models constrained by our measurements of the embryonic cerebellum , a large stiffness ratio was required of around 50 . To map the stiffness contrast in the cerebellum we used scanning acoustic microscopy ( SAM ) to measure the bulk modulus of the cerebellum daily from E16 . 5 to P18 . 5 ( Figure 2b–c , Figure 2—figure supplement 1 ) using established methods ( Rohrbach et al . , 2015; Rohrbach et al . , 2018; Rohrbach and Mamou , 2018 ) . For small deformations , the instantaneous bulk modulus should linearly relate to the stiffness and , therefore , the ratio of the instantaneous bulk moduli should scale similarly to the ratio of stiffnesses ( assuming the same Poisson’s ratio for the EGL and for the core , neither of which have been directly measured ) . While this qualitative approach and SAM tissue preparation protocols may not be able to produce the absolute values of the elastic properties of the tissues , it can give a reasonable indication of the relative stiffnesses of different parts in the cerebellum . Using this estimation , we found that the EGL has a slightly higher instantaneous bulk modulus than the core at all stages measured . Unsurprisingly , the ratio ( ~1 . 05:1 ) was not close to being sufficient to produce a folding wavelength similar to that in the cerebellum ( Figure 2d ) . Consistent with our finding , small modulus contrasts have been reported for other brain regions with multiple loading modes , such as shear , compression , and tension ( Xu et al . , 2010; Lejeune et al . , 2016; Budday et al . , 2017 ) . Elastic material models with graded growth profiles have been developed that predict folding of cerebral cortex without a large stiffness differential ( Tallinen et al . , 2014 ) . However , these models are still bound by other measurable requirements as discussed below . Elastic bi-layer wrinkling models predict compressive forces in the outer layer . Simulations performed of cuts through the outer layer and into the inner layer predict that upon relaxation the outer layer should not open ( Figure 2e ) . We tested whether this prediction reflects the biology using surgical dissection blades to make radial cuts across the meninges , EGL , and into the core of live E16 . 5 tissue slices . Time-lapse imaging revealed that , in contrast to the prediction , the EGL opens as well as part of the underlying cut in the core ( Figure 2f–h , Figure 2—figure supplement 2a–c , and Video 1 ) . This result indicates there is circumferential tension within the outer layers of the cerebellum . This finding also rules out the elastic models with graded growth profiles as they predict compressive forces in the outer region as well . The elastic bi-layer model requires the EGL to be thinnest at the base of each AC , which are the lowest parts of the cerebellar surface . Thus , the EGL should have an ‘in-phase’ thickness variation . Without this feature , a purely elastic model – bi-layer based or even graded growth profile based – cannot be in mechanical equilibrium ( in the quasistatic limit ) ( Engstrom et al . , 2018 ) . However , we previously reported that the embryonic EGL is thickest in the ACs when folding initiates , that is it has an ‘out-of-phase’ thickness variation ( Sudarov and Joyner , 2007 ) . To validate this observation , we quantified the thickness variations in the EGL centered at the ACs present at E16 . 5–18 . 5 . Not all cerebella have visible AC at E16 . 5 . However in the subset that do and in the three ACs present at E17 . 5 , the EGL was found to be 1 . 2–1 . 4 times thicker in the ACs than in the surrounding EGL ( Figure 2i–l and Figure 2—figure supplement 3 ) . Moreover , the thickness ratio increased to 1 . 7 times at E18 . 5 ( Figure 2l ) . As described above , the final thickness variations of the IGL ( as well as the molecular layer ) of the cerebellar cortex are in-phase , just as the layers of the adult cerebral cortex . These results further show that traditional elastic wrinkling models cannot capture the initiation of cerebellum folding , and highlight the importance of making biological measurements at the time of folding rather than when it is complete . As elastic bi-layer models do not align with the biology of cerebellar folding , we looked for other drivers of morphometric changes . Since the EGL drives the majority of cerebellar growth ( Leto et al . , 2016; Legué et al . , 2015; Legué et al . , 2016 ) , we first tested whether regional differences in EGL proliferation rates are present that could influence the folding pattern of the cerebellum . Proliferation rates ( S phase index ) were measured in the EGL during folding initiation ( E16 . 5 and E17 . 5 ) in the inbred FVB/N strain to reduce variation between samples . First we asked if the regions that will give rise to distinct sets of lobules have different rates of proliferation that could contribute to the larger and smaller sizes that the lobules ultimately attain . We focused on the anterior cerebellum that divides into a larger region with lobules 1–3 ( L123 ) and smaller region ( L45 ) , as well as the central area that comprises lobules 6–8 ( L678 ) of the cerebellum ( Figure 3a–b ) . The more posterior cerebellum does not consistently fold at this stage , thus measurements were not included . Interestingly , we found that the proliferation rates were similar in the three regions at E16 . 5 ( Figure 3c ) . The EGL proliferation rate at E17 . 5 in L678 was slightly reduced compared to the L123 region , but no other differences were found ( Figure 3d ) . Thus proliferation is uniform just before initiation of folding and the small difference found during folding does not correlate with lobule size . This result indicates that lobule size is not determined by modulating the levels of proliferation at the onset of folding . Rather , lobule size could be set by both the timing of invagination , and the distance between ACs as granule cell precursors in one lobule do not cross the surrounding ACs to contribute to an adjacent lobule ( Legué et al . , 2015 ) . Each AC is first detected as a regional inward thickening of the EGL ( Sudarov and Joyner , 2007 ) ( Figure 2i–l and Figure 2—figure supplement 3 ) . We measured the proliferation of the EGL specifically within the forming AC regions to test whether altered proliferation rates could explain the thickenings and therefore the initiation of an AC . We found the rate of proliferation within each forming AC region at E16 . 5 and E17 . 5 was the same as in the surrounding EGL ( Figure 3e , f ) , thus proliferation within all regions of the EGL at the initiation of folding is uniform . Furthermore , regional modulation of proliferation does not form or position the ACs . At E18 . 5 , after the initiation of folding , we found that the rate of proliferation was significantly lower in the L678 region compared with the L123 and L45 regions ( Figure 3—figure supplement 1a ) . However , proliferation within the ACs at E18 . 5 remained uniform with the surrounding regions ( Figure 3—figure supplement 1b ) . Since ACs compartmentalize the EGL , our results show that regional differences in proliferation rates arise in lobule regions after initiation of folding , which thus could be important for determining the ultimate size of the folds . Changes in cell size and shape have been shown to induce morphological changes ( Mammoto and Ingber , 2010; Harding et al . , 2014; Stemple , 2005; He et al . , 2014 ) . To test if regionally specific regulation of cell shape or size directs folding , we fluorescently labeled cell membranes of scattered granule cell precursors ( GCPs ) in the EGL using genetics ( Atoh1-CreER/+; R26MTMG/+ mice injected with tamoxifen two days prior to analysis ) . We then segmented the cells in 3D and quantified their sphericity ( Figure 3g ) . We discovered that GCPs in the EGL take on a large variation of shapes and sizes at E16 . 5 and E18 . 5 . However , we found no difference in cell shape in the different lobule regions of the EGL or between the AC areas and the surrounding EGL at each age ( Figure 3h , i ) . Cell size was uniform at both stages except for a slight reduction in L678 at E16 . 5 when compared with L123 and the AC regions . However , the size of cells is reduced at E18 . 5 compared to E16 . 5 ( Figure 3j , k and Figure 3—figure supplement 1c ) . Thus , the proliferating GCPs that drive expansion of the EGL have both uniform proliferation rates and similar shapes and sizes across the lobule regions defined by the first three ACs at folding initiation . The EGL is traversed by fibers of Bergmann glial and radial glial cells ( Leung and Li , 2018; Yuasa , 1996; Yamada and Watanabe , 2002 ) . We tested whether the fibers are distributed in patterns that could locally change the physical properties of the EGL and induce invaginations . Genetics was used to fluorescently label cell membranes of scattered glial cells ( nestin-creER/+;R26MTMG/+ mice injected with tamoxifen at E14 . 5 ) ( Figure 4a ) . Fibers crossing the EGL at E16 . 5 were counted in sagittal slices and aligned relative to the ACs ( Figure 4b ) . This analysis showed that the Bergmann glial and radial glial fibers are distributed evenly along the AP axis of the EGL , and therefore are not directing the positions where folding initiates based on an uneven regional distribution . Tension based folding models suggest constraints from axons and other fibers could direct folding ( Xu et al . , 2010; Van Essen , 1997 ) . Since the cerebellum is under circumferential tension , as demonstrated above , we examined evidence of radial tension between the EGL and the ventricular zone ( VZ ) at the initiation of folding . Cuts were made in live E16 . 5 tissue slices between the EGL and VZ running approximately parallel to them so that they cut across radial fibers in the anterior cerebellum ( Figure 4c ) . As predicted , after cutting the tissue relaxed revealing tension directed radially within the cerebellum ( Figure 4d , e and Figure 2—figure supplement 2 and Video 2 ) . Interestingly , quantification of how the radial and horizontal cuts open revealed that only the horizontal cuts opened along the full length of the cut although they opened more slowly than radial cuts ( Figure 2—figure supplement 2g–j ) , indicating different stress profiles in the two orientations . Taken together , at the time of folding initiation the EGL , which is driving the differential expansion , is itself growing uniformly and the cerebellum is under both radial and circumferential constraints . Finally , there is no evidence of any pre-patterning in the EGL in either cellular behaviors or fiber distribution . As the granule cells within the EGL have such varied shapes as shown above , we looked to see if the cells within the EGL were undergoing any rearrangement movements that may indicate fluid properties . A small , scattered fraction of nuclei in the EGL were fluorescently labeled ( Atoh1-CreER/+; R26ntdTom/+ injected with tamoxifen two days prior to imaging ) and ex vivo slice-culture time-lapse imaging was performed for up to five hours . Tracking the cell positions through time revealed that granule cells within the EGL are highly motile within the EGL . Furthermore , there was no obvious directionality or collectivity to the movement . However , the dynamic motility resulted in the constant exchanging of nearest neighbors over the course of tens to hundreds of minutes and shows that at the timescale of folding the EGL is more fluid-like than a solid epithelial layer ( Figure 5 and Videos 3 and 4 ) . We recently developed a model for folding from differentially expanding bi-layer tissues that takes into account the out-of-phase thickness of the outer layer of several systems and possible contribution of radial mechanical constraints present in neurological tissue ( Engstrom et al . , 2018 ) . We applied the model here to the initiation of cerebellar folding based on five primary assumptions . First , the core is an incompressible material ( μ ) as indicated by the bulk modulus measurements . Second the outer layer , that is the EGL , expands uniformly ( kt ) as shown by the proliferation rate . Third , the EGL is assumed to be a fluid-like material as demonstrated by the live-imaging of neighbor exchanges . Fourth , there is an elastic component radially to the entire cerebellum ( kr ) , seen in the cutting and relaxation experiment and possibly mediated by radial glia . Fifth , the EGL is constrained towards a uniform thickness ( β ) , possibly by Bergmann glia fibers spanning the EGL . Given the interplay between incompressible material , compressible fibrous material , and a proliferating non-elastic EGL , this model is multi-phase . An energy functional parameterized by both the inner and outer boundary of the EGL and incorporating the above five assumptions into three dimensionless parameters ( μ/kr , kr/kt , kt/β ) is minimized to yield an equation for a driven harmonic oscillator resulting in sinusoidal shapes for both the inner and outer boundary of the EGL given an initial elliptical shape . In contrast with the elastic bilayer wrinkling model , EGL thickness oscillations are found to be out-of-phase with the surface height ( radius ) oscillations when 0 < μ/kr <1 . Additionally , the model predicts that the ratio of the measured surface height amplitude ( Ar ) and the EGL thickness amplitude ( At ) is given byArAt=μkr1-μkr , which need not be ≫1 as is typical of elastic bilayer wrinkling , and the number of initial folds at E16 . 5 is determined byn=ktβ1+μkt1-μkr Note that in contrast with elastic wrinkling , the number of initial folds does not depend on the thickness ( a length scale ) of the EGL , but only on material properties . Given that our tissue cutting and relaxation experiment revealed circumferential tension in the cerebellum at folding initiation ( Figure 2f–h , Video 1 ) , we returned to the mathematics and found a previously unrealized geometric relationship in the circular limit of the model that in fact assumes circumferential tension in addition to the previously discussed radial tension given that the perimeter of a circle is determined by its radius . To rigorously test the shape prediction of the model , we first constrained 3 of the 5 parameters for a circular version of the model by using both the thickness amplitude , and average thickness of the EGL , as measured at E16 . 5 , and the number of initial folds . Secondly , the parameter μ/kr ( denoted as ε ) was assumed to scale linearly with time . Together , this allowed for the generation of shape predictions at later developmental stages ( E17 . 5 and E18 . 5 ) from the E16 . 5 starting approximation . Solving the analytical model as constrained by our measured embryonic data we found that it closely approximates the phase and amplitude behavior of EGL thickness and radius oscillations from E16 . 5 through E18 . 5 . ( Figure 6a–c ) . However , the model is not able to produce self-contacting folds or hierarchical folding , both of which are seen in the cerebellum at later stages . The cerebellum has hierarchical folding in which the initial folds become subdivided . Given that ACs hold their position during development and compartmentalize granule cells within lobules of the EGL ( Legué et al . , 2015 ) we reasoned that the ACs could be acting as mechanical boundaries enabling similar mechanics to drive the secondary folding . To test this possibility we measured the expansion of the EGL and the core of the individual lobule regions from E18 . 5 to P3 . We found that indeed in the lobule regions that undergo folding there is a temporal correlation between when the onset of sub-folding and differential expansion occur ( Figure 7a–d ) . In contrast , the region ( L45 ) that does not fold during the same time period has a different , more rectangular shape , and the ratio of EGL growth to core growth is proportional for a rectangle during the time measured ( Figure 7 ) .
Here we have provided experimental evidence that cerebellar folding emerges without obvious pre-patterning . Additionally , the outer layer has fluid-like properties and expands uniformly , and the growth creates a differential expansion between the outer layer and the core . Thus , traditional morphometric cellular behaviors such as changes in cell shape , size and proliferation do not direct where cerebellar folding initiates . Furthermore , our developmental interrogation revealed tissue moduli , mechanical constraints , and emergent thickness variations in the EGL that are fundamentally inconsistent with traditional elastic bilayer wrinkling models . Therefore our results call for a new understanding of brain folding . By applying a multi-phase model constrained by our measured data we were able to capture the correct shape variations and number of folds at the onset of folding . Our new framework accounts for: the rapidly expanding fluid EGL , whose thickness is proposed to be regulated by Bergmann glial fibers , the slower growing incompressible core , and fibrous material in the form of glial fibers and possibly axons as well as the meninges that potentially provide radial and circumferential tension ( Figure 8 ) . This multi-phase model of folding makes many new predictions . One such prediction is that adjusting the amount of tension spanning the cerebellum will change the degree of folding . Indeed , alterations of the cells that likely create tension-based forces could explain the dramatically disrupted folding seen in mouse mutants in which radial glia do not produce Bergmann glia ( Li et al . , 2014 ) . Without Bergmann glia , the EGL would be expected to not form a layer with regular thickness and it should be more sensitive to variations in radial glial tension . Consistent with this prediction , mutants without Bergmann glia have more localized and less regular folds ( Li et al . , 2014 ) Our combination of experimental studies and modeling thus provide new insights into cerebellar folding , including an underappreciated role for tension . Under the new framework revealed by our measurements made in the developing mouse cerebellum , to approximate the observed shape changes in the murine cerebellum from E17 . 5 to E18 . 5 the ratio of the core stiffness over the radial tension must increase . Yet , the measured bulk modulus of the core shows no increase during development . Therefore a second prediction is that radial tension must decrease during development . While the cerebellum is crossed by many fibers at folding initiation , radial glial fibers are an attractive candidate to mediate this change in radial tension ( Sillitoe and Joyner , 2007; Rahimi-Balaei et al . , 2015 ) . First , they span from the VZ to the surface of the cerebellum at E16 . 5 . Additionally , during folding initiation the radial glia undergo a transition into Bergmann glia where they release their basal connection to the VZ and the cell body migrates towards the surface ( Mota and Herculano-Houzel , 2015 ) . This transition could lead to a reduction in the global radial tension and thus would be consistent with our model prediction . The mechanics underlying hierarchical folding remain an open challenge . However , our developmental data may provide a way forward . As ACs maintain their spatial positions , and as they compartmentalize granule cells within the EGL into the lobule regions ( Legué et al . , 2015 ) , we propose that they create fixed mechanical boundaries that divide the cerebellum into self-similar domains . These domains , with their similar physical properties to the initial unfolded cerebellum , can then undergo additional folding . Furthermore since ACs compartmentalize granule cells within the lobule regions , once separated the lobule regions can develop distinct characteristics , like the observed differential proliferation rates at E18 . 5 . We speculate , therefore , that the folding patterns seen across cerebella in different species evolved by adjustment of global as well as regional levels of differential expansion and tension which ultimately mold the functionality of the cerebellum . Finally it is interesting to note the similarities and differences between the developing cerebellum and the cerebral cortex . Radial glia span the entire cerebral cortex just as in the cerebellum ( Götz et al . , 2002 ) . Furthermore , species with folded cerebrums have evolved outer radial glial cells for which the cell body leaves the ventricular zone to become positioned near the surface while retaining fibers anchored on the surface , similar to Bergmann glia in the cerebellum ( Leung and Li , 2018; Reillo et al . , 2011 ) . While we have emphasized the notion of tension via glial fibers in the developing cerebellum , axonal tension has been discussed in the context of shaping the developing cerebrum ( Van Essen , 1997 ) . Tissue cutting in the cerebral cortex of ferrets has revealed a similar tension pattern during folding as we found in the cerebellum ( Xu et al . , 2010 ) . We therefore submit that our work calls for a revival of the notion of how tension affects the shape of the developing cerebrum . Unlike the cerebellum , the cerebral cortex is not divided into a simple bilayer system . However , outer radial glial cells proliferate , much like the GCPs of the EGL , to drive the expansion of the outer regions of the cerebral cortex around the time of initiation of folding ( Hansen et al . , 2010; Heng et al . , 2017; Nowakowski et al . , 2016 ) . Moving the zone of proliferation out from the VZ gives more space for the increased proliferation required in folding systems . The cerebellum , housing 80% of the neurons in the human brain may be an extreme example requiring the region of proliferation to be completely on the outer surface ( Andersen et al . , 1992 ) . Constraining models of folding of different brain regions with developmental data will bring about a more accurate quantitative understanding of the shaping of the developing brain .
All experiments were performed following protocols approved by Memorial Sloan Kettering Cancer Center’s Institutional Animal Care and Use Committee . The inbred FVB/N stain was used for all proliferation rate , area , length , and expansion rate measurements . Atoh1-CreER ( Machold and Fishell , 2005 ) , Nestin-CreER ( Imayoshi et al . , 2006 ) , Rosa26MTMG ( Muzumdar et al . , 2007 ) , Rosa26Ai75 ( Daigle et al . , 2018 ) were used to quantify cell shape and size as well as fiber distribution and were maintained on the outbred Swiss Webster background . The Swiss Webster strain was used for scanning acoustic microscopy . Both sexes were used for the analysis . Animals were kept on a 12 hr light/dark cycle and food and water were supplied ad libitum . All experiments were performed following protocols approved by Memorial Sloan Kettering Cancer Center’s Institutional Animal Care and Use Committee . The appearance of a vaginal plug set noon of the day as Embryonic day 0 . 5 ( E0 . 5 ) . All animals were collected within two hours of noon on the day of collection . Tamoxifen ( Tm , Sigma-Aldrich ) was dissolved in corn oil ( Sigma-Aldrich ) at 20 mg/mL . Pregnant females carrying litters with Atoh1-CreER/+;R26MTMG/MTMG or NestinCER/+;R26MTMG/MTMG embryos were given one 20 μg/g dose of TM via subcutaneous injection two days prior to analysis . 25 μg/g of 5-ethynyl-2-deoxyruidine ( EDU; Invitrogen ) was administered via subcutaneous injection one hour prior to collection . For embryonic stages heads were fixed in 4% paraformaldehyde overnight at 4°C . For postnatal animals , the brain was dissected out first before fixation . Tissues were stored in 30% sucrose . For all proliferation , area , length , and thickness measurements brains were embedded in optimal cutting temperature ( OCT ) compound . Parasagittal sections were collect with a Leica cryostat ( CM3050s ) at 10 μm . Prior to IHC , EdU was detected using a commercial kit ( Invitrogen , C10340 ) . Following EdU reaction the following primary antibodies were used either overnight at 4°C or 4 hr at room temperature: mouse anti-P27 ( BD Pharmingen , 610241 ) , rabbit anti-GFP ( Life Technologies , A11122 ) , rat anti-GFP ( Nacalai Tesque , 04404–84 ) . All antibodies were diluted to 1:500 in 2% milk ( American Bioanalytical ) and 0 . 2% Triton X-100 ( Fisher Scientific ) . Alexa Fluor secondary antibodies ( 1:500; Invitrogen ) were used: Alexa Fluor 488 donkey anti-rabbit , A21206 , Alexa Fluor 488 donkey anti-rat , A21208 , Alexa Fluor 488 donkey anti-mouse , A21202 , Alexa Fluor 647 donkey anti-mouse , A31571 . EdU was detected using a commercial kit ( Invitrogen , C10340 ) . For cell size , shape and fiber density analysis 60 μm parasagittal sections were collected on a Leica vibratome ( VT100S ) . Primary and secondary antibodies were diluted 1:500 in 2% milk and incubated overnight at 4°C . For scanning acoustic microscopy brains were processed for paraffin embedding and parasagittal sections of 10 μm thick were collected on a microtome ( Leica RM2255 ) . Structured illumination and confocal Imaging was done with Zeiss Observer Z . 1 with Apatome or Zeiss LSM 880 respectively . Measurements for all analysis were taken from the three most midline sagittal sections and averaged . The most midline section was determined by dividing the distance in half between the lateral edges where the third ventrical and the mesencephalic vesicle are no longer connected . Quantifications were made using Imaris ( Bitplane ) and MATLAB ( Mathworks ) software . EGL Proliferation rate was calculated as EDU+/ ( Dapi+;P27- ) cells . All cells were counted within the lobule region to the midpoint of the Anchoring Centers . For proliferation measurements through the ACs and the surrounding EGL at E16 . 5 and E17 . 5 a 50 μm window measured from the outer surface of the EGL was centered at the AC . The measuring window was centered at every 25 μm anterior and posterior to the EGL for a total distance of 250 μm anterior and posterior to the AC . At E18 . 5 , when the AC is fully formed , everything proximal to the centroid of the cerebellum under the midpoint of the AC was counted as the AC . Non-overlapping regions of 50 μm also were measured in either direction for a total of 200 μM anterior and posterior to the AC . Proliferation was measured in three cerebella at E16 . 5 and E17 . 5 and in four cerebella at E18 . 5 . EGL length was measured from the outer surface of the EGL following the curvature of the EGL . Cerebellar area was calculated as the area within the outer surface of the EGL and the ventricular zone . A short strait edge was made perpendicular to the ventricular zone to close the area back upon to the anterior end of the EGL . The convex curvature of the cerebellum was measured by following only the positive curvature of the EGL . The folding index was determined as FI = 1 - ( Positive curvature/EGL length ) . Data collected for E16 . 5 , E17 . 5 , E18 . 5 and P0 came from 6 , 8 , 7 , and 9 cerebella respectively . EGL thickness was measured by defining the outer and inner curvature of the EGL . The shortest distance lines were drawn to the outer curvature from discrete points distributed at every 12 . 5 μm along the inner curvature of the EGL . Nine ACs and surrounding regions from five cerebella were quantified at E16 . 5 and 13 ACs from five cerebellar were analyzed at E17 . 5 . At E18 . 5 six ACs from two cerebellar were quantified . Midline sections were imaged with a Zeiss LSM 880 . Serial images were taken to cover the entire EGL of lobule regions L123 , L45 , and L678 and the ACs . Manual cell masks were created with Imaris software defining the curvature at every z-slice . Every cell that was completely included in the imaging window and that was distinguishable from surrounding cells was counted to reduce sampling bias . Cells from three brains were measured at each stage for a total of 131 at E16 . 5 and 201 at E18 . 5 . Shape was defined via sphericity , which is the surface area of a sphere having the same volume as the cell of interest divided by the surface area of the cell of interest . Midline sections were imaged with a Zeiss LSM 880 . Image tiling was used to cover the EGL . Using Python , a fourth or fifth order polynomial was fitted to the outer edge of the EGL in each image , and five scan lines were positioned at 12 . 2 μm intervals beneath the surface , and parallel to it . A bin width of 50 μm as measured along the polynomial contour was centered at the AC . Bins of equal distance were extended both anteriorly and posteriorly . Staining intensity was counted along each scan line at every z-slice of the confocal stack . Each image was normalized to the mean intensity and smoothed with a Gaussian filter . Peak counting was done using minimum and maximum filters , keeping neighborhood size and threshold parameters constant for all images . The results from the five scan lines were averaged . Live cerebella of E16 . 5 FVB/N mice were collected in dissection buffer as previously described ( Wojcinski et al . , 2017 ) and embedded in low-melting point agarose ( Invitrogen ) . Sagittal slices at a thickness of 250 μm were collected . Slices were removed from the agarose and place in petri-dishes coated with Poly ( 2-hydroxyethyl methacrylate ) ( Sigma-Aldrich ) . Tissue cuts ( eight horizontal , 10 radial ) were made with a 30° Premier Edge stab knife ( Oasis Medical ) . Slices were allowed to relax for 10 min . Time-lapse images were acquired on a Leica MZ75 dissection scope . Live cerebella of E16 . 5 Atoh1-CreER/+; R26Ai75/+mice were collected and slices of a thickness of 250 μm were cultured on Millicell cell culture inserts ( Millipore ) in glass bottom plates ( Matek ) as previously described ( Wojcinski et al . , 2017 ) . Image stacks were acquired on a Zeiss LSM 880 at intervals of around 3 . 5 min for up to 5 hr . Cell positions were tracked using Imaris ( Bitplane ) software . Three time-lapses were analzyed . Mechanical tissue properties were analyzed using a 250 MHz Scanning Acoustic Microscope ( SAM ) , described previously ( Rohrbach et al . , 2015; Rohrbach et al . , 2018; Rohrbach and Mamou , 2018 ) . Briefly , 12 μm paraffin sections of mouse embryonic brains were de-parafinized , hydrated in de-ionized water and raster scanned ( 2 µm steps in both direction ) on the SAM to acquire radio-frequency ( RF ) ultrasound data . At each scan location , signal processing was performed to compute the amplitude , sample thickness , speed of sound , acoustic impedance , attenuation , bulk modulus , and mass density ( Rohrbach and Mamou , 2018 ) . Two-dimensional maps of tissues properties were formed using the values obtained at each scan location . Bulk modulus was computed from the product of the acoustic impedance and the speed of sound . Co-registered histology and SAM amplitude images were used to identify regions-of-interest ( ROIs ) corresponding to the EGL layer and underlying core of the cerebellum in each sample . Bulk modulus was analyzed as a measure of tissue stiffness: ROI measurements were acquired from 3 sections from three embryos at each developmental stage . The wrinkle of a circular bilayer structure in Figure 3a was simulated with commercial software ABAQUS . Both film and substrate were modeled as incompressible neo-Hookean materials . The ratio between shear moduli of the film and substrate was 50 and the initial radius of the simulated structure was 16 times that of the film thickness . The differential growth of the EGL and core was modeled by an isotropic expansion of the film in the bilayer structure . To test the elastic wrinkling model , we conducted finite element ( FE ) simulations for bilayer structures with a film bonded on a substrate , which represents the EGL layer and core structure , respectively . The structures were assumed to be under 2D plane strain deformation to mimic the quasi-2D nature of cerebellum wrinkles . Neo-Hookean model was adopted to describe the elastic properties of both film and substrate , whose strain energy can be expressed asU=12μ ( I1-3 ) where μ is the shear modulus and I1 represents the first invariant of the right Cauchy-Green strain tensor . The Poisson’s ratios for the film and substrate were set to be 0 . 5 , based on experimental observations that the bulk modulus of EGL and core are in the order of GPa , much larger than the shear modulus of soft tissues ( ~ kPa ) . We carried out FE simulations through commercial software ABAQUS . A second order 6 node hybrid element ( CPE6MH ) was utilized to discretize the film and substrate . Very fine FE meshes were used to make sure the results independent of mesh size . To incorporate differential growth in real EGL layer and core , an isotropic growth deformation tension was applied to the modeled film by decoupling the deformation tenor F into elastic deformation part A and growth part G . F=A∙G For simplicity , we assume the growth part is isotropic and controlled by a scalar variable gG=g100010001where g>1 represents a faster growth in EGL than the core . To trigger instabilities in numerical simulations , random perturbations ( e . g . , White Gaussian noise with 0 . 001t mean magnitude ) were applied to the nodal positions at the top surface of the film and the interface between the film and substrate . To qualitatively understand the cut experiments we ran a FE simulation of a pre-cut circular bilayer structure and then assigned swelling strain to the film . This neglected the dynamical process in the real cut experiments and only focused on the final equilibrium of the cerebellum after long time relaxation . All the simulation parameters were the same as those in the wrinkling simulation . The initial cut length a is equal to 8 t . The minimum in-plane principal stress corresponds to the hoop stress in the film . For a full treatment of the mathematics please see Engstrom et al . ( 2018 ) . We , formulated a two-dimensional model based on the parameters of a midsagittal section of the cerebellum . The distance of the outer edge of the EGL and , hence , the outer edge of the cerebellum from the center of the cerebellum was defined as r ( θ ) with θ as the angular coordinate . We assumed that r ( θ ) was single-valued . The thickness of the EGL was defined as t ( θ ) . See model schematic in . Taking into account the four assumptions discussed in the main text , we constructed the following energy functional to be minimizedE[r , t , dtdθ ]=∫dθ{kr ( r−r0 ) 2−kt ( t−t0 ) 2+β ( dtdθ2 ) } , with kr as the stiffness modulus ( a spring constant in one-dimension ) of the radial glial fibers and the pial surface contained in the meninges surrounding the cerebellum since the cerebellar radius is proportional to its perimeter , r0 as the preferred radius of the cerebellum , kt denoting a growth potential due to cell proliferation , t0 as thickness of the EGL ( cortex ) , and , β quantified the mechanical resistance to changing the thickness of the EGL . Given our first assumption of an incompressible cerebellar core , we imposed the constraint12∫dθ ( r-t ) 2=A0 , with A0 as a preferred cerebellar area . We applied the variational principle to minimize the energy functional subject to the core constraint , that isδ ( E-μ∫dθ r-t2 ) =0 , where μ is a Lagrange multiplier . Assuming the preferred radius of the cerebellum is constant and the thickness of the EGL/cortex is also constant , then the preferred cerebellar shape was a circle and the EGL an annulus . The variational analysis yielded the following equation of shape for t ( θ ) ;d2tdθ2 + q2 t ( θ ) =ktβ ( t0+μr0kt1−μkr ) , with q2= ktβ1+μkt1-μkr . The solution to the equation of shape wast ( θ ) =Atsin ( qθ+ϕ ) +C1 ( r0 , t0 , kr , kt , μ ) , with C1 independent of θ and At= 21-μkrA0π-C2r0 , t0 , kr , kt , μ such that A0>πC2 . There was an additional equation of shape for r ( θ ) from the variational principle that depended on t ( θ ) and so was determinedr ( θ ) =−μkr1−μkrAtsin ( qθ+ϕ ) +C3 ( r0 , kr , μ ) . We used the measured data at E16 . 5 to set the parameters to make predictions for the shape of both the EGL and core ( and so the relationship between the two ) at later times . Because we are primarily interested in shape changes , rather than size changes , a nondimensionalized model solution was used , that is we chose units where r0 = 1 . This reduces the total number of parameters specifying the model to five dimensionless parameters . Plots assumed a circular preferred shape , and with other parameters as follows: ϵ=μ/kr is shown in Figure 6b , c , c=kr/kt=0 . 06/ϵ , At/r0=ϵ/9 . 6 , t0/r0=ϵ/4 . 8 , and q=6 . Note that for ϵ=0 . 3 , these parameters are numerically consistent with our E16 . 5 measurements: At/t0=0 . 5 and r0/t0=16 , as well as the observed number of invaginations in the half circle: q/2=3 . All of these parameters are either constant or depend on the time-like parameter ϵ . One of these dependencies has a functional form that is physically justifiable ( At~ϵ ) , another has a form that is biologically justifiable ( c~1/ϵ ) , owing to the decrease in the number of radial glia over time . We defined a dimensionless 'shape factor' as half of the perimeter divided by the square root of half of the area as appropriate for a semi-circle . To compare the model’s predictive deviation of this quantity form the semi-circular value we assumed a linear relationship between ϵ and time T measured in embryonic days: ϵ ( T ) = 0 . 3 ( T-15 . 5 ) . Statistical analyses were performed using Matlab software . Significance was determined at p<0 . 05 . Two-way ANOVA was used for proliferation analysis as two variables were tracked , mouse and region . Cell shape , volume , fiber distribution , EGL thickness and bulk modulus were run under a standard ANOVA . After ANOVA analysis a multiple comparison was run with Tukey’s honestly significant difference criterion . F-test for variance and two-tailed student’s paired t-test were used for slice cutting and relaxation quantifications . The degrees of freedom , where appropriate , and P values are given in the figure legends . All error bars are standard deviations . No statistical methods were used to predetermine the sample sizes . We used sample sizes aligned with the standard in the field . No randomization was used nor was data collection or analysis performed blind . | The human brain has a characteristic pattern of ridges and grooves that make up its folded shape . Folds in the outer layer of the brain , known as the cortex , increase the surface area and make more space for cells to connect and form complex circuitries . Different models have been put forward to explain how these folds form during development . Examples include tension from cells pulling areas of the cortex together , or layers of the cortex growing at different rates , causing the cortex to buckle and create folds . Discriminating between these different models requires biological information about the cells and tissue of the brain at the start of the folding process . However , it has been difficult to extract this information when considering the development of the human brain in three dimensions . Lawton et al . have overcome these difficulties by using a part of the mouse brain called the cerebellum as a simpler system . As in humans , the mouse cerebellum is a densely folded structure , sitting underneath the brain , that plays a major role in regulating movements , as well as cognition . The symmetrical structure of the mouse cerebellum means it can be analyzed in two dimensions , making it easier to track the mechanics of folding . By applying the extracted biological data onto a mathematical model , Lawton et al . showed folding was driven by a combination of previously unknown features . For instance , that cells in the outer layer of the cerebellum grow faster than cells in the center , with cells growing uniformly across the outer layer . Other features include the fluid-like composition of the outer layer , which allows cells to move freely and regularly change position , and tensions surrounding the cerebellum mechanically straining its growth . Notably , the pattern of cells and tissue fibers in the cortex had no influence over these mechanical properties and provided no pre-indication of where the sites of folding would occur . The data collected deviates from other models , and has led Lawton et al . to propose a new explanation for how the brain folds , incorporating these newly found features . Problems with brain folding during human development can lead to debilitating conditions . Applying this new model to folding disorders of the human brain could help scientists to understand how these folding defects arise . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"physics",
"of",
"living",
"systems"
] | 2019 | Cerebellar folding is initiated by mechanical constraints on a fluid-like layer without a cellular pre-pattern |
During meiosis , a single round of DNA replication is followed by two consecutive rounds of nuclear divisions called meiosis I and meiosis II . In meiosis I , homologous chromosomes segregate , while sister chromatids remain together . Determining how this unusual chromosome segregation behavior is established is central to understanding germ cell development . Here we show that preventing microtubule–kinetochore interactions during premeiotic S phase and prophase I is essential for establishing the meiosis I chromosome segregation pattern . Premature interactions of kinetochores with microtubules transform meiosis I into a mitosis-like division by disrupting two key meiosis I events: coorientation of sister kinetochores and protection of centromeric cohesin removal from chromosomes . Furthermore we find that restricting outer kinetochore assembly contributes to preventing premature engagement of microtubules with kinetochores . We propose that inhibition of microtubule–kinetochore interactions during premeiotic S phase and prophase I is central to establishing the unique meiosis I chromosome segregation pattern .
Cells have evolved intricate mechanisms to execute proper partitioning of the genetic material during cell division . This task is especially complex in meiosis , the cell division used by sexually reproducing organisms to generate gametes . The goal of meiosis is to reduce the genome content by half such that proper ploidy is maintained upon fusion of gametes . To achieve this , a single round of DNA replication is followed by two consecutive rounds of nuclear division called meiosis I and meiosis II . During meiosis I homologous chromosomes segregate . Meiosis II resembles mitosis in that sister chromatids segregate from each other . The establishment of this specialized chromosome segregation pattern requires three changes that modulate how chromosomes interact with each other and with the microtubule cytoskeleton: ( 1 ) reciprocal recombination between homologous chromosomes , ( 2 ) the way linkages between sister chromatids , known as sister-chromatid cohesion , are removed from chromosomes and ( 3 ) the manner in which chromosomes attach to the meiotic spindle . Homologous recombination is initiated by programmed double-strand breaks ( DSBs ) , which are catalyzed by Spo11 following premeiotic DNA replication ( Keeney et al . , 1997 ) . Subsequent repair of DSBs by crossover recombination generates physical linkages between homologous chromosomes . This , in turn , allows homologs to attach to the meiosis I spindle such that each homolog interacts with microtubules emanating from opposite spindle poles . As a result , homologous chromosomes biorient on the meiosis I spindle . The spindle assembly checkpoint prevents the onset of chromosome segregation until this process is completed . Once each pair of homologs is bioriented , checkpoint signaling ceases and anaphase entry ensues . A ubiquitin ligase known as the anaphase promoting complex/cyclosome and its specificity factor Cdc20 ( APC/C-Cdc20 ) targets Securin for degradation , relieving Separase inhibition ( Cohen-Fix et al . , 1996; Ciosk et al . , 1998 ) . Separase is a protease that cleaves the kleisin subunit of cohesin , the protein complex that mediates sister-chromatid cohesion ( Uhlmann et al . , 1999 , 2000; Schleiffer et al . , 2003 ) . In meiosis I , cleavage of cohesin at chromosome arms allows homologs to segregate ( Buonomo et al . , 2000 ) . However , cohesin around the centromeres is protected from cleavage during meiosis I , which is essential for the accurate segregation of sister chromatids during meiosis II . Protection of centromeric cohesin is accomplished by preventing phosphorylation of Rec8 , the meiosis-specific kleisin . This occurs , at least in part , by Sgo1 ( MEI-S332 ) -dependent recruitment of the protein phosphatase PP2A to centromeric regions where it antagonizes Rec8 phosphorylation ( Kerrebrock et al . , 1995; Katis et al . , 2004a; Kitajima et al . , 2004 , 2006; Riedel et al . , 2006 ) . The third modification necessary to bring about the meiotic chromosome segregation pattern is the manner in which kinetochores attach to microtubules during meiosis I and meiosis II . In meiosis I , kinetochores of sister chromatid pairs ( henceforth sister kinetochores ) attach to microtubules emanating from the same spindle pole , a process called sister kinetochore coorientation . During meiosis II , as during mitosis , sister kinetochores attach to microtubules emanating from opposite spindle poles and are thus bioriented ( reviewed in Marston and Amon , 2004 ) . In budding yeast , sister kinetochore coorientation is brought about by the monopolin complex , which consists of Mam1 , Lrs4 , Csm1 and the casein kinase 1 , Hrr25 ( Toth et al . , 2000; Rabitsch et al . , 2003; Petronczki et al . , 2006 ) . Lrs4 and Csm1 localize to the nucleolus during interphase . During exit from pachytene , a stage of prophase I , Lrs4 and Csm1 associate with Mam1 and Hrr25 at kinetochores , a process that requires the Polo kinase Cdc5 ( Clyne et al . , 2003; Lee and Amon , 2003; Matos et al . , 2008 ) . How the association of monopolin with kinetochores is coordinated with respect to kinetochore assembly and microtubule–kinetochore interactions during meiosis is not understood . Cyclin-dependent kinases ( CDKs ) are the central regulators of the mitotic and meiotic divisions . In budding yeast , a single CDK associates with one of six B-type cyclins ( Clb1-Clb6 ) ( reviewed in Morgan , 1997 ) . In meiosis , Clb5- and Clb6-CDKs drive DNA replication and recombination , whereas Clb1- , Clb3- and Clb4-CDKs promote the meiotic nuclear divisions ( reviewed in Marston and Amon , 2004 ) . Meiotic cyclin-CDK activity is regulated both at the transcriptional and translational level ( Grandin and Reed , 1993; Carlile and Amon , 2008 ) . Transcription of CLB1 , CLB3 and CLB4 occurs only after exit from pachytene ( Chu and Herskowitz , 1998 ) ; CLB3 is also translationally repressed during meiosis I , thus restricting Clb3-CDK activity to meiosis II ( Carlile and Amon , 2008 ) . The major mitotic cyclin , CLB2 , is not expressed during meiosis ( Grandin and Reed , 1993 ) . Here we investigate the importance of cyclin-CDK regulation in establishing the meiotic chromosome segregation pattern . We show that expression of a subset of cyclins during premeiotic S phase and early prophase I , defined as the prophase stages up to exit from pachytene , causes premature microtubule–kinetochore interactions . This , in turn , disrupts both sister kinetochore coorientation and protection of centromeric cohesin during meiosis I , revealing that the temporal control of microtubule–kinetochore interactions is essential for meiosis I chromosome morphogenesis . Furthermore , we define the mechanism by which premature microtubule–kinetochore interactions are prevented; through regulation of cyclin-CDK activity and of outer kinetochore assembly . Our results demonstrate that preventing premature microtubule–kinetochore interactions is essential for establishing a meiosis I-specific chromosome architecture and provide critical insights into how the mitotic chromosome segregation machinery is modulated to achieve a meiosis I-specific pattern of chromosome segregation .
We previously reported that CLB3 expression prior to meiosis I induces a change in the pattern of chromosome segregation such that sister chromatids , instead of homologous chromosomes , segregate during the first nuclear division ( Carlile and Amon , 2008 ) . To determine how Clb-CDKs impact meiotic chromosome segregation and whether Clb-CDKs play redundant or specific roles in regulating this process , we examined the consequences of prematurely expressing CLB1 , CLB3 , CLB4 or CLB5 . In our previous studies we expressed CLB3 from the GAL1-10 promoter driven by an estrogen inducible Gal4-ER fusion ( Carlile and Amon , 2008 ) . Expression from the GAL1-10 promoter led to Clb3 accumulation in meiosis I to levels that are comparable to those seen in meiosis II in wild-type cells ( Carlile and Amon , 2008 ) . However , estrogen interferes with meiotic progression when added during early stages of sporulation ( Figure 1A ) . To circumvent this problem we utilized the copper-inducible CUP1 promoter to drive Clb3 expression . Expression from the CUP1 promoter led to approximately fivefold higher levels of Clb3 protein compared to expression from the GAL1-10 promoter ( Figure 1B ) . To examine the consequences of the two CLB3 constructs on chromosome segregation we used GAL-CLB3 and CUP-CLB3 strains in which one of the two homologs of chromosome III was marked by integrating a tandem array of tetO sequences ∼20 kb from CENIII ( heterozygous LEU2-GFP dots ) . These cells also expressed a tetR-GFP fusion , which allowed visualization of the tetO arrays ( Michaelis et al . , 1997 ) . The analysis of GFP dot segregation during the first meiotic division revealed that despite the difference in Clb3 protein levels , the extent of sister chromatid segregation in meiosis I was similar between GAL-CLB3 and CUP-CLB3 cells ( Figure 1C ) . This finding indicates that expression of Clb3 from either the CUP1 or GAL1-10 promoter efficiently induces sister chromatid segregation during meiosis I . Furthermore , the timing of when Clb3 is expressed , rather than the amount of Clb3 present , appears to be the primary determinant of this phenotype . Based on this observation and the finding that all four cyclins showed equal expression when produced from the CUP1 promoter ( Figure 1D ) we utilized the CUP1 promoter for most subsequent analyses . Having established a system to effectively express various cyclins prior to meiosis I we next examined the consequences of their premature expression on meiosis I events . We first asked whether misexpression of various cyclins is sufficient to induce spindle formation in cells arrested in pachytene of prophase I , due to lack of the transcription factor Ndt80 ( Xu et al . , 1995; Chu and Herskowitz , 1998 ) . We induced cyclin expression from the CUP1 promoter 135 min after the induction of sporulation when typically 40–65% of the cells have replicated their DNA ( Figure 1E; Blitzblau et al . , 2012 ) and examined spindle pole body ( SPB , centrosome equivalent in budding yeast ) separation and spindle morphology following induction . As expected , wild-type cells did not form spindles in the absence of NDT80 function . Expression of CLB5 from the CUP1 promoter did not lead to SPB separation and spindle formation either , although expression of CLB5 in the prophase I arrest led to a significant increase in total CDK activity ( Figures 1F and 2A , Figure 2—figure supplement 1 ) . In contrast , CUP-CLB1 , CUP-CLB3 and CUP-CLB4 cells separated SPBs and formed bipolar spindles , shortly after copper addition ( Figure 2A and Figure 2—figure supplement 1 ) . Similar results were observed in cells with intact NDT80 ( data not shown ) . We conclude that expression of CLB1 , CLB3 or CLB4 is sufficient to promote bipolar spindle assembly in NDT80-depleted cells . 10 . 7554/eLife . 00117 . 004Figure 2 . Premature expression of CLB1 or CLB3 causes sister kinetochore biorientation during prophase I and sister chromatid segregation in meiosis I . Wild-type or CUP-CLB cells were induced to sporulate . After 2 hr 15 min , cyclins were induced by addition of CuSO4 ( 50 μM ) . Cells were either arrested during prophase I or released from an NDT80 block 4 hr 30 min after induction of sporulation . ( A ) Bipolar spindle formation determined in wild-type ( A22678 ) , CUP-CLB1 ( A27421 ) , CUP-CLB3 ( A22702 ) , CUP-CLB4 ( A27423 ) and CUP-CLB5 ( A27425 ) during prophase I ( n = 100 per time point ) . Images on left show spindle formation in CUP-CLB cells 4 hr after induction of sporulation; in this and all subsequent Figures microtubules are shown in green and DNA in blue . The dotted line depicts the cell membrane . ( B ) Microtubule–kinetochore engagement monitored during prophase I , starting at 1 hr after CuSO4 addition in wild-type ( A30700 ) , CUP-CLB1 ( A30702 ) , CUP-CLB3 ( A30704 ) , CUP-CLB4 ( A30707 ) and CUP-CLB5 ( A30708 ) by live cell microscopy . SPBs ( marked by arrow ) and heterozygous CENV-GFP dots are shown ( arrowheads mark separated CENV dots ) . In this and all subsequent figures SPBs are in red , GFP dots are in green . ( C ) Top panel: representative images of wild-type ( A30700 ) and CUP-CLB3 ( A30704 ) . Bottom panel: separation of heterozygous CENV-GFP dots in prophase I-arrested cells quantified in wild-type ( A22678 ) , CUP-CLB1 ( A27421 ) , CUP-CLB3 ( A22702 ) , CUP-CLB4 ( A27423 ) and CUP-CLB5 ( A27425 ) by live cell microscopy ( over the duration of 8 hr , n > 100 ) as described in the ‘Materials and methods’ . The fraction of nuclei that display sister kinetochores as separate or together for each CUP-CLB strain was compared to wild-type using a chi-square test ( df 1 ) : CUP-CLB1 , χ2 = 40 . 77 , p<0 . 0001; CUP-CLB3 , χ2 = 34 . 84 , p<0 . 0001; CUP-CLB4 , χ2 = 0 . 1163 , p=0 . 7330; CUP-CLB5 , χ2 = 1 . 418 , p=0 . 2337 . ( D ) Segregation of sister chromatids ( equational division ) using heterozygous CENV-GFP dots quantified in binucleates from wild-type ( A22678 ) , CUP-CLB1 ( A27421 ) , CUP-CLB3 ( A22702 ) , CUP-CLB4 ( A27423 ) and CUP-CLB5 ( A27425 ) ( n = 100 ) . The fraction of binucleates that display a reductional or equational division for each CUP-CLB strain was compared to wild-type using a chi-square test ( df 1 ) : CUP-CLB1 , χ2 = 45 . 13 , p<0 . 0001; CUP-CLB3 , χ2 = 48 . 22 , p<0 . 0001; CUP-CLB4 , χ2 = 1 . 020 , p=0 . 3124; CUP-CLB5 , χ2 = 0 , p=1 . ( E ) Wild-type ( A31019 ) and CUP-CLB3 ( A31021 ) cells monitored for segregation of heterozygous CENV-GFP dots with respect to Pds1 ( Securin , red ) degradation by live cell microscopy ( n > 17 ) . Time of Pds1 degradation set to t = 0 , percent cells were plotted as a Kaplan–Meier curve . Note that for A31021 , the analysis of cells that segregate sister chromatids in the first nuclear division is shown . Pds1 accumulation during prophase II is not observed using the Pds1-tdTomato construct , likely due to delayed maturation of the fluorophore ( Katis et al . , 2010 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 00410 . 7554/eLife . 00117 . 005Figure 2—figure supplement 1 . Spindle pole body separation in CUP-CLB cells . Wild-type or CUP-CLB cells also carrying the GAL4-ER and GAL-NDT80 fusions were induced to sporulate . After 2 hr 15 min , cyclins were induced by addition of CuSO4 ( 50 μM ) . Cells were arrested during prophase I and the percentage of cells with separated Spc42 foci ( red dots ) was determined at indicated time points in wild-type ( A29581 ) , CUP-CLB1 ( A29582 ) , CUP-CLB3 ( A29583 ) , CUP-CLB4 ( A29584 ) and CUP-CLB5 ( A29585 ) ( n > 100 for each time point ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 00510 . 7554/eLife . 00117 . 006Figure 2—figure supplement 2 . Homolog separation in CUP-CLB4 cells . Wild-type ( A22688 ) , CUP-CLB4 ( A32470 ) also carrying the GAL4-ER and GAL-NDT80 fusions or cdc20-mn ( A15163 ) cells all carrying homozygous CENV GFP dots were induced to sporulate . After 2 hr 15 min , cyclins were induced by addition of CuSO4 ( 50 μM ) . Cells were arrested either during prophase I ( A22688 , A324470 ) or metaphase I ( A15163 ) . Separated GFP foci ( homologs separate ) were analyzed 6 hr ( prophase I-arrest ) or 8 hr 30 min ( metaphase I-arrest ) after induction of sporulation ( n > 100 for each time point ) . Using a chi-square test ( df 1 ) , the fraction of mononucleates that display homologs as together or separate during a prophase I arrest was compared between wild-type and CUP-CLB4 χ2 = 0 . 4422 , p=0 . 5061 . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 00610 . 7554/eLife . 00117 . 007Figure 2—figure supplement 3 . Chromosome III sister chromatid segregation in CUP-CLB3 cells . Wild-type ( A18185 ) and CUP-CLB3 ( A22682 ) cells also carrying the GAL4-ER and GAL-NDT80 fusions were induced to sporulate . After 3 hr , CLB3 was induced by addition of CuSO4 ( 50 μM ) . At 6 hr , cells were released from the NDT80 block . Subsequently , segregation of sister chromatids ( equational division ) using heterozygous GFP dots integrated at LEU2 ( ∼20 kb from CENIII ) was quantified in binucleate cells . The appearance of segregated sister chromatids in wild-type is likely due to recombination between LEU2 and CEN3 . Using a chi-square test ( df 1 ) , the fraction of binucleates that display a reductional or equational division was compared between wild-type and CUP-CLB3 χ2 = 35 . 65 , p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 00710 . 7554/eLife . 00117 . 008Figure 2—figure supplement 4 . Sister chromatid segregation in CUP-CLB3 cells using dual-color marked chromosomes . Wild-type ( A27476 ) and CUP-CLB3 ( A27480 ) cells carrying the GAL4-ER and GAL-NDT80 fusions and CENV-LacO/LacI-GFP on one homolog of chromosome V ( green ) and CENV-tetO/tetR-RFP on the other homolog of chromosome V ( red ) were induced to sporulate and CuSO4 ( 50 μM ) was added at 2 hr 15 min . At 4 hr 30 min , cells were released from NDT80 block and monitored by live cell microscopy starting 30 min after estradiol addition , and monitored every 15 min for 8 hr . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 00810 . 7554/eLife . 00117 . 009Figure 2—figure supplement 5 . Recombination in CUP-CLB3 cells . Left panel: wild-type ( A21104 ) and GAL-CLB3 ( A21105 ) cells were induced to sporulate and estradiol ( 1 μM ) was added 3 hr after transfer into sporulation medium . Genomic DNA was prepared and digested with XhoI and MluI and hybridized with Probe A . See Storlazzi et al . ( 1995 ) for details . Right panel: recombination products were quantified as R2/P1 . Note: A21104 and A21105 contain auxotrophies and have reduced meiotic kinetics relative to prototrophic strains . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 00910 . 7554/eLife . 00117 . 010Figure 2—figure supplement 6 . Localization of Rad51 in CUP-CLB3 cells . Wild-type ( A22864 ) and CUP-CLB3 ( A22866 ) cells were induced to sporulate and CuSO4 ( 50 μM ) was added 3 hr after transfer into sporulation medium . Localization of the double-strand break repair protein Rad51 ( green ) was determined by nuclear spreads 4 hr after transfer to sporulation medium . DNA is shown in blue . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 01010 . 7554/eLife . 00117 . 011Figure 2—figure supplement 7 . Localization of Zip1 in CUP-CLB3 cells . Wild-type ( A22836 ) and CUP-CLB3 ( A22838 ) cells were induced to sporulate and CuSO4 ( 50 μM ) was added 3 hr after transfer into sporulation medium . Localization of the synaptonemal complex component Zip1 ( green ) and the cohesin subunit Rec8-13myc ( red ) was determined by nuclear spreads 5 hr after transfer to sporulation medium . DNA is shown in blue . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 01110 . 7554/eLife . 00117 . 012Figure 2—figure supplement 8 . Preventing homologous recombination does not affect the phenotypes caused by premature CLB3 expression . Wild-type ( A19396 ) , GAL-CLB3 ( A19400 ) , spo11∆ ( A21193 ) and spo11∆ GAL-CLB3 ( A21194 ) cells were induced to sporulate and estradiol ( 1 μM ) was added 3 hr after transfer into sporulation medium . Subsequently , segregation of sister chromatids ( equational division ) was quantified using heterozygous CENV GFP dots in binucleate cells ( n = 100 ) . Note that CLB3-induced meiosis I sister chromatid segregation is higher in GAL-CLB3 cells than in CUP-CLB3 cells . This is presumably due to the more homogenous expression of CLB3 in cells where expression is driven from the GAL1-10 promoter . Using a chi-square test ( df 1 ) , the fraction of binucleates that display a reductional or equational division was compared between GAL-CLB3 and GAL-CLB3 spo11∆ χ2 = 0 . 3072 , p=0 . 5794 . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 012 Next , we determined whether expression of CLB1 , CLB3 or CLB4 in pachytene-arrested cells also affects the manner in which chromosomes attach to the meiotic spindle using live-cell imaging . To this end we used strains carrying heterozygous CENV-GFP dots and an Spc42-mCherry fusion ( Spc42 is an SPB component ) to monitor the behavior of the marked centromere with respect to the spindle axis . In wild-type and CUP-CLB5 cells , sister kinetochores remained closely associated with each other and did not appear to be tightly associated with SPBs , consistent with the observation that these cells failed to form a spindle . In contrast , we observed dynamic separation of heterozygous CENV-GFP dots upon expression of CLB1 or CLB3 , with sister kinetochores frequently splitting and coming together ( Figure 2B , C ) . This observation is reminiscent of the behavior of bioriented sister chromatids during metaphase of mitosis ( Pearson et al . , 2001 ) . Cells expressing CLB4 did not show transient splitting of sister kinetochores in prophase I , indicating that chromosomes are either unable to attach to the spindle or that homologous chromosomes , instead of sister chromatids , are bioriented as occurs in wild-type cells during metaphase I . To distinguish between these possibilities , we examined the behavior of CUP-CLB4 cells in which both homologs of chromosome V harbor CENV-GFP dots ( henceforth homozygous CENV-GFP dots ) . Similar to wild-type , we observed that in CUP-CLB4 cells the two CENV-GFP dots remained tightly associated in prophase I , indicating that the homologous chromosomes are paired and not attached to the prematurely formed spindle ( Figure 2—figure supplement 2 ) . Together , these results indicate that CUP-CLB1 , CUP-CLB3 or CUP-CLB4 expression promotes bipolar spindle formation in pachytene-arrested cells , but only CLB1 and CLB3 expression can promote stable microtubule–kinetochore attachments sufficient to generate tension . To determine whether different amounts of total CDK activity were responsible for the phenotypic differences of prematurely expressing Clb1 or Clb3 compared to Clb4 , we measured total CDK activity ( Cdc28 in budding yeast ) using Histone H1 as a substrate . Cdc28-associated kinase activity was low during prophase I and increased more than 25-fold during metaphase I/anaphase I in wild-type cells ( Figure 1F ) . Expression of all four cyclins led to a significant increase in total CDK activity in prophase I ( Figure 1F ) , but importantly , the degree of increase did not correlate with the ability to induce sister chromatid splitting in the NDT80 arrest . For example , Clb1 expression led to a similar increase in Cdc28-associated kinase activity as expression of Clb4 , yet Clb1 induced sister chromatid splitting whereas Clb4 did not ( Figures 1F and 2B , C ) . We conclude that the ability to induce sister chromatid splitting does not correlate with total CDK activity produced by the various CUP-CLB fusions . Furthermore , SPB separation and spindle formation are not sufficient to induce microtubule–kinetochore interactions . Events that can be triggered by Clb1 and Clb3 , but not Clb4 are also necessary to promote attachments sufficient to generate tension . Determining why CLB4 expressing cells fail to form productive microtubule–kinetochore interactions could provide important insights into substrate specificity of cyclin-CDK complexes . To determine the consequences of premature cyclin expression on meiosis I chromosome segregation , we examined the segregation of heterozygous CENV-GFP dots in cells that were reversibly arrested in pachytene using the NDT80 block-release system . In this system , expression of NDT80 is controlled by the GAL1-10 promoter , which is regulated by an estrogen-inducible Gal4-ER fusion ( Benjamin et al . , 2003; Carlile and Amon , 2008 ) . Cells were induced to sporulate and after 135 min , copper was added to induce cyclin expression . 4 hr 30 min after sporulation induction , estrogen was added to allow cells to synchronously proceed through the meiotic divisions . In wild-type , CUP-CLB4 and CUP-CLB5 cells , sister chromatids cosegregated in the first division , resulting in binucleate cells with a GFP dot in one of the two nuclei . In contrast , 39% of CUP-CLB1 and 41% of CUP-CLB3 cells segregated sister chromatids in the first division , as judged by the presence of binucleate cells with a GFP dot in each nucleus ( Figure 2D ) . We observed a similar result for chromosome III and cells in which one copy of chromosome V was marked with a GFP dot and the other copy with an RFP dot ( Figure 2—figure supplements 3 and 4 ) . To confirm that sister chromatids indeed split during meiosis I in cells expressing CLB3 during prophase I , we examined when sister chromatid separation occurred with respect to Securin ( Pds1 in budding yeast ) degradation in CUP-CLB3 cells . In wild-type cells harboring heterozygous CENV-GFP dots , Pds1 degradation was immediately followed by movement of the single GFP dot to one side of the cell , indicating that homologous chromosomes had segregated . Subsequently , these cells underwent meiosis II and sister chromatids segregated ( median = 86 min after Pds1 degradation; Figure 2E ) . In contrast , CUP-CLB3 cells segregated sister chromatids immediately after Pds1 degradation ( median = 7 min after Pds1 degradation; Figure 2E ) . These results demonstrate that CUP-CLB3 cells segregate sister chromatids during the first meiotic division . Thus , CUP-CLB3 cells must be defective in two key aspects of meiosis I chromosome segregation: coorientation of sister kinetochores and maintenance of centromeric cohesion . We note that another essential aspect of meiosis I chromosome segregation , homologous recombination , was not affected by premature CLB3 expression . We observed no major defects in DSB formation , synaptonemal complex assembly and generation of recombination products , nor did preventing homologous recombination affect the phenotypes caused by premature CLB3 expression ( Figure 2—figure supplements 5–8 ) . The finding that CUP-CLB1 or CUP-CLB3 cells segregate sister chromatids during meiosis I indicates that sister kinetochore coorientation is defective . To investigate this further , we examined monopolin localization in cells that segregate sister chromatids in meiosis I ( CUP-CLB3 cells ) and cells that do not exhibit chromosome missegregation despite cyclin misexpression ( CUP-CLB4 cells ) . Colocalization of Lrs4 or Mam1 with the kinetochore component Ndc10 was dramatically reduced in CUP-CLB3 but not CUP-CLB4 cells ( Figure 3A and Figure 3—figure supplements 1 and 2 ) . Hyperphosphorylation of Lrs4 , which correlates with monopolin function ( Clyne et al . , 2003; Lee and Amon , 2003; Matos et al . , 2008 ) , was also significantly reduced in CUP-CLB3 , but not in CUP-CLB4 cells ( Figure 3B and Figure 3—figure supplement 3 ) . These results indicate that premature expression of CLB3 prevents monopolin association with kinetochores . 10 . 7554/eLife . 00117 . 013Figure 3 . Premature CLB3 expression disrupts monopolin function . ( A ) Lrs4-13myc ( green ) localization relative to Ndc10-6HA ( red ) was determined in spread nuclei from wild-type ( A9217 ) , CUP-CLB3 ( A26278 ) and CUP-CLB4 ( A29643 ) harboring a Cdc20 depletion allele ( cdc20-mn ) were induced to undergo sporulation and arrested in metaphase I due to depletion of Cdc20 . CuSO4 was added at 3 hr after induction of sporulation ( n > 40 ) . The fraction of spread nuclei that display colocalized , partial or mislocalized Lrs4 with respect to Ndc10 was compared to wild-type using a chi-square test ( df 2 ) : CUP-CLB4 , χ2 = 1 . 136 , p=0 . 5666; CUP-CLB3 , χ2 = 45 . 84 , p<0 . 0001 . ( B ) Western blots for Lrs4-13myc , Clb3 and Pgk1 from wild-type ( A9217 ) and CUP-CLB3 ( A26278 ) cells . Cells were sporulated as described in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 01310 . 7554/eLife . 00117 . 014Figure 3—figure supplement 1 . Monopolin association with kinetochores is disrupted in CUP-CLB3 but not in CUP-CLB4 cells . Wild-type ( A7450 ) , CUP-CLB3 ( A28673 ) and CUP-CLB4 ( A28674 ) cells carrying the cdc20-mn allele were induced to sporulate and CuSO4 ( 50 μM ) was added 2 hr 30 min after transfer into sporulation medium . Mam1-9myc ( green ) localization relative to Ndc10-6HA ( red ) was determined in spread nuclei from metaphase I-arrested cells ( n > 40 ) . DNA is shown in blue . Using a chi-square test ( df 2 ) the fraction of spread nuclei that display colocalized , partial or mislocalized Mam1 with respect to Ndc10 was compared to wild-type: CUP-CLB4 , χ2 = 2 . 554 , p=0 . 2788; CUP-CLB3 , χ2 = 39 . 31 , p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 01410 . 7554/eLife . 00117 . 015Figure 3—figure supplement 2 . Premature Clb3 expression does not interfere with Mam1 expression . Wild-type ( A7450 ) , CUP-CLB3 ( A28673 ) and CUP-CLB4 ( A28674 ) cells carrying the cdc20-mn allele were induced to sporulate and CuSO4 ( 50 μM ) was added 2 hr 30 min after transfer into sporulation medium . Mam1 protein levels were analyzed to determine whether premature Clb3 expression interferes with Mam1 expression . Pgk1 was used as a loading control . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 01510 . 7554/eLife . 00117 . 016Figure 3—figure supplement 3 . Lrs4 phosphorylation is not disrupted in CUP-CLB4 cells . Wild-type ( A26277 ) and CUP-CLB4 ( A29643 ) cells carrying the cdc20-mn allele were induced to sporulate and CuSO4 ( 50 μM ) was added 2 hr 30 min after transfer into sporulation medium . Levels of Lrs4 , Clb3 and Pgk1 from cells arrested in metaphase I were examined by Western blot analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 016 Sister chromatids segregate during meiosis I in CUP-CLB3 cells , indicating that centromeric cohesin either fails to associate with chromosomes or is lost prematurely . To test the first possibility , we examined chromosome association of the cohesin subunit Rec8 and the cohesion maintenance factor Pds5 with chromosomes . Chromatin immunoprecipitation ( ChIP ) and chromosome spreads revealed that association of both proteins with chromosomes in CUP-CLB3 cells was indistinguishable from that of wild-type cells during prophase I or metaphase I ( Figure 4A and Figure 4—figure supplements 1 and 2 ) . Thus , loading of cohesion factors onto chromosomes is not affected in CUP-CLB3 cells . 10 . 7554/eLife . 00117 . 017Figure 4 . CLB3 misexpression disrupts protection of centromeric cohesin . Cyclin expression was induced after 2 hr 15 min ( C ) and ( D ) , 2 hr 30 min ( A ) , ( B ) , ( E ) , ( F ) and ( H ) or 3 hr ( G ) and ( I ) of sporulation . ( A ) Chromosomal association of Rec8-13myc was monitored by ChIP-chip in wild-type ( A28716 ) and CUP-CLB3 ( A28718 ) during prophase I arrest . Centromere position is identified by a black circle . ( B ) Centromeric Rec8 localization was monitored in spread nuclei from wild-type ( A28684 ) , CUP-CLB3 ( A28685 ) and CUP-CLB4 ( A28686 ) cells carrying REC8-3HA ( red ) and NDC10-13myc ( green ) ( n > 40 ) . The fraction of spread nuclei that were Rec8 positive or negative was compared to wild-type using a chi-square test ( df 1 ) : CUP-CLB4 , χ2 = 0 . 001323 , p=0 . 9710; CUP-CLB3 , χ2 = 32 . 79 , p<0 . 0001 . ( C ) Rec8 cleavage monitored by Western blot after release from an NDT80 block ( 4 hr 30 min ) in wild-type and CUP-CLB3 carrying both a myc-tagged REC8 allele as well as either HA-tagged REC8 or rec8-29A allele ( left to right: A29957 , A29959 , A29961 , A29963 ) . ( D ) Percentage of cells with short bipolar spindles was determined at indicated times in wild-type ( A22804 ) , CUP-CLB3 ( A29965 ) , rec8-29A ( A22803 ) and CUP-CLB3 rec8-29A ( A29967 ) after release from an NDT80 block ( 4 hr 30 min ) ( n = 100 per time point ) . ( E ) ChIP analysis for total Rec8 , p-S179 Rec8 or p-S521 Rec8 from metaphase I-arrested ( cdc20-mn ) wild-type ( A28681 ) , CUP-CLB3 ( A28682 ) and Sgo1-depleted ( sgo1-mn; A29994 ) cells . Relative occupancy at a chromosome arm site ( c194 ) or at a centromeric site ( CENV ) was determined relative to a low binding region ( c281 ) . Error bars represent range ( n = 2 ) . ( F ) Chromosomal association of Sgo1-3V5 was monitored by ChIP-chip in wild-type ( A29795 ) and CUP-CLB3 ( A29799 ) cells during prophase I-arrest . Centromere position is identified by a black circle . ( G ) , ( H ) Localization of Sgo1-9myc ( G , green ) or Rts1-13myc ( H , green ) relative to Ndc10-6HA ( red ) determined by nuclear spreads in ( G ) wild-type ( A22868 ) and CUP-CLB3 ( A22870 ) or ( H ) wild-type ( A28329 ) and CUP-CLB3 ( A28330 ) during prophase I ( n > 40 ) . For ( G ) , the fraction of spread nuclei that display colocalized or mislocalized Sgo1 relative to Ndc10 was compared between wild-type and CUP-CLB3 using a chi-square test ( df 1 ) χ2 = 1 . 554 , p=0 . 2125 . For ( H ) , the fraction of spread nuclei that display colocalized , partial or mislocalized Rts1 relative to Ndc10 was compared between wild-type and CUP-CLB3 using a chi-square test ( df 2 ) χ2 = 3 . 712 , p=0 . 1563 . ( I ) Localization of Sgo1-9myc ( green ) in binucleates relative to Ndc10-6HA ( red ) determined by nuclear spreads from wild-type ( A22868 ) and CUP-CLB3 ( A22870 ) ( n > 40 ) . The fraction of spread nuclei that were Sgo1 positive or negative was compared between wild-type and CUP-CLB3 using a chi-square test ( df 1 ) χ2 = 23 . 92 , p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 01710 . 7554/eLife . 00117 . 018Figure 4—figure supplement 1 . Chromosomal association of Rec8 in CUP-CLB3 cells . Wild-type ( A26547 ) and CUP-CLB3 ( A26548 ) cells were induced to sporulate and CuSO4 ( 50 μM ) was added 3 hr after transfer into sporulation medium . Rec8-3HA localization ( red ) was determined in spread nuclei from prophase I cells . DNA is shown in blue . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 01810 . 7554/eLife . 00117 . 019Figure 4—figure supplement 2 . Chromosomal association of Pds5 in CUP-CLB3 cells . Wild-type ( A28681 ) and CUP-CLB3 ( A28682 ) cells carrying the cdc20-mn allele were induced to sporulate and CuSO4 ( 50 μM ) was added 2 hr 30 min after transfer into sporulation medium . Pds5 localization was determined by ChIP-chip from metaphase I-arrested cells . Black balls depict centromere positions . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 01910 . 7554/eLife . 00117 . 020Figure 4—figure supplement 3 . CUP-CLB3 cells partially bypass the nuclear division delay of mam1∆ cells . Wild-type ( A22678 ) , CUP-CLB3 ( A22702 ) , mam1∆ ( A31340 ) and mam1∆ CUP-CLB3 ( A31342 ) cells carrying the GAL4-ER and GAL-NDT80 fusions were induced to sporulate and CuSO4 ( 50 μM ) was added 2 hr 15 min after transfer into sporulation medium . Cells were released from the NDT80 block 4 hr 30 min after transfer into sporulation medium . The percentage of cells that had undergone one or two meiotic divisions was determined at the indicated time points ( n = 100 per time point ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 02010 . 7554/eLife . 00117 . 021Figure 4—figure supplement 4 . Meiotic progression of the cells analyzed for Rec8 cleavage in Figure 4C . REC8-myc/REC8-HA ( A29957 ) , REC8-myc/rec8-29A-HA ( A29961 ) , REC8-myc/REC8-HA CUP-CLB3 ( A29959 ) and REC8-myc/rec8-29A-HA CUP-CLB3 ( A29963 ) cells carrying the GAL4-ER and GAL-NDT80 fusions were induced to sporulate and CuSO4 ( 50 μM ) was added 2 hr 15 min after transfer into sporulation medium . Cells were released from the NDT80 block 4 hr 30 min after transfer into sporulation medium . The percentage of cells in metaphase I ( grey symbols ) , anaphase I ( violet symbols ) , metaphase II ( dark blue symbols ) and anaphase II ( green symbols ) was determined at the indicated times ( n = 100 per time point ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 02110 . 7554/eLife . 00117 . 022Figure 4—figure supplement 5 . Analysis of Rec8 cleavage in cells used for Figure 4D . Wild-type ( A22804 ) , CUP-CLB3 ( A29965 ) , rec8-29A ( A22803 ) and rec8-29A CUP-CLB3 ( A29967 ) cells carrying the GAL4-ER and GAL-NDT80 fusions were induced to sporulate and CuSO4 ( 50 μM ) was added 2 hr 15 min after transfer into sporulation medium . Cells were released from the NDT80 block 4 hr 30 min after transfer into sporulation medium . Levels of full-length Rec8 , cleaved Rec8 , Clb3 and Pgk1 were monitored by Western blot . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 02210 . 7554/eLife . 00117 . 023Figure 4—figure supplement 6 . Meiotic progression of the cells analyzed for Rec8 cleavage in Figure 4D . Wild-type ( A22804 ) , CUP-CLB3 ( A29965 ) , rec8-29A ( A22803 ) and rec8-29A CUP-CLB3 ( A29967 ) cells carrying the GAL4-ER and GAL-NDT80 fusions were induced to sporulate and CuSO4 ( 50 μM ) was added 2 hr 15 min after transfer into sporulation medium . Cells were released from the NDT80 block 4 hr 30 min after transfer into sporulation medium . The percentage of cells in metaphase I ( grey symbols ) , anaphase I ( violet symbols ) , metaphase II ( dark blue symbols ) and anaphase II ( green symbols ) was determined at the indicated times ( n = 100 per time point ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 02310 . 7554/eLife . 00117 . 024Figure 4—figure supplement 7 . Chromosomal association of Sgo1 in CUP-CLB3 cells . Wild-type ( A28712 ) and CUP-CLB3 ( A28713 ) cells carrying the cdc20-mn allele were induced to sporulate and CuSO4 ( 50 μM ) was added 2 hr 30 min after transfer into sporulation medium . Sgo1-3V5 localization was determined by ChIP-chip , 7 hr after transfer into sporulation medium when cells were arrested in metaphase I . Arm peaks for Sgo1 correspond to cohesin-associated regions . The basis for Sgo1 enrichment at these sites is currently unclear . Black balls depict centromere positions . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 02410 . 7554/eLife . 00117 . 025Figure 4—figure supplement 8 . Localization of Rts1 in CUP-CLB3 cells . Wild-type ( A28331 ) and CUP-CLB3 ( A28332 ) cells carrying the cdc20-mn allele were induced to sporulate and CuSO4 ( 50 μM ) was added 2 hr 30 min after transfer into sporulation medium . Rts1-13myc ( green ) localization relative to Ndc10-6HA ( red ) was determined in spread nuclei from metaphase I-arrested cells ( n > 40 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 02510 . 7554/eLife . 00117 . 026Figure 4—figure supplement 9 . Chromosomal association of Spo13 in CUP-CLB3 cells . ( Left panel ) wild-type ( A30856 ) , CUP-CLB3 ( A30858 ) and CUP-CLB4 ( A30860 ) cells carrying the GAL4-ER and GAL-NDT80 fusions were induced to sporulate and CuSO4 ( 50 μM ) was added 2 hr 15 min after transfer into sporulation medium . Spo13-3V5 localization was determined by ChIP from prophase I-arrested cells . Relative occupancy at a centromeric site ( CEN5 ) relative to a low binding region ( HMR ) was determined . Error bars represent the range ( n = 2 ) . ( Right panel ) wild-type ( A30743 ) , CUP-CLB3 ( A30745 ) and CUP-CLB4 ( A30747 ) cells carrying the cdc20-mn allele were induced to sporulate and CuSO4 ( 50 μM ) was added 2 hr 30 min after transfer into sporulation medium . Spo13-3V5 localization was determined by ChIP 7 hr after transfer into sporulation medium when cells were arrested in metaphase I . Relative occupancy at a centromeric site ( CENV ) relative to a low binding region ( HMR ) was determined . Error bars represent the range ( n = 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 02610 . 7554/eLife . 00117 . 027Figure 4—figure supplement 10 . Rts1 localization in binucleate CUP-CLB3 cells . Wild-type ( A28329 ) and CUP-CLB3 ( A28330 ) cells were induced to sporulate and CuSO4 ( 50 μM ) was added 2 hr 30 min after transfer into sporulation medium . Rts1-13myc ( green ) localization relative to Ndc10-6HA ( red ) was determined in spread nuclei from binucleates ( n > 40 ) . Using a chi-square test ( df 2 ) the fraction of spread nuclei that display strong , weak or negative Rts1 with respect to Ndc10 was compared between wild-type and CUP-CLB3 χ2 = 54 . 49 , p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 02710 . 7554/eLife . 00117 . 028Figure 4—figure supplement 11 . Analysis of Rts1 localization in Rec8 phosphomimetic mutants . Wild-type ( A29645 ) and rec8-S136D S179D S197D T209D ( A29647 ) cells were induced to sporulate and Rec8-3HA/rec8-4D-3HA or Rts1-3V5 localization relative to Ndc10-13myc was determined in spread nuclei from binucleates ( n > 40 ) . Characterization of rec8-S136D S179D S197D T209D has been described in Katis et al . ( 2010 ) . Note that strains carrying this allele fail to maintain centromeric cohesin beyond metaphase I ( bottom panel ) . These binucleates also have weak Rts1 staining ( top panel ) , suggesting that Rts1 maintenance at centromeric regions in anaphase I depends on cohesin . For top panel , using a chi-square test ( df 2 ) the fraction of spread nuclei that display strong , weak or negative Rts1 with respect to Ndc10 was compared between wild-type and CUP-CLB3 χ2 = 18 . 02 , p=0 . 0001 . For bottom panel , using a chi-square test ( df 2 ) the fraction of spread nuclei that display strong , weak or negative Rec8 with respect to Ndc10 was compared between wild-type and CUP-CLB3 χ2 = 121 . 2 , p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 028 To test the possibility that CUP-CLB3 cells fail to maintain centromeric cohesion beyond anaphase I , we first determined the localization of the cohesin subunit Rec8 in cells that had progressed past metaphase I . Rec8 colocalized with the kinetochore component Ndc10 in binucleate wild-type and CUP-CLB4 cells , demonstrating that centromeric cohesin is protected from removal until the onset of anaphase II . In contrast , Rec8 was not detected around centromeres in a substantial fraction of binucleate CUP-CLB3 cells ( Figure 4B ) . Functional assays confirmed the defect in centromeric cohesion maintenance in CUP-CLB3 cells . Although mam1∆ cells biorient sister chromatids during meiosis I , they delay nuclear division until meiosis II due to the presence of centromeric cohesin ( Toth et al . , 2000; Rabitsch et al . , 2003 ) . The delay in nuclear division of a mam1∆ was partially alleviated by the expression of CUP-CLB3 ( Figure 4—figure supplement 3 ) . This partial effect is likely due to not all CUP-CLB3 cells losing centromeric cohesion prematurely in meiosis I ( Figure 4B ) . We conclude that both centromeric and arm cohesin are lost from chromosomes at the onset of anaphase I in CUP-CLB3 cells . Next , we investigated the cause of premature centromeric cohesin removal in CUP-CLB3 cells . Cleavage of cohesin by separase requires the phosphorylation of Rec8 at multiple residues ( Brar et al . , 2006; Katis et al . , 2010 ) . A recessive allele of REC8 in which 29 in vivo phosphorylation sites were mutated to alanine ( rec8-29A ) ( Brar et al . , 2006 ) was not cleaved in CUP-CLB3 cells , but wild-type Rec8 was ( Figure 4C and Figure 4—figure supplement 4 ) . Furthermore , the rec8-29A allele caused a similar metaphase I delay in wild-type and CUP-CLB3 cells when expressed as the sole source of REC8 ( Figure 4D and Figure 4—figure supplements 5 and 6 ) . We noticed that the Rec8 cleavage product was detected at lower levels in CUP-CLB3 cells ( Figure 4C and Figure 4—figure supplement 5 ) . The cause of this reduction is currently unclear , but could indicate that in CUP-CLB3 cells , cohesin removal also relies on a separase-independent pathway , that is the prophase removal pathway ( Yu and Koshland , 2005 ) . Our results demonstrate that Rec8 phosphorylation is required for cohesin removal in CUP-CLB3 cells and suggest that the defect in centromeric cohesin protection may result from increased phosphorylation of centromeric Rec8 . To test this possibility , we used phospho-specific antibodies against two in vivo phosphorylation sites of Rec8 ( pS179 and pS521 ) ( Brar et al . , 2006; Katis et al . , 2010; M . Attner personal communication , October 2011 ) and analyzed the relative enrichment of total Rec8 and phospho-Rec8 at CENV or at an arm cohesin binding site by ChIP in metaphase I-arrested cells . The two phospho-specific antibodies immunoprecipitated similar amounts of Rec8 in wild-type and CUP-CLB3 cells at the arm site ( Figure 4E ) , which is consistent with arm cohesin being primed for Separase cleavage . However , the amount of phosphorylated Rec8 was increased at the centromere in CUP-CLB3 cells compared to wild-type cells , albeit not to the same extent as in cells depleted for Sgo1 ( sgo1-mn ) , in which meiosis I centromeric-cohesin protection is completely defective ( Figure 4E ) . We conclude that CUP-CLB3 cells are compromised in preventing centromeric Rec8 phosphorylation during meiosis I . Sgo1-PP2A and the meiosis-specific protein Spo13 prevent centromeric Rec8 phosphorylation during meiosis I to protect this cohesin pool from cleavage . All three proteins localize to kinetochores during meiosis I , which is thought to be critical for their cohesin-protective function ( Katis et al . , 2004a , 2004b; Kitajima et al . , 2004; Lee et al . , 2004; Kitajima et al . , 2006; Riedel et al . , 2006 ) . Surprisingly , Sgo1 , the PP2A regulatory subunit Rts1 and Spo13 localized normally in prophase I- and metaphase I-arrested CUP-CLB3 cells ( Figure 4F–H and Figure 4—figure supplements 7–9 ) . We noticed a moderate reduction of Sgo1 and Rts1 at centromeres in binucleate CUP-CLB3 cells ( Figure 4I and Figure 4—figure supplement 10 ) . However , this reduction during anaphase I is most likely a consequence rather than a cause of premature loss of centromeric cohesin . In cells expressing a phosphomimetic version of Rec8 ( rec8-4D ) that cannot be retained at centromeres beyond meiosis I , Rts1 localization is also reduced in anaphase I ( Figure 4—figure supplement 11 ) . It is thus unlikely that the reduction of Sgo1 and Rts1 at centromeres during anaphase I contributes to the premature loss of centromeric cohesin . These findings , together with our observation that centromeric Rec8 phosphorylation is increased in CUP-CLB3 cells , indicate that Sgo1-PP2A function , but not localization , is impaired in CUP-CLB3 cells . How does premature expression of CLB3 interfere with establishment of the meiosis I chromosome segregation pattern ? The comparison of the effects caused by CLB3 and CLB4 misexpression provided insight into this question . Both cyclins induce spindle formation in prophase I . However , chromosomes are able to attach to this spindle and experience pulling forces only in CUP-CLB3 cells . Thus , the ability to form tension-generating attachments ( i . e . CUP-CLB1 or CUP-CLB3 cells ) correlates with defects in meiosis I chromosome morphogenesis and segregation . This correlation suggests that premature microtubule–kinetochore engagement during premeiotic S phase/early prophase I is the underlying cause of chromosome missegregation in CUP-CLB3 cells and predicts that tension generating microtubule–kinetochore attachments should inhibit meiosis I chromosome morphogenesis . Conversely , preventing them should enable building a proper meiosis I chromosome architecture . We tested the first prediction using a previously described method in which monopolin-dependent sister kinetochore coorientation is induced during mitosis ( Monje-Casas et al . , 2007 ) . Overexpression of MAM1 and CDC5 upon a pheromone-induced G1 arrest is sufficient to induce cosegregation of sister chromatids in mitotic anaphase ( Monje-Casas et al . , 2007 , Figure 5A ) . However , when cells are allowed to form microtubule–kinetochore attachments prior to CDC5 and MAM1 expression , cosegregation of sister chromatids is prevented . We reversibly arrested cells in metaphase using a methionine repressible CDC20 allele ( MET-CDC20 ) and induced MAM1 and CDC5 expression after cells had arrested in metaphase and had formed microtubule–kinetochore interactions . Under these conditions , MAM1 and CDC5 expression did not induce sister chromatid cosegregation when cells were released into anaphase ( Figure 5A ) . Importantly , disrupting microtubule–kinetochore interactions by depolymerizing microtubules with nocodazole during the metaphase arrest resulted in robust cosegregation of sister chromatids in anaphase ( 48% cosegregation , Figure 5A ) . These results show that microtubule–kinetochore interactions modulate the ability of monopolin to induce sister chromatid cosegregation . 10 . 7554/eLife . 00117 . 029Figure 5 . Transient disruption of microtubule–kinetochore interactions suppresses the chromosome segregation defects in CUP-CLB3 cells . ( A ) Wild-type ( A10684 ) and GAL-CDC5 GAL-MAM1 ( A26546 ) cells , carrying a MET-CDC20 allele and CENIV-GFP dots , were monitored for chromosome segregation in anaphase ( see ‘Materials and methods’ for details ) . MT = microtubule , KT = kinetochore , ( n = 100 ) . The fraction of anaphase cells that segregate or cosegregate sister chromatids was compared between GAL-CDC5 GAL-MAM1 condition ( 2 ) and GAL-CDC5 GAL-MAM1 condition ( 3 ) using a chi-square test ( df 1 ) χ2 = 59 . 71 , p<0 . 0001 . ( B ) Schematic description of the experimental regime used for ( C ) through ( H ) see ‘Materials and methods’ for details . ( C ) Localization of Lrs4-13myc ( green ) in mononucleates relative to Ndc10-6HA ( red ) determined by nuclear spreads ( n > 40 ) and ( D ) phosphorylation of Lrs4-13myc determined by gel mobility shift in wild-type ( A29612 ) , ndc80-1 ( A29614 ) , CUP-CLB3 ( A29616 ) and CUP-CLB3 ndc80-1 ( A29618 ) . For ( C ) , using a chi-square test ( df 2 ) the fraction of spread nuclei that display colocalized , partial or mislocalized Lrs4 with respect to Ndc10 was compared between wild-type and ndc80-1 χ2 = 0 . 9668 , p=0 . 6167 and between CUP-CLB3 and CUP-CLB3 ndc80-1 χ2 = 56 . 34 , p<0 . 0001 . ( E ) Localization of Rec8-13myc ( green ) in binucleates relative to Ndc10-6HA ( red ) determined by nuclear spreads in wild-type ( A28716 ) , ndc80-1 ( A28720 ) , CUP-CLB3 ( A28718 ) and CUP-CLB3 ndc80-1 ( A28722 ) ( n > 40 ) . Using a chi-square test ( df 1 ) the fraction of spread nuclei that were Rec8 positive or negative was compared between wild-type and ndc80-1 χ2 = 1 . 185 , p=0 . 2764 and between CUP-CLB3 and CUP-CLB3 ndc80-1 χ2 = 23 . 96 , p<0 . 0001 . ( F ) Segregation of sister chromatids using heterozygous CENV-GFP dots quantified in binucleates ( n = 100 ) and ( G ) spore viability from wild-type ( A22678 ) , ndc80-1 ( A28621 ) , CUP-CLB3 ( A22702 ) and CUP-CLB3 ndc80-1 ( A28623 ) ( n = 40 tetrads for wild-type and ndc80-1 , n > 60 tetrads for CUP-CLB3 and CUP-CLB3 ndc80-1 ) ( nonpermissive temperature >36°C ) . Using a chi-square test ( df 1 ) the fraction of binucleates with a reductional or equational division was compared between CUP-CLB3 and CUP-CLB3 ndc80-1 χ2 = 24 . 18 , p<0 . 0001 . ( G ) Segregation of chromosome V using homozygous CENV-GFP dots quantified in tetranucleates from wild-type ( A22688 ) , ndc80-1 ( A28625 ) , CUP-CLB3 ( A22708 ) and CUP-CLB3 ndc80-1 ( A28627 ) . Top panel: cells kept at 25°C for the duration of the experiment . Bottom panel: Cells treated as in ( B ) but monitored after meiosis II ( n = 100 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 02910 . 7554/eLife . 00117 . 030Figure 5—figure supplement 1 . Sporulation efficiency of ndc80-1 mutants . Wild-type ( A22678 ) and ndc80-1 ( A28221 ) cells were induced to sporulate at 25°C . 2 hr 30 min after transfer into sporulation medium , cells were shifted to the indicated temperature and sporulation efficiency was determined after 24 hr . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 03010 . 7554/eLife . 00117 . 031Figure 5—figure supplement 2 . Sister kinetochore coorientation in ndc80-1 cells under a continuous inactivation regime at 34°C during a metaphase I arrest . Wild-type ( A7118 ) , CUP-CLB3 ( A23074 ) , ndc80-1 ( A29690 ) and ndc80-1 CUP-CLB3 ( A29692 ) cells also carrying the cdc20-mn allele , were induced to sporulate at 25°C . 2 hr 45 min after transfer into sporulation medium , CuSO4 ( 50 μM ) was added and concurrently , cultures were shifted to 34°C . The percentage of mononucleate cells with separated CENV-GFP dots was determined 7 hr 30 min after transfer into sporulation medium when cells were arrested in metaphase I ( n = 100 ) . The fraction of nuclei that display sister kinetochores as separate or together was compared between CUP-CLB3 and CUP-CLB3 ndc80-1 using a chi-square test ( df 1 ) χ2 = 7 . 228 , p=0 . 0072 . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 03110 . 7554/eLife . 00117 . 032Figure 5—figure supplement 3 . Sister kinetochore coorientation in ndc80-1 cells after transient inactivation regime at 34°C during a metaphase I arrest . Wild-type ( A20958 ) , CUP-CLB3 ( A23076 ) , ndc80-1 ( A29718 ) and ndc80-1 CUP-CLB3 ( A29720 ) cells also carrying the GAL4-ER and GAL-NDT80 fusions and the cdc20-mn allele were induced to sporulate at 25°C . 2 hr 45 min after transfer into sporulation medium CuSO4 ( 50 μM ) was added and concurrently , cultures were shifted to 34°C . After 5 hr , when cells had arrested in the NDT80 arrest , cells were released from the NDT80 block and transferred to 25°C . The percentage of mononucleate cells with separated CENV-GFP dots was determined 7 hr 30 min after transfer into sporulation medium when cells were arrested in metaphase I ( n = 100 ) . The fraction of nuclei that display sister kinetochores as separate or together was compared between CUP-CLB3 and CUP-CLB3 ndc80-1 using a chi-square test ( df 1 ) χ2 = 5 . 007 , p=0 . 0252 . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 03210 . 7554/eLife . 00117 . 033Figure 5—figure supplement 4 . Meiosis I chromosome segregation in ndc80-1 cells after a transient inactivation regime at 34°C . Wild-type ( A22678 ) , ndc80-1 ( A28621 ) , CUP-CLB3 ( A22702 ) and CUP-CLB3 ndc80-1 ( A28623 ) cells also carrying the GAL4-ER and GAL-NDT80 fusions were induced to sporulate at 25°C . 2 hr 45 min after transfer into sporulation medium CuSO4 ( 50 μM ) was added and concurrently , cultures were shifted to 34°C . After 5 hr , when cells had arrested in the NDT80 block , cells were released and transferred to 25°C . The percentage of binucleate cells with segregated heterozygous CENV-GFP dots was determined 7 hr 30 min after transfer into sporulation medium ( n = 100 ) . Note that a greater suppression of meiosis I sister chromatid segregation was observed in ndc80-1 CUP-CLB3 cells when cells were incubated at temperatures higher than 34°C ( Figure 5F and data not shown ) . The fraction of binucleates that underwent reductional or equational division was compared between CUP-CLB3 and CUP-CLB3 ndc80-1 using a chi-square test ( df 1 ) χ2 = 5 . 776 , p=0 . 0162 . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 03310 . 7554/eLife . 00117 . 034Figure 5—figure supplement 5 . Transient disruption of microtubule–kinetochore interactions using dam1-1 allele restores meiosis I chromosome segregation in CUP-CLB3 cells . Wild-type ( A22678 ) , dam1-1 ( A28311 ) , CUP-CLB3 ( A22702 ) and CUP-CLB3 dam1-1 ( A28341 ) cells also carrying the GAL4-ER and GAL-NDT80 fusions were induced to sporulate at 25°C . 2 hr 45 min after transfer into sporulation medium CuSO4 ( 50 μM ) was added and concurrently , cultures were shifted to 34°C . After 5 hr , when cells had arrested in the NDT80 block , cells were released and transferred to 25°C . The percentage of binucleate cells with segregated heterozygous CENV-GFP dots was determined 7 hr 30 min after transfer into sporulation medium ( n = 100 ) . The fraction of binucleates that underwent reductional or equational division was compared between CUP-CLB3 and CUP-CLB3 dam1-1 using a chi-square test ( df 1 ) χ2 = 16 . 77 , p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 03410 . 7554/eLife . 00117 . 035Figure 5—figure supplement 6 . Transient disruption of microtubule–kinetochore interactions by benomyl treatment restores meiosis I chromosome segregation in CUP-CLB3 cells . Wild-type ( A22678 ) and CUP-CLB3 ( A22702 ) cells also carrying the GAL4-ER and GAL-NDT80 fusions were induced to sporulate at 30°C . 2 hr 15 min after transfer into sporulation medium CuSO4 ( 50 μM ) was added and concurrently , cells were treated with DMSO or benomyl ( 120 μg/ml ) . Cells were subsequently released from NDT80 block 4 hr 30 min after transfer into sporulation medium and benomyl was washed out concomitant with NDT80-block release . The percentage of binucleate cells with segregated heterozygous CENV-GFP dots was determined 6 hr after transfer into sporulation medium ( n = 100 ) . See ‘Materials and methods’ for further details . The fraction of binucleates that underwent reductional or equational division was compared between CUP-CLB3 + DMSO and CUP-CLB3 + benomyl using a chi-square test ( df 1 ) χ2 = 32 . 12 , p<0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 03510 . 7554/eLife . 00117 . 036Figure 5—figure supplement 7 . Transient disruption of microtubule–kinetochore interactions during S phase/prophase I suppresses CUP-CLB3-induced meiosis I sister chromatid segregation in a spindle assembly checkpoint independent manner . ( Top panel ) wild-type ( A22678 ) , mad3∆ ( A30386 ) , ndc80-1 ( A28621 ) , ndc80-1 mad3∆ ( A30390 ) , CUP-CLB3 ( A22702 ) , CUP-CLB3 mad3∆ ( A30388 ) , CUP-CLB3 ndc80-1 ( A28623 ) and CUP-CLB3 ndc80-1 mad3∆ ( A30392 ) cells also carrying the GAL4-ER and GAL-NDT80 fusions were induced to sporulate at 25°C . 2 hr 45 min after transfer into sporulation medium CuSO4 ( 50 μM ) was added and concurrently , cultures were shifted to 36°C . Cells were subsequently released from NDT80 block at 5 hr and transferred to 25°C . Percent binucleates with segregated heterozygous CENV-GFP dots was determined ( n = 100 ) . ( Bottom panel ) wild-type ( A22688 ) , mad3∆ ( A30638 ) , ndc80-1 ( A28625 ) , ndc80-1 mad3∆ ( A30642 ) , CUP-CLB3 ( A22708 ) , CUP-CLB3 mad3∆ ( A30640 ) , CUP-CLB3 ndc80-1 ( A28627 ) and CUP-CLB3 ndc80-1 mad3∆ ( A30644 ) cells also carrying the GAL4-ER and GAL-NDT80 fusions were induced to sporulate at 25°C . 2 hr 45 min after transfer to sporulation medium CuSO4 ( 50 μM ) was added and concurrently , cultures were shifted to 36°C . After 5 hr , when cells had arrested in the NDT80 block , cells were released and transferred to 25°C . The percentage of binucleate cells with segregated homozygous CENV-GFP dots was determined 7 hr 30 min after transfer into sporulation medium . Binucleate cells with GFP signal in only one of the two nuclei were categorized as having experienced a meiosis I non-disjunction event ( n = 100 ) . For top panel , using a chi-square test ( df 1 ) the fraction of binucleates that underwent reductional or equational division was compared between CUP-CLB3 and CUP-CLB3 mad3∆ χ2 = 0 . 1800 , p=0 . 6714 and between CUP-CLB3 ndc80-1 and CUP-CLB3 ndc80-1 mad3∆ χ2 = 0 . 02454 , p=0 . 8755 . For bottom panel , using a chi-square test ( df 1 ) the fraction of binucleates that displayed MI nondisjunction or other was compared between CUP-CLB3 and CUP-CLB3 mad3∆ χ2 = 1 . 228 , p=0 . 2678 and between CUP-CLB3 ndc80-1 and CUP-CLB3 ndc80-1 mad3∆ χ2 = 0 . 6486 , p=0 . 4206 . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 03610 . 7554/eLife . 00117 . 037Figure 5—figure supplement 8 . Transient disruption of microtubule–kinetochore interactions during S phase/prophase I restores meiotic chromosome segregation in CUP-CLB3 cells in a spindle assembly checkpoint independent manner . Wild-type ( A22688 ) , ndc80-1 ( A28625 ) , CUP-CLB3 ( A22708 ) , CUP-CLB3 ndc80-1 ( A28627 ) , mad3∆ ( A30638 ) , ndc80-1 mad3∆ ( A30642 ) , CUP-CLB3 mad3∆ ( A30640 ) and CUP-CLB3 ndc80-1 mad3∆ ( A30644 ) cells also carrying the GAL4-ER and GAL-NDT80 fusions were induced to sporulate at 25°C . 2 hr 45 min after transfer to sporulation medium CuSO4 ( 50 μM ) was added and concurrently , cultures were shifted to 36°C . After 5 hr , when cells had arrested in the NDT80 block , cells were released and transferred to 25°C . Segregation of homozygous CENV-GFP dots was determined in tetranucleates 12 hr after transfer into sporulation medium ( n = 100 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 03710 . 7554/eLife . 00117 . 038Figure 5—figure supplement 9 . Transient ndc80-1 inactivation does not alter in vitro Cdk1 activity . Wild-type ( A25508 ) , ndc80-1 ( A33203 ) , CUP-CLB3 ( A33201 ) and CUP-CLB3 ndc80-1 ( A33205 ) cells also carrying the GAL4-ER and GAL-NDT80 fusions were induced to sporulate at 25°C . 2 hr 45 min after transfer to sporulation medium CuSO4 ( 50 μM ) was added and concurrently , cultures were shifted to 35°C . Samples were harvested 5 hr post transfer to sporulation medium , when cells were arrested in the NDT80 block . In vitro kinase assays were performed with Cdc28-3V5 ( Cdk1 ) immunoprecipitated from prophase I-arrested samples . Amounts of phosphorylated Histone H1 and immunoprecipitated Cdc28-3V5 are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 038 If the defects in sister kinetochore coorientation and centromeric cohesin maintenance of CUP-CLB3 cells are caused by premature microtubule–kinetochore interactions , proper meiosis I chromosome morphogenesis should be restored by transiently disrupting microtubule–kinetochore interactions . To test this , we used a temperature sensitive allele of NDC80 ( ndc80-1 ) , which encodes a component of the outer kinetochore . ndc80-1 cells grow and sporulate normally at 25°C , but fail to undergo any nuclear divisions at temperatures above 34°C ( Figure 5—figure supplement 1 ) . We first asked whether disrupting microtubule–kinetochore interactions suppresses the kinetochore localization defect of monopolin in CUP-CLB3 cells . Using the NDT80 block-release system , we induced cells to sporulate at 25°C . After 165 min , we induced cyclin expression and concurrently transferred cells to 34°C to inactivate the ndc80-1 allele . Cells were then incubated for an additional 135 min to arrest them in the NDT80-depletion block . We then transferred cells to the permissive temperature and released them from the NDT80 block into a metaphase I-arrest by depleting CDC20 ( cdc20-mn ) ( Figure 5B ) . Under these conditions , wild-type and ndc80-1 cells arrested in metaphase I with the monopolin subunit Lrs4 localized to kinetochores , while CUP-CLB3 cells showed a defect in Lrs4 localization ( Figure 5C ) . Remarkably , CUP-CLB3 ndc80-1 cells showed near wild-type levels of Lrs4 association with kinetochores ( Figure 5C ) . Transient inactivation of Ndc80 also restored Lrs4 phosphorylation in CUP-CLB3 cells ( Figure 5D ) . Our results demonstrate that premature microtubule–kinetochore interactions prevent sister kinetochore coorientation by disrupting proper localization of the monopolin complex . The finding that transient disruption of microtubule–kinetochore interactions also suppresses the Lrs4 phosphorylation defect of CUP-CLB3 cells , furthermore suggests that Lrs4 hyperphosphorylation occurs not at the time of nucleolar release , but once Lrs4 localizes to kinetochores . We next asked whether transient inactivation of microtubule–kinetochore interactions also suppresses the premature loss of centromeric cohesin observed in CUP-CLB3 cells . We used a similar protocol to the one described above , except cells were not arrested in metaphase I following release from the NDT80 block , but were allowed to proceed into anaphase I to examine Rec8 localization . Remarkably , disrupting microtubule–kinetochore interactions at the time of Clb3 expression caused a considerable increase in the percentage of CUP-CLB3 cells that retained Rec8 around centromeres during anaphase I ( Figure 5E ) . Finally , restoring centromeric cohesin protection and sister kinetochore coorientation to CUP-CLB3 cells by transient inactivation of Ndc80 restored homolog segregation during meiosis I ( Figure 5F and Figure 5—figure supplements 2–4 ) . Similar results were obtained with a temperature sensitive allele of the gene encoding the outer kinetochore component Dam1 ( dam1-1 ) or by disrupting microtubule–kinetochore interactions by benomyl treatment ( Figure 5—figure supplements 5 and 6 ) . We further observed a striking improvement in overall chromosome segregation and spore viability in CUP-CLB3 ndc80-1 compared to CUP-CLB3 cells ( Figure 5G , H ) . The suppression of chromosome missegregation in CUP-CLB3 ndc80-1 cells did not depend on the spindle assembly checkpoint , because deletion of MAD3 had no discernable effect on the extent of ndc80-1 mediated suppression ( Figure 5—figure supplements 7 and 8 ) , nor was it due to the ndc80-1 allele lowering Clb3-CDK activity ( Figure 5—figure supplement 9 ) . In summary , our results demonstrate that the defects associated with CUP-CLB3 cells are due to premature microtubule–kinetochore interactions . Our results further suggest that preventing microtubule–kinetochore interactions during premeiotic S phase and prophase I is necessary to establish a meiosis I-specific chromosome architecture . Our results demonstrate that preventing premature interactions of microtubules with kinetochores is essential for establishing a meiosis I chromosome architecture . This occurs , at least in part , by restricting Clb-CDK activity during premeiotic S phase and prophase I . Are additional mechanisms in place to prevent premature microtubule–kinetochore interactions ? Insight into this question came from the variability in CUP-CLB3-associated phenotypes . We initially noticed that the timing of CLB3 induction had an impact on the extent of sister chromatid segregation in meiosis I , especially in experiments that employed the NDT80 block-release system . To investigate this further , we expressed CLB3 at different times after induction of sporulation . We observed that the extent of meiosis I sister chromatid segregation declined as CLB3 was expressed later during the NDT80 block ( Figure 6A ) . One possibility is that CLB3-induced sister chromatid segregation depends on additional factors that become limiting . Kinetochore components are good candidates for such additional factors , because previous studies in fission yeast demonstrated that a subset of outer kinetochore components dissociates from the kinetochore during prophase I ( Asakawa et al . , 2005 ) . 10 . 7554/eLife . 00117 . 039Figure 6 . Meiosis I sister chromatid segregation correlates with presence of outer kinetochore components . ( A ) Schematic description of the experimental regime and segregation of sister chromatids using heterozygous CENV-GFP dots quantified in binucleates from wild-type ( A22678 ) and CUP-CLB3 ( A29406 ) after cyclin induction at 2 hr 15 min , 3 hr , 4 hr and 4 hr 30 min post transfer to sporulation medium . Cells released from NDT80-block at 4 hr 30 min ( n = 100 ) . Using a chi-square test ( df 1 ) , the fraction of binucleates that display a reductional or equational division was compared between wild-type and CUP-CLB3 for each induction time point: ( 2:15 ) , χ2 = 58 . 00 , p<0 . 0001; ( 3:00 ) , χ2 = 14 . 46 , p=0 . 0001; ( 4:00 ) , χ2 = 1 . 020 , p=0 . 3124; ( 4:30 ) , χ2 = 0 . 3384 , p=0 . 5607 . ( B ) Cluster analysis of kinetochore components from the indicated time points . Further details are in the ‘Materials and methods’ and in Brar et al . ( 2012 ) . Inner kinetochore = Cse4 nucleosomes , Cbf3 , Ctf19 complexes and Mif2 . Outer kinetochore = Spc105 , Mis12 , Ndc80 and DASH complexes . Fold induction is calculated by dividing the average expression from metaphase I—anaphase I by the average expression from DNA replication-prophase I . ( C ) Ordered plot for mRNA-seq and ribosome footprinting data for NDC80 and ( D ) HSK3 at the indicated stages . Dotted line indicates time of release from NDT80 block . Further details are in the ‘Materials and methods’ and in Brar et al . ( 2012 ) . ( E ) Western blot for Ndc80-3V5 and Pgk1 from A30340 cells and ( F ) Hsk3-3V5 and Pgk1 from A31861 cells . Cells induced to sporulate and released from NDT80 block at 4 hr 30 min . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 03910 . 7554/eLife . 00117 . 040Figure 6—figure supplement 1 . Schematic representation of the kinetochore–microtubule interface . Components of the kinetochore subcomplexes are grouped in color coded boxes . Note that the schematic representation is not drawn to scale . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 04010 . 7554/eLife . 00117 . 041Figure 6—figure supplement 2 . Meiotic cluster analysis of kinetochore components . Cluster analysis of kinetochore components from the indicated time points . Further details are in ‘Materials and methods’ and in Brar et al . ( 2012 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 04110 . 7554/eLife . 00117 . 042Figure 6—figure supplement 3 . Meiotic expression of DASH complex subunits . Ordered plot of mRNA-seq and ribosome footprinting data for the DASH complex components at indicated stages of sporulation . Dotted line indicates time of release from NDT80 block . See Brar et al . ( 2012 ) for details . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 04210 . 7554/eLife . 00117 . 043Figure 6—figure supplement 4 . Meiotic expression of Ndc80 complex subunits . Ordered plot of mRNA-seq and ribosome footprinting data for the Ndc80 complex at indicated stages . Dotted line indicates time of release from NDT80 block . See Brar et al . ( 2012 ) for details . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 04310 . 7554/eLife . 00117 . 044Figure 6—figure supplement 5 . Meiotic expression of Mif2 . Ordered plot of the mRNA-seq and ribosome footprinting data for Mif2 at the indicated stages . Dotted line indicates time of release from the NDT80 block . See Brar et al . ( 2012 ) for details . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 04410 . 7554/eLife . 00117 . 045Figure 6—figure supplement 6 . Meiotic expression of KNL-1 complex subunits . Ordered plot of the mRNA-seq and ribosome footprinting data for KNL-1 complex subunits ( Spc105 complex ) at the indicated stages . Dotted line indicates time of release from the NDT80 block . See Brar et al . ( 2012 ) for details . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 04510 . 7554/eLife . 00117 . 046Figure 6—figure supplement 7 . Meiotic expression of Mis12 complex subunits . Ordered plot of the mRNA-seq and ribosome footprinting data for Mis12 complex subunits at the indicated stages . Dotted line indicates time of release from NDT80 block . See Brar et al . ( 2012 ) for details . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 04610 . 7554/eLife . 00117 . 047Figure 6—figure supplement 8 . Meiotic expression of Ctf19 complex subunits . Ordered plot for mRNA-seq and ribosome footprinting data for Ctf19 complex at indicated stages . Dotted line indicates time of release from the NDT80 block . See Brar et al . ( 2012 ) for details . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 04710 . 7554/eLife . 00117 . 048Figure 6—figure supplement 9 . Meiotic expression of Cbf3 complex subunits . Ordered plot of the mRNA-seq and ribosome footprinting data for Cbf3 complex subunits at the indicated stages . Dotted line indicates time of release from the NDT80 block . See Brar et al . ( 2012 ) for details . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 04810 . 7554/eLife . 00117 . 049Figure 6—figure supplement 10 . Meiotic expression of Histone subunits . Ordered plot for mRNA-seq and ribosome footprinting data for the histones at indicated stages . Dotted line indicates time of release from NDT80 block See ( Brar et al . , 2012 ) for details . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 04910 . 7554/eLife . 00117 . 050Figure 6—figure supplement 11 . Meiotic expression of Dam1 and Ask1 . DAM1-3V5 ( A28898 ) cells carrying the GAL4-ER and GAL-NDT80 fusion and ASK1-13myc ( A29161 ) carrying ndt80∆ were induced to sporulate and were either released from ( left panel ) or arrested in ( right panel ) the NDT80 block . Levels of Dam1-3V5 , Ask1-13myc , Kar2 and Pgk1 were monitored by Western blot analysis . Pgk1 and Kar2 served as loading controls . DOI: http://dx . doi . org/10 . 7554/eLife . 00117 . 050 Using a high-resolution ribosome profiling dataset ( Brar et al . , 2012 ) , we examined the timing of synthesis of all kinetochore components during meiotic progression by cluster analysis . This analysis revealed two major expression classes , one included kinetochore components that peak in expression prior to or during prophase I ( early class ) , and the other contained components that instead show peak expression during the meiotic divisions ( late class ) . The early class was enriched for inner kinetochore components ( 16 of 23 ) , while the late class included primarily outer kinetochore components ( 13 of 18 ) ( Figure 6B , Figure 6—figure supplements 1–10 ) . The inner kinetochore binds to the centromere and generates a platform for the assembly of the outer kinetochore , which mediates microtubule attachments ( Tanaka , 2010 ) . The temporal difference in expression suggests that the inner kinetochore is assembled prior to the meiotic divisions , while the outer kinetochore is constructed only as cells enter the meiotic divisions . Among the outer kinetochore components that displayed peak synthesis during the divisions , NDC80 and a subunit of the DASH complex , HSK3 , displayed the most differential expression prior to meiosis I and during meiosis I , with a 9 . 02 and 2 . 64-fold induction , respectively ( Figure 6B–D ) . This decline in Ndc80 expression is consistent with a previous study in fission yeast , showing that Ndc80 becomes undetectable at kinetochores during prophase I ( Asakawa et al . , 2005 ) . Analysis of Ndc80 protein levels provided an explanation for why cells upregulate the synthesis of outer kinetochore components during entry into meiosis I . Ndc80 levels declined during premeiotic S phase and became undetectable during late prophase I ( Figure 6E ) . Importantly , the ability of CUP-CLB3 to induce sister-chromatid segregation during meiosis I tightly correlated with Ndc80 protein levels; as Ndc80 protein declines , so does CLB3-induced meiosis I sister chromatid segregation ( compare Figure 6A , E ) . Hsk3 protein levels were also low until meiosis I entry ( Figure 6F ) , but not all outer kinetochore components exhibited this decline in protein levels . For example , Ask1 , a subunit of the DASH complex , was present throughout prophase I and levels of another DASH complex component , Dam1 , increased during prophase I ( Figure 6—figure supplement 11 ) . Our findings indicate that the assembly of the outer kinetochore is restricted prior to NDT80 expression and pachytene exit due to low levels of a subset of outer kinetochore components . To determine whether reduced expression of the outer kinetochore components Ndc80 and Hsk3 contributes to preventing premature microtubule–kinetochore engagement , we examined the consequences of expressing the two genes from the CUP1 promoter ( Figure 7 ) . We first assessed whether expression of the two proteins allows for the recruitment of the DASH complex to kinetochores , which occurs via delivery through microtubules and can thus be used as a means of assessing end-on attachment of kinetochores ( Cheeseman et al . , 2001; Tanaka , 2010 ) . Cells were induced to sporulate and after 4 hr , a time when Ndc80 levels are normally diminished , we induced the expression of CLB3 , NDC80 and/or HSK3 . Whereas expression of either gene alone caused only a few cells to recruit Ask1 to kinetochores , cells simultaneously expressing NDC80 , HSK3 and CLB3 during prophase I showed colocalization between Ask1 and the inner kinetochore component Ndc10 , to an equal or greater extent than what was observed in metaphase I-arrested wild-type cells ( Figure 8A , B ) . The difference in Ask1 localization was not due to a difference in ASK1 expression ( Figure 8C ) . In addition , induction of CLB3 under the conditions mentioned above gave rise to bipolar spindles that appeared fragile with a weakened midzone . In contrast , consistent with stable microtubule–kinetochore interactions , coexpression of CLB3 , HSK3 and NDC80 resulted in the formation of robust bipolar spindles ( Figure 8D , E ) . Importantly , the expression of NDC80 and/or HSK3 during an NDT80 block caused a considerable increase in meiosis I sister chromatid segregation in CUP-CLB3 cells ( Figure 8F ) . Furthermore , under conditions in which CLB3 expression alone failed to induce meiosis I sister chromatid segregation , expression of CLB3 together with NDC80 and HSK3 caused a substantial increase in meiosis I sister chromatid segregation ( Figure 8G ) . This occurred even when cells were maintained in a prolonged NDT80 block prior to expression of CLB3 , NDC80 and HSK3 ( Figure 9 ) , ruling out the possibility that the expression of NDT80 targets , such as CDC5 , early during sporulation contributes to sister chromatid segregation during meiosis I . We conclude that limiting outer kinetochore assembly is an additional mechanism to prevent microtubule–kinetochore interactions during premeiotic S phase and prophase I .
Transcriptional and translational controls restrict CLB3 expression to meiosis II ( Carlile and Amon , 2008 ) . Eliminating both , by placing the gene under the control of the GAL1-10 promoter or the CUP1 promoter has dramatic effects on meiosis I chromosome segregation . CLB3 expression from the GAL1-10 promoter , which leads to Clb3 levels comparable to those seen for wild-type cells in meiosis II , causes a significant suppression of the meiosis I chromosome segregation pattern . This defect is not further enhanced by overexpression of the protein ( by expression from the CUP1 promoter ) , which further indicates that this phenotype does not emanate from expressing exceedingly high levels of the cyclin , but is a consequence of premature expression . The consequences of premature CLB3 expression are dramatic . It leads to premature microtubule–kinetochore interactions and prevents coorientation factors from associating with kinetochores . The observation that the transient disruption of microtubule–kinetochore interactions , either by inactivating the outer kinetochore or by microtubule depolymerization , allowed coorientation factors to associate with kinetochores , despite CLB3 misexpression , led us to conclude that it is premature microtubule–kinetochore interactions that interfere with the establishment of sister kinetochore coorientation during meiosis I . It is currently unclear how preexisting microtubule–kinetochore interactions prevent monopolin association with kinetochores . Precocious attachment of microtubules to kinetochores could occlude the monopolin complex from binding to kinetochores . Alternatively , tension between sister kinetochores generated from stable microtubule–kinetochore interactions could induce a conformational change at the kinetochore and/or pericentric chromatin such that coorientation factors can no longer associate with the kinetochore . In addition to preventing sister kinetochore coorientation , premature expression of CLB3 interferes with protecting centromeric cohesin from removal during meiosis I . The same logic as outlined for coorientation factors applies to the conclusion that it is Clb3-CDK mediated premature microtubule–kinetochore interactions that lead to loss of centromeric cohesin protection in CUP-CLB3 cells; disrupting microtubule–kinetochore interactions by various means restores stepwise loss of cohesin in CUP-CLB3 cells . A simple interpretation of this result is that the centromeric-cohesin protective domain can be disrupted by tension between sister kinetochores at any meiotic stage prior to anaphase I . This does not appear to be the case . In cells lacking the coorientation factor MAM1 , sister kinetochores come under tension in metaphase I , yet in these cells centromeric cohesin is not removed prematurely ( Toth et al . , 2000 and Figure 4—figure supplement 3 ) . Thus , the timing of microtubule–kinetochore interactions is of importance . It is tempting to speculate that the establishment of the centromeric-cohesin protective domain , which occurs during prophase I or perhaps even earlier , is sensitive to premature microtubule–kinetochore interactions and/or tension that promote biorientation of sister kinetochores . However , once this domain is established , its maintenance during meiosis I can no longer be disrupted by tension between sister kinetochores . How premature microtubule–kinetochore interactions affect the centromeric cohesin protection machinery is not yet known . A defect in localization of the protective machinery to kinetochores does not appear to be the cause of this defect . Sgo1 and PP2A localize normally to kinetochores in CUP-CLB3 cells . Therefore , lack of cohesin protection upon premature microtubule–kinetochore engagement must either result from a defect in an unknown cohesin protection pathway or from a decrease in the activity of Sgo1 and/or PP2A . Premature association of kinetochores with microtubules could result in the untimely recruitment of a factor ( e . g . Clb-CDKs themselves ) to the pericentromere that inhibits the cohesin protective machinery . Alternatively , microtubule–kinetochore engagement could directly affect the activity of the protective machinery . Two mechanisms have been previously proposed whereby tension modulates the activity of the cohesin protective machinery . In mammalian cells , tension spatially separates centromeric cohesin from Sgo1-PP2A , perhaps leading to loss of protection ( Lee et al . , 2008 ) . Tension has also been proposed to cause a deformation in PP2A , thus inhibiting its catalytic activity ( Grinthal et al . , 2010 ) . Irrespective of whether it is tension-dependent perturbation of Sgo1-PP2A and/or recruitment of inhibitory factors , it is clear that premature microtubule–kinetochore engagement is a bona fide modulator of the cohesin protective machinery . Cyclin-CDKs regulate multiple aspects of microtubule–kinetochore dynamics . Cyclin-CDKs promote centrosome separation and bipolar spindle assembly ( Fitch et al . , 1992 ) , kinetochore maturation ( Holt et al . , 2009 ) and chromosomal passenger complex localization ( Tsukahara et al . , 2010 ) . Given the importance of preventing premature microtubule–kinetochore engagement to meiosis I chromosome morphogenesis , it is not surprising that cyclin-CDK activity is regulated at multiple levels in budding yeast; transcription of CLB1 , CLB3 and CLB4 is not activated until cells exit pachytene ( Chu and Herskowitz , 1998 ) and CLB3 translation is restricted to meiosis II ( Carlile and Amon , 2008 ) . Cyclin-CDK activity is also tightly regulated in other eukaryotes . Metazoan oocytes arrest for an extended period of time in prophase I . Multiple mechanisms keep cyclin-CDK activity low to maintain this arrest ( reviewed in Von Stetina and Orr-Weaver , 2011 ) . Similar regulation is observed in D . melanogaster and C . elegans . Remarkably , inappropriate activation of Cyclin A or cyclin E during prophase I in fruit flies and worms , respectively , results in a mitosis-like division ( Sugimura and Lilly , 2006; Biedermann et al . , 2009 ) . Thus , restricting cyclin-CDK activity during premeiotic S phase and prophase I also appears to be required to establish a meiosis I-specific chromosome architecture in higher eukaryotes . Restriction of cyclin-CDK activity during premeiotic S phase and prophase I appears to be the major mechanism preventing premature microtubule–kinetochore interactions , but our data indicate that regulation of outer kinetochore assembly serves as an additional mechanism to prevent this from occurring . CUP-CLB3 can only induce meiosis I sister chromatid segregation when expressed during premeiotic S phase/early prophase I , but fails to do so when expressed during late prophase I . This difference is likely due to the outer kinetochore being present only until early prophase I . When Ndc80 , Hsk3 and Clb3 are coexpressed during late prophase I , sister chromatid segregation occurs in meiosis I . This result demonstrates that the presence of Clb3-CDKs alone during late prophase I is not sufficient to cause meiosis I sister chromatid segregation but that outer kinetochore components must also be expressed . Whether outer kinetochore disassembly solely occurs to prevent microtubule kinetochore interactions remains to be determined . Outer kinetochore disassembly could also serve additional purposes during prophase I such as enabling telomere-mediated chromosome movements . Further study of the kinetochore assembly/disassembly cycle during meiosis will provide insights into the full impact of kinetochore regulation on meiotic chromosome segregation . In budding yeast , two essential components of the outer kinetochore , Ndc80 and Hsk3 , are downregulated during prophase I . In S . pombe , Ndc80 and its binding partner Nuf2 dissociate from kinetochores in prophase I ( Asakawa et al . , 2005 ) raising the interesting possibility that deconstruction of the outer kinetochore is a conserved feature of meiotic prophase I . This dissociation depends on the mating pheromone signaling pathway ( Asakawa et al . , 2005 ) . Intriguingly , ectopic induction of meiosis without mating pheromone signaling ( i . e . in pat1 mutants ) , results in segregation of sister chromatids instead of homologous chromosomes in meiosis I ( Yamamoto and Hiraoka , 2003; Yamamoto et al . , 2004 ) . Perhaps this change in the pattern of chromosome segregation in pat1 mutants arises from premature microtubule–kinetochore interactions due to a defect in outer kinetochore disassembly . Interestingly , in mouse oocytes , the Ndc80 complex is recruited to chromosomes only after nuclear envelope breakdown ( Sun et al . , 2011 ) , raising the possibility that outer kinetochore assembly is also prevented in meiotic prophase I in vertebrates . Proper segregation of the genome during gametogenesis is critical for the proliferation of sexually reproducing species . Errors in chromosome segregation during meiosis result in aneuploidy , the leading cause of birth defects and miscarriages in humans ( Hassold and Hunt , 2001 ) . Thus , it is crucial to understand how accurate meiotic chromosome segregation is achieved . We discovered that the establishment of a meiosis-specific chromosome segregation pattern depends on the regulation of microtubule–kinetochore interactions . This is achieved by regulating cyclin-CDK activity as well as assembly of the outer kinetochore . There is evidence for similar regulatory events across different organisms ( Asakawa et al . , 2005; Sugimura and Lilly , 2006; Biedermann et al . , 2009; Von Stetina and Orr-Weaver , 2011 ) , suggesting that temporal restriction of microtubule-kinetochore interactions is an evolutionarily conserved event required to execute proper meiotic chromosome segregation .
Strains used in this study are described in Supplementary file 1 and are derivatives of SK1 ( all meiosis experiments ) or W303 ( Figure 5A ) . GAL-NDT80 and GAL4-ER constructs are described in Benjamin et al . ( 2003 ) . CUP-CLB1 , CUP-CLB3 , CUP-CLB4 , CUP-CLB5 , SPC42-mCherry , SGO1-3V5 , RTS1-13myc , RTS1-3V5 , HSK3-3V5 , NDC80-3V5 , ASK1-13myc , CUP-NDC80-3V5 , CUP-HSK3 , mam1∆ , SPO13-3V5 , mad3∆ , DAM1-3V5 , CUP-HSK3-3HA were constructed by PCR-based methods described in Longtine et al . ( 1998 ) . Primer sequences for strain constructions are available upon request . ndc80-1 and dam1-1 are described in Wigge et al . ( 1998 ) ; Jones et al . ( 1999 ) and SK1 strains carrying these alleles were constructed via backcrossing ( >9X ) . CENV-LacO was constructed by cloning a CENV homology region with XhoI restriction sites into the SalI cut plasmid pCM40 ( gift from Doug Koshland ) and integrated near CDEIII ( <1 kb ) by BamHI digest . pHG40 carrying CUP1 promoter was a gift from Hong-Guo Yu . 3V5 tagging plasmids were provided by Vincent Guacci . Strains were grown to saturation in YPD at room temperature , diluted in BYTA ( 1% yeast extract , 2% tryptone , 1% potassium acetate , 50 mM potassium phthalate ) to OD600 = 0 . 25 , and grown overnight at 30°C ( room temperature for ndc80-1 and dam1-1 experiments ) . Cells were resuspended in sporulation medium ( 0 . 3% potassium acetate [pH 7] , 0 . 02% raffinose ) to OD600 = 1 . 85 and sporulated at 30°C unless otherwise indicated . GAL-NDT80 GAL4-ER strains were released from the NDT80 block by the addition of 1 μM β-estradiol ( 5 mM stock in ethanol; Sigma E2758-1G , St . Louis , MO ) at 4 hr 30 min unless otherwise indicated . Note: strains released from NDT80 block at 4 hr 30 min are prototrophic and have accelerated meiotic kinetics relative to strains containing auxotrophies . Strains with CUP1 promoter driven alleles were induced by addition of CuSO4 ( 50 μM final concentration; 100 mM stock made from anhydrous powder [FW = 159 . 6 g/mol]; Mallinckrodt , Hazelwood , MO ) at indicated times . Wild-type , ndc80-1 or dam1-1 cells carrying GAL-NDT80 GAL4-ER were induced to sporulate at room temperature ( permissive temperature ) . After 2 hr 45 min , cyclin expression was induced by addition of 50 μM CuSO4 and cells were concurrently shifted to the semi-permissive ( 34°C ) or non-permissive ( >35 . 5°C ) temperature and allowed to arrest in pachytene . Cells were then transferred to the permissive temperature and released from the NDT80 block by addition of 1 μM β-estradiol into either a metaphase I arrest ( by depleting Cdc20 ) or allowed to proceed through the meiotic divisions . Wild-type or CUP-CLB3 cells carrying the GAL-NDT80 GAL4-ER constructs were induced to sporulate at 30°C . 2 hr 15 min after transfer into sporulation medium , cells were filtered and transferred to medium containing CuSO4 ( 50 μM ) and either 0 . 4% DMSO or benomyl ( 120 μg/ml ) . After an additional 2 hr 15 min incubation , benomyl was washed out by filtering and washing cells with 10 volumes of sterile dH20 containing 0 . 4% DMSO . Cells were subsequently resuspended in sporulation medium containing 1 μM β-estradiol to release from NDT80 block . The efficacy of benomyl treatment was confirmed by spindle morphology . See Hochwagen et al . ( 2005 ) for further technical details regarding benomyl resuspension in sporulation medium . MATa haploid cells carrying the MET-CDC20 or MET-CDC20 GAL-CDC5 GAL-MAM1 fusions and CENIV-GFP dots cultured in complete synthetic medium without methionine ( CSM-MET ) containing 2% raffinose were arrested in G1 with 5 μg/ml α-factor . For Figure 5A condition ( 1 ) , cells were treated with galactose ( to induce Cdc5 and Mam1 production ) for 1 hr prior to α-factor release . When arrest was complete , cells were released into rich medium ( YEP ) with 2% raffinose lacking pheromone and containing 2% galactose , 1% DMSO and 8 mM methionine ( to repress Cdc20 production ) . 8 mM methionine was added every hour to maintain metaphase arrest . When metaphase arrest was complete , cells were released into CSM-MET medium , containing 2% dextrose , 1% DMSO and 5 μg/ml α-factor . For condition ( 2 ) , G1 arrested cells were released into YEP medium with 2% raffinose , lacking pheromone , containing 8 mM methionine and 1% DMSO . 8 mM methionine was added every hour to maintain the metaphase arrest . After 2 hr , cells were treated with 2% galactose for 1 hr and were subsequently released into CSM-MET medium , containing 2% dextrose , 1% DMSO and 5 μg/ml α-factor . For condition ( 3 ) , G1 arrested cells were released into YEP medium with 2% raffinose , lacking pheromone , containing 8 mM methionine and 15 μg/ml nocodazole in DMSO . 8 mM methionine was added every hour to maintain the metaphase arrest . After 2 hr , cells were treated with 2% galactose for 1 hr and were subsequently released into CSM-MET medium , containing 2% dextrose , 1% DMSO and 5 μg/ml α-factor . Samples were taken every 15 min after release from metaphase arrest to determine GFP dot segregation in anaphase . Indirect immunofluorescence was performed as described in Kilmartin and Adams ( 1984 ) . Spindle morphologies were classified as follows: metaphase I or metaphase I-like spindles were defined as a short , bipolar spindle spanning a single DAPI mass . Anaphase I spindles were defined as an elongated spindle spanning two distinct DAPI masses . Metaphase II spindles were defined as two short , bipolar spindles , each spanning a DAPI mass . Anaphase II spindles were defined as two elongated spindles , each spanning two distinct DAPI masses ( four DAPI masses total ) . For Figure 8D , E , robust bipolar spindle was classified as a short , thick , bipolar spindle with equal intensity tubulin staining across the entire length of the spindle . A fragile spindle was classified as a short bipolar spindle with lower intensity tubulin staining in the middle of the spindle axis . Cells were induced to sporulate and CuSO4 was added at the indicated times . After 30–60 min post CuSO4 induction , cells were layered on a Concanavalin A ( 2 mg/ml; stock solution 20 mg/ml diluted in 50 mM CaCl2 , 50 mM MnSO4 ) coated cover slip and assembled into an FCS2 fluidic chamber ( Bioptechs Inc . Butler , PA ) . Sporulation medium was heated to 30°C , aerated using an aquarium air pump ( Petco Animal Supplies , Inc . Cambridge , MA ) and was perfused into the fluidic chamber using a peristaltic pump ( Gilson Inc . , Middleton , WI ) with a flow rate of 4–7 ml/h . Alternatively , cells were induced to sporulate as above and transferred to a microfluidic chamber ( CellASIC Corp . Hayward , CA ) . Cells were imaged using a Zeiss Axio Observer-Z1 with a 100× objective ( NA = 1 . 45 ) , equipped with a Hamamatsu ORCA-ER digital camera . 11 Z-stacks ( 1 micron apart ) were acquired and maximally projected . Metamorph software was used for image acquisition and processing . Images for Figure 2B was processed using Metamorph deconvolution software . For Figure 2C , a cell was scored as harboring a separated pair of sister kinetochores if the heterozygous CENV-GFP dot signal underwent transient splitting for at least two time points for the duration of the movie . An aliquot of cells was fixed with 3 . 7% formaldehyde in 100 mM phosphate buffer ( pH 6 . 4 ) for 10–15 min . Cells were washed once with 100 mM phosphate , 1 . 2 M sorbitol buffer ( pH 7 . 5 ) and permeabilized with 1% Triton X-100 stained with 0 . 05 μg/ml 4′ , 6-diamidino-2-phenylindole ( DAPI ) . Cells were imaged using a Zeiss Axioplan 2 microscope or a Zeiss Axio Observer-Z1 with a 100× objective ( NA = 1 . 45 ) , equipped with a Hamamatsu ORCA-ER digital camera . Openlab or Metamorph software was used for image acquisition and processing . 4 OD600 units of cells were harvested and spheroplasted with 0 . 1 mg/ml zymolyase 100T ( Seikagaku Corp , Japan ) and 15 mM DTT in solution 1 ( 2% potassium acetate , 0 . 8% sorbitol ) for 10–13 min at 37°C . Ice-cold solution 2 ( 100 mM MES [pH 6 . 4] , 1 mM EDTA , 0 . 5 mM MgCl2 , 1 M sorbitol ) was added to stop spheroplasting and cells were centrifuged at 2500 rpm for 2–3 min . The supernatant was discarded and the pellet was gently resuspended in 100–200 μl of solution 2 . 15 μl of the resuspension was spread onto a glass slide . Subsequently , 30 μl of fixative solution ( 4% paraformaldehyde , 3 . 4% sucrose ) , 60 μl of 1% lipsol and 60 μl of fixative solution were added on top of cell suspension and spread using a glass rod seven to ten times back and forth . The slides were dried for at least 2 hr at room temperature , rehydrated in PBS pH 7 . 4 , blocked with 0 . 2% gelatin , 0 . 5% BSA in PBS , and stained as described in the ‘Antibody’ section . For quantifications of spread nuclei , images were acquired using a Zeiss Axioplan 2 microscope or a Zeiss Axio Observer-Z1 with a 100× objective ( NA = 1 . 45 ) , equipped with a Hamamatsu ORCA-ER digital camera . Openlab or Metamorph software was used for image acquisition and processing . 40–100 spread nuclei were counted for each sample , except for strain A31955 in Figure 8B ( n = 28 ) . Two proteins were identified as colocalized in spread nuclei when more than 90% of foci overlapped . They were defined as partially colocalized when the overlap between foci was approximately 50% and as mislocalized when the overlap was negligible . In vitro kinase assays were performed as described in Carlile and Amon ( 2008 ) with the following modifications: 1 mg of total protein was incubated with 40 μl of 50% slurry anti-V5 agarose affinity gel ( Sigma , St . Louis , MO ) for 2 hr at 4°C . One half of the immunoprecipitate was used for the in vitro kinase assay , while the other half was used for Western blotting to detect Cdc28-3V5 . For immunoblot analysis , ∼10 OD600 units of cells were harvested and treated with 5% trichloroacetic acid for at least 10 min at 4°C . The acid was washed away with acetone and the cell pellet was subsequently dried . The cell pellet was pulverized with glass beads in 100 μL of lysis buffer ( 50 mM Tris–HCl at pH 7 . 5 , 1 mM EDTA , 2 . 75 mM DTT , complete protease inhibitor cocktail [Roche , Basel , Switzerland] ) using a bead-beater ( Biospec Products , Inc . Bartlesville , OK ) . 3× SDS sample buffer was added and the cell homogenates were boiled . Standard procedures for sodium dodecyl sulfate–polyacrylamide gel electrophoresis ( SDS-PAGE ) and Western blotting were followed ( Laemmli , 1970; Towbin et al . , 1979; Burnette , 1981 ) . A nitrocellulose membrane ( VWR International LLC , Radnor , PA ) was used to transfer proteins from polyacrylamide gels . Antibody dilutions are described in the ‘Antibody’ section . 1 ml aliquot of a meiotic culture was spun down and the pellet was re-suspended in 70% ethanol and fixed for at least 60 min . Ethanol was removed and the cell pellet was washed with 50 mM sodium citrate , pH 7 and sonicated for 6 s at 50% output . The sample was subsequently incubated with 0 . 25 mg/ml Ribonuclease A ( Sigma , St . Louis , MO ) in 50 mM sodium citrate overnight at 37°C , washed once with 50 mM sodium citrate and re-suspended in 50 mM sodium citrate with either 1 µM Sytox Green ( Molecular Probes , Carlsbad , CA ) or 16 µg/ml propidium iodide ( Sigma ) . Samples were analyzed using FACSCalibur ( Becton Dickenson Co . Franklin Lakes , NJ ) . 400 OD600 units of cells were fixed for 15 min at room temperature in 1% formaldehyde . The formaldehyde was quenched by addition of 125 mM glycine . Samples were processed as previously described ( Vader et al . , 2011 ) . Before immunoprecipitation , 120th of the sample was removed as the input sample . The antibodies used for immunoprecipitation are described in the ‘Antibody’ section . For ChIP-chip , samples were processed and analyzed as described in Vader et al . ( 2011 ) . For qPCR analysis , DNA was amplified using SYBR Premix ExTaq Perfect Real Time Kit ( Takara Bio Inc . Otsu , Shiga , Japan ) . PCR reactions were 40 cycles of 95°C , 20 s; 55°C , 30 s; 72°C , 30 s using a Roche LightCycler 480 II ( Roche , Basel , Switzerland ) . The following primers were used ( 5′–3′ ) :CENV F: CTT GTT TAG TGC AAG CCA CTG TTCENV R: CCG CAT TTC CTT GAT TTA CTG TCc281 F: CAA CGA ACC GTG GGA ACG TTA TAGc281 R: GAA ACT TTC CTG GTA CCT TCT GCc194 F: GCT GAA AGC ATG CCA CTG TAc194 R: GGT GTT CCT GCT TCG TTG TTA GHMR F: ACG ATC CCC GTC CAA GTT ATGHMR R: CTT CAA AGG AGT CTT AAT TTC CCT G ∼20 OD600 units of cells were harvested and treated with sodium azide ( 0 . 1% final concentration ) . Cells were pelleted and snap frozen in liquid nitrogen . Genomic DNA was extracted as follows: Cells were washed once in TE and spheroplasted with 1/100 volume of beta-mercaptoethanol and 250 μg/ml zymolyase T100 in spheroplasting buffer ( 1 M sorbitol , 42 mM K2HPO4 , 8 mM KH2PO4 , 5 mM EDTA ) for 30 min at 37°C on a rotating rack . 100 μl preheated ( 65°C ) lysis buffer ( 1:1 mix of 1 M Tris pH 8 and 0 . 5 M EDTA , 2 . 5–3% SDS ) was added and mixed by inverting . 15 μl proteinase K ( 18 ± 4 mg/ml PCR grade solution; Roche , Basel , Switzerland ) was added and incubated at 65°C for ∼1 . 5 hr . Subsequently , 150 μl 5 M potassium acetate was added , mixed by inverting and transferred to 4°C for 10 min . Samples were centrifuged at 4°C for 20 min and 650 μl of supernatant was transferred into a 2 ml tube containing 750 μl 100% ethanol , avoiding as much of the white fluff as possible . Samples were mixed by inverting and left at 4°C for 10 min . Nucleic acid was precipitated at 15 , 000 rpm for 10 min , 4°C . Samples were subsequently resuspended in TE and treated with RNase A ( 50 μg/ml; Roche ) , for 15–20 min at 37°C and kept at 4°C overnight . DNA was extracted with phenol/chloroform/isopropanol and was resuspended in 125 μl TE . XhoI-MluI digested DNA fragments were separated on 0 . 6% agarose gel in 1× TBE and transferred onto Hybond-XL plus membranes ( GE Healthcare Biosciences , Pittsburgh , PA ) by alkaline transfer . Southern blotting was performed as previously described ( Hunter and Kleckner , 2001 ) . Cluster analysis of the ribosome footprinting data for the kinetochore components listed in Figure 6—figure supplement 1 was performed using Cluster 3 . 0 . Genes were clustered by hierarchical average based on Spearman correlation using mean centered arrays . Clustering data ( Figure 6B , Figure 6—figure supplement 2 ) were visualized using Java Treeview . Note that ribosome footprints are normalized such that the sum of expression across the time course is equivalent for each gene . For plots in Figure 6C , D and Figure 6—figure supplements 3–10 , mRNA-seq and ribosome footprinting data were plotted for indicated genes based on the dataset from Brar et al . ( 2012 ) . The meiotic stages plotted on the x-axis are in the following order: vegetative ( gb15 exponential and A14201 exponential ) , meiotic entry ( 1 , A , B and D ) , DNA replication ( E and F ) , recombination ( G and I ) , prophase I ( 3 and 4 ) , metaphase I ( 5 and 6 ) , anaphase I ( 7 and 8 ) , metaphase II ( 9 and 10 ) , anaphase II ( 11 , 12 and 13 ) and spore formation ( 15 and 18 ) . The detailed explanation of the above letter and number codes can be found in Brar et al . ( 2012 ) . Chi-square ( χ2 ) tests were performed using GraphPad Prism 6 . 0 software with two-tailed P values and 95% confidence intervals . Corresponding degrees of freedom ( df ) , χ2 and P values are shown in the figure legends . | Diploid organisms contain two sets of chromosomes , one set inherited from the mother and the other from the father . Humans , for example , have 23 pairs of chromosomes , and the chromosomes within each pair are said to be homologous because they are similar to each other in a number of ways , including length and shape . When it comes time for one of these cells to duplicate , each chromosome is first replicated to generate a pair of identical chromosomes called sister chromatids , which subsequently separate in a cell division process known as mitosis to produce two identical daughter cells . While most cells proliferate via mitotic cell division , the germ cells that generate gametes in the form of sperm or eggs undergo a different cell division known as meiosis . This process reduces the number of chromosomes by a factor of two , so that the original number of chromosomes is restored by the fusion of gametes during sexual reproduction . During meiotic cell division , a single round of DNA replication is followed by two consecutive rounds of nuclear division called meiosis I and meiosis II . During meiosis I , homologous chromosomes are separated . Subsequently , during meiosis II , the sister chromatids separate to produce a total of four products , each with half the number of chromosomes as the original cell . The separation of homologous chromosomes or sister chromatids relies on them being pulled apart by microtubules . One end of each microtubule is attached to a protein-based structure called a kinetochore , which is assembled onto the centromere of each chromosome . The other end of each microtubule is attached to a structure that is called a centrosome in human cells and a spindle pole body in yeast cells . Human cells have two centrosomes , which reside on the opposite poles of the cell , and likewise for the spindle pole bodies in yeast cells . In mitotic cells and in meiosis II , microtubules attach to kinetochores in a way that means the sister chromatids are pulled apart . During meiosis I , on the other hand , they attach to kinetochores in a manner so the homologous chromosomes are pulled apart . Miller et al . now show how the timing of the interaction between the kinetochore and microtubules is critical to ensure that the homologous chromosomes are separated during meiosis I . They found that premature interactions resulted in the separation of sister chromatids ( as happens in mitosis ) rather than the separation of homologous chromosomes , as is supposed to happen in meiosis I . They also showed that cells prevent such premature interactions by dismantling the outer regions of the kinetochore and reducing the levels of enzymes called CDKs in the cell . These results demonstrate that preventing premature microtubule–kinetochore interactions is essential for establishing a meiosis I-specific chromosome architecture , and they also provide fresh insights into how the molecular machinery that is responsible for mitotic chromosome segregation can be modulated to achieve meiosis . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"cell",
"biology"
] | 2012 | Meiosis I chromosome segregation is established through regulation of microtubule–kinetochore interactions |
Calcium-activated chloride channel regulator 1 ( CLCA1 ) activates calcium-dependent chloride currents; neither the target , nor mechanism , is known . We demonstrate that secreted CLCA1 activates calcium-dependent chloride currents in HEK293T cells in a paracrine fashion , and endogenous TMEM16A/Anoctamin1 conducts the currents . Exposure to exogenous CLCA1 increases cell surface levels of TMEM16A and cellular binding experiments indicate CLCA1 engages TMEM16A on the surface of these cells . Altogether , our data suggest that CLCA1 stabilizes TMEM16A on the cell surface , thus increasing surface expression , which results in increased calcium-dependent chloride currents . Our results identify the first Cl− channel target of the CLCA family of proteins and establish CLCA1 as the first secreted direct modifier of TMEM16A activity , delineating a unique mechanism to increase currents . These results suggest cooperative roles for CLCA and TMEM16 proteins in influencing the physiology of multiple tissues , and the pathology of multiple diseases , including asthma , COPD , cystic fibrosis , and certain cancers .
The calcium-activated chloride channel regulator ( CLCA- previously known as chloride channel calcium activated ) proteins ( Cunningham et al . , 1995 ) are a family of secreted self-cleaving metalloproteases that activate calcium-dependent chloride currents ( ICaCC ) in mammalian cells ( Yurtsever et al . , 2012 ) . CLCA family members are highly expressed in mucosal epithelia where they play important roles in mucus homeostasis and related diseases ( Patel et al . , 2009 ) . For example , human CLCA1 plays a central role in interleukin ( IL- ) 13-induced mucus cell metaplasia , the main source of inflammatory mucus overproduction in chronic obstructive airway diseases , such as asthma and COPD ( Alevy et al . , 2012 ) . Both clinical and animal model studies suggest a compensatory role for CLCAs in the context of cystic fibrosis ( CF ) : the fatal intestinal disease , meconium ileus , arising in CFTR-deficient mice is corrected by overexpression of mCLCA3 ( an orthologue of human CLCA1 ) ( Young et al . , 2007 ) and , correspondingly , mutations in CLCA1 are found in a subset of CF patients with aggravated intestinal disease ( van der Doef et al . , 2010 ) . At the cellular level , overexpression of CLCA proteins leads to activation of calcium-dependent chloride currents ( Gandhi et al . , 1998; Britton et al . , 2002; Elble et al . , 2002; Greenwood et al . , 2002 ) , and this functional observation had caused CLCAs to be initially misidentified as calcium-activated chloride channels ( CaCCs ) themselves ( Cunningham et al . , 1995 ) . However , further bioinformatic and biochemical studies have demonstrated that CLCA proteins are secreted , soluble proteins and that they act to modulate CaCCs that are endogenous to mammalian cells ( Gibson et al . , 2005; Hamann et al . , 2009; Yurtsever et al . , 2012 ) . The molecular identity of these channels , the mechanism of CLCA activation , and their potential roles in CLCA-mediated diseases , remain unknown . TMEM16A ( also known as Anoctamin1/DOG1 ) was recently identified as the first genuine CaCC in mammals by three independent groups ( Caputo et al . , 2008; Schroeder et al . , 2008; Yang et al . , 2008 ) . 10 members of the TMEM16/Anoctamin family have been identified ( TMEM16A-K , or Ano1-10 ) ; these proteins , predicted to be transmembrane proteins with eight membrane-spanning helices , have been found to function predominantly as CaCCs ( TMEM16A and B ) or as phospholipid scramblases ( TMEM16C , D , F , G , and J ) ( Pedemonte and Galietta , 2014 ) . TMEM16A , the best-characterized member of the family to date , is expressed in airway epithelia and smooth muscle , and its activity recapitulates some of the airway disease traits associated with CLCA1 . Not only is TMEM16A expression significantly increased by IL-13 and IL-4 in primary cell models of chronic inflammatory airway disease ( Caputo et al . , 2008; Alevy et al . , 2012 ) , but TMEM16A overexpression is also linked to mucus cell metaplasia and airway hyperreactivity ( Huang et al . , 2012; Scudieri et al . , 2012 ) . In addition , TMEM16A-specific inhibitors decrease mucus secretion and airway hyperreactivity in cellular models ( Huang et al . , 2012 ) . Although experiments with purified TMEM16A protein reconstituted in liposomes indicate that it can form a functional channel on its own ( Terashima et al . , 2013 ) , several cytosolic modulators and interaction partners , such as calmodulin , phosphatidylinositol 4 , 5-bisphosphate ( PIP2 ) , ezrin , radixin , and moesin , have been described ( Tian et al . , 2011; Perez-Cornejo et al . , 2012; Pritchard et al . , 2014 ) . However , no secreted regulators of TMEM16A activity have been identified as of yet . Here we report that secreted CLCA1 modulates TMEM16A-dependent calcium-activated chloride currents , and that this activation can occur in a paracrine fashion . Furthermore , we show that CLCA1 and TMEM16A co-localize and physically interact on the surface of mammalian cells , and that CLCA1 increases the level of TMEM16A protein at the cell surface , representing a novel mechanism of channel regulation by a secreted protein . We thus demonstrate a first downstream target of CLCA proteins and provide the first example of a secreted protein modulator of TMEM16A activity . These findings have significant implications for the roles of CLCA1 and TMEM16A proteins as cooperative partners , not only in the physiology and pathophysiology of the airways , but also in those of other tissues and organs .
We previously demonstrated that ICaCC are activated in HEK293T ( 293T ) cells overexpressing human CLCA1 ( Yurtsever et al . , 2012 ) . Given that CLCA1 proteins are cleaved and secreted from these cells , we hypothesized that exogenous CLCA1 may activate ICaCC . In a first set of experiments to test this idea , GFP-expressing cells that had been co-cultured overnight with cells transfected with CLCA1-pHLsec plasmid ( CLCA1 ) or with empty pHLsec vector ( pHLsec ) were tested for ICaCC by means of whole-cell patch clamp electrophysiology ( Figure 1A ) . In the presence of 10 μM intracellular Ca2+ and physiological concentrations of extracellular Cl− , robust , slightly outward rectifying currents were activated in cells co-cultured with CLCA1-transfected cells , but only substantially smaller currents were detected in cells co-cultured with vector-transfected cells ( Figure 1B–D ) . In a complementary experiment , whole-cell ICaCC were measured in untransfected cells that had been cultured in medium obtained from CLCA1- or from pHLsec-transfected cells ( Figure 2A ) . We observed activation of large currents in cells exposed to CLCA1-conditioned medium that had the same Ca2+- and voltage-dependence properties as those induced in cells co-cultured with CLCA1-expressing cells ( Figure 2B–D ) . As shown in Figure 2B–C , outward rectification of the CLCA1-activated current decreases at higher Ca2+ concentrations . In addition , current reversal potential shifts positive upon lowering extracellular Cl− . These features are in agreement with the properties of the Ca2+-dependent Cl− conductance observed in CLCA1-expressing 293T cells ( Hamann et al . , 2009; Yurtsever et al . , 2012 ) , and consistent with those of CaCCs in native cells and heterologous expression systems ( Jeong et al . , 2005; Yamazaki et al . , 2005; Xiao et al . , 2011 ) . These data indicate that secreted CLCA1 can activate ICaCC in a paracrine fashion . 10 . 7554/eLife . 05875 . 003Figure 1 . Paracrine activation of calcium-dependent chloride currents in HEK293T cells by CLCA1 . ( A ) GFP-expressing cells were co-cultured with pHLsec- or CLCA1-transfected cells , and assayed for ICaCC by patch clamp electrophysiology . ( B–C ) Whole-cell currents measured in GFP-positive cells from experiments as in ( A ) , superfused with standard extracellular solution , and in the absence or presence of 10 μM free Ca2+ in the pipette ( respectively , [0 μM Ca2+]in or [10 μM Ca2+]in ) . ( B ) Representative current traces . The pulse protocol is shown at the top left . Outward currents are represented by upward deflections , and dotted lines indicate zero current . Membrane capacitance was similar in all cases at ∼25 pF . ( C ) Current–voltage relationships at the end of the 600-ms voltage steps . Membrane potential values were corrected off-line for the calculated liquid junction potentials , respectively −5 . 5 mV ( [0 μM Ca2+]in ) and −6 . 0 mV ( [10 μM Ca2+]in ) . Data are presented as means ± S . E . ( n = 5–9 ) . ( D ) Current density at +100 mV , from the same experiments as in ( C ) . Symbols represent data from individual patches; bars indicate the means ± S . E . of all experiments . *p < 0 . 01 ( one-way ANOVA , F = 30 . 3 and p = 1 . 2 × 10−7; followed by Tukey test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05875 . 00310 . 7554/eLife . 05875 . 004Figure 2 . Activation of calcium-dependent chloride currents by secreted CLCA1 . ( A ) Untransfected cells were cultured in medium from pHLsec- or CLCA1-expressing cells , and assayed by patch clamp electrophysiology . ( B–D ) Whole-cell currents measured in cells from experiments as in ( A ) , superfused with standard ( [154 mM Cl−]out ) or reduced Cl− ( [14 mM Cl−]out ) extracellular solution; and in the absence or presence of 1 μM or 10 μM free Ca2+ in the pipette ( respectively , [0 μM Ca2+]in , [1 μM Ca2+]in or [10 μM Ca2+]in ) . ( B ) Representative current traces obtained with the same pulse protocol and displayed as in Figure 1B . Membrane capacitance was similar in all cases at ∼25 pF . ( C ) Current-voltage relationships at the end of the 600-ms voltage steps . Membrane potential values were corrected off-line for the calculated liquid junction potentials , respectively −5 . 5 mV ( [0 μM Ca2+]in ) and −6 . 0 mV ( [1 μM Ca2+]in and [10 μM Ca2+]in ) for the experiments in [154 mM Cl−]out; and −20 mV for the experiments in [14 mM Cl−]out . Data are presented as means ± S . E . ( n = 5–20 ) . Inset , CLCA1-mediated currents right-shifted ∼ +15 mV upon reduction of extracellular Cl−; symbols have been removed for clarity . ( D ) Current density at +100 mV , from the same experiments as in ( C ) . Symbols represent data from individual patches; bars indicate the means ± S . E . of all experiments . *p < 0 . 01 ( one-way ANOVA , F = 10 . 4 and p = 2 . 1 × 10−8; followed by Tukey test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05875 . 004 We next focused on identifying the CaCC responsible for carrying the CLCA1-mediated currents . The CLCA1-modulated ICaCC in 293T cells are Ca2+-dependent , moderately outward rectifying in the presence of μM concentrations of intracellular Ca2+ , Cl−-selective , and blocked by gluconate ( Hamann et al . , 2009; Yurtsever et al . , 2012 ) ( Figures 1 , 2 ) , closely resembling the biophysical characteristics of those observed for TMEM16A currents in heterologous expression systems ( Schroeder et al . , 2008; Yang et al . , 2008; Xiao et al . , 2011 ) , proteoliposomes ( Terashima et al . , 2013 ) , and native tissues ( Caputo et al . , 2008 ) . Given the biophysical and pathophysiological parallels between TMEM16A currents , and those activated by CLCA1 , we hypothesized that CLCA1-activated currents may be carried by TMEM16A . Consistent with this idea , 293T cells were transfected with either TMEM16A siRNA or with non-specific , scrambled RNA ( siControl ) , and cultured in CLCA1-conditioned medium . Exposure to secreted CLCA1 led to the activation of ICaCC in siControl-transfected cells ( Figure 3A , B ) that were comparable to the activation recorded in untransfected cells ( Figure 2B , D ) , but these CLCA1-dependent currents were knocked down to essentially background levels in TMEM16A siRNA-transfected cells ( Figure 3A , B ) . The TMEM16A siRNA significantly decreased expression of TMEM16A protein , assessed by Western blot ( Figure 3C ) . These results demonstrate that CLCA1-dependent ICaCC in 293T cells are indeed carried by TMEM16A . 10 . 7554/eLife . 05875 . 005Figure 3 . Genetic knockdown of TMEM16A inhibits CLCA1-mediated calcium-dependent chloride currents . ( A–B ) HEK293T cells transfected with RNAi negative control ( siControl ) or TMEM16A siRNA were incubated in CLCA1-conditioned medium and assayed by patch-clamp electrophysiology , in standard extracellular solution ( [154 mM Cl−]out ) and 10 μM free Ca2+ in the pipette ( [10 μM Ca2+]in ) . ( A ) Representative current traces obtained with the same pulse protocol and displayed as in Figure 1B . Membrane capacitance was similar in all cases at ∼25 pF . ( B ) Current density at +100 mV . Symbols represent data from individual patches ( n = 14 ) ; bars indicate the means ± S . E . of all experiments . *p < 0 . 01 ( unpaired Student's t test ) . ( C ) Effect of CLCA1 and/or TMEM16A siRNA treatment on TMEM16A protein expression . Upper panel: top , TMEM16A; and bottom , actin ( loading control ) Western blot from solubilized HEK293T cells . Lanes are labeled as follows: pHLsec-t , pHLsec transfected cells; CLCA1-t , CLCA1-transfected cells; CLCA1-c , cells treated with CLCA1-conditioned medium; TMEM16A siRNA , cells transfected with TMEM16A siRNA; TMEM16A siRNA + CLCA1-c , cells transfected with TMEM16A siRNA and treated with CLCA1-conditioned medium . Bar graph: quantitation of TMEM16A band intensity normalized to actin band intensity using ImageJ ( NIH ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05875 . 005 Next , we used immunohistochemistry and confocal microscopy to examine CLCA1 and TMEM16A localization in non-permeabilized HEK293T cells . Cells transfected with pHLsec vector alone did not display noticeable staining for either CLCA1 or TMEM16A ( Figure 4A–D ) , consistent with lack of endogenous expression of CLCA1 and low endogenous levels of TMEM16A in these cells ( Kunzelmann et al . , 2009; Pritchard et al . , 2014 ) . However , cells transfected with CLCA1 stained strongly both for CLCA1 and , surprisingly , for TMEM16A ( Figure 4E–H ) , suggesting that CLCA1 increases TMEM16A protein levels . Furthermore , signal for both proteins clearly overlapped with the membrane stain ( WGA ) , consistent with a model in which CLCA1 and TMEM16A associate with and stabilize one another on the cell surface . Since secreted CLCA1 can activate TMEM16A-mediated ICaCC in a paracrine manner ( Figures 1–3 ) , we carried out similar imaging experiments to determine whether exogenously applied secreted CLCA1 also increased TMEM16A surface expression . Cells cultured in media from pHLsec-transfected cells again displayed no detectable staining for either CLCA1 or TMEM16A ( Figure 4I–L ) , but cells exposed to secreted CLCA1 displayed robust staining for TMEM16A . Signal for CLCA1 was also detected in a few cells , overlapping with TMEM16A and WGA staining ( Figure 4M–P ) . Surprisingly , although TMEM16A surface levels increased after exposure to CLCA1 , total TMEM16A in cells did not change ( Figure 3C ) . These results indicate that exogenous secreted CLCA1 colocalizes with and enhances the fraction of TMEM16A located at the cell surface . 10 . 7554/eLife . 05875 . 006Figure 4 . CLCA1 colocalizes with TMEM16A and increases TMEM16A surface expression . ( A–D ) Membrane ( WGA ) or immunostaining of HEK293T cells transfected with pHLsec vector; ( E–H ) , or with CLCA1 . Surface TMEM16A is greatly increased by expression of CLCA1 . ( I–L ) Membrane ( WGA ) or immunostaining of HEK293T cells cultured in conditioned media from cells transfected with pHLsec vector; ( M–P ) or cells cultured in conditioned media from cells transfected with CLCA1 . TMEM16A surface expression is greatly enhanced after cells are exposed to secreted CLCA1 . DOI: http://dx . doi . org/10 . 7554/eLife . 05875 . 006 To investigate whether CLCA1 and TMEM16A associate directly with one another on the cell surface , we adapted an assay commonly used to identify immunological receptor-ligand pairs ( Altman et al . , 1996 ) . We previously demonstrated that CLCA1 is cut into two fragments by self-cleavage and that the N-terminal fragment is necessary and sufficient to activate CaCCs in HEK293T cells ( Yurtsever et al . , 2012 ) . Thus , for these assays we developed a CLCA1 cell-staining reagent composed of the N-terminal fragment of CLCA1 ( N-CLCA1 ) containing a specific biotinylation motif on the C-terminus ( Figure 5A ) . Biotinylated N-CLCA1 was coupled to SA-PE ( streptavidin conjugated to phycoerythrin ) to produce a tetrameric fluorescent reagent with enhanced avidity toward its ligand . Cell-binding assays were carried out in the presence or absence of an anti-TMEM16A antibody raised against epitopes in the last extracellular loop and then analyzed by flow cytometry . The tetramerized N-CLCA1 displayed robust binding to intact HEK293T cells compared to background , and this binding was significantly reduced by pre-incubating the cells with the anti-TMEM16A antibody ( Figure 5B ) . Two control antibodies , one raised against an intracellular epitope of TMEM16A and the other an isotype control , did not affect N-CLCA1 binding ( Figure 5C ) . In order to validate that the biotinylation of N-CLCA1 did not adversely affect function , we carried out whole-cell patch clamp experiments where either purified N-CLCA1 or purified biotinylated N-CLCA1 was exogenously applied to HEK293T cells . We found that both of these proteins were able to robustly activate the observed currents ( Figure 5D , E ) . These results indicate that N-CLCA1 engages TMEM16A on the surface of HEK293T cells , and suggests that the enhanced TMEM16A level is a consequence of stabilization by CLCA1 . 10 . 7554/eLife . 05875 . 007Figure 5 . N-CLCA1 engages TMEM16A on the cell surface . ( A ) Schematic of CLCA1 N-terminal fragment ( N-CLCA1 ) construct with specific biotinylation site and resultant tetrameric cell-staining reagent created after complexation with SA-PE . ( B ) Flow cytometry of intact HEK293T cells stained with SA-PE alone ( black line ) , N-CLCA1/SA-PE ( green line ) , or N-CLCA1/SA-PE in the presence of anti-TMEM16A antibody S-20 ( red line ) . ( C ) Flow cytometry of intact HEK293T cells either stained with SA-PE alone ( black line ) , N-CLCA1/SA-PE ( green line ) , N-CLCA1/SA-PE in the presence of anti-TMEM16A antibody C-5 ( raised against an intracellular TMEM16A epitope; orange line ) , or N-CLCA1/SA-PE in the presence of anti-Aquaporin5 antibody G-19 ( blue line ) . ( D–E ) Cells were incubated in the absence ( [−] ) or presence ( [+] ) of purified N-terminal ( N-term ) CLCA1 protein before ( N-CLCA1 ) or after biotinylation ( N-CLCA1biotin ) , and assayed by patch-clamp electrophysiology , in standard extracellular solution ( [154 mM Cl−]out ) and 10 μM free Ca2+ in the pipette ( [10 μM Ca2+]in ) . ( D ) Representative current traces obtained with the same pulse protocol and displayed as in Figure 1B . Membrane capacitance was similar in all cases at ∼25 pF . ( E ) Current density at +100 mV . Symbols represent data from individual patches ( n = 8–11 ) ; bars indicate the means ± S . E . of all experiments . *p < 0 . 05 ( unpaired Student's t test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05875 . 007
It has been demonstrated that purified TMEM16A protein reconstituted in proteoliposomes recapitulates the permeation , pharmacological , voltage- and Ca2+-dependence properties of TMEM16A channels characterized in heterologous expression systems and native cell models ( Terashima et al . , 2013 ) , which implies that TMEM16A does not require association with other proteins for CaCC activity . However , a recent proteomics approach has identified a large number of endogenous proteins implicated in protein trafficking , surface expression , folding and stability that interact with TMEM16A , including SNAREs such as syntaxins and syntaxin-binding proteins , and the ezrin-radixin-moesin ( ERM ) scaffolding complex ( Perez-Cornejo et al . , 2012 ) . Our data identify CLCA1 as the first secreted direct regulator of TMEM16A , and our findings suggest that CLCA1 may modulate TMEM16A channel activity by stabilizing it at the cell surface , much like the SNARE and ERM protein networks of the TMEM16A interactome . So how does CLCA1 modulate TMEM16A currents ? We observe that CLCA1 increases TMEM16A surface expression without increasing expression of the protein ( Figure 3C ) . A model consistent with our data and the current literature is that CLCA1 engages and stabilizes dimeric TMEM16A on the surface of the cell . Previous studies have shown that TMEM16A can exist as a dimer ( Fallah et al . , 2011; Sheridan et al . , 2011 ) , dimerization being mediated by an intracellular region in the N-terminus of TMEM16A ( Tien et al . , 2013 ) . Mutations to this region abolish dimerization , prevent protein trafficking to the plasma membrane , and , consequently , ablate channel activity ( Tien et al . , 2013 ) . These observations indicate that dimerization regulates TMEM16A trafficking and channel activity . It is possible that TMEM16A dynamically shuttles between the cell surface and intracellular compartments . One possibility is that CLCA1 engages monomeric TMEM16A and drives dimerization; alternatively , CLCA1 may engage and stabilize dimeric TMEM16A on the cell surface , thereby preventing its removal and , consequently , increasing calcium-dependent chloride currents ( Figure 6 ) . This highlights an unprecedented mechanism for regulating ion channel currents by a secreted protein . 10 . 7554/eLife . 05875 . 008Figure 6 . Model for CLCA1 modulation of TMEM16A-mediated calcium-dependent chloride currents . Following secretion and self-cleavage of CLCA1 , the N-terminal fragment ( N-CLCA1 ) acts in paracrine fashion ( 1 ) . Dimerization appears to regulate surface trafficking of TMEM16A . N-CLCA1 engages TMEM16A on the cell surface ( 2 ) , stabilizing TMEM16A dimers , preventing internalization ( 3 ) and in turn , results in increased TMEM16A surface expression and calcium-dependent chloride current density . DOI: http://dx . doi . org/10 . 7554/eLife . 05875 . 008 The identification of CLCA1 as a modulator of TMEM16A activity raises the possibility of functional associations between other CLCA and TMEM16 family members . Four to eight CLCA ( Patel et al . , 2009 ) and ten TMEM16 family members ( Pedemonte and Galietta , 2014 ) are expressed in mammalian tissues . A number of these TMEM16 proteins have poorly defined functions and do not obviously traffic to the cell surface when expressed alone ( Duran et al . , 2012 ) . Future studies will be needed to determine whether other CLCA proteins can associate with other TMEM16 proteins and influence their function . CLCA1 ( Yang et al . , 2013 ) , CLCA2 ( Sasaki et al . , 2012; Walia et al . , 2012 ) , and CLCA4 ( Yu et al . , 2013 ) have all been implicated in various cancers as have a number of TMEM16 proteins ( West et al . , 2004; Dutertre et al . , 2010; Duvvuri et al . , 2012; Liu et al . , 2012; Qu et al . , 2014 ) , and such studies could have tremendous implications for cooperative CLCA/TMEM16 roles in cancer and other diseases . Here we report that secreted CLCA1 modulates TMEM16A-dependent ICaCC , and that this activation can occur in a paracrine fashion . Furthermore , we show that CLCA1 and TMEM16A colocalize and physically interact on the surface of mammalian cells , with CLCA1 increasing the level of TMEM16A protein at the cell surface . We thus demonstrate a first downstream target of CLCA proteins , solving the 20-year-old mystery regarding how CLCA proteins activate ICaCC , and provide the first example of a secreted protein modulator of TMEM16A activity . CLCA1 ( Alevy et al . , 2012 ) and TMEM16A ( Huang et al . , 2012; Scudieri et al . , 2012 ) have been separately observed to play critical roles in chronic inflammatory airway disease models . Our findings have significant implications for the roles of CLCA1 and TMEM16A proteins as cooperative partners , not only in the physiology and pathophysiology of the airways , but also in those of other tissues and organs .
The following commercial antibodies were used according to the manufacturer's specifications: mouse-anti-human-TMEM16A monoclonal antibody C-5 ( Santa Cruz Biotechnology , Dallas , TX ) ; goat-anti-human-TMEM16A polyclonal antibody S-20 ( Santa Cruz Biotechnology ) ; mouse-anti-actin monoclonal antibody C4 ( Millipore , Billerica , MA ) , goat-anti-human-Aquaporin5 polyclonal antibody G-19 ( Santa Cruz Biotechnology ) ; rabbit anti-6-His-antibody-HRP conjugate ( Bethyl Laboratories , Montgomery , TX ) , goat anti-mouse IgG antibody-HRP conjugate ( Santa Cruz Biotechnology ) ; wheat germ agglutinin ( WGA ) -Alexa Fluor 633 conjugate ( Life Technologies , Carlsbad , CA ) ; donkey anti-goat IgG-Alexa Fluor 488 conjugate ( Life Technologies ) ; donkey anti-rabbit IgG-Alexa Fluor 594 conjugate ( Life Technologies ) ; and rabbit anti-human CLCA1 polyclonal antibody 1228 ( Biosystems , Rockford , IL ) . Mouse anti-human CLCA1 monoclonal antibody 8D3 was produced in-house and used as previously described ( Alevy et al . , 2012; Yurtsever et al . , 2012 ) . Streptavidin conjugated to phycoerythrin ( SA-PE ) was purchased from BD Biosciences ( San Jose , CA ) . Hype-5 transfection reagent was purchased from OZ Biosciences ( San Diego , CA ) . Full length human CLCA1 ( 22–914 ) ( CLCA1 ) cloned into pHLsec vector was used throughout ( Yurtsever et al . , 2012 ) . HEK293T cells were cultured in 6-well dishes in Dulbecco's Modified Eagle Medium ( Life Technologies ) supplemented with 10% fetal bovine serum , 105 units/l penicillin and 100 mg/l streptomycin , at 37°C and 5% CO2 . Cells were transfected at 80% confluency using 293fectin transfection reagent ( Life Technologies ) at a 1:2 ratio ( µg DNA: µl 293fectin ) , using 1 μg of plasmid DNA per 1 million cells . Experiments were conducted in cells that were transiently transfected with CLCA1 , or in cells that were exposed to exogenous CLCA1 protein via means of two different experimental approaches: either co-culture with CLCA1 transfected cells; or treatment with CLCA1-conditioned medium . For co-culture experiments , cells transfected with CLCA1 , empty pHLsec vector ( pHLsec ) , or EGFP-pCDNA3 . 1 plasmid ( GFP ) were trypsinized 24 hr post-transfection , and GFP-expressing cells were mixed at a 1:1 ratio with either CLCA1 or pHLsec-transfected cells , and replated at low density on UV-sterilized , 8 mm round German glass coverslips ( Electron Microscopy Sciences , Hatfield , PA ) . Following trypsin treatment , all transfected cells were pelleted by centrifugation and washed with sterile PBS prior to replating to prevent carry-over of transfection medium . After 24 hr , the GFP-expressing cells were assayed for Ca2+-dependent Cl− currents by patch clamp electrophysiology . For conditioned medium experiments , cells were transfected with either CLCA1 or pHLsec for 6 hr , then transfection medium was removed , cells were washed with sterile PBS , and fresh medium was applied; following overnight incubation , medium from these cells was harvested and centrifuged gently ( 1500×g , 5 min ) to remove non-adherent cells . Untransfected cells were plated onto round coverslips and incubated for 24 hr in 2 ml of cleared CLCA1- or pHLsec-conditioned medium supernatants . The N-terminal fragment of CLCA1 ( 22–694; N-CLCA1 ) was cloned into pHL-Avitag3 vector ( Aricescu et al . , 2006 ) , which contains a BirA biotin ligase recognition motif and hexahistidine tag at the C-terminus . This secreted protein was expressed in 293F cells via transient transfection using Hype-5 at 1:1 . 5 µl ratio ( µg DNA: µl Hype-5 ) , using 1 µg of plasmid DNA per 1 million cells . Media supernatants were harvested after 72 hr . Protein was purified from media supernatant using Ni-NTA chromatography and eluted in 5 ml buffer A ( 50 mM K2HPO4 pH 8 , 300 mM NaCl and 250 mM imidazole ) . Purified N-CLCA1 was concentrated to a final volume of 300 µl in a centrifuge concentrator and protein concentration was calculated from absorbance at 280 nm . For the experiments reported in Figure 5D , E , protein was added onto the untransfected cells at 10 µg/ml and incubated for 24 hr prior to whole-cell patch clamp experiments . The same volume of buffer A was added onto cells as buffer control . For in vitro biotinylation , N-CLCA1 containing the specific biotinylation tag at the C-terminus was exchanged into buffer B ( 100 mM Tris pH 7 . 5 , 200 mM NaCl , and 5 mM MgCl2 ) and specifically biotinylated by addition of biotin and Escherichia coli BirA ligase ( produced and purified in-house ) at 4°C overnight . Excess biotin was removed using a 2 ml desalting column . Biotinylated N-CLCA1 was added onto the untransfected cells at 10–50 µg/ml and incubated for 24 hr prior to whole-cell patch clamp experiments . The same volume of buffer B was added onto cells as buffer control . To investigate the molecular identity of CLCA1-modulated CaCCs , a targeted approach was taken focusing on TMEM16A . For siRNA knockdown of TMEM16A , cells plated in 48-well plates were transfected with either 200 nM TMEM16A siRNA ( HSS123904; 5′-AAG UUA GUG AGG UAG GCU GGG AAC C-3′ , Life Technologies ) or 200 nM medium GC-content Stealth RNAi negative control ( 12935300; 5′-GGU UCC CAG CCU ACC UCA CUA ACU U-3′ , Life Technologies ) using Lipofectamine 2000 ( Life Technologies ) at a 20:2 ratio ( pmol siRNA: μl Lipofectamine 2000 ) ; 24 hr later , cells were plated onto round coverslips and incubated for an additional 24 hr in CLCA1- or pHLsec-conditioned medium as described above . TMEM16A knockdown was estimated at 60–70% as assayed by qPCR . Experiments were performed at 25°C , 24 hr after co-culture or incubation in conditioned medium . Micropipettes were prepared from non-heparinized hematocrit glass ( Kimble-Chase , Vineland , NJ ) on a horizontal puller ( Sutter Instruments , Novato , CA ) , and filled to a typical electrode resistance of 2 megaohms with pipette solution containing 150 mM N-methyl-D-glucamine ( NMDG ) chloride , 10 mM Hepes , 2 mM MgCl2 , 8 mM HEDTA , and 5 . 8 mM CaCl2 to attain 10 µM free Ca2+ , as calculated by means of the CaBuf program ( available through Katholieke Universiteit Leuven ) . Selected experiments were performed with a pipette solution containing ( mM ) 150 NMDG chloride , 10 Hepes and 2 MgCl2 , in the absence ( [0 μM Ca2+]in ) or presence of 5 mM EGTA and 4 mM CaCl2 to attain 1 μM free Ca2+ ( [1 μM Ca2+]in ) . The pH of all pipette solutions was adjusted to 7 . 1 with Tris . The bath solution was 10 mM Hepes , 1 mM CaCl2 and 1 mM MgCl2; plus 150 mM NaCl ( standard extracellular , [154 mM Cl−]out ) , or 140 mM Na-gluconate and 10 mM NaCl ( reduced extracellular Cl− [14 mM Cl−]out ) , and adjusted to pH 7 . 4 with Tris . After formation of a gigaohm seal and establishment of whole-cell configuration , cells were voltage-clamped at 0 mV . A pulse protocol was applied in which membrane potential was held at 0 mV for 50 ms and stepped to a test value for 600 ms before returning to the holding potential for an additional 400 ms . The test potential varied from −100 to +100 mV in 20 mV increments . Membrane capacitance was calculated from the integral of the current transient in response to 10 mV depolarizing pulses , and was monitored for stability throughout the experiment . Data were filtered at 2 kHz , and signals were digitized at 5 kHz with a Digidata 1322A ( Molecular Devices , Sunnyvale , CA ) . MultiClamp 700B Commander and pClamp software ( Molecular Devices ) were used for pulse protocol application and data acquisition . Data were analyzed using Clampfit 10 . 1 ( Molecular Devices ) . Liquid junction potentials were calculated using Clampex JPCalc software and command voltages were corrected a posteriori as specified in the figure legends . Results are presented as mean ± S . E . , differences between two groups were assessed by unpaired Student's t test with Welch's correction , and differences between more than two groups were evaluated by one-way ANOVA and post-hoc Tukey test ( Prism 5 . 0c , GraphPad Software , San Diego , CA ) . For staining experiments , cells were either transfected or exposed to conditioned medium as described above . Following 24 hr incubation , cells were fixed on glass slides with 4% paraformaldehyde ( PFA ) in PBS for 5 min and washed twice with PBS . Cells were blocked for 1 hr at room temperature with 1% blocking solution in PBS ( Life Technologies ) and then incubated with primary antibodies ( rabbit anti-human CLCA1 polyclonal antibody 1228 at 1:100 dilution and goat-anti-human-TMEM16A polyclonal antibody S-20 at 1:50 dilution ) overnight at 4°C . Slides were washed and incubated with WGA-Alexa Flour 633 conjugate ( 5 μg/ml ) for 30 min at room temperature , followed by secondary antibodies ( donkey anti-rabbit IgG-Alexa Fluor 594 conjugate at 1:250 dilution and donkey anti-goat IgG-Alexa Fluor 488 conjugate at 1:200 dilution ) for 2 hr at room temperature . Washed slides were then mounted in VECTASHIELD H-1200 Mounting Medium with DAPI ( Vector Laboratories , Burlingame , CA ) . Confocal microscopy was carried out using a Zeiss LSM 510 META Confocal Laser Scanning Microscope ( Carl Zeiss Microscopy , Thornwood , NY ) . The images were acquired with LSM 4 . 2 software and batch processed with AxioVision 4 . 8 . 2 ( Carl Zeiss Microscopy ) . Human CLCA1 was assayed for binding to cell-surface TMEM16A using a flow cytometry-based binding assay . Prior to staining , intact HEK293T cells were treated with human FcR blocking reagent at 1:100 dilution ( Miltenyi Biotec , San Diego , CA ) for 15 min . The biotinylated N-CLCA1 was pre-incubated with SA-PE at a 4:1 molar ratio for 15 min at room temperature to produce fluorescently labeled tetramers of N-CLCA1 ( N-CLCA1/SA-PE ) . Cells ( 4 × 105 cells/sample ) were either stained with SA-PE alone ( 1:50 ) or N-CLCA1/SA-PE ( 1:50 ) diluted in PBS containing 1% BSA ( FACS buffer ) at 4°C . In order to validate specific binding of N-CLCA1/SA-PE to cell surface TMEM16A , goat-anti-human-TMEM16A polyclonal antibody S-20 ( 1:10 ) , which was raised against a 15–20 amino acid peptide within residues 820–870 ( corresponding to the last extracellular loop; UniProt Q5XXA6 ) , was added prior to addition N-CLCA1/SA-PE . A goat polyclonal IgG antibody for human Aquaporin5 was used ( 1:10 ) as an isotype control for the blocking antibody S-20 . Mouse-anti-human-TMEM16A monoclonal antibody C5 , which binds a cytosolic epitope , was used as a second control antibody ( 1:10 ) . Following staining , cells were washed with FACS buffer , and then analyzed by flow cytometry ( BD FACScan ) . Data analysis was performed using FlowJo ( Tree Star , Ashland , OR ) . HEK293T cells were pelleted , lysed in lysis buffer ( 1 . 5 mM KH2PO4 , 2 . 7 mM KCl , 4 . 3 mM Na2HPO4 , 137 mM NaCl and 1% Triton X-100 in deionized water ) , and then diluted 1:2 in 2× SDS containing 2-mercaptoethanol . Samples were boiled for 5 min , and then loaded on a 4–12% bis-tris Nupage gel ( Life Technologies ) . The proteins were transferred to nitrocellulose membranes using an iBlot Gel Transfer Device ( Life Technologies ) . Membranes were blocked by 0 . 5% nonfat milk in PBS with 0 . 1% TWEEN . Primary antibodies ( mouse α-human CLCA1 8D3 , 1:4000; mouse-anti-human-TMEM16A monoclonal antibody C-5 , 1:1000; mouse-anti-actin monoclonal antibody C4 , 1:5000 ) in blocking buffer were incubated on the membrane for 15 min . Following three washes with PBS-TWEEN , secondary antibodies ( goat-anti-mouse IgG-HRP conjugate 1:5000 ) in blocking buffer were applied for 15 min . After three PBS-TWEEN washes , signal was detected using Pierce ECL Western Blotting Substrate ( Thermo Fisher Scientific , Rockford , IL ) . Developed films were scanned and converted to 8-bit tiff files . Protein bands were processed equally and the pixel intensities were quantified with ImageJ 1 . 48 ( http://imagej . nih . gov/ij ) . | Many biological processes that are important for our health involve the movement of ions into , and out of , our cells . For example , the flow of chloride ions out of cells controls the production of the sticky mucus that lines our windpipe and other airways . This mucus helps trap pollution and other foreign particles before they reach our lungs , and thus protects the lungs from harm . However in some diseases—such as cystic fibrosis and asthma—excessive amounts of thick mucus are produced; this can lead to breathing difficulties and an increased risk of infection . Proteins belonging to the CLCA protein family were first thought to act as channels that allow chloride ions to flow through cell membranes . Later studies then revealed that these proteins are not channels; instead they trigger the movement of chloride ions across cell membranes by activating other channel proteins . However , the identity of these channel proteins was unknown , and it was unclear how CLCA proteins might activate these channels . Sala-Rabanal , Yurtsever et al . have now shown that a member of the CLCA protein family , called CLCA1 , is released from human cells and causes nearby cells to release chloride ions when the channel detects calcium ions . The movement of chloride ions triggered by CLCA1 looked very similar to the way chloride ions flow through a channel protein called TMEM16A , and so Sala-Rabanal , Yurtsever et al . asked whether these two proteins interact . TMEM16A was discovered several years ago , but remains the only calcium-dependent chloride channel known in mammals . Sala-Rabanal , Yurtsever et al . showed that adding CLCA1 to cells caused more TMEM16A channels to appear in the cell surface membrane and thereby increased the flow of chloride ions . The CLCA protein also physically interacted with the chloride channel in the membrane to stabilize it; no other protein has been shown to regulate ion channels in this way before . The findings of Sala-Rabanal , Yurtsever et al . provide a much clearer understanding of how the CLCA protein and the chloride channel work . Both of these proteins are known to contribute to excess mucus production in airway diseases; and both have been linked to cardiovascular diseases and certain cancers . These new findings may therefore also help researchers to target these proteins and develop treatments for these diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2015 | Secreted CLCA1 modulates TMEM16A to activate Ca2+-dependent chloride currents in human cells |
Conductance-based models of neural activity produce large amounts of data that can be hard to visualize and interpret . We introduce visualization methods to display the dynamics of the ionic currents and to display the models’ response to perturbations . To visualize the currents’ dynamics , we compute the percent contribution of each current and display them over time using stacked-area plots . The waveform of the membrane potential and the contribution of each current change as the models are perturbed . To represent these changes over a range of the perturbation control parameter , we compute and display the distributions of these waveforms . We illustrate these procedures in six examples of bursting model neurons with similar activity but that differ as much as threefold in their conductance densities . These visualization methods provide heuristic insight into why individual neurons or networks with similar behavior can respond widely differently to perturbations .
Experimental and computational studies have clearly demonstrated that neurons and circuits with similar behaviors can , nonetheless , have very different values of the conductances that control intrinsic excitability and synaptic strength . Using a model of the crustacean stomatogastric ganglion ( STG ) , Prinz et al . ( 2004 ) showed that similar network activity can arise from widely different sets of membrane and synaptic conductances . Recent experimental measurements have shown two- to six-fold variability in individual components in the same identified neurons ( Schulz et al . , 2006; Schulz et al . , 2007; Roffman et al . , 2012; Swensen and Bean , 2005 ) . The use of RNA sequencing and other molecular measurements have shown significant cell-to-cell variability in the expression of ion channels ( Temporal et al . , 2012; Temporal et al . , 2014; Tobin et al . , 2009 ) . Together these results suggest that similar activities arise from different cellular and network mechanisms . Here , we use conductance-based models to explore how different these mechanisms are and how they respond to perturbation . Because of the intrinsic variability , canonical models that capture the mean behavior of a set of observations are not sufficient to address these issues ( Golowasch et al . , 2002; Balachandar and Prescott , 2018 ) . To incorporate intrinsic biophysical variability Prinz et al . ( 2004 ) introduced an ensemble modeling approach . They constructed a database with millions of model parameter combinations , analyzed their solutions to assess network function , and screened for conductance values for which the activity resembled the data ( Calabrese , 2018 ) . An alternative was used by Achard and De Schutter ( 2006 ) . They combined evolutionary strategies with a fitness function based on a phase-plane analysis of the models’ solutions to find parameters that reproduce complex features in electrophysiological recordings of neuronal activity , and applied their procedure to obtain 20 very different computational models of cerebellar Purkinje cells . Here , we adopt a similar approach and apply evolutionary techniques to optimize a different family of landscape functions that rely on thresholds or Poincaré sections to characterize the models’ solutions . In some respects , biological systems are a black-box because one cannot read out the values over time of all their underlying components . In contrast , computational models allow us to inspect how all the components interact and this can be used to develop intuitions and predictions about how these systems will respond to perturbations . Despite this , much modeling work focuses on the variables of the models that are routinely measured in experiments , such as the membrane potential . While in the models we have access to all state variables , this information can be hard to represent when many conductances are at play . Similarly , the effect of perturbations – such as the effect of partially or completely blocking or removing a particular channel – can be complex and also hard to display in a compact fashion . Here , we address these difficulties and illustrate two novel visualization methods . We represent the currents in a model neuron using stacked area plots: at each time step , we display the shared contribution of each current to the total current through the membrane . This representation is useful to visualize which currents are most important at each instant and allows the development of insight into how these currents behave when the system is perturbed . Perturbation typically results in drastic changes of the waveform of the activity and these changes depend on the kind of perturbation under consideration . Additionally , we developed a novel representation that relies on computing the probability of V ( t ) , which allows a visualization of these changes . We illustrate the utility of these procedures using models of single neuron bursters or oscillators .
The numerical exploration of conductance-based models of neurons is a commonplace approach to address fundamental questions in neuroscience ( Dayan and Abbott , 2001 ) . These models can display much of the phenomenology exhibited by intracellular recordings of single neurons and have the major advantage that many of their parameters correspond to measurable quantities ( Herz et al . , 2006 ) . However , finding parameters for these models so that their solutions resemble experimental observations is a difficult task . This difficulty arises because the models are nonlinear , they have many state variables and they contain a large number of parameters ( Bhalla and Bower , 1993 ) . These models are complex , and we are not aware of a general procedure that would allow the prediction of how an arbitrary perturbation in any of the parameters will affect their solutions . The problem of finding sets of parameters so that a nonlinear system will display a target behavior is ubiquitous in the natural sciences . A general approach to this problem consists of optimizing a score function that compares features of the models’ solutions to a set of target features . Consequently , landscape-based optimization techniques for finding parameters in compartmental models of neurons have been proposed before ( Achard and De Schutter , 2006; Druckmann et al . , 2007; Ben-Shalom et al . , 2012 ) . Here , we employ these ideas to develop a family of score functions that are useful to find parameters so that their activities reach a desired target . In this work , we started with a well-studied model of neural activity described previously ( Liu et al . , 1998; Goldman et al . , 2001; Prinz et al . , 2004; O'Leary et al . , 2014 ) . The neuron is modeled according to the Hodgkin-Huxley formalism using a single compartment with eight currents . Following Liu et al . ( 1998 ) , the neuron has a sodium current , INa; transient and slow calcium currents , ICaT and ICaS; a transient potassium current , IA; a calcium-dependent potassium current , IKCa; a delayed rectifier potassium current , IKd; a hyperpolarization-activated inward current , IH; and a leak current Ileak . We explored the space of solutions of the model using landscape optimization . The procedure consists of three steps . First , we generate voltage traces by integration of Equation 5 ( Materials and methods ) . We then score the traces using an objective or landscape function that defines a target activity . Finally , we attempt to find minima of the objective function . The procedures used to build objective functions whose minima correspond to sets of conductances that yield the target activities are shown in Figure 1 . Voltage traces were generated by integration of Equation 5 and were then scored according to a set of simple measures . The procedure is efficient in part because we chose measures that require little computing power and yet are sufficient to build successful target functions . For example , we avoid the use of Spike Density Functions ( SDF ) and Fourier transforms when estimating burst frequencies and burst durations . In this section , we describe target functions whose minima correspond to bursting and tonic activity in single compartment models . This approach can also be applied to the case of small circuits of neurons ( Prinz et al . , 2004 ) . We begin with the case of bursters ( Figure 1A ) . We targeted this type of activity by measuring the bursting frequency , the duty cycle , and the number of crossings at a threshold value to ensure that spiking activity is well separated from slow wave activity . To measure the burst frequency and duty cycle of a solution , we first compute the time stamps at which the cell spikes . Given the sequence of values V={Vn} we determine that a spike occurs every time that V crosses the spike detection threshold Tsp=−20mV ( red in Figure 1 ) . We build a sequence of spike times S={si} by going through the sequence of voltages {Vn} and keeping the values of n for which Vn≤TspandVn+1>Tsp ( we consider upward crossings ) . Each element si of the sequence S contains the time step at which the i-th spike is detected . Bursts are determined from the sequence of spike times S; if two spikes happen within a temporal interval shorter than δspt=100msec they are part of a burst . Using this criterion we can find which of the spike times in S correspond to the start and end of bursts . The starts ( bs ) and ends ( be ) of bursts are used to estimate the duty cycle and burst frequency . We loop over the sequence of spike times and determine that a burst starts at si if si+1−si<δsptandsi−si−1>δspt . After a burst starts , we define the end of the burst at sk if sk+1−sk>δsptandsk−sk−1<δspt . When a burst ends we can measure the burst duration as δb=sk-si and since the next burst starts ( by definition ) at sk+1 we also can measure the ‘period’ ( if periodic ) of the oscillation as τb=δb+ ( sk+1-sk ) . Every time a burst starts and ends we get an instance of the burst frequency fb=1τb and the duty cycle dc=δbτb . We build distributions of these quantities by looping over the sequence S and define the burst frequency and duty cycle as the mean values <fb> and <dc> . Finally , we count downward crossings in the sequence Vn with two slow wave thresholds #sw ( with tsw=-50±1mV ) and the total number of bursts #b in S . For any given set of conductances , we simulated the model for 20 s and dropped the first 10 s to mitigate the effects of transient activity . We then computed the burst frequency <fb> , the duty cycle <dc> , the number of crossings with the slow wave thresholds #sw and the number of bursts #b . We discard unstable solutions; a solution is discarded if std ( {fb} ) ≥ ( <fb>×0 . 1 ) or std ( {dc} ) ≥ ( <dc>×0 . 2 ) . If a solution is not discarded , we can use the following quantities to measure how close it is to the target behavior , ( 1 ) Ef= ( ftg−<fb>i ) 2Edc= ( dctg−<dc>i ) 2Esw= ( #sw2−#b ) 2 Here , Ef measures the mismatch of the bursting frequency of the model cell with a target frequency ftg and Edc accounts for the duty cycle . Esw measures the difference between the number of bursts and the number of crossings with the slow wave thresholds tsw=-50±1mV . Because we want a clear separation between slow wave activity and spiking activity , we ask that #sw=#b . Note that if during a burst V goes below tsw this solution would be penalized ( factor 12 accounts for using two slow wave thresholds ) . Let g denote a set of parameters , we can then define an objective function ( 2 ) E ( g ) =αEf+βEdc+γEsw , where the weights ( α , β , γ ) determine the relative importance of the different sources of penalties . In this work we used α=1 , β=100 , γ=1 , and the penalties Ei were calculated using T=10 seconds with dt=0 . 1 msecs . The target behavior for bursters was defined by dctg=0 . 2 ( duty cycle 20% ) ( dctg=0 . 2 ) and bursting frequency ftg=1Hz . We can use similar procedures to target tonic spiking activity . Note that the procedure we described previously to determine bursts from the sequence of spike times S is also useful in this case . If a given spike satisfies the definition of burst start and it also satisfies the definition of burst end then it is a single spike and the burst duration is zero . Therefore , we compute the bursts and duty cycles as before and ask that the the target duty cycle is zero . There are multiple ways to produce tonic spiking in this model and some solutions display very different slow wave activity . To further restrict the models , we placed a middle threshold at tmid=-35mV and detected downward crossings at this value . We defined Elag as the lag between the upward crossings at the spiking threshold ( tspk=-20mV ) and downward crossings at tmid . Elag is useful because it takes different values for tonic spikers than it does for single-spike bursters even though their spiking patterns can be identical . Finally , we found that the model attempts to minimize Elag at the expense of hyperpolarizing the membrane beyond -50mV and introducing a wiggle that can be different in different solutions . To penalize this we included additional thresholds between -35mV and -45mV , counted the number of downward crossings at these values #midi , and asked that these numbers are equal to the number of spikes #s . With these definitions , we define the partial errors as before , ( 3 ) Ef= ( ftg−<fb>i ) 2Edc= ( dctg−<dc>i ) 2Emid=∑i ( #midi−#s ) 2Esw= ( #sw ) 2 . The total error as a function of the conductances reads as follows , ( 4 ) E ( g ) =αEf+βEdc+γEmid+δEsw+ηElag . The values α=1000 , β=1000 , γ=100 , δ=100 and η=1 , produce solutions that are almost identical to the one displayed in Figure 1B . In all cases , evaluation of the objective functions requires that the models are simulated for a number of seconds and this is the part of the procedure that requires most computing power . Longer simulations will provide better estimations for the burst frequency and duty cycle of the cells , but will linearly increase the time it takes to evaluate the objective function . If the simulations are shorter , evaluations of the objective function are faster but the minimization may be more difficult due to transient behaviors and its minima may not correspond to stable solutions . In this work , we minimized the objective function using a standard genetic algorithm ( Holland , 1992; Goldberg and Holland , 1988 ) . The choice of the optimization routine and the choice of the numerical scheme for the simulations are independent of the functions . See Materials and methods for details on the how we performed this optimization . The same functions can be utilized to estimate parameters in models with different channel types . Most modeling work focuses on the variables of the models that are routinely measured in experiments such as the membrane potential as is shown in Figure 2A for a bursting neuron . While in the models we have access to all state variables , this information can be hard to represent when several current types are at play . One difficulty is that some currents like Na and Kd vary over several orders of magnitude , while other currents like the leak and H span smaller ranges . Additionally , the relative contribution of each current to the total flux through the membrane varies over time . Here , we introduce a novel representation that is simple and permits displaying the dynamics of the currents in a cohesive fashion . At any given time stamp , we can compute the total inward and outward currents . We can then express the values of each current as a percentage of this quantity . The normalized values of the currents at any time can be displayed as a pie chart representing the share of each current type ( Figure 2B ) . Because we want to observe how these percentages change in time , we display the shares in a bar instead of a disk . The currentscapes are constructed by applying this procedure to all time stamps and stacking the bars . These types of plots are known as stacked area plots and their application to this problem is novel . Figure 2C shows the currentscape of a periodically bursting model neuron over one cycle . The shares of each current type to the total inward and outward currents are displayed in colors , and the total inward and outward currents are represented by the filled black curves in logarithmic scale in the top and bottom . To visualize changes in the activity as a conductance is gradually removed we computed the distribution of membrane potential V values . This reduction contains information about the waveform of the membrane potential , while all temporal information such as frequency can no longer be recovered . The number of times that a given value of V is sampled is proportional to the time the system spends at that value . Figure 3A shows the distribution of V for a periodic burster with fb≈1Hz and dc≈20% sampled from 30 s of simulation . The count is larger than 104 for values between -52mV and -40mV , and smaller than 103 for V between -35mv and 20mV . The areas of the shaded regions are proportional to the probability that the system will be observed at the corresponding V range ( Figure 3B ) . Note that the area of the dark gray region is 105 while the light gray is 0 . 5×104 , so the probability that the cell is , at any given time , in a hyperpolarized state is more than 20 times larger than the probability that the cell is spiking . The distribution of V features sharp peaks . In many cases , the peaks in these distributions correspond to features of the waveform , such as the amplitudes of the individual spikes , or the minimum membrane potential ( see Figure 3—figure supplement 1 ) . This happens because every time the membrane potential reaches a maxima or minima ( in time ) the derivative dVdt is close to zero . The system spends more time close to values of V where the velocity dVdt is small than in regions where dVdt is large , as it occurs during the flanks of spikes . Therefore , when we sample a solution at a random instant , it is more likely that V corresponds to the peak of a spike than to either flank of the spike , while the most likely outcome is that V is in the hyperpolarized range ( <-40mV ) . In this particular burster , there are 12 spikes in the burst but there are only 7 peaks in the distribution ( between 10mV and 20mV ) ; some spikes have similar amplitudes so they add to a larger peak in the distribution . The overall or total amplitude of the oscillation can be read from the distribution since the count of V is zero outside a range ( -52mV to 20mV ) . These distributions can also be represented by a graded bar as shown in Figure 3B . As conductances are gradually removed the waveform of the activity changes and so does the distribution of V values . Figure 3C shows how the distribution of V changes as gNa is decreased . The waveforms at a few values of gNa are shown for reference . For each value in the range ( 100%gNa to 0%gNa with N=1001 values ) we computed the count p ( V , gNa ) and display log10 ( p ( V , gNa ) +1 ) in gray scales . In this example , the cell remains in a bursting regime up to ≈85%gNa and transitions abruptly into a single-spike bursting mode for further decrements ( %80gNa ) . The spikes produce thin ridges in the distribution that show how their individual amplitudes change . The colored symbols indicate the correspondence between features in the waveform and ridges in the distribution . In this example , the peak amplitudes of the spikes are similar for values of gNa greater than %85gNa . After the transition , the amplitudes of the spikes are very different; two spikes go beyond 0mV and the rest accumulate near -25mV . As gNa→0 the oscillations collapse onto a small band at ≈-20mV and only one spike is left . The distributions allow the visualization of the amplitudes of the individual spikes , the slow waves , and other features as the parameter gNa is changed . To highlight ridges in the distributions , the center panel in Figure 3D shows the derivative ∂Vlog10 ( p ( V ) ) in color . This operation is similar to performing a Sobel filtering ( Sobel and Feldman , 1968 ) of the image in Figure 3C . The traces on each side of this panel correspond to the control ( left ) and 80%gNa conditions . Notice how the amplitudes of each spike , features of the slow wave , and overall amplitude correspond to features in the probability distribution . This representation permits displaying how the features of the waveform change for many values of the perturbation parameter gNa . We explored the solutions of a classic conductance-based model of neural activity using landscape optimization and found many sets of parameters that produce similar bursting activity . Inspired by intracellular recording performed in the Pyloric Dilator ( PD ) neurons in crabs and lobsters we targeted bursters with frequencies fb≈1Hz and duty cycles dc≈20% . We built 1000 bursting model neurons and visually inspected the dynamics of their currents using their currentscapes . Based on this , we selected six models that display similar membrane activity via different current compositions for further study . Because the models are nonlinear , the relationship between the dynamics of a given current type and the value of its maximal conductance is non-trivial . Figure 4 shows the values of the maximal conductances in the models ( top ) and their corresponding activity together with their currentscapes ( bottom ) . It can be difficult to predict the currentscapes based on the values of the maximal conductances . In most cases , it appears that the larger the value of the maximal conductance , the larger the contribution of the corresponding current . However , this does not hold in all cases . For example , burster ( f ) shows the largest A current contribution , but bursters ( c ) and ( e ) have larger values of gA . The maximal conductance of the CaS current is low in model ( f ) but the contribution of this current to the total is similar to that in models ( a ) and ( b ) . The values of gKCa are similar for bursters ( e ) and ( f ) but the contribution of this current is visibly different in each model . The models produce similar activity with different current dynamics . To further reveal differences in how these activities are generated , we subjected the models to simple perturbations . We begin describing the response to constant current injections in Figure 5 . Figure 5A and Figure 5B show the membrane potential of model ( a ) for different values of injected current . In control , the activity corresponds to regular bursting and larger depolarizing currents result in a plethora of different regimes . The distributions of inter-spike intervals ( ISI ) provide a means to characterize these regimes ( Figure 5C ) . When the cell is bursting regularly such as in control and in the 0 . 8nA condition , the interspike interval distributions consist of one large value that corresponds to the interburst interval ( ≈640msec in control ) and several smaller values around 10msec which correspond to the ISI within a burst . There are values of current for which the activity appears irregular and correspondingly , the ISI values are more diverse . Figure 5B shows the response of the model to larger depolarizing currents . The activity undergoes a sequence of interesting transitions that result in tonic spiking . When Ie=3 . 45nA the activity is periodic and there are 4 ISI values . Larger currents result in 2 ISI values and tonic spiking produces one ISI value . Figure 5C shows the ISI distributions ( y-axis , logarithmic scale ) for each value of injected current ( x-axis ) . All these bursters transition into tonic spiking regimes for depolarizing currents larger than 5nA but they do so in different ways . To explore these transitions in detail , we computed the inter-spike interval ( ISI ) distributions over intervals of 60sec for different values of the injected current . Figure 6 shows the ISI distributions for the six models at N=1001 equally spaced values of injected current over the shown range . The y-axis shows the values of all ISIs on a logarithmic scale and the x-axis corresponds to injected current . In the control , the ISI distribution consists of a few small values ( <100msec ) that correspond to the ISIs of spikes within a burst , and a single larger value ( >100msec ) that corresponds to the interval between the last spike of a burst and the first spike of the next burst . When the cell fires tonically the ISI distributions consist of a single value . The ISI distributions exhibit complicated dependences on the control parameter that result in beautiful patterns . For some current values , the cells produce small sets of ISI values indicating that the activity is periodic . However , this activity is quite different across regions . Interspersed with the regions of periodicity there are regions where the ISI distributions densely cover a band of values indicating non-periodic activity . Overall the patterns feature nested forking structures that are reminiscent of classical period doubling routes to chaos ( Feigenbaum , 1978; Canavier et al . , 1990 ) . Detailed conductance-based models show complex and rich behaviors in response to all kinds of perturbations . There is a vast amount of information that can be seen in these models and their visualizations in Figures 7 - 15 . It is entirely impossible for us to point out even a fraction of what can be seen or learned from these figures . Nonetheless , we will illustrate a few examples of what can be seen using these methods , knowing that these details will be different for models that are constructed in the future and analyzed using these and similar methods . Figures 7 and 8 show the effects of gradually decreasing each of the currents in these bursters from 100% to 0% for all six models . This type of analysis might be relevant to some kinds of pharmacological manipulations or studies of neuromodulators that decrease a given current . The figures show 3 s of data for each condition . In all panels , the top traces correspond to the control condition ( 100% ) and the traces below show the activity that results from decreasing the maximal conductance . The dashed lines are placed for reference at -50mV and 0mV . Each panel shows the traces for 11 values of the corresponding maximal conductance equally spaced between 100% ( control ) and 0% ( completely removed ) . Each row of panels corresponds to a current type and the columns correspond to the different model bursters . Figure 7 displays the perturbations for the inward currents and Figure 8 shows the outward and leak currents . Taken together Figures 7 and 8 illustrate that each model ( a-f ) changes its behavior differently in response to decreases in each current . Additionally , decreases in some currents have only relatively small effects but decreases in others have much more profound effects . Because the description of all that can be seen in these figures is beyond the scope of this paper , we chose to focus on the effects of decreasing the CaT because it has rich and unexpected behaviors . The effect of decreasing the CaT conductance is quite diverse across models . The activities of the models at the intermediate values of gCaT shows visible differences . When gCaT→0 . 7gCaT models ( a ) , ( b ) and ( c ) show bursting activity at different frequencies and with different duty cycles . Models ( d ) , ( e ) and ( f ) become tonic spikers at this condition , but their frequencies are different . Note that in the case of model ( e ) the spiking activity is not regular and the ISIs take several different values . When gCaT→0 . 2gCaT most models spike tonically but now ( e ) is regular and ( f ) shows doublets . When CaT is completely removed , most models transition into a tonic spiking regime with the exception of model ( a ) , that displays a low frequency bursting regime with duty cycle ≈0 . 5 . Decreasing any conductance can trigger qualitative changes in the waveform of the membrane potential and in the contributions of each current to the activity . In Figure 9 we plot currentscapes for the effects of decreasing CaT in model ( f ) . This allows us to examine at higher resolution the changed contributions of currents that give rise to the interesting dynamics seen in Figure 7 . Each panel in Figure 9 corresponds to a different decrement value and shows the membrane potential on top , and the currentscapes at the bottom . The top panels show 1 second of data and correspond to the 100%gCaT ( control ) , 90%gCaT and 80%gCaT conditions . The center panels show 0 . 1 s of data for decrements ranging from 70% to 20% and the bottom panels show 2 s for the 10% and 0% conditions . As CaT is gradually removed the activity transitions from a bursting regime to a tonic spiking regime . When gCaT→90%gCaT the neuron produces bursts but these become irregular and their durations change . Decreasing the conductance to 80%gCaT results in completely different activity . The spiking pattern appears to be periodic but there are at least three different ISI values . It is hard to see changes in the CaT contribution across these conditions , but changes in other currents are more discernible . The contribution of the A current that is large in the control and 90%gCaT conditions , is much smaller in the 80%gCaT condition . Additionally , the Na and KCa currents show larger contributions , the CaS current contributes less and the H current is negligible . Further increments in simulated blocker concentration result in tonic spiking regimes with frequencies ranging from ≈20Hz to ≈10Hz . The center panels in Figure 9 show the currentscapes for these conditions on a different time scale to highlight the contributions of CaT . The leftmost panel shows the 70%gCaT condition . In this panel , we placed vertical lines indicating the time stamps at which the peak of the spike and the minimum occur . Notice the large contribution of the Na current prior to the peak of the spike , and the large contribution of the Kd current for the next ≈10msec . When the membrane potential is at its minimum value the CaT current dominates the inward currents and remains the largest contributor for the next ≈10msec . The CaT current reduces its share drastically by the time the Na current is visible and CaS takes over . The contribution of CaT remains approximately constant during repolarization and vanishes as the membrane becomes depolarized and the Na current becomes dominant . The effect of removing CaT is visible on this scale . The waveform of the contribution remains qualitatively the same: largest at the minimum voltage and approximately constant until the next spike . However , the contribution of CaT during repolarization becomes smaller , and for larger conductance decrements results in a thinner band . Finally , the bottom panels show the cases 10%gCaT and 0%gCaT which correspond to a two-spike burster and a tonic spiker , respectively . Note that even though the contribution of CaT is barely visible , complete removal of this current results in a very different pattern . The activity switched from bursting to spiking and the current composition is different; KCa disappeared in the 0% condition and the A current takes over . Notice also the larger contribution of the H current . There has been a great deal of work studying the effects of genetic and/or pharmacological deletions of currents . One of the puzzles is why some currents , known to be physiologically important , can have relatively little phenotype in some , or all individuals . For this reason in Figures 10 and 11 , we show the effects of deletion of each current in all six models . Each panel shows 2 seconds of data . The inward currents are portrayed in Figure 10 and the outward and leak currents are shown in Figure 11 . Removal of some currents has little obvious phenotype differences across the population although the currentscapes are different , such as seen for the gNa and gCaS cases . Removal of some currents produces similar phenotypes in most , but not all of the six models as seen in the gH and gA cases . Removal of Kd had virtually identical effects both on the phenotype and the currents . For other currents , such as KCa and the Leak , we find two types of responses with nearly half of the models for each case ( the exception is model ( d ) Leak ) . In the case of the CaT current both the phenotype and the currents composition are very diverse across models . A fuller description of the behavior/phenotype of all of the models for all values of conductance decrements can be seen in Figures 12 and 13 . These figures use the probability scheme described in Figure 3 and Figure 3—figure supplement 1 . Using these methods , it is possible to see exactly how the waveforms change and the boundaries of activity for each model and each conductance . The panels show the ridges of the probability distributions p ( V ) of the membrane potential V ( t ) for 1001 values of maximal conductance values ( see Materials and methods ) . The probability of V ( t ) was computed using 30 s of data after dropping a transient period of 120 s . It was estimated using Nb=1001 bins in the range ( -70 , 35 ) mv and N≈2×106 samples for each maximal conductance value . The system spends more time in regions where dVdt≈0 and is sampled more at those values . Therefore , features such as the amplitudes of the spikes appear as sharp peaks in the probability distributions . To highlight these peaks and visualize how they change as currents are gradually decreased , we plot the derivative or sharpness of the distribution in colors ( see color scale in Figure 3D ) . Overall , these plots show that for any given current , there are ranges of the conductance values where a small change results in a smooth deformation of the waveform , and there are specific values at which abrupt transitions take place . As before there is too much detail to describe everything in these figures so we will discuss a subset of the features highlighted by this representation . The top rows in Figure 12 correspond to removing the Na current in the models . Note that the minimum value of V in control ( left ) is close to -50mV and a small decrement in gNa results in larger amplitude . The colored curves inside the envelopes correspond to the spikes’ amplitudes and features of the slow waves . For instance , when the Na current is completely removed ( right ) the amplitude of the oscillation is ≈40mV and the activity corresponds to a single-spike bursting mode . The spike amplitude is given by the top edge of the colored region and the curve near ≈-20mV indicates the burst ‘belly’: the membrane hyperpolarizes slowly after spike termination and there is a wiggle at this transition . Removing CaT in model ( a ) does not disrupt bursting activity immediately . Notice that the amplitude of the bursts remains approximately constant over a range of gCaT values . The dim red and yellow lines at ≈20mV show that the amplitudes of the spikes are different and have different dependences with gCaT . When the model transitions into a tonic spiking regime , the amplitude of the spikes is the same and there is only one amplitude value . This value stays constant over a range but the minimum membrane potential decreases and the overall amplitude therefore increases . The model returns to a bursting regime for values of gCaT smaller than 30%gCaT . Notice that in model ( a ) the membrane potential during bursts goes below -50mV , unlike in the control condition . Notice that the waveform of the membrane potential changes abruptly as gCaT is reduced and the models transition into a spiking regime . Model ( f ) is less resilient to this perturbation and this transition takes place at lower conductance values . Removing CaS does not much change the waveform , but it alters the temporal properties of the activity . The models remain bursting up to a critical value and the amplitude of the spikes was changed little . The features of the slow wave do not much change either except in model ( f ) . Model ( c ) is less resilient to this perturbation since it becomes quiescent for lower decrements of the maximal conductance than the other models . The effect of gradually removing H appears similar to CaS in this representation . In this case again , the morphology of the waveform is less altered than its temporal properties ( except in model ( e ) where a transition takes place ) . Figure 13 shows the same plots for the outward and leak currents . The A current in model ( a ) is very small ( gA≈10μS ) and its removal has little effect on the activity . This translates into curves that appear as parallel lines indicating spikes with different amplitudes that remain unchanged . The rest of the models exhibit a transition into a different regime . The waveforms of this regime appears similar to the waveforms which result from removing gNa ( see Figure 7 ) but in this representation it is easier to observe differences such as the overall amplitude of the oscillation . The amplitude decreases as gNa is decreased and increases as gA is decreased . Removing KCa has a similar effect to removing gCaT in that the models transition into tonic spiking regimes . The difference is that the spiking regimes that result from removing KCa have smaller amplitudes and also correspond to more depolarized states . All models are very sensitive to removing Kd and low values result in single-spike bursting modes with large amplitudes . Model ( c ) is least fragile to this perturbation and exhibits a visible range ( ∼100% to ∼90% ) with bursting modes . These oscillations break down in a similar way to the Na case and display similar patterns . However , an important difference is that unlike in the gNa case , the overall amplitude of the oscillation increases as gKd is decreased . As before , the top edge corresponds to the amplitude of the large spike and the curves in the colored region correspond to extrema of the oscillation . After spiking , the membrane remains at a constant depolarized value ( ≈-20mV ) for a long period and produces a high-frequency oscillation before hyperpolarization . The amplitude of this oscillation increases as Kd is further decreased , and this results in a white curve that starts above 0mV and ends above 0mV . The beginning of this curve corresponds to a high-frequency oscillation that occurs after spike termination . This type of activity is termed plateau oscillations and was reported in models of leech heart interneurons ( Cymbalyuk and Calabrese , 2000 ) and in experiments in lamprey spinal neurons ( Wang et al . , 2014 ) . These features are hardly visible in the traces in Figure 8 and are highlighted by this representation . Finally , the Leak case appears similar to mixture of the Na and A cases . The cells remain bursting over a range of values and some of them transition into a single-spike bursting mode that is different from the KCa case . The key to the visualization method in Figures 12 and 13 is to consider V ( t ) not as a time series but as a stochastic variable with a probability distribution ( see Figure 3 and supplement ) . The same procedure can be applied to the time series of each current . However , because the contributions of the currents are different at different times , and at different decrements of conductance values , it is not possible to display this information using the same scale for all channels . To overcome this , we proceed as in the currentscapes and instead focus on the normalized currents or shares to the total inward and outward currents ( the rows of matrices C^+ and C^- , see Materials and methods ) . The current shares Ci^ ( t ) correspond to the width of the color bands in the currentscapes and can also be represented by a time series that is normalized to the interval [0 , 1] . The probability distribution of Ci^ ( t ) permits displaying changes in the contributions of each current to the activity as one current is gradually removed . Interpreting these distributions is straightforward as before: the number of times the system is sampled in a given current share configuration is proportional to the time the system spends there . The aim of plotting these distributions is to visualize how the currentscapes would change for all values of the conductance decrement . To illustrate this procedure , we return to CaT to explore further the causes of the complex behavior of model ( f ) ( see Figure 9 ) . Figure 14 shows the probability distributions of the current shares as CaT is gradually decreased in model ( f ) ( see also Figure 9 and Figure 14—figure supplement 1 ) . The panels show the share of each current as CaT is gradually decreased and the probability is indicated in colors . In control the Na and CaT current shares are distributed in a similar way . Both currents can at times be responsible for ≈90% of the inward current , but most of the time they contribute ≈20% . The Na current is larger right before spike repolarization and the CaT amounts to ≈90% of the small ( ≈5nA ) total inward current . For larger decrements , the system transitions into tonic spiking and the contribution of the Na current is more evenly distributed over a wider range . The contribution of the CaT current is predominantly ≈15% and trends to zero as gCaT→0 . Note also that as the contribution of CaT decreases , the contribution of CaS increases to values larger than 75% while in control it contributes with ≈50% . The contribution of the H current is small ( ≤25% ) between 100%gCaT and ≈80%gCaT; it becomes negligible between ≈80%gCaT and ≈20%gCaT and becomes dominant after 20%gCaT . The A current behaves similarly to the H . It contributes ≈90% of the ( small ≈2nA ) total outward current before burst initiation and its contribution decreases drastically when the system transitions into tonic spiking . As CaT is removed further the A current is more likely to contribute with a larger share . The contribution of the KCa current decreases as gCaT is decreased and some of it persists even when gCaT is completely removed . In contrast , the contribution of the Kd current does not appear to change much and nor does its role in the activity . Performing the same analysis for all conductances results in a large amount of information . Despite this and because we are plotting the normalized currents or current shares , our representation allows us to display this information in a coherent fashion . As an example , in Figure 15 we show the effect of gradually decreasing each current on all the currents in model ( c ) . The rows indicate which conductance is decreased and the columns show the effect of this perturbation on the corresponding current . The first row shows how the shares of each current change as the Na current is decreased . For instance , the effect of decreasing gNa on the Na current ( indicated by * ) is as expected , with the maxima of the distribution trending to zero as gNa→0 . The effect of removing gNa on the other currents is non-trivial and is displayed along the same row . Notice that while the effect of removing a current on that same current ( diagonal panels ) is relatively predictable , the rest of the currents become rearranged in complicated ways . Again , a full description of these diagrams is beyond the scope of this work so we will only make some observations . When the pertubations are negligible or weak ( 100% to ≈90% ) all currents play a role because there are periods of time in which they contribute to at least ≈20% of the total current . There are ranges of the conductances over which small changes result in smooth transformations of the current configuration , there are specific values at which sharp transitions take place , and these values are different depending on the current that is decreased . While some of this information can also be extracted from Figures 12 and 13 , the diagrams in Figure 15 show how the currents get reorganized at these transitions . In addition , this arrangement is convenient for comparing the effect of decreasing each conductance on a given current . For example , the contributions of the Na and Kd currents change little for most perturbations ( except when these conductances are decreased ) . In contrast , the contributions of CaT , CaS , H , KCa , and the leak change more noticeably . Finally , the contribution of the A current increases for most conductance decrements of any type , except at the transition values where it can grow or shrink in an abrupt manner .
There is an ever larger availability of experimental data to inform detailed models of identified neuron types ( McDougal et al . , 2017 ) . Experimenters have determined the kinetics of many channel types , both in vertebrate and invertebrate neurons . There are also model databases with thousands of parameters which permit the development of large scale models of neural tissue ( Bezaire et al . , 2016 ) . One difficulty in ensemble modeling is the necessity of incorporating the biological variability observed in some of the parameters – such as the conductances – at the same time that we require the models to capture some target activity . In other words , we may be interested in modeling a type of cell that displays some sterotypical behavior , and would like to obtain many different versions of such models . Two main approaches to this problem were introduced in the past . One consists of building a database of model solutions over a search domain and screening for target solutions: this considers all possible value combinations within an allowed range up to a numerical resolution and then applies quantitative criteria to determine which solutions correspond to the target activity ( Prinz et al . , 2004 ) . An alternative approach consists of designing a target function that assigns a score to the models’ solutions in such a way that lower scores correspond to solutions that meet the targets , and then optimizing these functions ( Achard and De Schutter , 2006; Druckmann et al . , 2007; Ben-Shalom et al . , 2012 ) . Both approaches have advantages and shortcomings . In the case of the database approach , trying all posible parameter combinations in a search range becomes prohibitively expensive as more parameters are allowed to vary . One advantage of this approach is that it provides a notion of how likely it is to find conductances within a search range that will produce the activity . In the landscape approach , we find solutions by optimization and – without further analysis – we do not know how likely a given solution type is . This approach has the advantage that it can be scaled to include large numbers of parameters . Additionally , if a particular solution is interesting , we can use genetic algorithms on successful target functions to ‘breed’ as many closely related models as desired . Ultimately , any optimization heuristic requires blind testing random combinations of the parameters , and developing quantitative criteria for screening solutions in a database results in some sort of score function , so the two approaches are complementary . A successful target function can determine if a random perturbation results in disruption of the activity and this can be used to perform population-based sensitivity analyses ( Devenyi and Sobie , 2016 ) . Regardless of the optimization approach , most work is devoted to the design of successful target functions . Different modeling problems require different target functions ( Roemschied et al . , 2014; Fox et al . , 2017; Migliore et al . , 2018 ) and one challenge in their design is that sometimes we do not know a priori if the model contains solutions that will produce good minima . In addition , a poorly constrained target function can feature multiple local minima that could make the optimization harder , so even if there are good minima they may be hard to find . One difference between the landscape functions in Achard and De Schutter ( 2006 ) and the ones utilized here is that in their setup model solutions are compared to a target time series via a phase-plane method . The functions introduced in this work use an analysis based on Poincaré sections or thresholds to characterize the waveform and to define an error or score . Instead of targeting a particular waveform , we ask that some features of the waveform – such as the frequency and the burst duration – are tightly constrained , while other features – such as the concavity of the slow waves – can be diverse . This is motivated by the fact that across individuals and species , the activity of the pyloric neurons can be diverse but the neurons always fire in the same sequence and the burst durations have a well-defined mean . Our approach is successful in finding hundreds of models that display a target activity in minutes using a commercially available desktop computer . Application of evolutionary techniques to optimize these functions provides a natural means to model the intrinsic variability observed in biological populations . One of the main benefits of computational modeling is that once a behavior of interest is successfully captured we then possess a mechanistic description of the phenomena that can be used to test ideas and inform experiments ( Coggan et al . , 2011; Lee et al . , 2016; Devenyi and Sobie , 2016; Gong and Sobie , 2018 ) . As the models gain biophysical detail these advantages wane in the face of the complexity imposed by larger numbers of variables and parameters . Conductance-based models of neural activity generate large amounts of data that can be hard to visualize and interpret . The development of novel visualization procedures has the potential to assist intuition into the details of how these models work ( Gutierrez et al . , 2013 ) . Here , we introduced a novel representation of the dynamics of the ionic currents in a single compartment neuron . Our representation is simple and displays in a concise way the contribution of each current to the activity . This representation is easily generalizable to multi-compartment models and small networks , and to any type of electrically excitable cell , such as models of cardiac cells ( Britton et al . , 2017 ) . We employed these procedures to build many similar bursting models with different conductance densities and to study their response to perturbations . The responses of the models to current injections and gradual decrements of their conductances can be diverse and complex . Inspection of the ISI distributions revealed wide ranges of parameter values for which the activity appears irregular , and similar regimes can be attained by gradually removing some of the currents . Period doubling routes to chaos in neurons have been observed experimentally and in conductance-based models ( Hayashi et al . , 1982; Hayashi and Ishizuka , 1992; Szücs et al . , 2001; Canavier et al . , 1990; Xu et al . , 2017 ) . The sort of bifurcation diagrams displayed by these models upon current injection are qualitatively similar to those exhibited by simplified models of spiking neurons for which further theoretical insight is possible ( Touboul and Brette , 2008 ) . Period doubling bifurcations and low dimensional chaos arise repeatedly in neural models of different natures including rate models ( Ermentrout , 1984; Alonso , 2017 ) . The bursters studied here are close ( in parameter space ) to aperiodic or irregular regimes suggesting that such regimes are ubiquitous and not special cases . We showed that in these model neurons similar membrane activities can be attained by multiple mechanisms that correspond to different current compositions . Because the dynamical mechanisms driving the activity are different in different models , perturbations can result in qualitatively different scenarios . Our visualization methods allow us to gather intuition on how different these responses can be and to explore the contribution of each current type to the neural activity . Even in the case of single compartment bursters , the response to perturbations of a population can be diverse and hard to describe . To gain intuition into the kind of behaviors the models display upon perturbation , we developed a representation based on the probability of the membrane potential V . This representation permits displaying changes in the waveform of V as each current is blocked . This representation shows that the models respond to perturbations in different ways , but that there are also similarities among their responses . A concise representation of the effect of a perturbation is a necessary step towards developing a classification scheme for the responses .
The membrane potential V of a cell containing N channels and membrane capacitance C is given by: ( 5 ) CdVdt=Ie-∑i=18Ii . Each term in the sum corresponds to a current Ii=gimpihqi ( V-Ei ) and Ie is externally applied current . The maximal conductance of each channel is given by gi , m and h are the activation and inactivation variables , the integers pi and qi are the number of gates in each channel , and Ei is the reversal potential of the ion associated with the i-th current . The reversal potential of the Na , K , H and leak currents were kept fixed at ENa=30mV , EK=-80mV , EH=-20mV and Eleak=-50mV while the calcium reversal potential ECa was computed dynamically using the Nernst equation assuming an extracellular calcium concentration of 3×103μM . The kinetic equations describing the seven voltage-gated conductances were modeled as in Liu et al . ( 1998 ) , ( 6 ) τmi ( V ) dmidt=m∞i ( V ) −miτhi ( V ) dhidt=h∞i ( V ) −hi . The functions τmi ( V ) , m∞i ( V ) , τhi ( V ) and h∞i ( V ) are based on the experimental work of Turrigiano et al . , 1995 and are listed in refs . ( Liu et al . , 1998; Turrigiano et al . , 1995 ) . The activation functions of the KCa current require a measure of the internal calcium concentration [Ca+2] ( Liu et al . , 1998 ) . This is an important state variable of the cell and its dynamics are given by , ( 7 ) τCad[Ca+2]dt=-CaF ( ICaT+ICaS ) -[Ca+2]+Ca0 . Here , CaF=0 . 94μMnA is a current-to-concentration factor and Ca0=0 . 05μM . These values were originally taken from Liu et al . and were kept fixed . Finally , C=10nF . The number of state variables or dimension of the model is 13 . We explored the solutions of this model in a range of values of the maximal conductances and calcium buffering time scales . The units for voltage are mV , the conductances are expressed in μS and currents in nA . Voltage traces were obtained by numerical integration of Equation 5 using a Runge-Kutta order 4 ( RK4 ) method with a time step of dt=0 . 1msec ( Press et al . , 1988 ) . We used the same set of initial conditions for all simulations in this work V=-51mV , m , hi=0 and [Ca+2]=5μM . For some values of the parameters , the system ( Equation 5 ) can display multistability ( Cymbalyuk et al . , 2002; Shilnikov et al . , 2005 ) . Optimization of the objective function Equation 2 is useful to produce sets of parameters g that result in bursting regimes . In this work , the optimization was performed over a search space of allowed values listed here: we searched for gNa∈[0 , 2×103] ( [μS] ) , gCaT∈[0 , 2×102] , gCaS∈[0 , 2×102] , gA∈2×[0 , 102] , gKCa∈[0 , 2×103] , gKd∈[0 , 2×102] , gH∈[0 , 2×102] , gL∈[0 , 2×10] , τCa∈[0 , 103] ( [msecs] ) . We minimized the objective function using a standard genetic algorithm Holland ( 1992 ) . This is optimization technique is useful to produce large pools of different solutions and is routinely utilized to estimate parameters in biophysical models ( see for example Assaneo and Trevisan , 2010 ) . The algorithm was started with a population of 1000 random seeds that were evolved for ≈10000 generations . The mutation rate was 5% . Fitter individuals were chosen more often to breed new solutions ( elitism parameter was 1 . 2 with 1 corresponding to equal breeding probability ) . The computation was performed on a multicore desktop computer ( 32 threads ) and takes about ≈1 hr to produce good solutions . The currentscapes are stacked area plots of the normalized currents . Although it is easy to describe their meaning , a precise mathematical definition of the images in Figure 2 can appear daunting in a first glance . Fortunately , the implementation of this procedure results in simple python code . The time series of the 8 currents can be represented by a matrix C with 8 rows and nsecs×1dt=N columns . For simplicity , we give a formal definition of the currentscapes for positive currents . The definition is identical for both current signs and is applied independently for each . We construct a matrix of positive currents C+ by setting all negative elements of C to zero , Ci , j+=Ci , j∣Ci , j>0 and Ci , j+=0∣Ci , j≤0 . Summing C+ over rows results in a normalization vector n+ with N elements nj+=∑iCi , j+ . The normalized positive currents can be obtained as C^+=C+/n+ ( element by element or entry-wise product ) . Matrix C^+ is hard to visualize as it is . The columns of C^+ correspond to the shares of each positive current and can be displayed as pie charts ( see Figure 2 ) . Here , instead of mapping the shares to a pie we map them to a segmented vertical ‘churro’ . The currentscapes are generated by constructing a new matrix CS whose number of rows is given by a resolution factor R=2000 , and the same number of columns N as C . Each column j of C^+ produces one column j of CS . Introducing the auxiliary variable pi , j=C^i , j+*R we can define the currentscape as , ( 8 ) CSi , j=k∣∑mkpm , j≤i<pk+1 , j+∑mkpm , j . The current types are indexed by k∈[0 , 7] and we assume ∑mk=0pm , j=0 . The black filled curve in Figure 2B corresponds to the normalization vector n+ plotted in logarithmic scale . We placed dotted lines at 5nA , 50nA and 500nA for reference throughout this work . The currentscapes for the negative currents are obtained by applying definition ( Equation 8 ) to a matrix of negative C- currents defined in an analogous way as C+ . Finally , note that matrices C^+ and C^- are difficult to visualize as they are . The transformation given by definition ( Equation 8 ) is useful to display their contents . We inspected the effects of injecting currents in our models by computing the inter-spike interval ISI distributions . For this , we started the models from the same initial condition and simulated them for 580 s . We dropped the first 240 s to remove transient activity and kept the last 240 s for analysis . Spikes were detected as described before . We collected ISI values for N=1001 values of injected current equally spaced between -1nA and 5nA . To sample the distributions of V we simulated the system with high temporal resolution ( dt=0 . 001msec ) for 30 s , after dropping the first 120 s to remove transients . We then sampled the numerical solution at random time stamps and kept 2×106 samples V={Vi} for each percent value . We took 1001 values between 1 and 0 . Model parameters used in this study are listed in Table 1 . | The nervous system contains networks of neurons that generate electrical signals to communicate with each other and the rest of the body . Such electrical signals are due to the flow of ions into or out of the neurons via proteins known as ion channels . Neurons have many different kinds of ion channels that only allow specific ions to pass . Therefore , for a neuron to produce an electrical signal , the activities of several different ion channels need to be coordinated so that they all open and close at certain times . Researchers have previously used data collected from various experiments to develop detailed models of electrical signals in neurons . These models incorporate information about how and when the ion channels may open and close , and can produce numerical simulations of the different ionic currents . However , it can be difficult to display the currents and observe how they change when several different ion channels are involved . Alonso and Marder used simple mathematical concepts to develop new methods to display ionic currents in computational models of neurons . These tools use color to capture changes in ionic currents and provide insights into how the opening and closing of ion channels shape electrical signals . The methods developed by Alonso and Marder could be adapted to display the behavior of biochemical reactions or other topics in biology and may , therefore , be useful to analyze data generated by computational models of many different types of cells . Additionally , these methods may potentially be used as educational tools to illustrate the coordinated opening and closing of ion channels in neurons and other fundamental principles of neuroscience that are otherwise hard to demonstrate . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"tools",
"and",
"resources",
"neuroscience"
] | 2019 | Visualization of currents in neural models with similar behavior and different conductance densities |
In many cell types , lateral diffusion barriers compartmentalize the plasma membrane and , at least in budding yeast , the endoplasmic reticulum ( ER ) . However , the molecular nature of these barriers , their mode of action and their cellular functions are unclear . Here , we show that misfolded proteins of the ER remain confined into the mother compartment of budding yeast cells . Confinement required the formation of a lateral diffusion barrier in the form of a distinct domain of the ER-membrane at the bud neck , in a septin- , Bud1 GTPase- and sphingolipid-dependent manner . The sphingolipids , but not Bud1 , also contributed to barrier formation in the outer membrane of the dividing nucleus . Barrier-dependent confinement of ER stress into the mother cell promoted aging . Together , our data clarify the physical nature of lateral diffusion barriers in the ER and establish the role of such barriers in the asymmetric segregation of proteotoxic misfolded proteins during cell division and aging .
Cellular diversity in eukaryotes relies in many instances on the asymmetric partition of cell fate determinants between daughter cells at cell division . In such instances , asymmetric cell division generates daughters with different division potentials . For example , most somatic cells exhibit a limited division potential , whereas the stem cells that generate them can divide indefinitely . Understanding the basis for these differences is critical for cell biology , aging research and regenerative medicine . The yeast Saccharomyces cerevisiae divides in an asymmetric manner through the budding of daughters from the surface of the mother cell . While these daughters are born young and form eternal lineages , the mother cells divides only a limited number of times ( 20–50 ) before stopping and dying . This process , termed replicative aging ( Egilmez and Jazwinski , 1989; Kennedy et al . , 1994; Steinkraus et al . , 2008 ) , is a consequence of the retention and accumulation of aging factors in the mother cell . A large variety of cellular features have been implicated in limiting the life span of yeast mother cells , including DNA-repair by-products called extra-chromosomal ribosomal DNA circles ( ERCs ) , carbonylated proteins , oxidized lipids ( Nyström , 2005; Steinkraus et al . , 2008 ) , multi drug transporters ( Eldakak et al . , 2010 ) , vacuolar pH and mitochondrial integrity ( Hughes and Gottschling , 2012 ) . How many more factors contribute to aging , whether and how these factors influence each other , which of them are early and primary causes of aging , and which of them actually kill the cell at the end of its life remain unclear . We also know little about how the segregation of these factors is biased towards the mother cell during mitosis . Recent data indicated that a lateral diffusion barrier in the outer nuclear membrane compartmentalizes the dividing nucleus and promotes the retention of DNA circles in the mother compartment ( Shcheprova et al . , 2008 ) and ERC accumulation ( Lindstrom et al . , 2011 ) . Accordingly , barrier defective cells are long-lived while their successive daughters become progressively shorter lived as they are born to mothers of increasing age . However , these mothers still age , indicating that they still accumulate some aging factors . Furthermore , the retention of old multi drug transporters in the mother cell is independent of the diffusion barriers ( Eldakak et al . , 2010 ) . Thus , several mechanisms control the segregation of aging factors towards the mother cell . However , what these mechanisms are and what their respective contribution to age segregation is remain unclear . Lateral diffusion barriers have been described in a number of eukaryotic membranes , including the initial segment of axons , dendritic spines , tight junctions of epithelial cells , the base of primary cilia , and the neck of budding yeast cells ( Myles et al . , 1984; Winckler and Mellman , 1999; Barral et al . , 2000; Takizawa et al . , 2000; Matter and Balda , 2003; Nakada et al . , 2003; Luedeke et al . , 2005; Vieira et al . , 2006; Shcheprova et al . , 2008; Caudron and Barral , 2009 ) . However , we still know very little about their physical nature and their mechanisms of action . The membrane systems of budding yeast cells are compartmentalized by at least three lateral diffusion barriers , one in the plasma membrane ( Barral et al . , 2000; Takizawa et al . , 2000 ) , one in the cortical ER ( cER , Luedeke et al . , 2005 ) and one in the outer membrane of the dividing nucleus ( Shcheprova et al . , 2008; Boettcher et al . , 2012 ) . Their assembly at the bud neck depends on a family of filament-forming GTPases , the septins ( Faty et al . , 2002; Weirich et al . , 2008; Hu et al . , 2010; Kim et al . , 2010; Saarikangas and Barral , 2011 ) , and on the actin- and formin-interacting protein Bud6 ( Amberg et al . , 1995 , 1997; Luedeke et al . , 2005; Shcheprova et al . , 2008 ) . Numerous questions remain concerning their molecular composition , their assembly , and their respective roles in cellular physiology . The ER is the site of folding and maturation of secretory proteins and protein complexes . A significant fraction of nascent secretory proteins fail to fold , are not correctly glycosylated , or are unable to find their destined partners ( Turner and Varshavsky , 2000; Ellgaard and Helenius , 2003; Princiotta et al . , 2003 ) . When accumulating , these misfolded proteins activate the unfolded protein response ( UPR ) and are recognized by the ER-associated degradation ( ERAD ) machinery , retrotranslocated to the cytoplasm , polyubiquitinated and targeted for degradation by the 26S proteasome ( Turner and Varshavsky , 2000; Ron and Walter , 2007; Brodsky and Skach , 2011 ) . These quality control pathways play an important role in preventing or responding to ER stress , which can otherwise lead to cell death ( Tabas and Ron , 2011 ) . However , whether ER stress contributes to aging is unknown . To address the nature and function of the different diffusion barriers in the yeast ER , we investigated the mechanisms underlying the compartmentalization of the cortical ER and their potential contribution to the retention of misfolded ER proteins in the mother cell .
To determine how misfolded secretory proteins distribute between mother and bud during yeast cell division , we detected them using a recently developed live cell microscopy assay based on fluorescence recovery upon photobleaching ( FRAP ) . Briefly , the endogenous copy of the ER Hsp70 chaperone BiP/Kar2 is expressed as a fusion protein with the super-folding green fluorescent protein ( sfGFP ) to form Kar2-sfGFP . The sfGFP is followed by the amino acids HDEL which are the yeast equivalent of the ER retention signal ( KDEL in animal cells ) , such as to maintain it in the ER-lumen . This protein is fully functional as a chaperone ( Lajoie et al . , 2012 ) . The diffusion of this Kar2-sfGFP , as recorded by FRAP , slows down as the levels of unfolded and aggregated proteins , with which it interacts , increase ( Lajoie et al . , 2012 ) . We exploited this assay to investigate the occupancy of Kar2-sfGFP with clients in the mother and bud compartments . In homeostatic wild type cells , the initial half time ( t1/2 ) of Kar2-sfGFP fluorescence recovery was 8 . 5 ± 2 . 3 s in the mother compartment and similarly , 7 . 7 ± 1 . 9 s in the bud ( Figure 1A , C ) . Consistent with the previous reports ( Lajoie et al . , 2012 ) , the mobility of Kar2-sfGFP dramatically decreased upon treating cells with a lethal dose of tunicamycin ( Tm , 1 µg/ml ) , an inhibitor of N-linked glycosylation , in both mother and bud ( t1/2 = 25 . 6 ± 9 . 8 s and 23 . 4 ± 10 . 1 s , respectively , after 2 hr , data not shown ) . However , when cells where treated with a non-lethal dose of the drug but for a longer time ( 0 . 5 µg/ml for 2 hr ) , to let misfolded proteins progressively accumulate , the situation was quite different ( Figure 1B ) . Kar2-sfGFP diffusion slowed down with the incubation time in tunicamycin ( t1/2 = 15 . 1 ± 6 . 8 after 1 hr and t1/2 = 20 . 7 ± 9 . 7 s after 2 hr , Figure 1C ) . The fraction of fast particles derived from the FRAP curves dropped from 65 . 7% in untreated mother cells to 19 . 0% in mother cells treated for 2 hr with Tm . Whereas Kar2-sfGFP diffusion was clearly slowed down in the mother cell , it remained quick in the bud ( t1/2 = 11 . 4 ± 4 . 8 s after 1 hr and t1/2 = 11 . 1 ± 4 . 3 s after 2 hr , Figure 1C ) , indicating that misfolded proteins accumulated slowly in the mother cell and did not invade the bud . 10 . 7554/eLife . 01883 . 003Figure 1 . Misfolded proteins are retained in the mother cell during budding yeast division . ( A ) FRAP analysis of Kar2-sfGFP in wild type mothers ( n = 29 cells ) and buds ( n = 24 cells ) under normal growth condition and ( B ) in the presence of 0 . 5 µg/ml of Tm . Representative cells are shown . Rectangles indicate the bleached areas . Image show cell before bleaching ( pre-bleach ) , immediately after bleaching ( 0 s ) and 10 s after the initial bleach ( 10 s ) . T1/2 represents the time it took to recover 50% of fluorescence of the reached plateau . Graphs of fluorescence recovery of the corresponding cells are shown . ( C ) Graph indicates the distribution of t1/2 for individual cells of each mutant tested in the presence ( + ) or absence ( − ) of 0 . 5 µg/ml Tm ) n >25 cells . n . s . = not significantly different ***p<0 . 001 ( t test ) . ( D ) Z-stack projections of CPY*-GFP and corresponding DIC images in wild type cells are shown . More images are shown in Figure 6 . ( E ) Quantification of the FRAP of CPY*-GFP . ( F ) Single focal plane of Sec61-GFP and Sec61-2-GFP in wild type or ubc7Δ mutant cells . Note that the gray values have been scaled such as to be able to see both proteins . Otherwise , Sec61-2-GFP values are 5–8-fold lower than wild type Sec61-GFP . ( G ) Distribution of asymmetry indices of Sec61-GFP and Sec61-2-GFP in wild type or ubc7Δ mutant cells ***p<0 . 0001 ( t test ) . Average ± SD are indicated ( A , B , C , E , G ) . Scale bars = 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01883 . 003 Similar results were obtained when we performed the same assay in cells lacking either the ER quality control lectin Yos9 or its binding partner , the membrane protein Hrd3 . T1/2 in the mother part of yos9Δ mutant cells was 13 . 3 ± 4 s and 18 . 8 ± 7 . 4 s in hrd3Δ mutant cells ( Figure 1C ) , showing that Kar2-sfGFP dynamics were significantly slower than in the mother part of unstressed wild type cells ( t1/2 = 8 . 1 ± 2 . 2 s ) . As observed with tunicamycin ( Tm ) , t1/2 of Kar2-sfGFP was comparable in the buds of the yos9Δ ( 7 . 3 ± 2 . 9 s ) and hrd3Δ mutant cells ( 8 . 4 ± 2 . 3 s ) and those of untreated wild type cells ( 8 . 0 ± 2 . 0 s , Figure 1C ) . Therefore , impairing protein quality control in the ER by deleting YOS9 or HRD3 leads to a measurable increase in misfolded proteins in the mother part of the cell over time , but little in the bud . Since the yos9Δ and hrd3Δ mutant cells are also defective in ER associated degradation of misfolded proteins ( ERAD ) , this data excludes the possibility that the low abundance of misfolded proteins in the bud compared to the mother is due to a more potent ERAD pathway in the bud . Since the data above suggested that misfolded proteins are retained in the mother part of budded yeast cells , we wanted to test this possibility more directly through visualization of such proteins . Thus , we next characterized the localization of CPY* , a mutated form of the vacuolar carboxypeptidase Y ( CPY ) that fails to properly fold and exit the ER . For visualization , we overexpressed GFP-tagged CPY* using the GAL1-promoter and monitored the fate of the protein by time-lapse microscopy . Over time , CPY*-GFP accumulated up to high levels in mother cell , but not in their buds . Even cells that underwent budding with high levels of CPY*-GFP in the ER in the mother cell showed no or very low levels of GFP signal in their growing bud ( Figure 1D , see also Figure 6C , upper panel ) . Prolonged over-expression of CPY*-GFP lead to the formation of strongly fluorescent dots in the mother cell that progressively became static . In order to monitor the dynamics of these aggregates , we performed a FRAP experiment on these cells . The aggregates did recover some fluorescence ( from 37% ± 5 . 4% 10s after bleaching to 53% ± 10 . 4% 5 min after bleaching , n = 10 cells , Figure 1E ) . Therefore , the dynamics of the aggregates themselves is rather slow . Following the cells that underwent budding , we observed that the CPY*-GFP dots remained in the mother cell ( Figure 6C ) . The buds started to form dots themselves only after they had separated from their mother . Thus , no aggregate or otherwise misfolded material appeared to diffuse from the mother cell into its bud . As a second reporter , we chose the mutant protein Sec61-2 , a resident ER protein inserted in the ER-membrane . At high temperature , Sec61-2 misfolds and is rapidly degraded . We tagged this allele with GFP and followed its localization . Its levels were already low at permissive temperature , and rapidly decreased upon incubation at 37°C as already shown for the untagged mutant protein ( Bordallo et al . , 1998 ) . We rationalized that even at permissive temperature , Sec61-2 is significantly misfolded . Therefore , we compared the segregation of Sec61-2 protein to the wild type , correctly folded , Sec61 protein in cells grown at 30°C ( Figure 1F ) . We measured the cortical ER mean fluorescence intensity in the mother cell ( Im ) and its growing bud ( Ib ) , choosing the frame before anaphase as a reference point and calculated its asymmetry index ( ( Im−Ib ) / ( Im + Ib ) ) . This index had a value of 0 . 00 ± 0 . 09 ( n = 52 cells ) for wild type Sec61-GFP showing that Sec61-GFP distributes equally between the cortical ER of the mother cell and its bud ( Figure 1G ) . However , in cells expressing Sec61-2-GFP the asymmetry index reached a significantly higher value ( 0 . 12 ± 0 . 09 , n = 49 cells; p<0 . 0001 t test , Figure 1F–G ) . We then used a strain in which Sec61-2-GFP is a bit more stable , since the ERAD gene UBC7 is deleted . In ubc7Δ mutant cells the asymmetry index of Sec61-2-GFP reached an even higher value ( 0 . 27 ± 0 . 14 , n = 71 cells ) than in the single Sec61-2-GFP mutant cells ( Figure 1F , G , p<0 . 0001 t test ) , consistent with a stabilization of misfolded Sec61-2 proteins and their specific retention in the mother cell . Unlike for CPY*-GFP , no large dots or aggregates were observed . Therefore , we conclude that misfolded Sec61-2 is asymmetrically retained in the mother compartment , and that retention does not require aggregation . Collectively , these results indicate that misfolded proteins are indeed not shared equally between mother and bud , but retained in the mother cell . We next wondered whether retention of misfolded ER-proteins in the mother cell relied on the lateral diffusion barrier in the ER membrane at the bud neck ( Luedeke et al . , 2005 ) . To determine whether confinement of ER stress to the mother cell required ER compartmentalization , we next set out to characterize in more details how the ER diffusion barrier is established . Based on its published genetic interaction with the septin mutation cdc3-1 ( Costanzo et al . , 2010 ) , we discovered that the sphinganine C4-hydroxylase Sur2 is required for proper barrier formation . Evidence for ER compartmentalization was obtained using fluorescence loss in photobleaching ( FLIP ) ( Ellenberg et al . , 1997 ) as previously described ( Luedeke et al . , 2005 ) , using the ER membrane protein Sec61-GFP as a reporter . Upon repeated photobleaching of a small area of the cortical ER in the mother cell , fluorescence loss elsewhere in the cell is proportional to the exchange rate of the reporter protein between the considered and the photobleached area . As reported for other ER-membrane proteins ( Luedeke et al . , 2005 ) , Sec61-GFP fluorescence is rapidly depleted throughout the wild type mother cell ( Figure 2 , red curves ) , but not in her bud ( green curves ) , indicating that exchange through the bud neck is reduced . In contrast , in the same assay the luminal reporter protein GFP-HDEL is lost with nearly the same kinetics in the bud as in the mother cell ( Luedeke et al . , 2005; Boettcher et al . , 2012 ) , demonstrating that the barrier is specific for membrane proteins . 10 . 7554/eLife . 01883 . 004Figure 2 . Sec61-GFP compartmentalization depends on sphingolipids and Bud1 module . ( A–C ) FLIP of Sec61-GFP in wild type and indicated mutant cells . Fluorescence level decay over time in the mother ( red ) and daughter part ( green ) are shown . BI = barrier Index , corresponding to the time of 50% fluorescence loss in the bud divided by the time of 50% fluorescence loss in the mother cell . ( D ) Graph indicates BI + SD of tested strains . n > 20 cells . ***p<0 . 001 , *p<0 . 05 ( t test ) . Scale bars = 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01883 . 004 To quantify this phenomenon , we used the ‘barrier index’ ( BI , Luedeke et al . , 2005 ) , defined as the t50 ( time to lose 50% of fluorescence ) in the bud divided by the t50 in the mother ( Figure 2A , D ) . The value of this index increases when exchange between mother and bud becomes limited . In wild type cells , the BI was 7 . 5 ± 2 . 3 . In contrast , in sur2Δ mutant cells Sec61-GFP compartmentalization was reduced , with a BI of 3 . 5 ± 1 . 1 ( Figure 2A , D ) . Based on this , we rationalized that sphingolipids may contribute to barrier formation . To test this hypothesis , we used the same assay to characterize the involvement of other genes required for sphingolipid biosynthesis . We disrupted the serine palmitoyl-transferase Lcb1 ( required for the first step of sphingolipid biosynthesis ) by either using a temperature sensitive allele ( lcb1-100 ) or the Lcb1 specific inhibitor Myriocin , the ceramide phosphoinositol transferase Aur1 using the inhibitor Aureobasidin A and the Inositolphosphotransferase Ipt1 ( involved in the synthesis of mannose- ( inositol-P ) 2-ceramide ) through deletion of its gene . In all these cases , Sec61-GFP compartmentalization was strongly affected ( BI between 3 . 6 and 4 , Figure 2D ) . In contrast , the ER diffusion barrier was not significantly affected upon inhibition of the glycerolipid biosynthesis pathway ( gpt2Δ , sct1Δ , BI = 6 and 5 . 8 , respectively , Figure 2B , D ) , the phospholipid biosynthesis pathway ( cho2Δ BI = 5 . 8 ) , the lipid-linked oligosaccharide biosynthesis pathway ( alg3Δ BI = 5 . 9 , Figure 2B , D ) or the ergosterol biosynthesis pathway ( erg3Δ , erg5Δ BI = 6 . 2 and 6 . 3 , respectively , Figure 2D ) . Thus , we concluded that formation of the diffusion barrier in the cortical ER at the bud neck specifically required the sphingolipid biosynthesis pathway to be functional . Previous work on the ER diffusion barrier showed that it is established as soon as the bud emerges ( Luedeke et al . , 2005 ) . It is therefore important that factors required for its establishment localize early on to the bud site . Therefore , we speculated that factors acting in early polarization events , leading to bud emergence , might also promote barrier formation . The earliest event in bud formation consists in the selection of the future site of bud emergence and the recruitment of the polarization machinery . This is achieved by the GTPase Bud1/Rsr1 ( Bender and Pringle , 1989 ) , which is activated at the future site of bud emergence and recruits and activates Cdc24 , the guanine-nucleotide exchange factor ( GEF ) for the Rho-related GTPase Cdc42 , a broadly conserved regulator of cellular polarity in eukaryotes ( Chant et al . , 1991; Powers et al . , 1991; Bender , 1993; Park et al . , 1993 , 2002; Zheng et al . , 1995; Shimada et al . , 2004 ) . At the plasma membrane , Bud1 is activated by its own GEF , Bud5 , which is deposited at the site of the previous cytokinesis through its interaction with the septin cytoskeleton . Upon bud emergence Bud5 re-localizes to the neck of the newly growing bud , where it stays throughout bud growth and where its function is unknown . The Bud2 protein forms the GTPase activating protein ( GAP ) involved in Bud1 inactivation . Due to the ability of the Cdc24/Cdc42 module to auto-activate , bud1Δ cells still form buds , but at random positions . To determine whether Bud1 is involved in the formation of the ER diffusion barrier , we assayed Sec61-GFP compartmentalization using FLIP in bud1Δ , bud2Δ or bud5Δ mutant cells ( Figure 2C , D ) . Deletion of either of BUD1 and BUD5 strongly abolished the compartmentalization of Sec61-GFP ( BI bud1Δ = 2 . 4 ± 1 . 3 , bud5Δ = 3 . 5 ± 1 . 3 ) . The deletion of BUD2 had a milder effect ( BI = 4 . 8 ± 1 . 9 ) . Consistent with these results , when Sec61 was tagged with the photoconvertible protein EOS very little of the protein photoconverted in the mother cell diffused into the bud of wild type cells within a 2 min chase ( Figure 3A , B ) . In contrast , up to three times more of the protein leaked into the bud already within the first 40 s of the experiment when photoconversion was carried out in the mother of bud1Δ mutant cells ( Figure 3A , B ) . We concluded that Bud1-GTP contributes to the formation of the diffusion barrier in the ER-membrane at the bud neck cortex . 10 . 7554/eLife . 01883 . 005Figure 3 . Sec61-EOS is compartmentalized in a Bud1-dependent manner . Photoconversion of Sec61-tdEOS in wild type and bud1Δ mutant mother cells . ( A ) Images show the red and green fluorescent micrograph before conversion and after 20 s and 60 s . Scheme indicates the converted area . ( B ) Graph of normalized red fluorescence intensity measured in the bud compartment . n > 15 cells , error bars depict the standard error of the mean ( SEM ) . **p<0 . 01 , *p<0 . 05 ( t test ) . Scale bars = 2 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 01883 . 005 To test whether Bud1 promoted barrier formation through its interaction with Cdc24 or through another effector , we next investigated whether interrupting the interaction between Bud1 and Cdc24 was sufficient to mimic the effect of Bud1 inactivation . A CDC24 mutant allele , cdc24-4 , abolishes this interaction at room temperature ( Shimada et al . , 2004 ) , a condition under which it is otherwise fully functional , that is , promotes normal cellular polarization and bud emergence , albeit at a random position . Thus , we tested whether the cdc24-4 mutant cells still compartmentalized the ER membrane properly under this regime . In addition , we used the cdc42-1 temperature sensitive allele to test whether Cdc42 contributed to barrier formation . FLIP experiments were carried out using Sec61-GFP as a reporter in the cdc24-4 mutant cells at room temperature and in the cdc42-1 mutant cells shifted to the restrictive temperature ( 37°C ) after bud emergence ( Figure 2C , D ) . Both , cdc24-4 ( BI = 2 . 8 ± 1 . 1 ) and cdc42-1 mutations ( BI = 2 . 5 ± 1 . 4 ) abolished barrier assembly in the cortical ER-membrane to the same extent as BUD1 inactivation . Taken together , our results indicate that Bud1 mediates the formation of the diffusion barrier in the cortical ER essentially through regulation of Cdc24 and activation of the GTPase Cdc42 . After identifying the signaling GTPases Bud1 and Cdc42 and sphingolipids as major players in the assembly of the diffusion barrier in the cortical ER , we next wondered whether these factors generally functioned in the assembly of diffusion barriers . Thus , we asked whether these molecules also contributed to the assembly of the diffusion barrier in the nuclear envelope ( Shcheprova et al . , 2008; Boettcher et al . , 2012 ) . We performed FLIP on either the nuclear pore complex component Nup49 or the outer nuclear membrane protein Nsg1 , each tagged with GFP . For both reporters , repeated photobleaching in a small area of early anaphase nuclei in the mother cell caused the rapid and complete depletion of fluorescence in the entire mother part of the nucleus . In contrast , the fluorescence in the bud part of the anaphase nucleus of wild type cells remained high ( BI = 23 . 2 ± 5 . 4 for Nup49 , Figure 4A and 9 . 5 ± 3 . 7 for Nsg1 , Figure 4B ) as previously reported ( Shcheprova et al . , 2008 ) . In the sur2Δ mutant cells , the fluorescence loss in the bud was significantly faster than in wild type cells ( BI = 12 ± 1 . 4 for Nup49 , Figure 4A and 4 . 9 for Nsg1 , Figure 4B ) . Remarkably , the bud1Δ mutation affected the compartmentalization of neither Nup49 nor Nsg1 ( BI = 26 . 1 ± 13 . 1 and Nsg1-GFP 10 . 9 ± 2 . 6 , respectively , Figure 4A , B ) . Similarly , the cdc24-4 mutation did not affect nuclear envelope compartmentalization at room temperature ( BI = 9 . 1 ± 2 . 4 for Nsg1 , Figure 4B ) . Therefore , the Bud1 signaling module functions in the assembly of the barrier in the cortical ER specifically , whereas sphingolipids are involved in the formation of barriers in both the cortical ER and the nuclear envelope . 10 . 7554/eLife . 01883 . 006Figure 4 . Compartmentalization of Nup49-GFP and Nsg1-GFP depends on the sphingolipids , but not on BUD1 . ( A and B ) FLIP experiments and BI values for the indicated markers of the nuclear envelope during early anaphase , in the cells of indicated genotype . Fluorescence level decay over time in the mother ( red ) and daughter part ( green ) are shown . White lines indicate cell outlines . Representative experiments are shown . n > 20 cells . ***p<0 . 001 ( t test ) . Scale bars = 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01883 . 006 Taking advantage of this information , we asked whether the retention of ER stress in the mother cell required the ER diffusion barrier . In order to determine this , we first analysed the dynamics of Kar2-sfGFP in sur2Δ and bud1Δ mutant cells . The t1/2 of Kar2-sfGFP in mother and buds of these mutant cells was very similar to wild type values ( 8 . 2 ± 2 . 1 s and 7 . 7 ± 2 . 6 s in sur2Δ; 7 . 6 ± 2 . 6 s and 7 . 4 ± 2 . 5 s in bud1Δ mutant cells; Figure 5A , B ) , indicating that neither of these mutations caused misfolding or ER stress on their own . As reported earlier , treatment with Tm ( 2 hr , 0 . 5 µg/ml ) slowed down Kar2 dynamics over time in the mother cell much more than in the daughter cell ( Figure 1B ) . Similarly , Kar2-sfGFP mobility strongly decreased in the sur2Δ and bud1Δ mutant cells treated with Tm ( in mother cells t1/2 = 26 . 5 ± 8 . 1 and 25 ± 10 . 6 , respectively; Figure 5A , B ) . However , the asymmetry of Kar2-sfGFP between mother and bud was lost , the mobility of Kar2-sfGFP being considerably slowed down in the bud as well ( sur2Δ = 19 . 7 ± 7 . 4 and bud1Δ =18 . 7 ± 8 . 8 , Figure 5A , B ) . Thus , inactivation of Sur2 and Bud1 did not affect the appearance of misfolded proteins and the emergence of ER stress in the presence of Tm , but caused misfolded proteins to distribute more equally between mother and bud . 10 . 7554/eLife . 01883 . 007Figure 5 . Retention of Kar2 clients in the mother cell depends on the ER diffusion barrier . ( A and C ) FRAP of Kar2-sfGFP in the mother or bud of mutants of indicated genotype in the presence or absence of 0 . 5 µg/ml tunicamycin . ( B and D ) Graphs indicate the distribution of individual t1/2 values measured in mutant cells of indicated genotype . Representative cells are shown . n > 20 cells . Averages ± SD are indicated . n . s . = not significantly different , ***p<0 . 0001 , *p<0 . 05 ( t test ) . Scale bars = 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01883 . 007 Similarly , deleting SUR2 or BUD1 in the yos9Δ mutant cells lead to the mobility of Kar2-sfGFP to increase symmetrically , that is in both mother and bud ( t1/2 = 11 . 5 ± 4 . 2 and 11 . 6 ± 3 . 5 respectively , in the yos9Δ sur2Δ double mutant cells; 11 . 8 ± 3 . 5 and 12 . 2 ± 3 . 5 in the yos9Δ bud1Δ double mutant cells , Figure 5D ) . Inactivation of BUD1 in the hrd3Δ mutant cells had the same effect ( t1/2 mother: 12 . 8 ± 3 . 1 , t1/2bud: 14 . 2 ± 4 . 7 , Figure 5C , D ) . To test whether weakening of the diffusion barrier affected the distribution of the misfolded protein CPY*-GFP between mother and bud as well , we expressed CPY*-GFP under the GAL1-promoter in the bud1Δ mutant cells and monitored the number of aggregates present in the buds ( Figure 6A ) . In the majority of wild type buds ( 53% , n = 32 cells ) , we did not detect any CPY*-GFP aggregates ( Figure 6B ) . In contrast , only 5% of bud1Δ mutant buds ( n = 36 cells ) were free of aggregates . More than 85% of the bud1Δ mutant buds had 1 to 5 dots , whereas 90% of wild type buds had maximum 2 dots ( Figure 6B ) . The increasing number of aggregates in the buds of these barrier-defective cells was not due to the diffusion of aggregates into the bud , since we never observed such an event , but to the fact that the buds started to form aggregates themselves much before they separated from their mother cell , that is , much earlier than wild type buds ( Figure 6C ) . These results indicate that wild type cells retained both CPY*-GFP aggregates and non-aggregated but misfolded CPY*-GFP in the mother cell . Whereas retention of the aggregates was not barrier-dependent , the retention of smaller , not yet aggregated entities was . 10 . 7554/eLife . 01883 . 008Figure 6 . The cER diffusion barrier retains misfolded CPY*-GFP in the mother cell . ( A ) Z-stack projections of CPY*-GFP and corresponding DIC images in wild type and bud1Δ mutant cells . ( B ) Graph shows the frequency of buds with indicated number of CPY*-GFP dots in wild type and bud1Δ mutant cells . ( C ) DIC and CPY*-GFP fluorescence images of time lapse microscopy of wild type ( up ) and bud1Δ mutant cells ( bottom ) . The arrow indicates the mother cell and the arrowhead the growing bud . Scale bars = 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01883 . 008 We then asked whether the retention of Sec61-2-GFP in the mother cell also depended on an intact diffusion barrier in the cortical ER . We followed single cells expressing Sec61-2-GFP by time-lapse microscopy . Upon UBC7 deletion Sec61-2-GFP is highly asymmetrically retained in the mother part of the cortical ER ( Figure 1G , Figure 7A , B ) . This retention was decreased upon the deletion of BUD1 ( asymmetry index of 0 . 07 ± 0 . 12 , n = 46 ubc7Δ bud1Δ Sec61-2-GFP mutant cells , p<0 . 0001 t test compared to ubc7Δ Sec61-2-GFP mutant cells , Figure 7A , B ) . Therefore , misfolded Sec61-2-GFP proteins were also retained in the mother cell in a Bud1-dependent manner . Together , our data indicate that aggregation and the compartmentalization of the ER by the diffusion barriers at the bud neck are required for the confinement of misfolded proteins into the mother cell . 10 . 7554/eLife . 01883 . 009Figure 7 . Misfolded Sec61-2-GFP asymmetry depends on the cER diffusion barrier . ( A ) Images of Sec61-2-GFP of a budding ubc7Δ ( up ) and a budding ubc7Δ bud1Δ ( bottom ) mutant cell . The arrow indicates the mother cell and the arrowhead the growing bud . Scale bar = 2 µm . ( B ) Distribution of asymmetry indices measured at the last frame prior to anaphase in individual cells of indicated genotype ( note that the quantification of Sec61-2-GFP ubc7Δ is the same as in Figure 1G ) . ***p<0 . 0001 ( t test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01883 . 009 The accumulation of misfolded proteins is toxic for the cell . Therefore , we wondered whether a mild ER stress may act as an aging factor . When pedigree analyses were carried out on plates containing very low and non-lethal levels of Tm ( 0 . 2 µg/ml ) , we observed dramatic effects on cellular longevity , which caused most cells to die after 2–4 generations . We therefore decided not to use Tm to study the replicative life span of cells with ER stress . Instead , we performed pedigree analysis of yos9Δ mutant cells . A clear shortening of cellular longevity was observed in theses cells ( median life span of 19 generations , Figure 8A ) compared to wild type cells ( median life span of 26 generations ) . Therefore , the accumulation of ER stress in the cells leads to the shortening of their replicative potential and ER stress is indeed an aging factor . 10 . 7554/eLife . 01883 . 010Figure 8 . Yos9 deletion shortens the life span of cells . ( A and B ) Pedigree analysis of wild type and mutant strains . Graph indicates the number of surviving cells after the indicated number of divisions . The median survival age for the tested strains is indicated n > 50 mother cells . p***<0 . 001 , **p<0 . 01 , *p<0 . 05 ( Gehan-Breslow-Wilcoxon test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01883 . 010 We then asked whether the diffusion barrier contributed to any substantial extent to the accumulation of ER stress in the mother cell with age , and the effect of ER stress on aging . Supporting this view , the yos9Δ bud1Δ double mutant cells ( 22 generations ) lived similarly long as the bud1Δ ( 23 generations ) and significantly longer than the yos9Δ single mutant cells ( Figure 8B ) . Likewise , the sur2Δ yos9Δ double mutant cells ( 28 generations ) lived nearly as long as the sur2Δ ( 30 generations ) and longer than the yos9Δ single mutant cells ( Figure 8B ) . Thus , weakening the cER diffusion barrier is sufficient to suppress the toxic effect of the YOS9 deletion on life span . However , we note that sur2Δ mutant cells are longer lived than wild type cells , whereas bud1Δ mutant cells are slightly shorter lived ( Figure 8A ) . We previously showed that weakening the ER and the nuclear envelope diffusion barriers through the deletion of the BUD6 gene extended life span , probably by passing on aging factors to the mother's cell progeny ( Shcheprova et al . , 2008 ) . Here , Sur2 is required for the formation of both barriers , hence for the retention of aging factors from both the ER and the nucleus , whereas Bud1 is only required for the ER barrier . Since sur2Δ mutant cells are long-lived and bud1Δ mutant cell slightly short lived , it seems that the ER diffusion barrier is not required for the retention of the aging factors limiting the life span of cells growing under optimal conditions . The ER diffusion barrier seems to become critical for aging only upon induction of ER stress . Together our data indicate that both the Bud1 module and sphingolipid biosynthesis contribute to the formation of the lateral diffusion barrier in the cortical ER , together with septins and Bud6 ( Luedeke et al . , 2005 ) . In order to determine how these players function together , we first investigated whether they formed a single genetic pathway or defined several pathways by determining whether their inactivation had synergistic effects on barrier strength . Measuring the barrier indices of the sur2Δ bud1Δ , sur2Δ bud6Δ and bud6Δ bud1Δ double mutants indicated that all of them showed a similar defect and the BI was not lower than the lowest BI due to the individual mutations ( BI = 2 . 6 ± 1 . 2 in bud1Δ , 3 . 9 ± 1 . 1 in sur2Δ , 3 . 8 ± 1 . 5 in bud6Δ , 3 . 3 ± 1 . 3 in sur2Δ bud1Δ , 3 . 5 ± 1 . 7 in sur2Δ bud6Δ , 3 . 5 ± 1 . 7 in bud6Δ bud1Δ single and double mutant cells , Figure 9 ) . Thus , sphingolipids , Bud1 and Bud6 functioned in the same genetic pathway . 10 . 7554/eLife . 01883 . 011Figure 9 . Bud1 , Bud6 and Sur2 act in the same pathway for barrier formation . Graph shows the BI measured by FLIP analysis of Sec61-GFP exchange between mother and bud in cells of indicated genotype , as in Figure 2 . BI + SD of tested strains are shown . n > 20 cells . ***p<0 . 001 , ( t test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01883 . 011 In order to position this pathway relative to septin function , we next investigated whether any of these proteins interfered with septin assembly . We characterized the localization of the septin Shs1 , fused to GFP , in sur2Δ , bud1Δ and bud6Δ single mutant cells . In all these cells , the localization and intensity of Shs1-GFP at the bud neck was comparable to wild type ( Figure 10A ) , suggesting that none of these factors acts upstream of septins . 10 . 7554/eLife . 01883 . 012Figure 10 . Epistasis analysis of the factors required for the assembly of the diffusion barrier in the cortical ER . ( A–C ) Fluorescence intensity of Shs1-GFP ( A ) , Bud5-GFP ( B ) and Bud6-GFP ( C ) was measured at the bud neck by taking z-stacks of the cells after deconvolution with Softworx software . n > 100 cells . Representative images ( right ) and average fluorescence intensities are shown ± SD ( arbitrary units , left ) . n > 20 cells . ***p<0 . 001 , ( t test ) . Scale bars = 2 µm . ( D ) Schematic drawing of the pathway required for the establishment of the diffusion barrier . DOI: http://dx . doi . org/10 . 7554/eLife . 01883 . 012 Similarly , we asked whether the septins , Bud6 and sphingolipids functioned upstream of the Bud1 module . Since Bud1 localization is diffuse , the localization of its GEF , Bud5 , is probably what determines where this GTPase is active . Therefore , we investigated whether any of the corresponding mutants affected the localization of Bud5 . Bud5-GFP fluorescent intensity at the bud neck was clearly reduced in shs1Δ mutant cells but was not affected in bud1Δ , sur2Δ and bud6Δ single mutant cells ( Figure 10B ) . To test whether Bud6 and Sur2 were acting redundantly on the localization of Bud5-GFP , we created the sur2Δ bud6Δ BUD5:GFP mutant strain . In these cells , the level of Bud5-GFP at the bud neck was comparable to the level observed in wild type cells ( Figure 10B ) . The localization of Bud5-GFP was also intact in the bud6Δ bud1Δ and sur2Δ bud1Δ double mutant cells ( Figure 10B ) . Thus , while septins control the localization of Bud5 and hence , the activation of Bud1 at the bud neck , Bud6 and Sur2 act downstream of Bud1 . Consistent with this interpretation , analysis of Bud6-GFP distribution indicated that proper Bud6 localization to the bud neck depends on Shs1 , as reported ( Luedeke et al . , 2005 ) , and on Bud1 function , but not on Sur2 ( Figure 10C ) . Double deletion of BUD1 and SUR2 did not influence the fluorescence intensity of Bud6-GFP more than the deletion of BUD1 alone , suggesting that Sur2 is downstream of Bud6 ( Figure 10C ) . Taken together , these studies place septins at the top of a signaling pathway involved in the assembly of the diffusion barrier in the ER-membrane at the bud neck , and Bud6 and sphingolipids at the bottom . Bud1/Cdc24/Cdc42 are likely to target Bud6 , which in turn regulates the sphingolipid-dependent formation of the diffusion barrier at the bud neck ( Figure 10D; ‘Discussion’ ) . Thus , Sur2 and hence the biosynthesis of sphingolipids , seemed to act most downstream in barrier formation . Furthermore , sphingolipid biosynthesis was among the only requisites conserved in both the cortical and the nuclear diffusion barriers . These observations suggested that the barrier might consist of a specialized lipid domain in the ER-membranes at the bud neck . We rationalized that the exclusion of the proteins Sec61 and Hmg1 from the bud neck region ( Luedeke et al . , 2005 ) might indicate that the ER-membrane composition is indeed distinct in this area . Therefore , we next wondered whether all ER membrane proteins were excluded from the ER-membrane at the bud neck . A set of such proteins , namely Lcb1 , Sur2 , Get1 and Ste24 was tagged with GFP in cells expressing the ER lumen marker dsRed-HDEL ( Onischenko et al . , 2009 ) , and their localization to the bud neck was characterized . Lcb1 and Sur2 are both required for the biosynthesis of sphingolipids ( Dickson , 2008 ) , Get1 is involved in the insertion of tail anchored proteins into the ER membrane ( Schuldiner et al . , 2008 ) and Ste24 is a conserved zinc metalloprotease required for alpha-factor peptide maturation and CAAX-box protein processing ( Boyartchuk et al . , 1997; Fujimura-Kamada et al . , 1997; Tam et al . , 1998; 2001 ) . All of these proteins were excluded from the bud neck region , whereas dsRed-HDEL was not ( Figure 11A ) . Therefore , all ER membrane proteins tested so far are excluded from the bud neck region . As previously shown ( Luedeke et al . , 2005 ) , dsRed-HDEL localization showed that the ER is a continuous organelle between the mother and bud . These results suggest that the ER-membrane at the bud neck differs from the rest of the ER . 10 . 7554/eLife . 01883 . 013Figure 11 . Exclusion of ER-transmembrane proteins at the cER bud neck . ( A ) localization of the indicated ER membrane resident proteins ( green ) and the ER luminal marker dsRed-HDEL ( red ) at the bud neck of the same cells . ( B ) Classification of Sec61-GFP exclusion from the bud neck in three groups: total exclusion ( green ) , partial exclusion ( grey ) and no exclusion ( red ) . ( C ) Evaluation of Sec61-GFP exclusion from the bud neck in wild type and mutant cells of indicated genotype , using the classification above . Arrowheads indicate exclusion , arrows indicate no exclusion . Representative cells are shown . n > 100 cells . Scale bars = 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01883 . 013 We next rationalized that sphingolipids might help barrier formation by forming this specialized membrane domain . To test this possibility , we asked whether disruption of sphingolipid biosynthesis affected the exclusion of trans-membrane proteins from the ER membrane at the bud neck . Therefore , we characterized the localization of Sec61-GFP at the bud neck in various barrier mutants . The phenotypes observed for individual cells were sorted in three categories ( Figure 11B ) . In the group 1 ( green ) Sec61-GFP was totally excluded from the bud neck . At the other extreme , Sec61-GFP localization was continuous at the bud neck periphery in at least one side of the neck in the cells from the group 3 ( red ) . The group 2 ( grey ) consisted of the cells with intermediate phenotypes . In wild type cells , 81% of cells totally excluded Sec61-GFP from the bud neck ( Figure 11B ) . The shs1Δ mutant cells showed the most prominent loss of Sec61-GFP exclusion from the bud neck ( ∼5%; Figure 11C ) . The bud1Δ , bud5Δ , cdc24-4 and lcb1-100 single mutant cells formed the second strongest set of mutants , showing a very strong defect in Sec61-GFP exclusion ( between 27–35% , Figure 11C ) , closely followed by the sur2Δ and bud6Δ mutant cells ( 44–47% exclusion ) . In contrast , the lipid synthesis mutations alg3Δ and gpt2Δ did not affect Sec61-GFP localization much ( 69–70% exclusion , Figure 11C ) . Therefore , formation of the exclusion zone in the ER-membrane at the bud neck is the most downstream event in the pathway of barrier formation , and the only event to depend on sphingolipid biosynthesis . We concluded that sphingolipids contribute to building a specialized ER membrane domain at the bud neck , in a Bud1- , Cdc42- and Bud6-dependent manner . Taken together , our data suggested that the barrier may consist of a specialized lipid domain in the ER membrane at the bud neck , possibly involving sphingolipids . Thus , we next wondered whether we could observe a difference in lipid composition for the ER-membrane at the bud neck . Unfortunately , we were unable to identify a dye able to reliably visualize sphingolipids in yeast membranes . Instead we used the DiOC6 ( 3 , 3′-dihexyloxacarbocyanine iodide ) fluorescence dye , a cell-permeant , green-fluorescent dye that stains initially the ER membrane by binding to the hydrophilic head group of lipids before being transported to the mitochondria ( Terasaki , 1989 ) . Strikingly , DiOC6 was excluded from the bud neck region of metaphase cells similarly to ER-membrane proteins ( Figure 12A ) . Similar to Sec61-GFP , total exclusion of DiOC6 from the bud neck was most strongly affected in the shs1Δ septin mutant cells ( wt = 77% , shs1Δ = 10% exclusion ) , followed by the lcb1-100 , and bud6Δ single mutant cells ( 38% , 40% exclusion , respectively , Figure 12B ) . The sur2Δ , bud1Δ , bud5Δ and cdc24-4 mutations caused a slightly weaker phenotype ( between 46–50% exclusion ) , whereas the alg3Δ and gpt2Δ mutations showed the weakest or no phenotype ( between 64–73% exclusion , Figure 12B ) . Thus , the entire barrier pathway was required for the exclusion of the lipid dye DiOC6 from the ER membrane at the bud neck , including sphingolipids , supporting the conclusion that barrier formation involves the assembly of a specialized membrane domain of unique lipid composition in the ER membrane at the bud neck . 10 . 7554/eLife . 01883 . 014Figure 12 . The composition of the cER and Nuclear membrane at the bud neck differs from elsewhere . ( A ) A representative metaphase cell stained with the dye DiOC6 and images at different z-positions through its bud neck are shown . ( B ) Evaluation of DiOC6 exclusion at the bud neck in wild type and mutant cells according to the classification used for Sec61-GFP in Figure 11B . ( C–D ) Characterization of the exclusion of Nup49-GFP ( C ) and DiOC6 ( D ) from the bud neck in early anaphase nuclei of cells of indicated phenotype using the same classification principle as in Figure 11B . Representative cells illustrated . n > 100 cells . Scale bars = 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01883 . 014 Since our data established that formation of the barrier in the nuclear envelope also depends on sphingolipid biosynthesis , we wondered whether a specialized lipid domain of the outer nuclear membrane formed this second barrier as well . Analysis of nuclear pore distribution , using Nup49-GFP as a reporter , indicated that in wild type , anaphase cells the abundance of NPCs was indeed reduced in the nuclear envelope at its intersection with the plane of the bud neck ( 67% exclusion , Figure 12C ) . Exclusion depended on Bud6 and Sur2 ( 27% and 35% exclusion , respectively ) , but not on Bud1 ( 67% exclusion , Figure 10C ) . Similarly , a fainter , but significant effect was also observed for DiOC6 staining in wild type , anaphase cells ( Figure 12D ) . Reduced DiOC6 staining in the nuclear envelope at the bud neck ( 64% of wild type cells ) partially depended on Bud6 and Sur2 function ( 38 and 50% respectively ) , but not on Bud1 ( 68% , Figure 12D ) . These observations are consistent with Bud6 and sphingolipids being required for the formation of a specialized membrane domain at the location of the barrier in the nuclear envelope . They are also fully consistent with our data establishing that Bud1 is not involved in the assembly of the diffusion barrier in the outer nuclear membrane .
While mother cells age with each division , the buds are born with a full life-span expectancy . Replicative aging of the mother is thought to be caused by the accumulation of damages , which they do not share with their daughters . However , what these different factors are and how they are retained in the mother cell is unclear . Here , we show that misfolded ER-proteins are retained in the mother cell by two distinct mechanisms: the formation of nearly immobile aggregates , and the retention of smaller entities by the diffusion barrier in the cortical ER membrane at the bud neck . Interestingly , the fact that the barrier in the ER membrane helps retaining individual , misfolded proteins in the mother cell , although many of them , including CPY* , are luminal , indicates that an unidentified receptor may anchor them in the membrane . Under optimal growth conditions , wild type cells appear to contain only little misfolded proteins and ER stress does not appear to be the main cause of their demise with age . Indeed , if ER stress were a constitutive cause of aging , disruption of the barrier in the cortical ER should systematically increase the longevity of the cells , which is not the case . However , low concentrations of tunicamycin or affecting protein quality control increased the load of misfolded proteins , and accelerated aging . Allowing mothers to pass stress on to their daughters by down regulating the ER diffusion barrier diluted this stress , as shown by the fact that it reduced the load in the mother cells and restored the life span of the yos9Δ mother cells . Thus , retention of ER-misfolded proteins in the mother cell caused stress and accelerated aging . ER stress may therefore be considered as a conditional aging factor . This might be particularly the case in highly secreting cells , such as beta-pancreatic cells in mammals ( Kennedy et al . , 2010 ) . Interestingly , affecting all barriers by disrupting the general barrier assembly factors Bud6 and Sur2 extends the longevity of the cell even under optimal growth conditions . Since this effect is not due to loss of the barrier in the cortical ER , we assume that it could be due to the loss of the nuclear barrier , as previously suggested ( Shcheprova et al . , 2008 ) . In this study we confirm that the septin ring at the bud neck lies upstream in the barrier formation pathway , at least in the cortical ER . But it was unclear how septins , which are closely apposed to the plasma membrane , promote formation of a barrier on both sides of the ER lumen . Two observations may help solving this conundrum . First , the role of septins in barrier assembly is partly indirect . The fact that the GTPases Bud1 and Cdc42 contribute to barrier formation downstream of septins suggests that the localization signal originating at the plasma membrane spreads from there to specialize the ER-membrane at the bud neck ( Figure 13 ) . This signal regulates the assembly of a specialized membrane domain , which itself functions as a barrier . However , none of the mutants characterized shows as strong a phenotype as the septin mutant shs1Δ . Therefore , septins might function above more than only one pathway in barrier assembly . Future studies will be needed to address this possibility . Together , our findings suggest that the pathway for barrier formation consists of three successive layers . At the top , the septins provide a spatial cue for where the barrier should form , functioning as a scaffold for regulatory elements . Downstream of septins , a second layer transduces the localization signal to effectors . This module depends at least in part on GTPases , namely Bud1 and Cdc42 in the case of the cortical ER barrier . Unlike during bud-site selection , Bud1’s role in barrier formation might not be restricted to the early times of bud emergence . Indeed , Bud1’s activator , Bud5 , remains at the bud neck throughout bud growth . Thus , barrier assembly might be a continuous process , also ensuring barrier maintenance . The second layer is specific to the barrier under consideration . Whereas the Bud1/Cdc42 module controls the cortical barrier , we do not know at this time which signaling molecules govern the formation of the barrier in the nuclear envelope . Downstream of this second layer , a third component of the pathway is more directly involved in barrier formation . This layer appears to be conserved between the different barriers in the ER , supporting the idea that it plays a structural role . This is where we find the protein Bud6 , a target of Cdc42 ( Jaquenoud and Peter , 2000 ) , and the sphingolipids . Future characterization of this layer will provide biophysical understanding about how barriers function . Recent electron tomography studies characterized the yeast ER in detail and confirmed that ER is present in the bud-neck and that the yeast ER is physically continuous throughout the cell ( West et al . , 2011 ) . However , using morphological criteria , these studies failed to observe ER at the cortex of the bud neck . Although the luminal marker dsRed-HDEL confirms the presence of the ER at the cortex of the bud neck , Sec61 ( Luedeke et al . , 2005 ) and all the ER-membrane protein that we visualized , as well as the lipid dye DiOC6 are excluded from the ER in this area . The poor staining of the ER membrane at the bud neck by DiOC6 , relative to the rest of the ER , suggests that the lipid composition of the membrane differs at that position . This could explain why the ER-membrane at the cortex of the bud neck was not identified in the tomography data . The fact that sphingolipids play such an important , specific role at the bottom of the pathway for the formation of all ER barriers makes them prime candidates for structural elements of the barrier . Complex sphingolipids are thought to reside primarily in the plasma membrane , yet many of the enzymes involved at least in the first steps of their biogenesis localize to the ER , including Lcb1 and Sur2 . Thus , it would be in principle possible that an ER-specific pool of sphingolipids is directly involved in barrier formation . If this were the case , it would explain why DiOC6 is excluded from the ER-membrane in the barrier areas . DiOC6 stains membranes essentially through interaction with hydrophilic heads of lipids , and shows little affinity for sphingolipid-rich membranes . Thus , DiOC6 exclusion from the ER membrane at the bud neck would be compatible with sphingolipid accumulation at this location . Sphingolipids are characterized by long side chains . The idea that such lipids form a core component of lateral diffusion barriers is attractive for several reasons . The long fatty acid chains would promote the formation of a thicker lipid bilayer . This would explain why proteins with transmembrane domains tailored to fit the average thickness of the ER membrane are excluded from the barrier area . Furthermore , long-chain sphingolipids are thought to be prone to form liquid ordered domains ( Aresta-Branco et al . , 2011 ) , promoting their separation from other lipids . In this area , the membrane would be predicted to be less fluid and to disperse only slowly , two ideal conditions for the formation of a localized barrier . However , it is very unlikely that this domain would be stable by itself . Therefore , it is likely that the barrier must be continuously reassembled and thus would comprise protein components to stabilize it . Bud6 is an excellent candidate for such a function . Indeed , its localization to the bud neck does not depend on sphingolipids , yet Bud6 , which is a peripheral membrane protein ( Jin and Amberg , 2000 ) , acts in forming the specialized membrane domain . Thus , one of Bud6 function may be to stabilize a sphingolipid domain in the ER membrane at the bud neck , probably together with other not yet identified proteins . Actin would be here an excellent candidate , given the role of Bud6 in actin cable formation and actin binding . However attractive this model is , we have so far no dye that stains sphingolipids in yeast , such that we are not in the position yet to determine whether sphingolipids indeed accumulate in the ER-membrane at the bud neck . Identifying sphingolipid-binding proteins will probably be necessary in order to solve this issue . Future studies will also need to identify the exact identity of the sphingolipids involved . Taken together , our data provide molecular insights into how lateral diffusion barrier might assemble and function in eukaryotic membranes . These insights set the stage for investigating how barriers function at the biophysical level , and whether their mechanism is conserved in other organisms . Indeed , it is remarkable that diffusion barriers of the plasma membrane are highly conserved , on a molecular level as demonstrated by the involvement of septins in many of them . However , we still know little about whether the ER is also compartmentalized in other eukaryotes . Based on the function of ER barriers in yeast , namely the asymmetric segregation of damages and aging determinants , it is tempting to assume that similar barriers will be found to play similar roles in a wide variety of other cell types , and particularly in cells that show strong self renewal .
All yeast strains were constructed according to standard genetic techniques and are isogenic to S288C , unless stated otherwise . The different deletions were obtained from the EUROSCARF deletion collection ( Frankfurt , Germany ) and provided to us by M Peter ( ETH , Zurich , Switzerland ) . All experiments were performed with three independent clones . All cultures or plates were grown at 30°C . For all FLIP and FRAP experiments , fresh cells were grown at 30°C on YPD ( yeast , peptone , and 2% dextrose ) plates , resuspended in synthetic complete medium , and immobilized on a 2% agar pad containing synthetic complete medium . The cells were imaged on a confocal microscope ( LSM 510; Carl Zeiss , Jena , Germany ) with a Plan Apochromat 63 × /1 . 4 NA oil immersion objective , using 2% of argon laser intensity ( 488 nm line ) at 40% output . The ZEN 2010 software ( Carl Zeiss ) was used to control the microscope . GFP emission was detected with a 505 nm long pass filter . Photobleaching was applied on a region of interest ( ROI ) as indicated in the figures . For FLIP experiments , the bleaching regions were irradiated with 80 iterations at 80% laser power and 40% transmission over a period of 40 frames . For FRAP experiments , bleaching was applied once with 60 iterations in the bud compartment or 70 iterations in the mother compartment at 80% laser power and 40% transmission . All photobleaching experiments were performed at 30°C , unless stated otherwise . FLIP quantification were performed using ImageJ 1 . 42q ( National Institutes of Health ) . The mean fluorescence signal was quantified in the entire mother cell , the entire bud and in the entire mother of five neighboring cells . After background subtraction , the fluorescence signals of the mother and bud were normalized to the mean of the five control cells and set to 100% at beginning of the experiment . All experiments were pooled and transferred to Prism 5 . 0b ( GraphPad Software , La Jolla , California ) , in which a one-phase decay curve constraining the first bleaching point to 100% was fitted . The Barrier Index ( BI ) was defined as the ratio of the times needed to lose 50% of the fluorescence signal in the bud over the mother . All images shown in the figures were processed using ImageJ 1 . 42q . FRAP quantification were also performed using ImageJ 1 . 42q . The mean fluorescence recovery signal was quantified in the bleached ROI in either the mother or bud compartment and as a control the same size region was measured in five neighboring cells . After background subtraction , the fluorescence signals of the mother and bud were normalized to the mean of the five control cells and set to 100% at beginning of the experiment . All experiments were pooled and transferred to Prism 5 . 0b and on an exponential FRAP curve , the mobile fraction was measured by determining the half time ( t50 ) of fluorescence recovery to reach a plateau level . To determine the exclusion of Sec61-GFP , dsRed-HDEL , DiOC6 and Nup49-GFP , we took 30 z-stacks ( 0 . 2 µm steps ) on a Deltavision microscope ( Applied Precision , GE Healthcare company , Issaquah , Washington ) equipped with a CoolSNAP HQ2 camera ( Photometrics ) and an Olympus ( Tokyo , Japan ) 100 × oil immersion objective and the GFP/Cherry filter set . A GC400 filter was used to eliminate UV . The microscope was controlled by the SoftWorks software . All quantifications were performed using ImageJ software . To determine the fluorescence intensity of bud neck protein for the epistasis experiments , we took 30 z-stacks ( 0 . 2 µm steps ) on the Deltavision microscope ( Applied Precision ) equipped with a CoolSNAP HQ2 camera ( Photometrics , Tucson , Arizona ) and an Olympus 100 × oil immersion objective and the GFP/Cherry filter set . A GC400 filter was used to eliminate UV . The microscope was controlled by the SoftWorks software . All quantifications were performed using ImageJ software . Cells containing the GAL-CPY*-GFP gene were grown at 30°C in synthetic media containing the appropriate amino acids and 2% galactose for 6 hr . To visualize the expression of CPY*-GFP , we took 30 z-stacks ( 0 . 2 µm steps ) on the Deltavision microscope ( Applied Precision ) equipped with a CoolSNAP HQ2 camera ( Photometrics ) and an Olympus 100 × oil immersion objective and the GFP/Cherry filter set . A GC400 filter was used to eliminate UV . The microscope was controlled by the SoftWorks software . All quantifications were performed using ImageJ software . Time-lapse microscopy was performed in a Y04C microfluidic chamber ( Cellasic , Hayward , California ) controlled by an ONIX microfluidic perfusion platform . Synthetic media was flowed in the chamber at 1 Psi . Sec61 was endogenously tagged with tdEOS as described ( McKinney et al . , 2009 ) . Cells were imaged using a laser scanning confocal microscope LSM 780 ( Zeiss ) equipped with a high sensitive multiarray 32PMT GaAsP detector . Photoconversion was applied on a ROI as indicated in the figures ( half of the mother cell ) . For photoconversion a 405 nm laser at 10% laser intensity was used . The ROI was converted once with 20 iterations . After conversion an image was taken every 20 s for 2 min . Images were analyzed using ImageJ . Integrated fluorescence intensity of the converted fluorophore ( red ) was measured in the bud for every time point and after subtraction of the background normalized to 10 for the first image ( before conversion ) . An equally normalized unconverted control cell was then subtracted . The average normalized fluorescence intensity of all cells and the standard error of the mean was plotted in the graph . Lifespan analyses were carried out as previously described ( Kennedy et al . , 1994 ) . | Cell division isn't always about splitting a cell into two identical parts . The diversity of many of our own cells relies on asymmetric cell divisions . The yeast used to make bread rely on a process called ‘budding’ that involves a small daughter cell emerging from the surface of the mother cell . Mother cells can only produce around 20–50 daughter cells before dying from old age . However , their daughters are always born rejuvenated , and not aged like their mothers . Budding involves part of the plasma membrane that surrounds the mother cell being pinched off to produce the daughter cell . This part of the membrane contains diffusion barriers that prevent various factors—including factors that cause aging—from entering the daughter cell . The barriers are known to contain several layers , but the details of how they work were not understood . Inside the budding cell , the membrane of the endoplasmic reticulum ( ER ) also contains lateral diffusion barriers . The ER is the structure in the cell responsible for folding newly made proteins correctly . Any misfolded , toxic proteins are kept in the ER to be refolded or destroyed . However , if there are too many misfolded proteins , the ER gets stressed and triggers a mechanism that in extreme cases causes the cell to self-destruct . Clay , Caudron et al . have now shown that ER stress causes yeast cells to age . Moreover , when the ER is stressed , the ER diffusion barrier prevents the stress that causes aging entering the daughter cells . Clay , Caudron et al . also established that the diffusion barrier in the ER is made up of three layers . A layer of fatty molecules called sphingolipids is found at the bottom of the barrier , and such a layer is also present in other diffusion barriers . This could therefore act as the skeleton on which diffusion barriers form . Further investigation of this layer should provide a better understanding of how diffusion barriers work . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"methods"
] | [
"cell",
"biology"
] | 2014 | A sphingolipid-dependent diffusion barrier confines ER stress to the yeast mother cell |
Although it is currently understood that the exon junction complex ( EJC ) is recruited on spliced mRNA by a specific interaction between its central protein , eIF4AIII , and splicing factor CWC22 , we found that eIF4AIII and the other EJC core proteins Y14 and MAGO bind the nascent transcripts of not only intron-containing but also intronless genes on Drosophila polytene chromosomes . Additionally , Y14 ChIP-seq demonstrates that association with transcribed genes is also splicing-independent in Drosophila S2 cells . The association of the EJC proteins with nascent transcripts does not require CWC22 and that of Y14 and MAGO is independent of eIF4AIII . We also show that eIF4AIII associates with both polysomal and monosomal RNA in S2 cell extracts , whereas Y14 and MAGO fractionate separately . Cumulatively , our data indicate a global role of eIF4AIII in gene expression , which would be independent of Y14 and MAGO , splicing , and of the EJC , as currently understood .
Pre-mRNA splicing is typically co-transcriptional , yet by modulating the protein composition and conformation of the ribonucleoprotein complex consisting of mature mRNA and associated proteins ( mRNP ) , it can affect cytoplasmic processes such as mRNA localization , translation and nonsense-mediated mRNA decay ( NMD ) ( Dreyfuss et al . , 2002; Le Hir et al . , 2016; Moore and Proudfoot , 2009 ) . The main distinctive feature of spliced mRNAs appears to be the presence of the exon junction complex ( EJC ) , a four-protein core-complex consisting of eIF4AIII , Y14 , MAGO and MLN51 , which is typically deposited 20–24 nucleotides ( nt ) upstream of the exon junctions in mammalian cells ( Kataoka et al . , 2001; Le Hir et al . , 2000; Saulière et al . , 2012; Singh et al . , 2012 ) . Such sequence-independent but position-specific deposition is initiated by the recruitment of the EJC core protein eIF4AIII to the spliceosome via its association with spliceosome component CWC22 ( Alexandrov et al . , 2012; Barbosa et al . , 2012; Steckelberg et al . , 2012 ) . While associated with CWC22 , eIF4AIII is in an inactive conformation which is incompatible with ATP and RNA binding , and therefore , with EJC assembly ( Buchwald et al . , 2013 ) . The current model predicts that following exon ligation , eIF4AIII dissociates from CWC22 , binds the spliced mRNA , typically at the −24 nt position of exon junction , and then triggers EJC assembly by recruiting the three remaining EJC core proteins , Y14 , MAGO and MLN51 ( Ballut et al . , 2005; Bono et al . , 2006; Steckelberg et al . , 2015 ) . The association of MAGO and Y14 stabilizes binding of the complex to mRNA by inhibiting eIF4AIII ATPase activity ( Ballut et al . , 2005 ) . The EJC core proteins are well conserved in Drosophila; where Y14 , MAGO and MLN51 are also known as Tsunagi , Mago Nashi and Barentsz , respectively ( Macchi et al . , 2003; Mohr et al . , 2001; Palacios et al . , 2004 ) . The genes encoding these four proteins are all required for oskar mRNA localization to the posterior end of the oocyte ( Macchi et al . , 2003; Mohr et al . , 2001; Palacios et al . , 2004; van Eeden et al . , 2001 ) . Specifically , similar to that reported in mammalian cells , the EJC consisting of eIF4AIII-MAGO-Y14 ( MLN51 is mostly cytoplasmic ) is deposited at the canonical position 20–24 nt upstream of exon junctions on in vitro spliced mRNAs , and both splicing and deposition of the EJC appear to be required for oskar mRNA localization during oogenesis in Drosophila ( Ghosh et al . , 2010 , 2012; Hachet and Ephrussi , 2004 ) . Only some introns appear to trigger EJC-dependent nonsense-mediated mRNA decay ( NMD ) in Drosophila ( Gatfield et al . , 2003; Saulière et al . , 2010 ) . These observations in Drosophila and recent reports that the EJC is not present at all exon junctions or solely at canonical positions in mammalian cells , raise the possibility that either deposition or stability of the EJC on spliced mRNA might be a regulated process ( Mühlemann , 2012; Saulière et al . , 2012; Singh et al . , 2012 ) . Additionally , EJC deposition on partially spliced pre-mRNA might modulate splicing of flanking introns in Drosophila , yet only for a subset of transcripts ( Ashton-Beaucage et al . , 2010; Hayashi et al . , 2014; Malone et al . , 2014; Roignant and Treisman , 2010 ) . With the aim to understand the mechanism that regulates EJC deposition , we used the giant polytene chromosomes from Drosophila salivary glands . Similar to other eukaryotes , pre-mRNA splicing occurs co-transcriptionally in Drosophila ( Khodor et al . , 2011; LeMaire and Thummel , 1990; Osheim et al . , 1985 ) . Therefore , the polytene chromosomes provide an ideal system to visualize and analyze the mechanism of the association of EJC proteins with both pre-mRNA and nascent spliced transcripts . The data we show here reveal that deposition of the EJC proteins eIF4AIII , Y14 and MAGO on nascent transcripts , neither depends on the presence of introns nor requires the spliceosomal protein CWC22 in this organism . Additionally , ChIP-seq analysis of Y14 similarly indicates that this protein associates with transcriptionally active genes in Drosophila S2 cells independently of splicing .
Using antibodies against eIF4AIII , MAGO and Y14 , which detect the proteins with minimal cross-reactivity in Western blotting ( Figure 1—figure supplement 1A and B ) , we found that EJC proteins are present in both nuclear and cytoplasmic fractions ( Figure 1—figure supplement 1C ) . The absolute amounts of these proteins are comparable between the nucleus and cytoplasm , but as indicated by whole salivary gland immunostaining , they are more concentrated in the nucleus ( Figure 1—figure supplement 1D ) . On the polytene chromosomes , the signals of all three proteins are most prevalent at transcriptionally active sites , which correspond to distinct cytologically decondensed segments ( interbands ) of the chromosome ( Figure 1A ) . This localization is apparent by simultaneously inspecting the intensity profile of the EJC proteins and DAPI signals along the same chromosome segment ( Figure 1A rightmost panel ) ; these line profile plots show apparent complementarity of eIF4AIII , Y14 and MAGO signals with DAPI signal ( Figure 1A , III , VI and IX ) . The signals of the EJC proteins colocalize with those of RNA Pol II ( Figure 1—figure supplement 2A , the antibody H5 , a marker of transcription elongation recognizes the Ser2 hyperphosphorylated CTD of the largest Pol II subunit ( Buratowski , 2009; O'Brien et al . , 1994 ) . The association of the EJC proteins with polytene chromosomes is predominantly with nascent transcripts as the signals are RNase sensitive ( Figure 1B ) . There is some residual signal of EJC , which cannot be attributed to incomplete RNA digestion , since that of Hrb87F , the hnRNPA1 homologue in Drosophila ( Lakhotia et al . , 2012 ) , is completely removed from the chromosomes ( Figure 1B ) . The banding pattern of eIF4AIII at polytene chromosomes is different from that of Y14 and MAGO . While the eIF4AIII signal is detected at every Pol II transcription site ( Figure 1—figure supplement 2A ) , there are sites at which Y14 and MAGO signals are either absent or just detectable ( yellow arrows in Figure 1—figure supplement 2B , IV-IX ) . As all three EJC antibodies used in the present study were raised in rabbits , we used transgenic flies expressing Y14 and eIF4AIII double tagged with HA and FLAG to analyze further the extent of colocalization of the EJC proteins ( Material and methods ) . The tagged proteins are of the predicted size ( Figure 1—figure supplement 3A and B ) and show a chromosomal banding pattern very similar to that of the endogenous proteins ( Figure 1—figure supplement 3C ) . Double immunostaining of tagged Y14 , and endogenous MAGO shows complete colocalization along the chromosomes ( Figure 1—figure supplement 4 ) , suggesting that the two proteins are in close association at transcription sites , most likely forming a stable heterodimer , as previously observed in vitro ( Lau et al . , 2003 ) . In agreement with the staining patterns of endogenous Y14 and MAGO differing from that of eIF4AIII , Y14 is either absent from , or just detectable , at many sites at which there is a very strong signal for tagged eIF4AIII ( Figure 1—figure supplement 5 ) . Cumulatively , our data indicate that the chromosomal binding pattern of Y14 and MAGO differs from that of eIF4AIII , which instead appears to bind at all Pol II transcription sites . In view of the EJC model , at this stage the data could therefore have been interpreted to signify that either Y14 and MAGO associate with eIF4AIII at a later stage , perhaps post-transcriptionally , or as envisaged by previous reports ( Saulière et al . , 2010 ) , that the EJC is not as stable in Drosophila as in mammalian cells . 10 . 7554/eLife . 19881 . 003Figure 1 . The EJC core proteins associate with nascent transcripts at polytene chromosomes . ( A ) Immunolocalization of EJC proteins ( red ) , eIF4AIII ( I-III ) , Y14 ( IV-VI ) and MAGO ( VII-IX ) , on salivary gland polytene chromosomes of wandering third instar larvae . Chromosomes were counter-stained with DAPI ( blue ) . Intensity profiles of EJC protein and DAPI signals ( III , VI , IX ) over a segment at the tip of chromosome 3L ( box in merged images ) show accumulation of these proteins at interband regions . Scale bar represents 20 µm length . ( B ) Parallel immunolocalization of EJC proteins ( green ) and Hrb87F ( hnRNP A1 ) ( red ) on polytene chromosomes spread from salivary gland without treatment ( Control ) or after incubation with RNase ( RNase treated ) . Scale bar represent 20 µm length . DOI: http://dx . doi . org/10 . 7554/eLife . 19881 . 00310 . 7554/eLife . 19881 . 004Figure 1—figure supplement 1 . Characterization of EJC antibodies . ( A ) Western blot of S2 cell protein extracts treated with antibodies against eIF4AIII ( lane I ) , MAGO ( lane II ) and Y14 ( lane III ) proteins of 46 , 17 and 19 kDa expected molecular weight , respectively . Asterisks in lanes 1 and 3 indicate bands of unexpected size , but possibly specific as their intensity is also reduced by RNAi of the target transcript , not shown . ( B ) Western blot showing that EJC antibodies , used in present study , specifically detect down-regulation of their antigen protein in its RNAi samples . ( C ) Western blot showing levels of Y14 , MAGO and eIF4AIII in nuclear and cytoplasmic fractions of S2 cells . RNA Pol II is shown at the bottom ( detected using 8WG16 antibody ) as fractionation control; the two expected bands Pol IIo and Pol IIa are indicated . Asterisk indicates a nonspecific band detected in the cytoplasmic fraction . ( D ) Immunolocalization of eIF4AIII ( I-III ) , Y14 ( IV-VI ) and MAGO ( VII-IX ) in whole mount salivary glands of wild-type third instar larvae . Chromosomes were counterstained with DAPI ( blue ) . Scale bar represents 50 µm length . DOI: http://dx . doi . org/10 . 7554/eLife . 19881 . 00410 . 7554/eLife . 19881 . 005Figure 1—figure supplement 2 . EJC protein signals co-localize with active Pol II . ( A ) Magnified view of a chromosome segment of a chromosome showing eIF4AIII ( green , I ) and Pol II Ser2 ( red , II ) . The split image ( III ) shows eIF4AIII ( green ) above and Pol II Ser2 ( red ) below on same arm . Right panel shows a line drawn on the same segment ( IV ) and corresponding intensity profile of both of the proteins ( V ) . ( B ) Double immunostaining of RNA Pol II Ser2 ( red ) and EJC proteins ( green ) ; eIF4AIII ( top row ) , Y14 ( middle row ) and MAGO ( bottom row ) . Yellow arrows in the middle and bottom row indicate absence of Y14 and MAGO , respectively , at transcription sites marked with intense Pol II Ser2 signal ( red ) . Scale bar represents 20 µm length . DOI: http://dx . doi . org/10 . 7554/eLife . 19881 . 00510 . 7554/eLife . 19881 . 006Figure 1—figure supplement 3 . Characterization of transgenic flies expressing tagged Y14 or eIF4AIII . ( A ) Western blotting of protein extracts of salivary glands from transgenic flies expressing double tagged eIF4AIII- ( HA-FLAG ) 2 detected either with anti-FLAG ( lanes , 1 and 2 ) or with anti-HA antibody ( 3 , 4 ) . ( B ) Western blotting of protein extracts of salivary glands from transgenes expressing Y14- ( HA-FLAG ) 2 by using anti-FLAG ( 1 , 2 ) and anti-HA antibodies ( 3 , 4 ) . ( C ) Immunolocalization of tagged Y14 ( red , I , III ) and eIF4AIII ( red , IV , VI ) by using anti-HA antibody , in genotypes indicated on the left of each row . Chromosomes were counterstained with DAPI ( blue ) . Scale bar represents 20 µm length . DOI: http://dx . doi . org/10 . 7554/eLife . 19881 . 00610 . 7554/eLife . 19881 . 007Figure 1—figure supplement 4 . Y14 and MAGO strictly colocalize at transcription sites . ( A ) Parallel immunostaining of tagged Y14 ( red ) detected with anti-FLAG and endogenous MAGO ( green ) on polytene chromosomes squash of fkhGAL4>Y14 ( HA-FLAG ) 2 . ( B ) Similar chromosomes immunostaining as in A using anti HA instead of anti-FLAG . Scale bar represents 20 µm length and applies for both panels . DOI: http://dx . doi . org/10 . 7554/eLife . 19881 . 00710 . 7554/eLife . 19881 . 008Figure 1—figure supplement 5 . Y14 banding pattern differs from that of eIF4AIII . Parallel immunostaining of tagged eIF4AIII ( red ) and endogenous Y14 ( green ) on polytene chromosomes squash of fkhGAL4>eIF4AIII ( HA-FLAG ) 2 larvae . Arrows are indicating sites with strong eIF4AIII but with weak or absent Y14 signal . Chromosomes were counterstained with DAPI ( blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19881 . 008 The presence of eIF4AIII at all transcription sites was particularly surprising , as many of these sites must correspond to intronless genes; ~20% of Drosophila genes do not contain introns ( De Renzis et al . , 2007 ) . Initially , we reasoned that perhaps Y14 and MAGO associate with eIF4AIII selectively , possibly on nascent transcripts of intron-containing genes . To test this hypothesis , we examined the distribution of the EJC proteins at heat shock genes on polytene chromosomes , some of which characteristically do not carry introns ( Lis et al . , 1981 ) . Remarkably , following heat shock , clear accumulation of the EJC proteins was observed at heat-shock puffs ( classically denominated by their chromosomal map position ) of both intron-containing genes ( 63B , encoding Hsp90; and 93D , encoding hsrω lncRNAs ) as well as that of intronless genes ( 87A and 87C , encoding Hsp70; and 95D , encoding Hsp68 ) ( Figure 2 ) . These observations indicated that interaction of the core EJC proteins with nascent RNA might be splicing independent in Drosophila . To assess further whether these proteins associate independently of introns under standard growth conditions at non-heat-shock genes , we constructed a novel inducible expression vector and used it to generate two transgenes , one carrying an intron ( S118 ) and another without intron ( S136 ) ( Material and methods ) . A distinctive feature of these transgenes is that they are flanked by an inducible promoter regulated by an ecdysone responsive element ( ERE ) ( Material and methods ) . Another feature is that they carry lacO repeats at the upstream of the promoter; these repeats can be easily visualized and cytologically mapped by GFP-lacI immunolocalization on polytene chromosomes ( Robinett et al . , 1996 ) . The transgenes were mapped at position 3B on the X chromosome ( S118 ) and at 63B on 3L chromosome ( S136 ) ( Figure 3A ) . The genes are transcriptionally silent in in vitro cultured salivary glands in the absence of ecdysone , but upon ecdysone treatment , a large puff is formed at both loci , indicative of strong transcriptional induction ( Figure 3B ) . Notably , immunostaining revealed that eIF4AIII ( Figure 3C ) as well as Y14 ( Figure 3—figure supplement 1 ) associate with both transcription puffs , and as expected , there is a strong Pol II Ser2 signal at these loci . These observations therefore indicate that association of the EJC proteins with nascent transcripts can occur independently of splicing in Drosophila salivary glands . 10 . 7554/eLife . 19881 . 009Figure 2 . EJC proteins accumulate at heat shock transcription puffs . Parallel immunolocalization of hyperphosphorylated RNA Pol II Ser2 ( red , II , III , V , VI , VIII , IX ) and EJC proteins ( green ) eIF4AIII ( I , III ) , Y14 ( IV , VI ) , and MAGO ( VII , IX ) , on polytene chromosomes following heat shock at 37°C for 1 hr . Yellow arrowheads indicate accumulation of EJC proteins at the induced intron-containing heat-shock genes , while white arrowheads indicate their accumulation at intronless heat-shock genes . Genes are identified by their map position ( details in Results ) , which are labeled in the merged images on the right panels . Corresponding chromosome arms are mentioned in left column of panel . Scale bar represents 20 µm length . DOI: http://dx . doi . org/10 . 7554/eLife . 19881 . 00910 . 7554/eLife . 19881 . 010Figure 3 . EJC proteins associate with nascent transcripts of both intron and intronless genes . ( A ) Immunolocalization of GFP-LacI ( red band in boxed areas ) at transgene insertion site: S118 ( intron plus ) on the X ( panel I ) and of S136 ( intron minus ) on the 3L ( III ) chromosome arm . Bands were mapped at 3B for S118 ( II ) and 63B for S136 ( IV ) using a standard polytene chromosome map shown above . ( B ) DAPI-stained ( gray ) segments of the X chromosome encompassing 3B ( I , II ) and 3L chromosome encompassing 63B ( III , IV ) , without ( I , III ) or with ( II , IV ) ecdysone treatment , which produces a distinct puff at the transgene insertion locus . ( C ) Immunolocalization of RNA Pol II Ser2 ( red ) and eIF4AIII ( green ) at 3B ( I-IV and IX-XII ) and 63B ( V-VIII and XIII-XVI ) loci in S118 ( I-VIII ) and S136 ( IX-XVI ) transgene following ecdysone treatment . As there is no insert at locus 63B in S118 ( V-VIII ) and at 3B in S136 ( IX-XII ) , these are used as ecdysone-unresponsive control loci for the transgene at 63B in S136 ( XIII-XVI ) and for 3B in S118 ( I-IV ) , respectively . Scale bar represents 20 µm length . DOI: http://dx . doi . org/10 . 7554/eLife . 19881 . 01010 . 7554/eLife . 19881 . 011Figure 3—figure supplement 1 . Y14 associates with nascent transcript of both intron and intronless gene reporters . Immunolocalization of RNA Pol II Ser2 ( red ) and Y14 ( green ) at an intro-containing gene reporter ( S118 construct at 3B ) ( I-IV and IX-XII ) and intronless counterpart ( S136 at 63B ) ( V-VIII and XIII-XVI ) . Both genes were induced by ecdysone treatment as described in Material and methods . As there is no insert at locus 63B in S118 ( V-VIII ) and at 3B in S136 ( IX-XII ) , these are used as ecdysone-unresponsive control loci for the transgene at 63B in S136 ( XIII-XVI ) and for 3B in S118 ( I-IV ) , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 19881 . 011 Next , we used chromatin immunoprecipitation ( ChIP ) followed by high-throughput DNA sequencing ( ChIP-seq ) to examine the association of the EJC proteins with gene loci in S2 cells . ChIP experiments were carried out for all three EJC proteins; however , only the Y14 antibody worked well in this assay and produced a clear enrichment profile relative to input DNA; this is most apparent at transcription start sites ( TSS ) ( Figure 4A ) , and selectively with genes that are expressed ( Figure 4B ) . Y14 enrichment does not appear to correlate with expression level though . While genes with a very low expression ( RPKM between 1 and 10 ) show lower enrichment , genes ranging from low ( RPKM 11–50 ) to very high expression levels are comparably enriched ( Figure 4—figure supplement 1 ) . The result of this analysis is therefore consistent with the chromosome immunostaining data shown above , which demonstrated that Y14 and MAGO , unlike eIF4AIII , are either absent or very weak at transcription sites with strong active Pol II signal . Notably Y14 appears to bind both intron-containing and intronless genes ( Figure 4C ) , and the enrichment is only slightly higher for genes with introns . When all genes , expressed and unexpressed , are included in the analysis , there is a positive correlation between Y14 binding and intron number ( Figure 4D ) ; however , in view that Y14 associates selectively with expressed genes , this must signify that many intronless genes are either not or very weakly expressed . A representative genome browser example of this intron-independent association is shown in Figure 4E , which corresponds to a region of chromosome three where apparent enrichment is detected at different genes . The highest association is at the Big brother ( Bgb ) and RpL23A gene loci ( indicated by blue and red arrows respectively ) , both genes are highly expressed in S2 cells , but only RpL23A encodes introns . The results of this ChIP-seq analysis are therefore consistent with our observations with polytene chromosomes described above , and further indicate that the association of Y14 with gene loci is dependent on whether they are transcribed , yet unlinked to intron presence . Given that Y14 association with transcription sites is RNA-dependent at the chromosomes ( Figure 1B ) , the prediction is that ChIP is detecting mostly Y14 associated with nascent transcript and that the enrichment at TSSs mirrors that of RNA Pol II , which globally in Drosophila , as in other metazoan , is detected by ChIP , and other biochemical assays , mostly around TSSs , rather than the gene body ( Adelman and Lis , 2012; Core et al . , 2012 ) . However , following up on our observation that association of Y14 ( and the other EJC proteins ) at the chromosomes is not as sensitive to RNase as hnRNPA1 ( Figure 1B ) , we examined whether the EJC proteins might also associate with Pol II directly . We immunoprecipitated ( IP ) elongating RNA Pol II using a Ser2 CTD antibody ( Material and methods ) . The result was that none of the three EJC proteins make an interaction with Pol II which is stable enough to be detect by IP , instead the association with general elongation factor Spt6 ( Kaplan et al . , 2000 ) , which was used as a positive control , was readily detected ( Figure 4—figure supplement 2 ) . 10 . 7554/eLife . 19881 . 012Figure 4 . Y14 associates with expressed genes independent of introns in S2 cells . ( A ) Genome-wide average Y14 ChIP-seq enrichment 3000 bases around transcription start sites ( TSS ) of ChIP ( blue trace ) versus input DNA ( green ) . ( B ) Y14 ChIP-seq enrichment ( after background subtraction ) 2000 bases around transcription start sites ( TSS ) of expressed ( blue trace ) or unexpressed ( green trace ) genes in S2 cells ( see Material and methods ) . ( C ) Average Y14 enrichment of expressed genes with a different number of introns: 0 to >=5 , indicated by traces of different colour ( see legend on right of the plot ) . ( D ) Average Y14 enrichment across genes as in C but including both expressed and unexpressed genes . ( E ) Representative chromosome region showing Y14 ChIP-seq enrichment profile ( blue ) , versus that of input DNA ( grey ) . Genes are labeled following Flybase nomenclature . DOI: http://dx . doi . org/10 . 7554/eLife . 19881 . 01210 . 7554/eLife . 19881 . 013Figure 4—figure supplement 1 . Y14 association with transcribed genes does not correlate with mRNA levels . Average Y14 enrichment of genes either not expressed ( cyan colored line; RPKM = 0 ) or expressed at different levels ( RPKM ranging from 1 to more than 100; indicated by traces of different color ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19881 . 01310 . 7554/eLife . 19881 . 014Figure 4—figure supplement 2 . EJC proteins do not co-purify with RNA Pol II . Immunoprecipitation ( IP ) with RNA Pol II Ser2 antibody following either RNase or mock treatment . Proteins were assayed by Western blotting for EJC proteins eIF4AIII , Y14 and MAGO , along with elongation factor Spt6 and Hrb87F ( hnRNPA1 ) . Goat IgG was used in a parallel IP as background control . DOI: http://dx . doi . org/10 . 7554/eLife . 19881 . 014 To examine further the connection to splicing , we tested whether the association of the EJC proteins with mRNA depends on CWC22 in Drosophila . The prediction from current models is that depletion of CWC22 ( known as Nucampholin or NCM in Drosophila ) would impair association of eIF4AIII with nascent transcripts . RT-PCR quantification shows that CWC22-RNAi efficiently reduces the CWC22 mRNA level ( Figure 5A ) . Its depletion in the early stages of development ( fkh-Gal4>NCM-RNAi ) strongly inhibits salivary gland developmental growth and polytenization ( Figure 5B ) , to an extent that we were unable to spread the polytene chromosomes for immunolocalization . Due to unavailability of an antibody , the extent of protein depletion could not be assessed directly . The presence of residual CWC22 could not be ruled out . However , the striking growth phenotype indicates that RNAi is efficient at depleting CWC22 essential function . There are no sequences in the Drosophila genome which could encode similar proteins , the strong phenotype should therefore indicate that there are also no distant-related proteins that could complement CWC22 function in its absence . Therefore , to asses whether CWC22 might have a role in the association of EJC proteins at the chromosomes , the small-gland phenotype was partially circumvented by inhibiting expression of CWC22-RNAi till early third instar larval stage , by co-expression of the temperature-sensitive Gal80 protein , which represses Gal4 activity when larvae are grown at 18°C ( tub-Gal80ts; Fkh-Gal4>NCM-RNAi ) . Larvae were then transferred to 29°C and by the late third instar stage the chromosomes were big enough to proceed for immunostaining . Notably , both eIF4AIII and Y14 remain localized in the nucleus in CWC22 depleted cells ( Figure 5C , I-IV vs . IX-XII ) , and polytene chromosome immunostaining demonstrates that eIF4AIII can still associate with chromosome interbands . Although chromosome morphology is not as well defined in these glands , the banding pattern and relative intensities of the signals are comparable to wild type ( Figure 5C , V-VIII ) . Additionally , Y14 can also still associate with interbands , producing a staining pattern as intense as , and very similar to , that in wild type ( Figure 5C , XIII-XVI ) . These observations demonstrate that while CWC22 has an essential function in salivary gland development , it does not appear to be required for the association of eIF4AIII and Y14 ( and by inference MAGO ) with the nascent transcripts in Drosophila . 10 . 7554/eLife . 19881 . 015Figure 5 . EJC proteins associate with nascent transcripts independently of CWC22 ( NCM ) . ( A ) Real-time PCR quantification of NCM RNA level in salivary glands of tubGAL80ts; +; FkhGAL4>NCM-RNAi ( right ) relative to that in wild type ( left ) . ( B ) Thirst instar larva and their salivary glands of different genotypes mentioned above each lane . The line indicates a fragment of fat body ( fb ) adhering to the glands . ( C ) Immunolocalization of EJC proteins eIF4AIII and Y14 in whole salivary gland cells ( I-IV , IX-XII ) and at polytene chromosomes ( V-VIII , XIII-XVI ) . The two panels on the left are from wild-type larvae and the two on the right are from tubGAL80ts; +; fkhGAL4/NCM-RNAi larvae . In the mutant ( III , IV , XI , XII ) white arrows indicate relatively large-sized nuclei while yellow arrows indicate small-sized nuclei . The white line ( panel III ) indicates a fat body ( fb ) cell nucleus attached with salivary gland . Chromosomes were counterstained with DAPI ( blue ) . Scale bars represent 20 µm length for corresponding set of images . DOI: http://dx . doi . org/10 . 7554/eLife . 19881 . 015 Since the interaction of Y14/MAGO with eIF4AIII drives RNA binding in mammalian cells and stabilizes the EJC in vitro ( Ballut et al . , 2005 ) , a prediction is that Y14/MAGO may not associate , or does so to a lesser extent , with nascent transcripts in the absence of eIF4AIII . We tested this by visualizing the distribution of Y14 on the polytene chromosomes from salivary glands depleted of eIF4AIII by RNAi ( Figure 6A , B and C ) . As for CWC22 , depletion of eIF4AIII from an early stage of development ( Fkh-Gal4>eIF4AIII-RNAi ) strongly inhibits salivary gland growth ( Figure 5B ) ; therefore , indicating that RNAi is effectively depleting an essential function of the protein . Again , the phenotype prevented polytenization of the chromosomes . However , by restricting RNAi activation until early third instar stage , ( using tub-Gal80ts; Fkh-Gal4>eIF4AIII-RNAi ) glands were big enough to prepare satisfactory chromosome spreads for immunostaining ( Figure 6A ) . Remarkably , we observed that despite the drastic reduction in chromosomal eIF4AIII signal ( Figure 6A , VII and VIII vs . V and VI ) , and despite the low level of polytenization , Y14 could still be detected at cytologically distinguishable interbands , similar to wild type ( Figure 6A , XV and XVI vs . XIII and XIV ) . In contrast to depletion of CWC22 and eIF4AIII which strongly inhibited salivary glands growth , depletion of either Y14 or MAGO resulted in visually normal salivary glands ( Figure 5B ) , despite their depletion being very efficient , based on salivary gland Western blotting ( Figure 6D and E ) . Although some Y14 signal persists in the nucleolus ( which often remains attached to the chromosomes during the spreading procedure ) , the respective chromosomal signals are completely absent from the chromosome arms ( 6F , I and II ) . Particularly , depletion of neither Y14 nor MAGO affects the association of eIF4AIII with the chromosomes ( Figure 6F , V and XI ) ; we observed instead that depletion of Y14 eliminates MAGO from the chromosomes ( Figure 6F , III and IV ) and depletion of MAGO also reduces the Y14 signal ( Figure 6F , IX and X ) . In summary , these observations indicate that association of Y14 ( and by inference MAGO ) with nascent transcripts does not require eIF4AIII and that of eIF4AIII is independent of Y14 or MAGO . Additionally , the RNAi phenotype of eIF4AIII is drastically different: while eIF4AIII knockdown impairs salivary gland development , neither that of Y14 nor MAGO has any apparent visual phenotype ( Figure 5B ) . 10 . 7554/eLife . 19881 . 016Figure 6 . eIF4AIII is not required for the association of Y14 and MAGO with nascent transcript . ( A ) Immunolocalization of EJC proteins ( red ) eIF4AIII and Y14 in whole salivary gland ( I-IV , IX-XII ) and polytene chromosomes ( V-VIII , XIII-XVI ) , in wild type ( left panels ) and tubGAL80ts; +; fkhGAL4/eIF4AIII-RNAi ( right panels ) larvae . Insets ( XV , XVI ) are showing a magnified view of the areas enclosed by the white boxes . Chromosomes were counterstained with DAPI ( blue ) . Scale bar represents 20 µm length . ( B ) Real-time PCR quantification of eIF4AIII and Y14 mRNA levels in fkhGAL4>eIF4AIIIRNAi relative to wild-type larval salivary glands . ( C ) Western blotting showing eIF4AIII protein levels in tubGAL80ts; +; fkhGAL4/eIF4AIII-RNAi and wild-type glands . ( D ) Western blotting showing Y14 protein level in fkhGAL4>Y14RNAi and wild-type salivary glands . ( E ) Western blotting showing level of MAGO protein in fkhGAL4>MAGO-RNAi and wild-type salivary glands . Tubulin was detected as loading control . ( F ) Immunolocalization of Y14 ( red , I , II , IX , X ) , MAGO ( red , III , IV , VII , VIII ) and eIF4AIII ( red , V , VI , XI , XII ) on polytene chromosomes from fkhGAL4>Y14RNAi ( left panels ) and fkhGAL4>MAGO-RNAi ( right panels ) larvae . Chromosomes were counterstained with DAPI ( blue ) . Yellow arrows ( II , X ) indicate accumulation of corresponding proteins at the nucleolus . Scale bar represents 20 µm length . DOI: http://dx . doi . org/10 . 7554/eLife . 19881 . 01610 . 7554/eLife . 19881 . 017Figure 6—figure supplement 1 . Depletion of eIF4AIII does not affect the pattern of recruitment of Y14 at Pol II transcription sites . Coimmunolocalization of Y14 ( green ) and RNA Pol II Ser2 ( red ) on salivary gland polytene chromosomes of tubGAL80ts; +; fkhGAL4/eIF4AIII-RNAi larvae . Line profile ( V ) is showing the intensity of both signals across a line drawn on a defined chromosome segment ( indicated by the white box , IV ) , along the DAPI signal ( in blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19881 . 017 The current understanding is that the EJC remains associated with spliced mRNAs until it is removed by the ribosome in the first round of translation . A key observation supporting this model is that Y14 is present in monosomal but not polysomal fractions in mammalian cells ( Diem et al . , 2007; Dostie and Dreyfuss , 2002; Gehring et al . , 2009 ) . We therefore analyzed the distribution of the EJC proteins in polysomal fractions of Drosophila S2 cells . Surprisingly , we detected neither Y14 nor MAGO in monosomal fractions; both proteins remain on the top of the gradient , corresponding to free proteins or small molecular weight complexes ( Figure 7A ) . In contrast , eIF4AIII was detected in all the fractions , but it was least abundant in the fraction containing Y14 and MAGO ( Figure 7A ) . There is a small correlation between eIF4AIII levels and the ribosomal trace; however , eIF4AIII is no more abundant in either monosomal or lighter fractions , it is therefore unlikely that more of it is associated with mRNAs at or before the first round of translation when they are loaded with just one ribosome or with the 40S ribosomal subunit . Such distribution suggests that eIF4AIII is not removed by translation and remains associated actively translated mRNA; this interpretation is further substantiated by the fact that mild RNase treatment , which breaks down polysomes by digesting the mRNA ( Material and methods ) , shifts most of eIF4AIII into lighter sub-ribosomal fractions ( Figure 7B ) . The sedimentation of Y14 and MAGO is not affected by the RNase treatment . In vitro dissociation of polysomes by EDTA treatment also shifts most of eIF4AIII in the lightest fraction ( Figure 7C , lane 12 ) , indicative of both polysome breakdown and perhaps release of the protein from mRNA . Notably , a portion of eIF4AIII apparently co-fractionates with the 60S ribosomal subunit ( Figure 7C , lanes 9 and 10 ) , and seemingly to a lesser extent with the 40S ( fraction 11 in Figure 7C; the red arrow indicates a faster migrating form of the protein in lane 12 , which possibly was generated by proteolysis during the in-vitro incubation ) . Some of eIF4AIII might fractionate with ribosomal subunits also in the untreated sample ( Figure 7A , fractions 9 and 10 ) . We investigated further whether there is an association with ribosomal subunits by pre-treatment of the cell culture with puromycin , followed by dissociation of the subunits by incubation at either 4°C or room temperature in presence of high-salt in vitro prior to loading on the gradient ( Material and methods ) . This treatment also delocalized eIF4AIII to lighter fractions , consistent with most of the protein being associated with actively translated mRNA . The majority of the protein was detected in the top two light fractions ( Figure 7D , lanes 11 and 12; as seen in Figure 7C , a large fraction of the protein runs as a faster migrating photolytic cleavage product ) . Cumulatively , the data demonstrate that eIF4AIII cosediments with both monosomes and polysomes , yet Y14 and MAGO fractionate separately in normal conditions , and therefore , it is not likely they form a stable complex with eIF4AIII on translating mRNA in S2 cells . Although some eIF4AIII is detected in the light fraction containing Y14 and MAGO , this is a relatively very minor amount of what is found in the heavier fractions ( possibly the result of some unavoidable mixing during fractionation ) . 10 . 7554/eLife . 19881 . 018Figure 7 . eIF4AIII but not Y14-MAGO associates with ribosome-loaded mRNA in S2 cells . ( A ) Polysome profiling of cytoplasmic extracts of S2 cells , and Western blotting of corresponding fractions , show distribution of eIF4AIII ( upper lane ) , Y14 ( middle lane ) and MAGO ( lower lane ) . ( B ) Polysome profiling and Western blotting as shown in A , following RNase treatment . ( C ) Polysome profiling and Western blotting following EDTA treatment . ( D ) Polysome profiling and Western blotting following puromycin treatment at either 4°C ( black line ) or at room temperature ( red line ) . The red arrows in C and D point to a faster migrating double band of eIF4AIII . DOI: http://dx . doi . org/10 . 7554/eLife . 19881 . 018 As the signals for Y14 and MAGO completely overlap at the chromosomes ( Figure 1—figure supplement 4 ) , and depletion of either protein affects the association of the other ( Figure 6F ) , we predicted that dimerization is most likely stabilizing these two proteins . To further investigate this , we assayed protein levels following RNAi-mediated knockdown of Y14 , MAGO and eIF4AIII in Drosophila S2 cells . Notably , consistent with what we observed at the chromosomes , knockdown of Y14 drastically reduces MAGO , and conversely , MAGO knockdown reduces the level of Y14 , although not completely ( Figure 8A ) . In both cases , the level of eIF4AIII was not affected . Furthermore , while knockdown of eIF4AIII led to an overall reduction in protein levels , it did not noticeably affect Y14 or MAGO . The inter-dependence is most likely the result of a change in stability of the proteins , as depletion of Y14 does not affect MAGO mRNA level and that of MAGO does not affect Y14 mRNA level ( Figure 8B ) . These results indicate that Y14 and MAGO heterodimerization is essential for their stability in cells– consistent with what was reported by one of the initial EJC studies ( Le Hir et al . , 2001 ) - yet it appears such effect is independent of eIF4AIII . 10 . 7554/eLife . 19881 . 019Figure 8 . Y14 and MAGO are required for the stability of each other . ( A ) Western blotting of total cell lysate from S2 cells to detect eIF4AIII , Y14 , MAGO , following eIF4AIII-RNAi , MAGO-RNAi , Y14-RNAi and in untreated cultures . Tubulin was detected as loading control . ( B ) Real-time PCR quantification of the mRNA level of indicated transcripts was carried out in MAGO-RNAi ( left ) , Y14-RNAi ( middle ) and eIF4AIII-RNAi ( right ) along with untreated S2 cells . DOI: http://dx . doi . org/10 . 7554/eLife . 19881 . 019
Our investigation reveals that the three proteins that are thought to form the core of the EJC across organisms , associate with nascent transcripts independently of intron presence and of CWC22 in Drosophila , the spliceosome component predicted to drive the assembly of the EJC on spliced mRNAs in mammalian cells ( Alexandrov et al . , 2012; Barbosa et al . , 2012; Saulière et al . , 2012; Singh et al . , 2012; Steckelberg et al . , 2012 ) . Our data indicate that Y14 and MAGO form an obligatory heterodimer in vivo , as observed previously ( Lau et al . , 2003; Le Hir et al . , 2001 ) . Y14/MAGO associates with nascent transcripts on polytene chromosomes , yet this association neither correlates with , nor requires eIF4AIII , suggesting that the Y14/MAGO heterodimer engages with nascent mRNAs independently of eIF4AIII . Moreover , Y14 and MAGO do not cosediment with eIF4AIII by sucrose gradient fractionation of S2 cell cytoplasmic extracts , indicating that these proteins are not forming a stable complex with eIF4AIII . As Y14 and MAGO stay on the top of the gradient , it is most likely that they are not bound to mRNA in the cytoplasm . In contrast , eIF4AIII associates with mRNA loaded with multiple ribosomes; it does not therefore appear to be removed , at least not irreversibly , by translocating ribosomes . Perhaps eIF4AIII primarily binds mRNA UTRs in Drosophila cytoplasmic extracts , and , to a smaller extent , as indicated by our sucrose fractionation data , ribosomal subunits , rather than coding regions , as reported for mammalian whole-cell lysates ( Saulière et al . , 2012; Singh et al . , 2012 ) . Particularly , our data indicate that Y14-MAGO dissociates from mRNA before translation initiation , unlike in mammalian cells ( Diem et al . , 2007; Dostie and Dreyfuss , 2002; Gehring et al . , 2009 ) . This interpretation is consistent with the report that Partner of Y14-MAGO ( PYM ) does not engage with the ribosome and that PYM-mediated dissociation of EJC proteins from oskar mRNA is independent of translation in ovarian extracts in Drosophila melanogaster ( Ghosh et al . , 2014 ) . We speculate that in Drosophila eIF4AIII , similar to related RNA helicases , dynamically binds mRNA in an ATP-driven reaction rather than making a stable interaction with a specific location on the RNA ( Linder and Jankowsky , 2011 ) . Our data suggest that eIF4AIII is a general mRNP component , and it may therefore have a global role in pre-mRNA processing and translation , independently of splicing and EJC assembly in Drosophila . The protein shares extensive sequence and domain similarity with the essential eukaryotic translation initiation factor eIF4A across eukaryotes ( Parsyan et al . , 2011 ) , which is its closest homolog in Drosophila with 71% aminoacid sequence identity . The protein eIF4A is the prototype of an RNA helicase which is required for unwinding 5’UTR secondary structures during the initial recruitment of the ribosome to mRNA 5’ end ( Linder and Jankowsky , 2011 ) . Consistent with this role , eIF4A was reported to sediment in lighter sucrose fraction corresponding to mRNAs not yet engaged in translation elongation ( Bordeleau et al . , 2006 ) . However , as we find eIF4AIII in heavy polysomal fractions , we predict that the protein might mostly bind 3’UTRs rather than 5’UTRs of polysomal mRNA . What the significance is of the partial cosedimentation of eIF4AIII with ribosomal subunits will need to be investigated . In this respect , eIF4AIII might be similar to VASA , another eIF4A-related protein that is required for germline differentiation in Drosophila; VASA appears to regulate translation via its association with specific 3’UTRs ( Liu et al . , 2009 ) . While it has long been understood that human eIF4AIII is functionally distinct from eIF4A ( Li et al . , 1999; Parsyan et al . , 2011 ) , it was recently reported that the protein enhances translation of mRNAs associated with the nuclear cap binding complex ( CBC ) in human cells , independently of the presence of introns ( Choe et al . , 2014 ) . As depletion of eIF4AIII impairs salivary gland growth , we predict that its essential role in Drosophila development might primarily reflect a global requirement in translation , rather than localization of specific mRNAs ( Palacios et al . , 2004 ) . Depletion of the EJC proteins causes changes in splicing patterns in Drosophila and other organisms ( Ashton-Beaucage et al . , 2010; Hayashi et al . , 2014; Malone et al . , 2014; Michelle et al . , 2012; Roignant and Treisman , 2010; Wang et al . , 2014 ) . These changes can be interpreted in two ways: either canonical EJC deposition can influence splicing of neighbouring introns or the EJC proteins can affect splice site recognition by being functional components of pre-spliceosome intermediates . Both interpretations are based on the assumption that EJC assembly is either constitutively coupled to splicing in Drosophila , as reported ( Ghosh et al . , 2012 ) , or that it is at least deposited on a subset of spliced mRNAs or pre-mRNAs . Our data are inconsistent with these interpretations . We cannot exclude the possibility that the EJCs may be deposited at specific junctions on some mRNAs , and that these are stable enough to persist through nuclear export and cytoplasmic mRNA localization in some cells , oocytes in particular ( Ghosh et al . , 2010 , 2012; Hachet and Ephrussi , 2004 ) . While the Y14 ChIP-assay shows slightly higher association with genes containing multiple introns ( Figure 4D ) , this might be a consequence of their higher transcription rate , rather than being driven by a direct interaction with spliceosome components . The most parsimonious explanation is that the so-called EJC proteins bind nascent transcripts independently of splicing and eIF4AIII , and therefore may not form a stable complex which can tag splice junctions , or even spliced mRNPs , on Drosophila mRNAs . Therefore , the mechanisms by which these proteins regulate pre-mRNA splicing , transposon activity , mRNA localization and translation possibly need to be re-examined in the context of the absence of an EJC . A well-characterized function of the EJC is its role in linking pre-mRNA splicing to translation and NMD in mammalian cells ( Chazal et al . , 2013; Le Hir and Séraphin , 2008; Saltzman et al . , 2008 ) . While this might apply to some Drosophila introns ( Saulière et al . , 2010 ) , it has instead been reported that splicing does not affect NMD of well-characterized mRNA reporters in Drosophila ( Gatfield et al . , 2003 ) . Additionally , deposition of the EJC , or a similar complex which would tag splice junctions , does not appear to be the mechanism that links pre-mRNA splicing to NMD in fission yeast ( Wen and Brogna , 2010 ) . Therefore , the important question of how splicing can affect translation and NMD remains to be understood ( Brogna et al . , 2016; Brogna and Wen , 2009 ) .
Flies were reared in standard fly food medium at 24°C . The yw strain was used as wild type ( DGGR_108736 ) . Double tagged ( 2X HA-FLAG ) Y14 and eIF4AIII transgenes were generated by cloning the corresponding cDNA sequence into pUAST-attB vector and insertion at the PhiC31 recombination site of the yw , P{CaryPattP3} strain using germline injection ( BestGene , USA ) . Transgenes expressing the lacO-tagged and ecdysone-regulated S118 and S136 constructs ( see below ) were generated by random P-element transformation . RNAi lines targeting Y14 ( 36585 ) and eIF4AIII ( 32444 ) were obtained from the Bloomington stock centre while that of MAGO ( 28132 ) was obtained from the VDRC stock centre . The GFP-LacI stock was previously described ( Vazquez et al . , 2002 ) . The forkhead ( fkh ) -Gal4 used in present study has a salivary-gland-specific expression from early stage of development ( Henderson and Andrew , 2000 ) . The tubGal80ts line was previously described ( McGuire et al . , 2003 ) . Flies with either the tubGAL80ts/+; +; fkhGAL4/eIF4AIII-RNAi or tubGAL80ts/+; +; fkhGAL4/NCM-RNAi genotype were obtained by selecting tubby female larvae from crossing tubGAL80ts; +; fkhGAL4/TM6B male with either +; eIF4AIII-RNAi or +; NCM-RNAi female flies , respectively . These larvae were initially cultured at 18°C until early third instar stage and then transferred at 29°C until late third instar larval stage when salivary glands were dissected out . Affinity purified antibodies against Drosophila EJC proteins , anti-MAGO , anti-Y14 and anti-eIF4AIII were produced in rabbit using recombinant proteins expressed in E . coli by YenZym ( USA ) . Their dilution used in immunostaining was typically 1:100 while it was 1:1000 in Western blotting . Other primary antibodies used in present study , and their dilution in immunostaining were mouse anti-Pol II ( AB_10143905 , H5 , Covance , 1:250 ) , mouse anti-HA ( AB_514505 , 12CA5 , 1:500 ) , mouse anti-FLAG ( AB_259529 , M2 , Sigma-Aldrich , 1: 200 ) and goat anti-GFP ( Bio-Rad AbD Serotec , 1:250 ) . The antibodies used in Western blotting were diluted as follow: anti-RNA Pol II ( AB_10013665 , 8WG16 , Covance , 1:2000 ) , anti-Pol II Ser2 ( 3E10 , Merck Millipore , 1:1000 ) , Anti-Spt6 ( 1:1000 ) , mouse anti-HA ( AB_514505 , 12CA5 , 1:5000 ) , mouse anti-FLAG ( AB_259529 , M2 , Sigma-Aldrich , 1: 2000 ) , mouse anti-Hrb87F ( hnRNPA1 ) ( 1:200 ) and mouse anti-β-tubulin ( Sigma-Aldrich , T5168 1:4000 ) . Secondary antibodies were from Jackson Immuno Research Technologies , Sigma-Aldrich or Life Technologies . The intron-containing construct ( S118 ) was generated by cloning medfly adh1 coding region including the intron ( Brogna et al . , 2006 ) in the EcoRI site of the pERE expression and germline transformation vector described below . This construct carries six repeats of the MS2 binding site in the 3’ UTR , subcloned from pIG-bs6-mix ( Golding and Cox , 2004 ) . The intronless construct ( S136 ) encodes a fusion transcript coding for 3 HA repeats , enhanced green fluorescent protein ( EGFP ) and beta-galactosidase ( LacZ ) . In pERE ecdysone-dependent transcription is driven by a promoter cassette consisting of approximately seven repeats of the ecdysone responsive element ( ERE ) from hsp27 ( Riddihough and Pelham , 1987 ) ; this was cloned upstream of the Adh distal promoter , both from Drosophila ( a clone of this cassette which was provided by Carl Thummel , University of Utah , has been generated and previously described ( Steve Stowers , PhD thesis , Stanford University ) . To generate pERE , the cassette was cloned into the HindIII and EcoRI sites of pUAST , therefore replacing the UAS promoter with the ERE . Additionally , eight lacO repeats [Robinett et al . , 1996] ) were cloned into the HindIII and PstI sites located just upstream of the ERE cassette to allow visualization of the locus using GFP-LacI . For heat-shock treatment , actively wandering third instar larvae of desired genotypes were transferred in batches to microfuge tubes lined with moist filter paper and submerged in a water bath maintained at 37°C for 45 min . Control samples of larvae of comparable age and genotypes were kept in microfuge tubes containing moist filter papers at 24°C , in parallel . For ecdysone treatment , salivary glands of actively wandering third instar larvae were dissected out and incubated in Shields and Sang M3 insect media ( M3 , Sigma ) with or without 1 µM ecdysone , for 1 hr at room temperature . Drosophila embryo driven S2 cells ( CVCL_Z232 ) were cultured in Insect–XPRESS medium ( Lonza ) supplemented with 10% fetal bovine serum ( FBS ) and 1% Penicillin-Streptomycin-Glutamine mix ( P/S/G ) ( Invitrogen ) at 27°C . Y14 or eIF4AIII sequence was PCR amplified with corresponding primer pairs from available plasmid constructs , while that of MAGO was amplified from adult male flies ( Supplementary file 1 ) . Along with the desired gene sequence , all these primer pairs carried the T7 promoter sequence at their 5’ end ( 5’-TTAATACGACTCACTATAGGGGAGA-3’ ) . The amplified PCR fragments were purified with a QIAquick PCR Purification Kit ( QIAGEN ) and dsRNA was synthesized using the T7 RiboMAX express RNAi system ( Promega ) . Typically , a six-well culture dish was seeded with 106 cells/well in serum-free media followed by addition of 15 µg of dsRNA into each well and by 1 hr incubation at room temperature . After the incubation , 2 mL of complete media was added to each well and the cells were incubated for four days to knockdown the corresponding RNA . RNA extraction from S2 cells was carried out using the RNeasy Mini Kit ( QIAGEN ) . Total RNA ( 700 ng ) was used for cDNA synthesis using qScript cDNA synthesis Kit ( Quanta Biosciences ) . Quantitative real-time PCR was carried out using the SensiFAST SYBR Hi ROX Kit ( Bioline ) in 96-well plates on a ABI PRISM 7000 ( Applied Biosystems ) . RNA isolation from salivary gland and real time quantification of desired gene transcripts was done with a Power SYBR Green Cells-to-Ct Kit ( Thermo Fisher Scientific ) . The Ct value for the desired transcript level was normalized by the RpL32 or 18 s rRNA transcripts as reference ( Supplementary file 1 ) . Nuclear and cytoplasmic fractions were purified following a published procedure ( Parker and Topol , 1984 ) . S2 cells were grown in a T75 tissue culture flask at 27°C containing 15 mL of media as described above for 2 days . Cells were harvested by centrifugation at 2500 rpm for 5 min and washed by resuspension in ice cold PBS twice . The washed pellet was then resuspended in five times its volume of buffer A ( 15 mM KCL , 10 mM HEPES ( pH 7 . 6 ) , 2 mM MgCl2 , 0 . 1 mM EDTA ) and was centrifuged at 5000 rpm at 4°C for 5 min . The cell pellet was then resuspended in the same volume of buffer A supplemented with 1 mM DTT and homogenized with the tight pestle ( B ) in a Dounce glass homogenizer until most of the cells appeared visually broken-down under the microscope . The cell suspension was then mixed with 1/10 vol of buffer B ( 1 M KCl , 50 mM HEPES ( pH 7 . 6 ) , 30 mM MgCl2 , 0 . 1 mM EDTA , 1 mM DTT ) to increase ionic strength and centrifuged at 10 , 000 rpm for 10 min at 4°C . The supernatant and pellet were resolved as cytoplasmic and nuclear fractions , respectively . S2 cells nuclei were isolated as previously described ( Khodor et al . , 2011 ) , and lysed in 5 vol of lysis buffer ( 20 mM HEPES pH 7 . 4 , 110 mM potassium acetate , 0 . 5% Triton X-100 , 0 . 1% Tween-20 , 10 mM MnCl2 , 1X EDTA-free Complete Protease Inhibitor Cocktail ( Roche ) , 1X PhosStop ( Roche ) and 50 U/ml Ribolock RNase inhibitor ) with 110 U/ml RNase-free DNase ( Roche ) and incubated for 1 hr at 4°C on rotator . The suspension was centrifuged at 13 , 000 rpm for 15 min , and the supernatant was transferred in fresh tube and incubated with 20 µL of DynabeadsProtein-G ( Thermo Fisher Scientific ) , which were coated with 5 µg of RNA Pol II Ser2 antibody ( Merck Millipore ) , for 1 hr at 4°C on a rotator . The beads were washed with wash buffer ( 20 mM HEPES pH 7 . 4 , 110 mM potassium acetate , 0 . 5% Triton X-100 , 0 . 1% Tween-20 , 50 U/mL RNase inhibitor , 4 mM MnCl2 ) and mixed with cleared cell supernatant and incubated for 1 hr at 4°C on rotator . Beads were washed 3X with wash buffer . For RNase treatment , the beads were incubated with 1 mg/ml RNase in wash buffer , while control sample was incubated in same buffer without RNase at 4°C for 20 min on rotator . The beads were further washed three times with wash buffer and bound proteins were then eluted in 40 µL SDS-sample buffer and boiled 7 min prior loading on the SDS-PAGE gel . S2 cells ( 107 ) were fixed in 1% formaldehyde ( EM grade , Polyscience ) for 15 min at room temperature and then transferred in 125 mM glycine for 5 min to stop the cross-linking reaction . Following centrifugation at 3000 rpm for 5 min at 4°C , the cell pellet was resuspended in 500 µL of PBS containing EDTA-free Complete Protease Inhibitor Cocktail ( Roche ) and washed twice . The cell pellet was resuspended in 100 µL of SDS lysis buffer ( 1% SDS , 10 mM EDTA ) and sonicated at high intensity for 12 min with 15 s on/off cycles in a Bioruptor sonicator ( Diagenode ) . Samples were diluted in 1 ml ChIP dilution buffer ( 0 . 01% SDS , 1 . 1% Triton X-100 , 1 . 2 mM EDTA , 16 . 7 mM Tris-HCl pH 8 . 1 , 167 mM NaCl ) and centrifuged at 13 , 000 rpm for 10 min at 18°C . An aliquot ( 100 µL ) of soluble chromatin was kept to extract input DNA . For each ChIP , typically 2 µg of antibody was incubated with 25 µL of DynabeadsProtein-G beads at 4°C overnight . Beads were washed four times with 1 mL of PBS with 5 mg/ml BSA , and resuspended in 40 µL of PBS with 5 mg/mL BSA . The coated beads were then added to the remaining chromatin sample and incubated for two and a half hours at room temperature with rotation . Beads were then sequentially washed with 1 mL of low-salt wash ( 0 . 1% SDS , 1% Triton X-100 , 2 mM EDTA , 20 mM Tris-HCL pH 8 . 1 , 150 mM NaCl ) ; high-salt wash ( as above but with 500 mM NaCl ) ; LiCl wash ( 0 . 25M LiCl , 1% IGEPAL-CA630 , 1% deoxycholic acid , 1 mM EDTA , 10 mM Tris-HCL ( pH 8 . 1 ) ; two 5 min washes in 1 mL of TE buffer ( 10 mM Tris-HCL ( pH 8 . 0 ) , 1 mM EDTA ) . The beads were then incubated with elution buffer ( 0 . 1M NaHC03 , 1% SDS ) at a room temperature for 15 min and reverse crosslinks overnight at 65°C in presence of Proteinase K ( 2 µL of 50 mg/ml ) . Chromatin was purified using AMPure XP beads ( Beckman Coulter ) . ChIP and input DNA were further fragmented to 250 bp fragment size using a Covaris S2 focused ultrasonicator prior to library preparation , sequencing and genome mapping was as previously described using SOLiD four genome analyser ( Life Technologies ) ( Kwon et al . , 2016 ) . Reads were aligned to the Drosophila genome ( BDGP R5/dm3 assembly ) . Average enrichments across gene regions were calculated and plotted prior or after background ( input DNA ) subtraction . Genes were divided as expressed ( RPKM >= 1 ) or unexpressed ( RPKM = 0 ) based on expression level calculated using Drosophila S2 cell expression data ( GEO accession no . GSM410195 ) . Average Y14 enrichment was also plotted at expressed gene further divided based on the number of introns ( 0 , 1 , 2 , 3 , 4 , >= 5 ) as annotated in R5/dm3 assembly . All our Chip-seq data were deposited in the GEO repository ( accession no . GSE84595 ) . Unless otherwise stated , polytene chromosomes squashes were prepared from salivary glands , dissected in 1X PBS ( 13 mM NaCl , 0 . 7 mM Na2HPO4 , 0 . 3 mM NaH2PO4 , pH 7 . 4 ) and transferred to 1% Triton X-100 in 1X PBS for 30 s . Glands were then transferred to 3 . 7% formaldehyde in 1X PBS followed by 3 . 7% formaldehyde with 45% acetic acid for 1 min each . Finally , glands were transferred to 45% acetic acid for 1 min and squashed under a coverslip as previously described ( Singh and Lakhotia , 2012 ) . In some of the experiments , salivary glands were dissected out in a different dissection buffer ( 15 mM HEPES pH 7 . 4 , 60 mM KCl , 15 mM NaCl , 1 . 5 mM Spermine , 1 . 5 mM Spermidine and 1% Triton ) and then processed as previously described ( Al-Jubran et al . , 2013; Rugjee et al . , 2013 ) . Both procedures produce similar immunostaining results . For RNase treatment , salivary glands were dissected in M3 media and incubated in a solution containing 1% Triton X-100 in M3 media for 2 min at room temperature . Glands were then transferred to M3 media with or without 1 mg/ml RNase A ( Invitrogen , California , USA ) and incubated for 20 min at room temperature . For whole salivary gland immunostaining , larvae were dissected in M3 media , glands were immediately fixed in ice cold PBS with 4% formaldehyde ( 10% EM grade , Polyscience ) and processed as described ( Al-Jubran et al . , 2013 ) . All the glands were mounted in PromoFluor Antifade Reagent ( PromoKine ) . Images were taken using either a Nikon Eclipse Ti epifluorescence microscope , equipped with ORCA-R2 camera ( Hamamatsu Photonics ) or a Leica TCS SP2-AOBS confocal microscope . S2 cells were grown in a T75 tissue culture flask as described above until they reached 80% confluence . Cultures were briefly treated with 25 µg/mL cycloheximide for 5 min , cells were harvested by centrifugation at 3000 rpm for 5 min and washed by resuspension in ice cold buffer ( 10 mM HEPES pH 7 . 4 , 2 mM magnesium chloride , 2 mM magnesium acetate , 100 mM potassium acetate , prepared in DEPC treated H2O ) . Cells were pelleted again , resuspended and left to lyse by incubating on ice for 10 min in 600 µL lysis buffer ( HEPES buffer with 1 mM PMSF , 1 mM DTT , 1 . 2 µL Ribolock RNase inhibitor , 250 µg/mL heparin , 0 . 6% Triton X-100 and 1X complete EDTA-free protease inhibitor cocktail ) . The lysate was cleared twice by centrifugation at maximum speed in a microfuge for 10 min . For RNase-treated samples , RNase inhibitor was excluded from the lysis buffer and 10 µL of 1 mg/mL RNase A was added to the cleared lysate and incubated on ice for 10 min . For EDTA treatment , the cleared lysate was supplemented with 30 mM EDTA for 30 min on ice prior fractionation . For puromycin treatment , 100 µg/mL puromycin was added to the cleared lysate and incubated either on ice or at room temperature for 30 min . The lysis buffer used for the puromycin treatment was as described above , but supplemented with 375 mM KCl and lacked magnesium as previously reported ( Al-Jubran et al . , 2013 ) . The equivalent of about 20 absorbance units at OD260 ( NanoDrop readings , blanked with water ) was loaded onto a 10–50% ( w/v ) sucrose gradient; gradients were prepared using a gradient mixer which mixed 50% and 10% stock sucrose solutions prepared in polysome buffer ( 10 mM Tris acetate pH7 . 4 , 70 mM ammonium acetate , 4 mM magnesium acetate , 25 µg/mL cycloheximide in DEPC treated H2O ) , dispensing directly into SW41 rotor tubes ( Beckman Coulter ) . Puromycin treated samples were loaded on 10–30% sucrose gradients instead of 10–50% prepared as described above but lacking magnesium . Samples were centrifuged at 38 , 000 rpm for 160 min at 4°C . Fractions were collected by inserting a steel capillary needle to the bottom of the tube , through which the gradient was pumped using a peristaltic pump through an ISCO UA-6 absorbance detector ( 254 nm ) connected to a plotter . Gradients were dispensed as ~800 µL aliquots using a fraction collector . Proteins were extracted using the methanol/chloroform method . A 200 µL aliquot of each fraction was mixed with 800 µL of methanol in a 1 . 5 mL tube to which 200 µL chloroform was then added and vortexed , followed by addition of 400 µL of H2O . The sample was then vortexed followed by centrifugation at maximum speed for 5 min ( proteins/RNA form a white disc at the aqueous/organic interface ) . The upper phase was discarded and 900 µL of methanol was added , mixed by inversion and centrifuged again for 5 min . The protein-containing pellets were air dried and analyzed by standard SDS-PAGE and Western blotting . | Cells and organisms survive and thrive in large part due to the activities of thousands of proteins . The instructions for making these proteins are found in the DNA sequences of genes . However , these genes also tend to contain large sections called introns that do not encode protein . To make a protein , the gene’s full sequence is first copied to a temporary molecule called pre-messenger RNA ( pre-mRNA for short ) . The introns are then removed from the pre-mRNA in a process known as splicing in the cell nucleus , during which the remaining regions of the molecule , called exons , are joined together to form a mature mRNA molecule . This mature mRNA can then move out of the cell nucleus and be used as a template to build proteins around the cell . Intriguingly , splicing of the pre-mRNAs in the nucleus affects how the mRNA is used to make proteins in the cytoplasm of the cell . This nucleus-cytoplasm connection is currently explained by the so-called exon junction complex , which is thought to attach to mature mRNAs at the junction between two exons and stay bound until the mRNA moves to the cytoplasm . Evidence suggests the exon junction complex affects how the mRNA is used to make protein , yet little is known about how it would do so . Choudhury , Singh et al . examined how exon junction complex proteins bind to newly made RNA in salivary gland cells of fruit flies and in cultured cells . Contrary to expectations , the three proteins thought to make the central part of the exon junction complex were found on different mRNAs and regardless of whether the mRNAs derived from genes with introns . Specifically , one of these proteins – eIF4AIII – can remain on the mRNA independently of the two other exon junction complex proteins or CWC22 , a protein required for splicing . CWC22 is also thought to be required for the complex to be deposited precisely at exon junctions in human cells . Overall , it appears that our current understanding of the exon junction complex needs to be revised . The findings presented by Choudhury , Singh et al . predict alternative roles for these proteins , particularly eIF4AIII , which will be independent of any deposition of the exon junction complex . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2016 | Exon junction complex proteins bind nascent transcripts independently of pre-mRNA splicing in Drosophila melanogaster |
Fireflies and their luminous courtships have inspired centuries of scientific study . Today firefly luciferase is widely used in biotechnology , but the evolutionary origin of bioluminescence within beetles remains unclear . To shed light on this long-standing question , we sequenced the genomes of two firefly species that diverged over 100 million-years-ago: the North American Photinus pyralis and Japanese Aquatica lateralis . To compare bioluminescent origins , we also sequenced the genome of a related click beetle , the Caribbean Ignelater luminosus , with bioluminescent biochemistry near-identical to fireflies , but anatomically unique light organs , suggesting the intriguing hypothesis of parallel gains of bioluminescence . Our analyses support independent gains of bioluminescence in fireflies and click beetles , and provide new insights into the genes , chemical defenses , and symbionts that evolved alongside their luminous lifestyle .
Fireflies ( Coleoptera: Lampyridae ) represent the best-studied case of bioluminescence . The coded language of their luminous courtship displays ( Figure 1A; Video 1 ) has been long studied for its role in mate recognition ( Lloyd , 1966; Lewis and Cratsley , 2008; Stanger-Hall and Lloyd , 2015 ) , while non-adult bioluminescence is likely a warning signal of their unpalatable chemical defenses ( De Cock and Matthysen , 1999 ) , such as the cardiotoxic lucibufagins of Photinus fireflies ( Meinwald et al . , 1979 ) . The biochemical understanding of firefly luminescence: an ATP , Mg2+ , and O2-dependent luciferase-mediated oxidation of the substrate luciferin ( Shimomura , 2012 ) , along with the cloning of the luciferase gene ( de Wet et al . , 1985; Ow et al . , 1986 ) , led to the widespread use of luciferase as a reporter with unique applications in biomedical research and industry ( Fraga , 2008 ) . With >2000 species globally , fireflies are undoubtedly the most culturally appreciated bioluminescent group , yet there are at least three other beetle families with bioluminescent species: click beetles ( Elateridae ) , American railroad worms ( Phengodidae ) and Asian starworms ( Rhagophthalmidae ) ( Martin et al . , 2017 ) . These four closely related families ( superfamily Elateroidea ) have homologous luciferases and structurally identical luciferins ( Shimomura , 2012 ) , implying a single origin of beetle bioluminescence . However , as Darwin recognized in his ‘Difficulties on Theory’ ( Darwin , 1872 ) , the light organs amongst the luminous beetle families are clearly distinct ( Figure 1B ) , implying independent origins . Thus , whether beetle bioluminescence is derived from a single or multiple origin ( s ) remains unresolved . To address this long-standing question , we sequenced and analyzed the genomes of three bioluminescent beetle species . To represent the fireflies , we sequenced the widespread North American ‘Big Dipper Firefly’ , P . pyralis ( Figure 1A , C ) and the Japanese ‘Heike-botaru’ firefly Aquatica lateralis ( Figure 1B ) . P . pyralis was used in classic studies of firefly bioluminescent biochemistry ( Bitler and McElroy , 1957 ) and the cloning of luciferase ( de Wet et al . , 1985 ) , while A . lateralis , a species with specialized aquatic larvae , is one of the few fireflies that can be reliably cultured in the laboratory ( Oba et al . , 2013a ) . These two fireflies represent the two major firefly subfamilies , Lampyrinae and Luciolinae , which diverged from a common ancestor over 100 Mya ( Figure 1B ) ( Misof et al . , 2014; Mckenna et al . , 2015 ) . To facilitate evolutionary comparisons , we also sequenced the ‘Cucubano’ , Ignelater luminosus ( Figure 1B ) , a Caribbean bioluminescent click beetle , and member of the ‘Pyrophorus’ used by Raphaël Dubois ( 1849-1929 ) to first establish the enzymatic basis of bioluminescence in the late 1800s ( Dubois , 1885; Dubois , 1886 ) . Comparative analyses of the genomes of these three species allowed us to reconstruct the origin ( s ) and evolution of beetle bioluminescence .
Photinus pyralis adult males were collected from the Great Smoky Mountains National Park , USA ( GSMNP ) and Mercer Meadows New Jersey , USA ( MMNJ ) ( Figure 1C ) , and sequenced using short-insert , mate-pair , Hi-C , and long-read Pacific Biosciences ( PacBio ) approaches ( Appendix 4—table 1 ) . These datasets were combined in a MaSuRCA ( Zimin et al . , 2013 ) hybrid genome assembly ( Appendix 1 . 5 ) . The Aquatica lateralis genome was derived from an ALL-PATHs ( Butler et al . , 2008 ) assembly of short insert and mate-pair reads from a single adult female from a laboratory-reared population , whose lineage , dubbed ‘Ikeya-Y90’ , was first collected 25 years ago from a now extinct population in Yokohama , Japan ( Appendix 2 . 5 ) . A single Ignelater luminosus adult male , collected in Mayagüez Puerto Rico , USA , was used to produce a high-coverage Supernova ( Weisenfeld et al . , 2017 ) linked-read draft genome ( Appendix 3 . 5 ) , which was further manually scaffolded using low-coverage long-read Oxford Nanopore MinION sequencing ( Appendix 3 . 5 . 4 ) . The gene completeness and contiguity statistics of our P . pyralis ( Ppyr1 . 3 ) and A . lateralis ( Alat1 . 3 ) genome assemblies are comparable to the genome of the model beetle Tribolium castaneum ( Figure 2F; Appendix 4 . 1 ) . The I . luminosus genome assembly ( Ilumi1 . 2 ) is less complete , but is comparable to other published insect genomes ( Figure 2F; Appendix 4 . 1 ) . Protein-coding genesets for our study species were produced via an EvidenceModeler-mediated combination of homology alignments , ab initio predictions , and de novo and reference-guided RNA-seq assemblies followed by manual gene curation for gene families of interest ( Appendix 1 . 10; 2 . 8; 3 . 8 ) . These coding gene annotation sets for P . pyralis , A . lateralis , and I . luminosus are comprised of 15 , 773 , 14 , 285 , and 27 , 557 genes containing 94 . 2% , 90 . 0% , and 91 . 8% of the Endopterygota Benchmarking Universal Single-Copy Orthologs ( BUSCOs ) ( Simão et al . , 2015 ) , respectively . Protein clustering via predicted orthology indicated 77% of genes were found in orthogroups with at least one other species ( Figure 2E; Appendix 4—figure 1 ) . We found the greatest orthogroup overlap between the P . pyralis and A . lateralis genesets , as expected given the more recent phylogenetic divergence of these species . Remaining redundancy in the P . pyralis assembly and annotation , as indicated by duplicates of the BUSCOs and the assembly size ( Figure 2F; Appendix 4—table 2 ) is likely due to the heterozygosity of the outbred input libraries ( Appendix 1 ) . The higher BUSCO completeness of the assemblies as compared to the genesets ( Appendix 4—table 3 ) , suggests that future manual curation efforts will lead to improved annotation completeness . To enable the characterization of long-range genetic structure , we super-scaffolded the P . pyralis genome assembly into 11 pseudo-chromosomal linkage groups using a Hi-C proximity-ligation linkage approach ( Figure 2A; Appendix 1 . 5 . 3 ) . These linkage groups contain 95% of the assembly ( 448 . 8 Mbp ) . Linkage group LG3a corresponds to the X-chromosome based on expected adult XO male read coverage and gene content ( Appendix 1 . 6 . 4 . 1 ) and its size ( 22 . 2 Mbp ) is comparable to the expected X-chromosome size based on sex-specific genome size estimates using flow cytometry ( ~26 Mbp ) ( Lower et al . , 2017 ) . Homologs to T . castaneum X-chromosome genes were enriched on LG3a over every other linkage group , suggesting that the X-chromosomes of these distantly related beetles are homologous , and that their content has been reasonably conserved for >200 MY ( Appendix 1 . 6 . 4 . 1 ) ( Mckenna et al . , 2015 ) . We hypothesized that the P . pyralis orthologs of known bioluminescence genes , including the canonical luciferase Luc1 ( de Wet et al . , 1985 ) and the specialized luciferin sulfotransferase LST ( Fallon et al . , 2016 ) , would be located on the same linkage group to facilitate chromosomal looping and enhancer assisted co-expression within the light organ . We , however , found these genes on separate linkage groups ( Figure 2A ) . In addition to nuclear genome assembly and coding gene annotation , we also assembled the complete mitochondrial genomes ( mtDNA ) of P . pyralis ( Figure 2C; Appendix 1 . 8 ) and I . luminosus ( Appendix 3 . 10 ) , while the mtDNA sequence of A . lateralis was recently published ( Maeda et al . , 2017 ) . These mtDNA assemblies show high conservation of gene content and synteny , with the exception of the variable ~1 Kbp tandem repeat unit ( TRU ) found in the firefly mtDNAs . As repetitive elements are common participants and drivers of genome evolution ( Feschotte and Pritham , 2007 ) , we next sought to characterize the repeat content of our genome assemblies . Overall , 42 . 6% , 19 . 8% , and 34 . 1% of the P . pyralis , A . lateralis , and I . luminosus assemblies were found to be repetitive , respectively ( Appendix 1 . 11; 2 . 9; 3 . 9 ) . Of these repeats 66 . 7% , 39 . 4% , and 55% could not be classified as any known repetitive sequence , respectively . Helitrons , DNA transposons that transpose through rolling circle replication ( Kapitonov and Jurka , 2001 ) , are among the most abundant individual repeat elements in the P . pyralis assembly . Via in situ hybridization , we identified that P . pyralis chromosomes have canonical telomeres with telomeric repeats ( TTAGG ) ( Figure 2B; Appendix 1 . 13 ) . DNA methylation is common in eukaryotes , but varies in degree across insects , especially within Coleoptera ( Bewick et al . , 2017 ) . Furthermore , the functions of DNA methylation across insects remain obscure ( Bewick et al . , 2017; Glastad et al . , 2017 ) . To examine firefly cytosine methylation , we characterized the methylation status of P . pyralis DNA with whole genome bisulfite sequencing ( WGBS ) . Methylation at CpGs ( mCG ) was unambiguously detected at ~20% within the genic regions of P . pyralis and its methylation levels were at least twice those reported from other holometabolous insects ( Figure 2D; Appendix 1 . 12 ) . Molecular evolution analyses of the DNA methyltransferases ( DNMTs ) show that direct orthologs of both DNMT1 and DNMT3 were conserved in P . pyralis , A . lateralis , and I . luminosus ( Appendix 4—figure 2; Appendix 4 . 2 . 3 ) , implying that our three study species , and inferentially likely most firefly lineages , possess mCG . Corroborating this claim , CpG[O/E] analysis of methylation indicated our three study species had DNA methylation ( Appendix 4—figure 3 ) . Two luciferase paralogs have been previously described in fireflies ( Oba et al . , 2013a; Bessho-Uehara et al . , 2017 ) . P . pyralis Luc1 was the first firefly luciferase cloned ( de Wet et al . , 1985 ) , and its direct orthologs have been widely identified from other fireflies ( Oba , 2014 ) . The luciferase paralog Luc2 was previously known only from a handful of Asian taxa , including A . lateralis ( Oba et al . , 2013a; Bessho-Uehara et al . , 2017 ) . Previous investigations of these Asian taxa have shown that Luc1 is responsible for light production from the lanterns of adults , larvae , prepupae and pupae , whereas Luc2 is responsible for the dim glow of eggs , ovaries , prepupae and the whole pupal body ( Bessho-Uehara et al . , 2017 ) . From our curated genesets ( Appendix 1 . 10; 2 . 8 ) , we unequivocally identified two firefly luciferases , Luc1 and Luc2 , in both the P . pyralis and A . lateralis genomes . Our RNA-Seq data further show that in both P . pyralis and A . lateralis , Luc1 and Luc2 display expression patterns consistent with previous reports . While Luc1 is the sole luciferase expressed in the lanterns of both larvae and adults , regardless of sex , Luc2 is expressed in other tissues and stages , such as eggs ( Figure 3C ) . Notably , Luc2 expression is detected in RNA libraries derived from adult female bodies ( without head or lantern ) , suggesting detection of ovary expression as described in previous studies ( Bessho-Uehara et al . , 2017 ) . Together , these results support that since their divergence via gene duplication prior to the divergence of Lampyrinae and Luciolinae , Luc1 and Luc2 have established different , but conserved roles in bioluminescence throughout the firefly life cycle . Firefly luciferase is hypothesized to be derived from an ancestral peroxisomal fatty acyl-CoA synthetase ( PACS ) ( Figure 3A ) ( Oba et al . , 2003; Oba et al . , 2006a ) . We found that , in both firefly species , Luc1 is genomically clustered with its closely related homologs , including PACSs and non-peroxisomal acyl-CoA synthetases ( ACSs ) , enzymes which can be distinguished by the presence/absence of a C-terminal peroxisomal-targeting-signal-1 ( PTS1 ) . We also found nearby microsomal glutathione S-transferase ( MGST ) family genes ( Figure 3D ) that are directly orthologous between both species , Genome-wide phylogenetic analysis of the luciferases , PACSs and ACSs genes indicates that Luc1 and Luc2 form two orthologous groups , and that the neighboring PACS and ACS genes near Luc1 form three major clades ( Figure 3C ) : Clade A , whose common ancestor and most extant members are ACSs , and Clades B and C whose common ancestors and most extant members are PACSs . Luc1 and Luc2 are highly conserved at the level of gene structure—both are composed of seven exons with completely conserved exon/intron boundaries ( Appendix 4—figure 4; Appendix 4—figure 5 ) , and most members of Clades A , B , and C also have seven exons . The exact syntenic and orthology relationships of the ACS and PACS genes adjacent to the Luc1 locus remains unclear , likely due to subsequent gene divergence and shuffling ( Figure 3C , D ) . Luc2 is located on a different linkage-group from Luc1 in P . pyralis and on a different scaffold from Luc1 in A . lateralis , consistent with the interpretation that Luc1 and Luc2 lie on different chromosomes in both firefly species . No PACS or ACS genes were found in the vicinity of Luc2 in either species . These data support that tandem gene duplication in a firefly ancestor gave rise to several ancestral PACS paralogs , one of which neofunctionalized in place to become the ancestral luciferase ( AncLuc ) ( Figure 3B ) . Prior to the divergence of the firefly subfamilies Lampyrinae and Luciolinae around 100 Mya ( Appendix 4 . 3 ) , this AncLuc duplicated , possibly via a long-range gene duplication event ( e . g . transposon mobilization ) , and then subfunctionalized in its transcript expression pattern to give rise to Luc2 , while the original AncLuc subfunctionalized in place to give rise to Luc1 ( Figure 3B ) . From the shared Luc gene clustering in both fireflies , we infer the structure of the pre Luc1/Luc2 duplication AncLuc locus contained one or more ACS genes ( Clade A ) , one or more PACS genes ( Clade B/C ) , and one or more MGST family genes ( Figure 3B ) . To resolve the number of origins of luciferase activity , and therefore bioluminescence , between fireflies and click beetles , we first identified the luciferase of I . luminosus luciferase ( IlumLuc ) , and compared its genomic context to the luciferases of P . pyralis and A . lateralis ( Figure 3D ) . Unlike some other described bioluminescent Elateridae , which have separate luciferases expressed in the dorsal prothorax and ventral abdominal lanterns ( Oba et al . , 2010a ) , we identified only a single luciferase in the I . luminosus genome which was highly expressed in both of the lanterns ( Figure 3C; Appendix 3 . 8 ) . The exon number and exon-intron splice junctions of IlumLuc are identical to those of firefly luciferases , but unlike the firefly luciferases which have short introns less than <100 bp long , IlumLuc has two long introns ( Appendix 4—figure 4 ) . We found several PACS genes in the I . luminosus genome which were related to IlumLuc and formed a clade ( Clade D ) specific to the Elateridae ( Figure 3C , D ) . IlumLuc lies on a 366 Kbp scaffold containing 18 other genes , including three related Clade D PACS genes ( Scaffold 13255; Figure 3D; Figure 4 ) ; however , the Clade D genes that are most closely related to IlumLuc are found on a separate 650 Kbp scaffold ( Scaffold 9864; Figure 3D ) . We infer that the IlumLuc locus is not orthologous to the extant firefly Luc1 locus , as IlumLuc is not physically clustered with Clade A , B or C ACS or PACS genes ( Figure 3C , D ) . We instead identified a different scaffold in I . luminosus that is likely orthologous to the firefly Luc1 locus ( Scaffold 9654; Figure 3D ) . This assessment is based on the presence of adjacent Clade A and B ACS and PACS genes , as well as orthologous exoribonuclease family ( PRNT ) and inositol monophosphatase family ( IMP ) genes , both of which were found adjacent to the A . lateralis Luc1 locus , but not the P . pyralis Luc1 locus ( Figure 3D ) . Interestingly , IlumPACS11 , the most early-diverging member of Clade D , was also found on Scaffold 9654 ( Figure 3D ) . This finding is consistent with an expansion of Clade D following duplication of the IlumPACS11 syntenic ancestor to a distant site . Overall , these genomic structures are consistent with independent origins of firefly and click beetle luciferases . We then carried out targeted molecular evolution analyses including the known beetle luciferases and their closely related homologs . Ancestral state reconstruction of luminescent activity on the gene tree using Mesquite ( Maddison and Maddison , 2017 ) recovered two independent gains of luminescence as the most parsimonious and likely scenario: once in click beetles , and once in the common ancestor of firefly , phengodid , and rhagophthalmid beetles ( Figure 4A; Appendix 4 . 3 . 3 ) . In an independent molecular adaptation analysis utilizing the coding nucleotide sequence of the elaterid luciferases and their close homologs within Elateridae , 35% of the sites of the branch leading to the ancestral click beetle luciferase showed a statistically significant signal of episodic positive selection with dN/dS > 1 ( ω or max dN/dS = 3 . 98 ) as compared to the evolution of its paralogs using the aBSREL branch-site selection test ( Smith et al . , 2015 ) ( Figure 4B; Appendix 4 . 3 . 4 ) . This implies that the common ancestor of the click beetle luciferases ( EAncLuc ) underwent a period of accelerated directional evolution . As the branch under selection in the molecular adaptation analysis ( Figure 4B ) is the same branch of luciferase activity gain via ancestral reconstruction ( Figure 4A ) , we conclude that the identified selection signal represents the relatively recent neofunctionalization of click beetle luciferase from a non-luminous ancestral Clade D PACS gene , distinct from the more ancient neofunctionalization of firefly luciferase . Based on the constraints from our tree , we determine that this neofunctionalization of EAncLuc occured after the divergence of the elaterid subfamily Agrypninae . In contrast , we cannot determine if the original neofunctionalization of AncLuc occurred in the ancestral firefly , or at some point during the evolution of ‘cantharoid’ beetles , an unofficial group of beetles including the luminous Rhagophthalmidae , Phengodidae and Lampyridae among other non-luminous groups , but not the Elateridae ( Branham and Wenzel , 2003 ) . There is evidence for a subsequent luciferase duplication event in phengodids , but not in rhagophthalmids , that is independent of the duplication event that gave rise to Luc1 and Luc2 in fireflies ( Figures 3C and 4 ) . Altogether , our results strongly support the independent neofunctionalization of luciferase activity in click beetles and fireflies , and therefore at least two independent gains of luciferin-utilizing luminescence in beetles . Beyond luciferase , we sought to characterize other metabolic traits which might have co-evolved in fireflies to support bioluminescence . Of particular importance , the enzymes of the de novo biosynthetic pathway for firefly luciferin remain unknown ( Oba et al . , 2013b ) . We hypothesized that bioluminescent accessory enzymes , either specialized enzymes with unique functions in luciferin metabolism or enzymes with primary metabolic functions relevant to bioluminescence , would be highly expressed ( HE: 90th percentile; Appendix 4 . 2 . 2 ) in the adult lantern , and would be differentially expressed ( DE; Appendix 4 . 2 . 2 ) between luminescent and non-luminescent tissues . To determine this , we performed RNA-Seq and expression analysis of the dissected P . pyralis and A . lateralis adult male lantern tissue compared with a non-luminescent tissue ( Appendix 4 . 2 . 2 ) . We identified a set of predicted orthologous enzyme-encoding genes conserved in both P . pyralis and A . lateralis that met our HE and DE criteria ( Figure 5 ) . Both luciferase and luciferin sulfotransferase ( LST ) , a specialized enzyme recently implicated in luciferin storage in P . pyralis ( Fallon et al . , 2016 ) , were recovered as candidate genes using four criteria ( HE , DE , enzymes , direct orthology across species ) , confirming the validity of our approach . While a direct ortholog of LST is present in A . lateralis , it is absent from I . luminosus , suggesting that LST , and the presumed luciferin storage it mediates , is an exclusive ancestral firefly or cantharoid trait . This finding is consistent with previous hypotheses of the absence of LST in Elateridae ( Fallon et al . , 2016 ) , and with the overall hypothesis of independent evolution of bioluminescence between the Lampyridae and Elateridae . Moreover , we identified several additional enzyme-encoding HE and DE lantern genes that are likely important in firefly lantern physiology ( Figure 5 ) . For instance , adenylate kinase likely plays a critical role in efficient recycling of AMP post-luminescence , and cystathionine gamma-lyase supports a key role of cysteine in luciferin biosynthesis ( Oba et al . , 2013b ) and recycling ( Okada et al . , 1974 ) . We also detected a combined adenylyl-sulfate kinase and sulfate adenylyltransferase enzyme ( ASKSA ) among the lantern-enriched gene list ( Appendix 4—figure 8 ) , implicating active biosynthesis of 3'-phosphoadenosine-5'-phosphosulfate ( PAPS ) , the cofactor of LST , in the lantern . This finding highlights the importance of LST-catalyzed luciferin sulfonation for bioluminescence . These firefly orthologs of ASKSA are the only members amongst their paralogs to contain a PTS1 ( Appendix 4—figure 8 ) , suggesting specialized localization to the peroxisome , the location of the luminescence reaction . This suggests that the levels of sulfoluciferin and luciferin may be actively regulated within the peroxisome of lantern cells in response to luminescence . Overall , our findings of several directly orthologous enzymes that share expression patterns in the light organs of both P . pyralis and A . lateralis suggests that the enzymatic physiology and/or the gene expression patterns of the photocytes were already fixed in the Luciolinae-Lampyrinae ancestor . We also performed a similar expression analysis for genes not annotated as enzymes , yielding several genes with predicted lysosomal function ( Appendix 4—table 6; Appendix 4 . 4 ) . This suggests that the abundant but as yet unidentified ‘differentiated zone granule’ organelles of the firefly light organ ( Ghiradella and Schmidt , 2004 ) could be lysosomes . Interestingly , we found a HE ( TPM value ~300 ) and DE opsin , Rh7 , in the light organ of A . lateralis , but not P . pyralis ( Appendix 4—figure 9; Appendix 4 . 5 ) , suggesting a potential light perception role for Rh7 in the A . lateralis lantern , akin to the light perception role described for Drosophila Rh7 ( Ni et al . , 2017 ) . Firefly bioluminescence is postulated to have first evolved as an aposematic warning of larval chemical defenses ( Branham and Wenzel , 2003 ) . Lucibufagins are abundant unpalatable defense steroids described from certain North American firefly species , most notably in the genera Photinus ( Meinwald et al . , 1979 ) , Lucidota ( Gronquist et al . , 2005 ) , and Ellychnia ( Smedley et al . , 2017 ) , and hence are candidates for ancestral firefly defense compounds . To test whether lucibufagins are widespread among bioluminescent beetles , we assessed the presence of lucibufagins in P . pyralis , A . lateralis , and I . luminosus by liquid-chromatography high-resolution accurate-mass mass-spectrometry ( LC-HRAM-MS ) . While lucibufagins were found in high abundance in P . pyralis adult hemolymph , they were not observed in A . lateralis adult hemolymph , nor in I . luminosus metathorax extract ( Figure 6B; Appendix 4 . 6 ) . Since chemical defense is presumably most critical in the long-lived larval stage , we next tested whether lucibufagins are present in all firefly larvae even if they are not present in the adults of certain species . We found lucibufagins in P . pyralis larval extracts; however , they were not observed in A . lateralis larval extracts ( Figure 6B; Appendix 4 . 6 ) . Together , these results suggest that the lucibufagin biosynthetic pathway is either a derived trait only found in particular firefly taxa ( e . g . subfamily: Lampyrinae ) , or that lucibufagin biosynthesis was an ancestral trait that was lost in A . lateralis . Consistent with the former hypothesis , the presence of lucibufagins in non-North-American Lampyrinae has been previously reported ( Tyler et al . , 2008 ) , but to date there are no reports of lucibufagins in the Luciolinae . The lucibufagin biosynthetic pathway is currently unknown . However , their chemical structure suggests a biosynthetic origin from cholesterol followed by a series of hydroxylations , -OH acetylations , and the side-chain oxidative pyrone formation ( Figure 6A ) ( Meinwald et al . , 1979 ) . We hypothesized that cytochrome P450s , an enzyme family widely involved in metabolic diversification of organic substrates ( Hamberger and Bak , 2013 ) , could underlie several oxidative reactions in the proposed lucibufagin biosynthetic pathway . We therefore inferred the P450 phylogeny among our three bioluminescent beetle genomes to identify any lineage-specific genes correlated with lucibufagin presence . Our analysis revealed a unique expansion of one P450 family , the CYP303 family , in P . pyralis . While 94/97 of currently sequenced winged-insect genomes on OrthoDB ( Zdobnov et al . , 2017 ) , as well as the A . lateralis and I . luminosus genomes , contain only a single CYP303 family gene , the P . pyralis genome contains 11 CYP303 genes and two pseudogenes ( Figure 6C ) , which expanded via tandem duplication on the same linkage group ( Figure 6D ) . The CYP303 ortholog of D . melanogaster , CYP303A1 , has been shown to play a role in mechanosensory bristle development ( Willingham and Keil , 2004 ) . Although the exact biochemical function and substrate of D . melanogaster CYP303A1 is unknown , its closely related P450 families operate on an insect steroid hormone ecdysone ( Willingham and Keil , 2004 ) . As ecdysone and lucibufagins are structurally similar , CYP303 may operate on steroid-like compounds . Therefore , the lineage-specific expansion of the CYP303 family in P . pyralis is a compelling candidate in the metabolic evolution of lucibufagins as chemical defenses associated with the aposematic role of bioluminescence . Alternatively , this CYP303 expansion in P . pyralis may be associated with other lineage-specific chemical traits , such as pheromone production . Given the increasingly recognized contributions of symbionts to host metabolism ( Newman and Cragg , 2015 ) , we characterized the hologenome of all three beetles as potential contributors to metabolic processes related to bioluminescence . Whole genome sequencing of our wild-caught and laboratory reared fireflies revealed a rich microbiome . Amongst our firefly genomes , we found various bacterial genomes , viral genomes , and the complete mtDNA for a phorid parasitoid fly , Apocephalus antennatus , the first mtDNA reported for genus Apocephalus . This mtDNA was inadvertently included in the P . pyralis PacBio library via undetected parasitization of the initial specimens , and was assembled via a metagenomic approach ( Appendix 5 . 2 ) . Independent collection of A . antennatus which emerged from field-collected P . pyralis adults and targeted COI sequencing later confirmed the taxonomic origin of this mtDNA ( Appendix 5 . 3 ) . We also sequenced and metagenomically assembled the complete circular genome ( 1 . 29 Mbp , GC: 29 . 7%; ~50x coverage ) for a P . pyralis-associated mollicute ( Phylum: Tenericutes ) , Entomoplasma luminosum subsp . pyralis ( Appendix 5 . 1 ) . Entomoplasma spp . were first isolated from the guts of North American fireflies ( Hackett et al . , 1992 ) and our assembly provides the first complete genomic assembly of any Entomoplasma species . Broad read coverage for the E . luminosus subsp . pyralis genome was detected in 5/6 of our P . pyralis DNA libraries , suggesting that Entomplasma is a highly prevalent , possibly vertically inherited , P . pyralis symbiont . It has been hypothesized that these Entomoplasma mollicutes could play a role in firefly metabolism , specifically via contributing to cholesterol metabolism and lucibufagin biosynthesis ( Smedley et al . , 2017 ) . Within our unfiltered A . lateralis genomic assembly ( Alat1 . 2 ) , we also found 43 scaffolds ( 2 . 3 Mbp; GC:29 . 8% , ~64x coverage ) , whose taxonomic annotation corresponded to the Tenericutes ( Appendix 2 . 5 . 2 ) , suggesting that A . lateralis may also harbor a mollicute symbiont . Alat1 . 2 also contains 2119 scaffolds ( 13 . 0 Mbp , GC:63 . 7% , ~25x coverage ) annotated as of Proteobacterial origin . Limited Proteobacterial symbionts were detected in the I . luminosus assembly ( 0 . 4 Mbp; GC:30–65% ~10x coverage ) ( Appendix 3 . 5 . 2 ) , suggesting no stable symbiont is present in adult I . luminosus . Lastly , we detected two species of novel orthomyxoviridae-like ssRNA viruses , which we dub Photinus pyralis orthomyxo-like virus 1 and 2 ( PpyrOMLV1/2 ) , that were highly prevalent across our P . pyralis RNA-Seq datasets , and showed multi-generational transovarial transmission in the laboratory ( Appendix 5 . 4 ) . We also found several endogenous viral elements ( EVEs ) for PpyrOMLV1/2 in P . pyralis ( Appendix 5 . 5 ) . These viruses are the first reported in any firefly species , and represent only the second report of transgenerational transfer of any Orthomyxoviridae virus ( Marshall et al . , 2014 ) , and the second report of Orthomyxoviridae derived EVEs ( Katzourakis and Gifford , 2010 ) . Together , these genomes from the firefly holobiont provide valuable resources for the continued inquiry of the symbiotic associates of fireflies and their biological and ecological significance .
Here , we generated genome assembles , diverse tissue and life-stage RNA-Seq data , and LC/MS data for three evolutionarily informative and historically well-studied bioluminescent beetles , and used a series of comparative analyses to illuminate long-standing questions on the origins and evolution of beetle bioluminescence . By analyzing the genomic synteny and molecular evolution of the beetle luciferases and their extant and inferred-ancestral homologs , we found strong support for the independent origins of luciferase , and therefore bioluminescence , between fireflies and click beetles . Our approaches and analyses lend molecular evidence to the previous morphology-phylogeny based hypotheses of parallel gain proposed by Darwin and others ( Darwin , 1872; Branham and Wenzel , 2003; Costa , 1975; Sagegami-Oba et al . , 2007; Bocakova et al . , 2007; Oba , 2009; Day , 2013 ) . While our elaterid luciferase selection analysis strongly supports an independent gain , we did not perform an analogous selection analysis of luciferase homologs across all bioluminescent beetles , due to the lack of genomic data from key related beetle families . Additional genomic information from early-diverged firefly lineages , other luminous beetle taxa ( e . g . Phengodidae and Rhagophthalmidae ) , and non-luminous elateroid taxa ( e . g . Cantharidae and Lycidae ) , will be useful to further develop and test models of luciferase evolution , including the hypothesis that bioluminescence also originated independently in the Phengodidae and/or Rhagophthalmidae . As some phylogenetic relationships of fireflies and other lineages of superfamily Elateroidea remain uncertain , continued efforts to produce reference phylogeny for these taxa are required ( Martin et al . , 2017; Bocak et al . , 2018 ) . Toward this goal , the recently published Pyrocoelia pectoralis Lampyrinae firefly genome is an important advance which will contribute to future phylogenetic and evolutionary studies ( Fu et al . , 2017 ) . The independent origins of the firefly and click beetle luciferases provide an exemplary natural model system to understand enzyme evolution through parallel mutational trajectories and the evolution of complex metabolic traits generally . The abundance of gene duplication events of PACSs and ACSs at the ancestral luciferase locus in both fireflies and I . luminosus suggests that ancestral promiscuous enzymatic activities served as raw materials for the selection of new adaptive catalytic functions ( Weng , 2014 ) . But while parallel evolution of luciferase implies evolutionary independence of bioluminescence overall , the reality may be more complex , and the other subtraits of bioluminescence amongst the bioluminescent beetles likely possess different evolutionary histories from luciferase . While subtraits presumably dependent on an efficient luciferase , such as specialized tissues and neural control , almost certainly arose well after luciferase specialization , and thus can be inferred to also have independent origins between fireflies and click beetles , luciferin , which was presumably a prerequisite to luciferase neofunctionalization , may have been present in their common ancestor . Microbial endosymbionts , such as the tenericutes detected in our P . pyralis and A . lateralis datasets , are intriguing candidate contributors to luciferin metabolism and biosynthesis . Alternatively , recent reports have shown that firefly luciferin is readily produced non-enzymatically by mixing benzoquinone and cysteine ( Kanie et al . , 2016 ) , and that a compound resulting from the spontaneous coupling of benzoquinone and cysteine acts as a luciferin biosynthetic intermediate in A . lateralis ( Kanie et al . , 2018 ) . Benzoquinone is known to be a defense compound of distantly related beetles ( Dettner , 1987 ) and other arthropods ( e . g . millipedes ) ( Shear , 2015 ) . Therefore , the evolutionary role of sporadic low-level luciferin synthesis through spontaneous chemical reactions , either in the ancestral bioluminescent taxa themselves , or in non-bioluminescent taxa , and dietary acquisition of luciferin by either the ancestral or modern bioluminescent taxa , should be considered . To decipher between these alternative evolutionary possibilities , the discovery of genes involved in luciferin metabolism in fireflies and other bioluminescent beetles will be essential . Here , as a first step toward that goal , we identified conserved , enriched and highly expressed enzymes of the firefly lantern that are strong candidates in luciferin metabolism and the elusive luciferin de novo biosynthetic pathway . Ultimately focused experimentation will be needed to decipher the biochemical function of these enzymes . The early evolution of firefly bioluminescence was likely associated with an aposematic role . The adaptive light production of the primordial firefly ( or alternatively , a primordial bioluminescent cantharoid beetle ) that enabled the selection and neofunctionalization of luciferase was perhaps linked to a response to predators by a primitive whole-body oxygen-gated luminescence , where a startle-response mediated increase in hemolymph oxygenation through spiracle opening and escape locomotion caused a concomitant increase in luminescence ( Buck and Case , 2002; Case , 2004 ) . Alternatively , an early role for firefly luminescence in mate attraction has not been ruled out ( Buck and Case , 2002 ) . The presence of particular unpalatable defense compounds in all extant fireflies would be consistent with an ancestral role and the former hypothesis , and the chemical analysis of tissues across species and life stages presented in this work provides new insights into the evolutionary occurrence of lucibufagins , the most well-studied defense compounds associated with fireflies . Our results reject lucibufagins as ancestral defense compounds of fireflies , but rather suggest them as a derived metabolic trait associated with Lampyrinae . Additional chemical analyses across more lineages of fireflies are needed , however , to further support or falsify this hypothesis . Toward this goal , the high sensitivity of our LC-HRAM-MS and MS2 molecular networking-based lucibufagin identification approach is particularly well suited to broadened sampling in the future , including those of rare taxa and possibly museum specimens . Combined with genomic data showing a concomitant expansion of the CYP303 gene family in P . pyralis , we present a promising path toward elucidating the biosynthetic mechanism underlying these potent firefly toxins . Overall , the resources and analyses generated in this study shed valuable light on the evolutionary questions Darwin first pondered , and will enable future studies of the ecology , behavior , and evolution of bioluminescent beetles . These resources will also accelerate the discovery of new enzymes from bioluminescent beetles that could enhance biotechnological applications of bioluminescence . Finally , we hope that the genomic resources shared here will facilitate the development of effective population genomic tools to monitor and protect wild bioluminescent beetle populations in the face of changing climate and habitats .
Genomic assemblies ( Ppyr1 . 3 , Alat1 . 3 , and Ilumi1 . 2 ) , associated official geneset data , a SequenceServer ( Priyam et al . , 2015 ) BLAST server , and a JBrowse ( Skinner et al . , 2009 ) genome browser are available at www . fireflybase . org . Raw genomic and RNA-Seq reads for P . pyralis , A . lateralis , and I . luminosus , are available under the NCBI/EBI/DDBJ BioProjects PRJNA378805 , PRJDB6460 , and PRJNA418169 respectively . Raw WGBS reads can be found on the NCBI Gene Expression Omnibus ( GSE107177 ) . Mitochondrial genomes for P . pyralis and I . luminosus and A . antennatus are available on NCBI GenBank with accessions KY778696 , MG242621 , and MG546669 . The complete genome of Entomoplasma luminosum subsp . pyralis is available on NCBI GenBank with accession CP027019 . The viral genomes for Photinus pyralis orthomyxo-like virus 1 and 2 are available on NCBI Genbank with accessions MG972985-MG972994 . LC-MS data is available on MetaboLights ( Accession MTBLS698 ) . Other supporting datasets are available on FigShare ( Appendix 6 . 1 ) . | Glowing fireflies dancing in the dark are one of the most enchanting sights of a warm summer night . Their light signals are ‘love messages’ that help the insects find a mate – yet , they also warn a potential predator that these beetles have powerful chemical defenses . The light comes from a specialized organ of the firefly where a small molecule , luciferin , is broken down by the enzyme luciferase . Fireflies are an ancient group , with the common ancestor of the two main lineages originating over 100 million years ago . But fireflies are not the only insects that produce light: certain click beetles are also bioluminescent . Fireflies and click beetles are closely related , and they both use identical luciferin and similar luciferases to create light . This would suggest that bioluminescence was already present in the common ancestor of the two families . However , the specialized organs in which the chemical reactions take place are entirely different , which would indicate that the ability to produce light arose independently in each group . Here , Fallon , Lower et al . try to resolve this discrepancy and to find out how many times bioluminescence evolved in beetles . This required using cutting-edge DNA sequencing to carefully piece together the genomes of two species of fireflies ( Photinus pyralis and Aquatica lateralis ) and one species of click beetle ( Ignelater luminosus ) . The genetic analysis revealed that , in all species , the genes for luciferases were very similar to the genetic sequences around them , which code for proteins that break down fat . This indicates that the ancestral luciferase arose from one of these metabolic genes getting duplicated , and then one of the copies evolving a new role . However , the genes for luciferase were very different between the fireflies and the click beetles . Further analyses suggested that bioluminescence evolved at least twice: once in an ancestor of fireflies , and once in the ancestor of the bioluminescent click beetles . More results came from the reconstituted genomes . For example , Fallon , Lower et al . identified the genes ‘turned on’ in the bioluminescent organ of the fireflies . This made it possible to list genes that may be involved in creating luciferin , and enable flies to grow brightly for long periods . In addition , the genetic information yielded sequences from bacteria that likely live inside firefly cells , and which may participate in the light-making process or the production of potent chemical defenses . Better genetic knowledge of beetle bioluminescence could bring new advances for both insects and humans . It may help researchers find and design better light-emitting molecules useful to track and quantify proteins of interest in a cell . Ultimately , it would allow a detailed understanding of firefly populations around the world , which could contribute to firefly ecotourism and help to protect these glowing insects from increasing environmental threats . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"genetics",
"and",
"genomics"
] | 2018 | Firefly genomes illuminate parallel origins of bioluminescence in beetles |
Paranoia is the belief that harm is intended by others . It may arise from selective pressures to infer and avoid social threats , particularly in ambiguous or changing circumstances . We propose that uncertainty may be sufficient to elicit learning differences in paranoid individuals , without social threat . We used reversal learning behavior and computational modeling to estimate belief updating across individuals with and without mental illness , online participants , and rats chronically exposed to methamphetamine , an elicitor of paranoia in humans . Paranoia is associated with a stronger prior on volatility , accompanied by elevated sensitivity to perceived changes in the task environment . Methamphetamine exposure in rats recapitulates this impaired uncertainty-driven belief updating and rigid anticipation of a volatile environment . Our work provides evidence of fundamental , domain-general learning differences in paranoid individuals . This paradigm enables further assessment of the interplay between uncertainty and belief-updating across individuals and species .
Paranoia is excessive concern that harm will occur due to deliberate actions of others ( Freeman and Garety , 2000 ) . It manifests along a continuum of increasing severity ( Freeman et al . , 2005; Freeman et al . , 2010; Freeman et al . , 2011; Bebbington et al . , 2013 ) . Fleeting paranoid thoughts prevail in the general population ( Freeman , 2006 ) . A survey of over 7000 individuals found that nearly 20% believed people were against them at times in the past year; approximately 8% felt people had intentionally acted to harm them ( Freeman et al . , 2011 ) . At a national level , paranoia may fuel divisive ideological intolerance . Historian Richard Hofstadter famously described catastrophizing , context insensitive political discourse as the ‘paranoid style’: “The paranoid spokesman sees the fate of conspiracy in apocalyptic terms—he traffics in the birth and death of whole worlds , whole political orders , whole systems of human values . He is always manning the barricades of civilization . He constantly lives at a turning point [emphasis added] . ” ( Hofstadter , 1964 ) . At its most severe , paranoia manifests as rigid beliefs known as delusions of persecution . These delusions occur in nearly 90% of first episode psychosis patients ( Freeman , 2007 ) . Psychostimulants also elicit severe paranoid states . Methamphetamine evokes new paranoid ideation particularly after repeated exposure or escalating doses ( 86% and 68% , respectively , in a survey of methamphetamine users ) ( Leamon et al . , 2010 ) . Paranoia has thus far defied explanation in mechanistic terms . Sophisticated Game Theory driven approaches ( such as the Dictator Game [Raihani and Bell , 2018; Raihani and Bell , 2017] ) have largely re-described the phenomenon — people who are paranoid have difficulties in laboratory tasks that require trust ( Raihani and Bell , 2019 ) . However , this is not driven by personal threat per se , but by negative representations of others ( Raihani and Bell , 2018; Raihani and Bell , 2017 ) . We posit that such representations are learned ( Fineberg et al . , 2014; Behrens et al . , 2008 ) , via the same fundamental learning mechanisms ( Cramer et al . , 2002 ) that underwrite non-social learning in non-human species ( Heyes and Pearce , 2015 ) . We hypothesize that aberrations to these domain-general learning mechanisms underlie paranoia . One such mechanism involves the judicious use of uncertainty to update beliefs: Expectations about the noisiness of the environment constrain whether we update beliefs or dismiss surprises as probabilistic anomalies . The higher the expected uncertainty ( i . e . , ‘I expect variable outcomes’ ) , the less surprising an atypical outcome may be , and the less it drives belief updates ( ‘this variation is normal’ ) . Unexpected uncertainty , in contrast , describes perceived change in the underlying statistics of the environment ( Yu and Dayan , 2005; Payzan-LeNestour and Bossaerts , 2011; Payzan-LeNestour et al . , 2013 ) ( i . e . ‘the world is changing’ ) , which may call for belief revision . Since excessive unexpected uncertainty is a signal of change , it might drive the recategorization of allies as enemies , which is a tenet of evolutionary theories of paranoia ( Raihani and Bell , 2019 ) . We tested the hypothesis that this drive to flexibly recategorize associations extends to non-social , domain-general inferences . We dissected learning mechanisms under expected and unexpected uncertainty – probabilistic variation and changes in underlying task structure ( volatility ) . Here , volatility is a property of the task . Unexpected uncertainty is the perception of that volatility . Participants completed a non-social , three-option learning task which challenged them to form and revise associations between stimuli ( colored card decks ) and outcomes ( points rewarded and lost ) , in addition to their beliefs about the volatility of the task environment . They encountered expected uncertainty as probabilistic win or loss feedback ( ‘each option yields positive and negative outcomes , but in different amounts’ ) , and unexpected uncertainty as reassignment of reward probabilities between options ( ‘sometimes the best option may change , ’ reversal events ) . Although reversal events elicit unexpected uncertainty by driving re-evaluation of the options , participants increasingly anticipate reversals and develop expectations about the stability of the task environment . We implemented an additional task manipulation: a shift in the underlying probabilities themselves ( contingency transition , unsignaled to the participants ) , that effectively changes task volatility . Armed with the task structure and participants’ choices , we applied a Hierarchical Gaussian Filter ( HGF ) model ( Mathys et al . , 2011; Mathys et al . , 2014 ) which allowed us to infer participants’ initial beliefs ( i . e . , priors ) about task volatility , their readiness to learn about changes in the task volatility itself ( meta-volatility learning rate ) and learning rates that captured their expected and unexpected uncertainty regarding the task . We examined the behavioral and computational correlates of paranoia both in-person and in a large online sample , spanning patients and healthy controls with varying degrees of paranoia . We also undertook a pre-clinical replication in rodents exposed chronically to saline or methamphetamine to determine whether a drug known to elicit paranoia in humans might induce similar perceptions of unexpected uncertainty , without contingency transition ( Groman et al . , 2018 ) . We predicted that people with paranoia and rats administered methamphetamine would exhibit stronger priors on volatility , facilitating aberrant learning through unexpected uncertainty . We further hypothesized that this learning style would manifest as frequent and unnecessary choice switching ( increased choice stochasticity and ‘win-switch’ behavior ) rather than increased sensitivity to negative feedback ( increased ‘lose-switch’ behavior/decreased ‘lose-stay’ behavior ) .
First , we explored trans-diagnostic associations between paranoia and reversal-learning in-person . Participants with and without psychiatric diagnoses ( mood disorders: anxiety , depression , bipolar disorder , n = 8; schizophrenia spectrum: schizophrenia or schizoaffective disorder , n = 8; and healthy controls , n = 16 ) , completed questionnaire versions of the Structured Clinical Interview for DSM-IV Axis II Personality Disorders ( SCID-II ) screening assessment ( Ryder et al . , 2007 ) , Beck’s Anxiety Inventory ( BAI ) ( Beck et al . , 1988 ) , Beck’s Depression Inventory ( BDI ) ( Beck et al . , 1961 ) , and demographic assessments ( Table 1 ) . Approximately two-thirds of participants endorsed three or fewer items on the SCID-II paranoid personality subscale ( median = 1 item ) . Participants who endorsed four or more items were classified as high paranoia ( n = 11 ) , consistent with the diagnostic threshold for paranoid personality disorder . Low paranoia ( n = 21 ) and high paranoia groups did not differ significantly by age , nor were there significant group associations with gender , educational attainment , ethnicity , or race , although a larger percentage of paranoid participants identified as racial minorities or ‘not specified’ ( Table 1 ) . Diagnostic category ( i . e . , healthy control , mood disorder , or schizophrenia spectrum ) was significantly associated with paranoia group membership , χ2 ( 2 , n = 32 ) =12 . 329 , p=0 . 002 , Cramer’s V = 0 . 621 , as was psychiatric medication usage , χ2 ( 1 , n = 32 ) =9 . 871 , p=0 . 003 , Cramer’s V = 0 . 555 . These differences were due to the higher proportion of healthy controls in the low paranoia group . As expected , paranoia , BAI , and BDI scores were significantly elevated in the high paranoia group relative to low paranoia controls ( Table 1; paranoia: mean difference ( MD ) = 0 . 536 , CI=[0 . 455 , 0 . 618] , t ( 30 ) =13 . 476 , p=2 . 92E-14 , Hedges’ g = 5 . 016; BAI: MD = 0 . 585 , CI=[0 . 239 , 0 . 931] , t ( 30 ) =3 . 453 , p=0 . 002 , Hedges’ g = 1 . 285 , MD = −0 . 585; BDI: MD = 0 . 427 , CI=[0 . 078 , 0 . 775] , t ( 11 . 854 ) = 2 . 67 , p=0 . 021 , Hedges’ g = 1 . 255 ) . Participants completed a three-option reversal-learning task in which they chose between three decks of cards with hidden reward probabilities ( Figure 1a and b ) . They selected a deck on each turn and received positive or negative feedback ( +100 or −50 points , respectively ) . They were instructed to find the best deck with the caveat that the best deck may change . Undisclosed to participants , reward probabilities switched among decks after selection of the highest probability option in nine out of ten consecutive trials ( ‘reversal events’ ) . Thus , the task was designed to elicit expected uncertainty ( probabilistic reward associations ) and unexpected uncertainty ( reversal events ) , requiring participants to distinguish probabilistic losses from change in the underlying deck values . In addition , reward contingencies changed from 90% , 50% , and 10% chance of reward to 80% , 40% , and 20% between the first and second halves of the task ( ‘contingency transition’; block 1 = 80 trials , 90-50–10%; block 2 = 80 trials , 80-40–20% , unsignaled to the participants ) . This transition altered the volatility of the task environment , thereby making it more difficult to achieve reversals and often delaying their occurrence . Successful achievement of reversals was contingent upon adapting stay-vs-switch strategies , thereby testing subjects’ abilities to update beliefs about the overall task volatility ( ‘metavolatility learning’ ) . High paranoia subjects achieved fewer reversals ( MD = −2 . 31 , CI=[−4 . 504 , –0 . 111 , ] , t ( 30 ) =-2 . 145 , p=0 . 04 , Hedges’ g = 0 . 798 ) , but total points earned did not significantly differ , suggesting that there was no penalty for the different behaviors expressed by the more paranoid subjects ( Table 1 ) . We predicted that paranoia would be associated with unexpected uncertainty-driven belief updating . We aimed to replicate and extend our investigation of paranoia and reversal-learning in a larger online sample . We administered three alternative task versions to control for the contingency transition ( Figure 1c ) . Version 1 ( n = 45 low paranoia , 20 high paranoia ) provided a constant contingency of 90-50–10% reward probabilities ( Easy-Easy ) ; version 2 ( n = 69 low paranoia , 18 high paranoia ) provided a constant contingency of 80-40–20% ( Hard-Hard ) ; version 3 ( n = 56 low paranoia , 16 high paranoia ) served to replicate Experiment 1 with a contingency transition from 90-50–10% to 80-40–20% ( Easy-Hard ) ; version 4 ( n = 64 low paranoia , 19 high paranoia ) provided the reverse contingency transition , 80-40–20% to 90-50–10% ( Hard-Easy ) . The stable contingencies ( versions 1 and 2 ) lacked contingency transitions . Versions 3 and 4 manipulated task volatility mid-way , although the contingency transition was not signalled to participants . We predicted that high paranoia participants would find versions 3 and 4 particularly challenging . Given that version 3 is easier to learn initially , we expected participants to develop stronger priors and thus be more confounded by the contingency transition , compared to version four participants . Participants’ demographic and mental health questionnaire responses did not differ significantly across task version experiments ( Table 2 ) . Total points and reversals achieved suggest variations in task difficulty ( Table 2 , version effects: points earned , F ( 3 , 299 ) =32 . 288 , p=4 . 16E-18 , ηp2=0 . 245; reversals achieved , F ( 3 , 299 ) =4 . 329 , p=0 . 005 , ηp2=0 . 042 ) , but there was no significant association between task version and attrition rate ( 52 . 7% , 52 . 9% , 54 . 6% , and 53 . 1% attrition , respectively; χ2 ( 3 , n = 752 ) =0 . 167 , p=0 . 983 , Cramer’s V = 0 . 015 ) . Across task versions , high paranoia participants endorsed higher BAI and BDI scores ( n = 73 high paranoia , 234 low paranoia; BAI: F ( 1 , 299 ) =38 . 752 , p=1 . 63E-09 , ηp2=0 . 115; BDI: F ( 1 , 299 ) =74 . 528 , p=3 . 62E-16 , ηp2=0 . 20; Table 2 ) . Both correlated with paranoia ( BAI: Pearson’s r = 0 . 450 , p=1 . 09E-16 , CI=[0 . 348 , 0 . 55]; BDI: Pearson’s r = 0 . 543 , p=6 . 26E-25 , CI=[0 . 448 , 0 . 638] ) . Trial-by-trial reaction time did not differ significantly between low and high paranoia ( Table 2 ) , but high paranoia participants earned fewer total points ( F ( 1 , 299 ) =6 . 175 , p=0 . 014 , ηp2=0 . 020 ) and achieved fewer reversals ( F ( 1 , 299 ) =5 . 762 , p=0 . 017 , ηp2=0 . 019; Table 2 ) . Deck choice perseveration after negative feedback ( lose-stay behavior ) did not significantly differ by paranoia group , but choice switching after positive feedback ( win-switch behavior ) was elevated in high paranoia ( block 1: F ( 1 , 299 ) =7 . 117 , p=0 . 008 , ηp2=0 . 023; block 2: F ( 1 , 299 ) =9 . 918 , p=0 . 002 , ηp2=0 . 032; Table 2 ) . To translate across species , we performed a new analysis of published data from rats exposed to chronic methamphetamine ( Groman et al . , 2018 ) . Rats chose between three operant chamber noseports with differing probabilities of sucrose reward ( 70% , 30% , and 10%; Figure 1d and e ) . Contingencies switched between the 70% and 10% noseports after selection of the highest reinforced option in 21 out of 30 consecutive trials ( Figure 1e ) . This task was most similar in structure to the first blocks of online versions 2 and 4 . There was no increase in unexpected volatility mid-way through the task . Rats were tested for 26 within-session reversal blocks ( Pre-Rx , n = 10 per group ) , administered saline or methamphetamine according to a 23 day schedule mimicking the escalating doses and frequencies of chronic human methamphetamine users ( Groman et al . , 2018 ) , and tested once per week for four weeks following completion of the drug regimen ( Post-Rx; n = 10 saline , seven methamphetamine ) ( Groman et al . , 2018 ) . Relative to rats exposed to saline , those rats exposed to methamphetamine exhibited increased win-switch behavior , similar to what we has observed in the high paranoia human participants , and additionally , unlike humans , they perseverated after negative feedback ( Groman et al . , 2018 ) . We employed hierarchical Gaussian filter ( HGF ) modeling to compare belief updating across individuals with low and high paranoia , as well as across human participants and rats exposed to methamphetamine ( Table 3 ) . We paired a three-level perceptual model with a softmax decision model dependent upon third level volatility ( Figure 2a ) . We inverted the model from subject data ( trial-by-trial choices and feedback ) to estimate parameters for each individual ( Figure 2b ) . Level 1 ( x1 ) characterizes trial-by-trial perception of task feedback ( win or loss in humans , reward or no reward in rats ) , Level 2 ( x2 ) distinguishes stimulus-outcome associations ( deck or noseport values ) , and Level 3 ( x3 ) renders perception of the overall task volatility ( i . e . , frequency of reversal events , changes in the stimulus-outcome associations ) . Belief trajectories were unique to each subject due to the probabilistic , performance-dependent nature of the task , so we estimated initial beliefs ( priors ) for x2 and x3 ( μ20 and μ30 , respectively ) . We also estimated ω2 , the tonic volatility of stimulus-outcome associations . Lower ω2 indicates that subjects are slower to adjust beliefs about the value of each option; they maintain rigid beliefs about the underlying probabilities . The κ parameter captures the impact of phasic volatility on updating stimulus-outcome associations . In the setting of our experiments , κ approximates the influence of unexpected uncertainty . Higher κ implies faster updating of stimulus-outcome associations – that is , participants are more likely perceive volatility as reversal events . Our final parameter of interest , ω3 , characterizes perception of ‘meta-volatility , ’ such as changes in the frequency of reversal events ( Lawson et al . , 2017 ) . The lower ω3 , the slower a subject is to adjust their volatility belief; they adhere more rigidly to their volatility prior ( μ30 ) . Priors did not differ between groups at x2 ( Table 3 ) but paranoid individuals and rats exposed to methamphetamine exhibited elevated μ30 , they expected greater task volatility ( Figure 2b , blue ) . In Experiment 1 , we observed an interaction between task block and paranoia group ( F ( 1 , 30 ) =5 . 344 , p=0 . 028 , ηp2=0 . 151; Table 1 ) : μ30 differed between high and low paranoia in both blocks ( block 1 , F ( 1 , 30 ) =4 . 232 , p=0 . 048 , ηp2=0 . 124 , MD = 0 . 658 , CI=[0 . 005 , 1 . 312]; block 2 , F ( 1 , 30 ) =7 . 497 , p=0 . 010 , ηp2=0 . 20 , MD = 1 . 598 , CI=[0 . 406 , 2 . 789] ) , but only low paranoia subjects significantly updated their priors between block 1 and block 2 ( F ( 1 , 30 ) =39 . 841 , p=5 . 85E-07 , ηp2=0 . 570 , MD = 1 . 504 , CI=[1 . 017 , 1 . 99] ) . In Experiment 2 , the analogous task design ( version 3 ) demonstrated significant effects of block ( F ( 1 , 70 ) =64 . 652 , p=1 . 54E-11 , ηp2=0 . 480 , MD = 1 . 303 , CI=[0 . 980 , 1 . 627] ) and paranoia ( F ( 1 , 70 ) =6 . 366 , p=0 . 014 , ηp2=0 . 083 , MD = 0 . 909 , CI=[0 . 191 , 1 . 628]; Table 1 ) . Rats showed a similar effect following methamphetamine exposure with a significant time ( Pre-Rx , Post-Rx ) by treatment ( methamphetamine , saline ) interaction ( F ( 1 , 15 ) =5 . 159 , p=0 . 038 , ηp2=0 . 256; pre versus post methamphetamine effect: F ( 1 , 15 ) =12 . 186 , p=0 . 003 , MD = 1 . 265 , CI=[−0 . 493 , 2 . 037]; Pre-Rx mean [standard error]=−1 . 25 [0 . 56] saline , −0 . 77 [0 . 80] methamphetamine; Post-Rx: m = −0 . 69 [0 . 74] saline , 0 . 58 [0 . 73] methamphetamine ) . Random effects meta-analyses confirmed significant cross-experiment replication of elevated μ30 in human participants with paranoia ( in laboratory and online version 3; MDMETA = 1 . 110 , CI=[0 . 927 , 1 . 292] , zMETA = 11 . 929 , p=8 . 356E-33 ) and across humans with paranoia and rats exposed to methamphetamine ( MDMETA = 2 . 090 , CI=[0 . 123 , 4 . 056] , zMETA = 2 . 083 , p=0 . 037 ) . Both paranoid humans and rats administered chronic methamphetamine had strong beliefs that the task contingencies would change rapidly and unpredictably – in other words , they expected frequent reversal events . Methamphetamine exposure made rats behave like humans with high paranoia ( Figure 2b , Post-Rx condition , orange ) . This is particularly striking when compared to human data from the first task block ( before contingency transition ) , when task designs are most similar across experiments . Paranoid participants and methamphetamine exposed rats updated stimulus-outcome associations more strongly in response to perceived volatility ( e . g . , correctly or incorrectly inferred reversals; Figure 2b ) . κ showed significant paranoia group and block effects across the in laboratory experiment and online version 3 ( Table 1; paranoia effects , in laboratory: F ( 1 , 30 ) =7 . 599 , p=0 . 010 , ηp2=0 . 202 , MD = 0 . 081 , CI=[0 . 021 , 0 . 140]; online version 3: F ( 1 , 70 ) =13 . 521 , p=0 . 0005 , ηp2=0 . 162 , MD = 0 . 068 , CI=[0 . 031–0 . 104]; MDMETA = 0 . 079 , CI=[0 . 063 , 0 . 095] , zMETA = 9 . 502 p=2 . 067E-21 ) ; see Table 3 for block effects ) . κ increased from baseline in rats on methamphetamine , yielding significant effects of treatment ( F ( 1 , 15 ) =13 . 356 , p=0 . 002 , ηp2=0 . 471 , MD = 0 . 045 , CI=[0 . 019 , 0 . 072] ) and time ( F ( 1 , 15 ) =9 . 132 , p=0 . 009 , ηp2=0 . 378 , MD = 0 . 041 , CI=[0 . 012 , 0 . 069] ) ; however , the interaction between time and treatment did not reach statistical significance ( Table 3; Pre-Rx m = 0 . 499 [0 . 015] saline , 0 . 523 [0 . 040] methamphetamine; Post-Rx: m = 0 . 518 [0 . 053] saline , 0 . 585 [0 . 029] methamphetamine ) . Replication of group effects was significant across all three experiments ( MDMETA = 2 . 063 , CI=[0 . 341 , 3 . 785] , zMETA = 2 . 348 , p=0 . 019 ) . Thus , learning was more strongly driven by unexpected uncertainty in high paranoia participants and rats chronically administered methamphetamine; they were faster to interpret volatility as reversal events than their low paranoia and saline exposed counterparts . Expected uncertainty ( ω2 ) was decreased in paranoid participants and rats exposed to methamphetamine ( Figure 2b ) . In laboratory and online ( version 3 ) , paranoid individuals were slower to update stimulus-outcome associations in response to expected uncertainty ( Table 1; ω2 paranoia effect , in laboratory: F ( 1 , 30 ) =4 . 186 , p=0 . 050 , ηp2=0 . 122 , MD = −1 . 188 , CI=[−2 . 375 , –0 . 002]; online version 3: F ( 1 , 70 ) =8 . 7 , p=0 . 004 , ηp2=0 . 111 , MD = −0 . 993 , CI=[−1 . 665 , –0 . 322]; MDMETA = −1 . 154 , CI=[−1 . 455 , –0 . 853] , zMETA = −7 . 521 , p=5 . 450E-14 ) . The effects of methamphetamine exposure in rats were consistent ( MDMETA = −1 . 992 , CI=[−3 . 318 , –0 . 665] , zMETA = −2 . 943 , p=0 . 003 ) yet more striking , with a strongly negative ω2 accounting for the more pronounced lose-stay behavior or perseveration in rats ( time by treatment interaction , F ( 1 , 15 ) =18 . 454 , p=0 . 001 , ηp2=0 . 552; pre versus post methamphetamine: F ( 1 , 15 ) =42 . 242 , p=1 . 0E-522 , ηp2=0 . 738 , MD = −1 . 604 , CI=[−2 . 130 , –1 . 078]; Pre-Rx m = 0 . 198 [0 . 33] saline , −0 . 036 [0 . 42] methamphetamine; Post-Rx: m = −0 . 023 [0 . 56] saline , −1 . 640 [0 . 71] methamphetamine ) . High paranoia humans and rats exposed to methamphetamine maintained rigid beliefs about the underlying option probabilities relative to low paranoia and saline controls . This was associated with perseverative behavior in the rats but not in humans . Meta-volatility learning ( ω3 ) was similarly decreased across paranoia and methamphetamine exposed groups ( in laboratory , online version 3 , and rats: MDMETA = −1 . 155 , CI=[−2 . 139 , –0 . 171] , zMETA = −2 . 3 , p=0 . 021 ) , suggesting more reliance on expected task volatility ( i . e . , anticipated frequency of reversal events ) than on actual task feedback . In laboratory , we observed a block by paranoia group interaction ( Table 1 , F ( 1 , 30 ) =6 . 948 , p=0 . 010 , ηp2=0 . 188 ) . Post-hoc tests differentiated first and second blocks for the low paranoia group only ( F ( 1 , 30 ) =26 . 640 , p=1 . 5E-5 , ηp2=0 . 470 , MD = −0 . 876 , CI=[−1 . 222 , –0 . 529] ) . The paranoia effect did not reach statistical significance for online version 3 ( block effect only , F ( 1 , 70 ) =14 . 932 , p=0 . 0002 , ηp2=0 . 176 , MD = −0 . 692 , CI=[−1 . 050 , –0 . 335]; Table 3 ) , but meta-analytic random effects analysis confirms a significant paranoia group difference ( in laboratory and online version 3: MDMETA = −0 . 341 , CI=[−0 . 522 , –0 . 159] , zMETA = −3 . 68 , p=0 . 0002 ) . Methamphetamine exposure rendered ω3 more negative in rats ( time by treatment interaction , ( F ( 1 , 15 ) =9 . 058 , p=0 . 009 , ηp2=0 . 376; pre versus post methamphetamine: F ( 1 , 15 ) =30 . 668 , p=5 . 7E-5 , ηp2=0 . 672 , MD = −1 . 210 , CI=[−1 . 676 , –0 . 745]; Pre-Rx m = −0 . 692 [0 . 44] saline , −0 . 607 [0 . 51] methamphetamine; Post-Rx: m = −1 . 044 [0 . 44] saline , −1 . 817 [0 . 32] methamphetamine ) . These data indicate that paranoia and methamphetamine are associated with slower learning about changes in task volatility , suggesting greater reliance on volatility priors than task feedback . In summary , our modeling analyses suggest the following about paranoia in humans and methamphetamine exposed animals: they expect the task to be volatile ( high μ30 ) , their expectations about task volatility are more rigid ( low ω3 ) , and they confuse probabilistic errors and task volatility as a signal that the task has fundamentally changed ( high κ , low ω2 ) . We applied False Discovery Rate ( FDR ) correction for multiple comparisons of each model parameter ( Hochberg and Benjamini , 1990 ) . κ group effects survived corrections within each experiment ( Table 4 ) . In addition to κ , μ30 survived for experiment 1; μ30 and ω2 survived in online version 3; and μ30 , ω2 , and ω3 survived in experiment three as group effects . Such correction is not yet standard practice with this modeling approach ( Lawson et al . , 2017; Powers et al . , 2017; Sevgi et al . , 2016 ) but we believe it should be , and when effects survive correction we should increase our confidence in them . To examine the relationship between beliefs about contingency transition and paranoia within our HGF parameters , we performed split-plot , repeated measures ANOVAs across all four task versions . Paranoia group effects were specific to versions of the task in which we explicitly manipulated uncertainty via contingency transition which increased volatility ( Figure 3 , Table 5 , versions 3 and 4 ) . Specifically , we observed paranoia by version interactions for κ ( F ( 3 , 299 ) =4 . 178 , p=0 . 006 , ηp2=0 . 040 ) and ω2 ( F ( 3 , 299 ) =2 . 809 , p=0 . 040 , ηp2=0 . 027; Table 2 ) . Post-hoc tests confirmed that significant paranoia group effects were restricted to version 3 ( κ: F ( 1 , 299 ) =12 . 230 , p=0 . 001 , ηp2=0 . 039 , MD = 0 . 068 , CI=[0 . 03 , 0 . 106]; ω2: F ( 1 , 299 ) =8 . 734 , p=0 . 003 , ηp2=0 . 028 , MD = −0 . 993 , CI=[−1 . 655 , –0 . 332] ) and a trend for version 4 ( ω2: F ( 1 , 299 ) =2 . 909 , p=0 . 089 , ηp2=0 . 010 , MD = −0 . 528 , CI=[−1 . 138 , 0 . 081] , Figure 3a ) . μ30 also exhibited a paranoia by version trend ( Table 2 , F ( 3 , 299 ) =2 . 329 , p=0 . 075 , ηp2=0 . 023 ) , largely driven by version 3 ( F ( 1 , 299 ) =6 . 206 , p=0 . 013 , ηp2=0 . 020 , MD = 0 . 909 , CI=[0 . 191 , 1 . 628]; Figure 3a ) . There were no significant paranoia effects or interactions for ω3 ( Table 5 ) . In sum , our contingency shift manipulation – from easily discerned options to underlying probabilities that are closer together – increased unexpected uncertainty the most , particularly in highly paranoid participants , compared to the other task versions . We completed three ANCOVAs for each HGF parameter derived from Experiment 2: demographics ( age , gender , ethnicity , and race ) ; mental health factors ( medication usage , diagnostic category , BAI score , and BDI score ) ; and metrics and correlates of global cognitive ability ( educational attainment , income , and cognitive reflection; Tables 6 and 7 ) . For κ , our metric of unexpected uncertainty , the paranoia by version interaction remained robust across all three ANCOVAs ( demographics: F ( 3 , 294 ) =3 . 753 , p=0 . 011 , ηp2=0 . 037; mental health: F ( 3 , 257 ) =4 . 417 , p=0 . 005 , ηp2=0 . 049; cognitive: F ( 3 , 290 ) =4 . 304 , p=0 . 005 ηp2=0 . 043 ) . The paranoia by version trend of μ30 diminished with inclusion of demographic , mental health , and cognitive covariates ( demographic: F ( 3 , 294 ) =1 . 997 , p=0 . 119 , ηp2=0 . 020; mental health: F ( 3 , 257 ) =1 . 942 , p=0 . 123 , ηp2=0 . 022; cognitive: F ( 3 , 290 ) =2 . 193 , p=0 . 089 , ηp2=0 . 022 ) . The paranoia by version interaction for ω2 was robust to mental health and cognitive factors ( F ( 3 , 257 ) =3 . 617 , p=0 . 014 , ηp2=0 . 041; F ( 3 , 290 ) =3 . 017 , p=0 . 030 , ηp2=0 . 030 ) . A paranoia group effect and paranoia by version trend remained with inclusion of demographics ( ω2 , paranoia effect: F ( 1 , 294 ) =4 . 275 , p=0 . 040 , ηp2=0 . 014; interaction: F ( 3 , 294 ) =2 . 507 , p=0 . 059 , ηp2=0 . 025 ) . Thus κ – participants’ perception of unexpected uncertainty – was the only parameter whose main effect of paranoia ( higher κ in high paranoia participants ) and paranoia-by-version interaction ( higher κ in high paranoia participants as a function of increasing unexpected volatility in version 3 ) survived covariation for demographic , mental health and cognitive covariates . We are most confident that high paranoia participants have higher unexpected uncertainty which drives their excessive updating of stimulus-outcome associations . We found a significant correlation between κ and paranoia scores ( Figure 4 ) . However , depression and anxiety were also related to κ , and indeed , paranoia and depression correlate with one another , in our data and in other studies ( Na et al . , 2019 ) . In order to explore commonalities among the rating scales in the present data , we performed a principle components analysis ( Figure 5 ) , identifying three principle components . The first principle component ( PC 1 ) explained 82 . 3% of the variance in the scales and loaded similarly on anxiety , depression , and paranoia . It correlated significantly with kappa ( r = 0 . 272 , p=0 . 021 ) . Depression , anxiety and paranoia all contribute to PC1 . We suggest that this finding is consistent with the idea that depression and anxiety represent contexts in which paranoia can flourish and likewise , harboring a paranoid stance toward the world can induce depression and anxiety . In order to make the case that our observations were most relevant to paranoia , we examined the effects of paranoia , anxiety , and depression on κ within the online version three dataset with multiple regression . A significant regression equation was found ( F ( 3 , 68 ) =3 . 681 , p=0 . 016 ) , with an R ( Freeman et al . , 2005 ) of 0 . 140 . Participants’ predicted κ equaled 0 . 486 + 0 . 062 ( PARANOIA ) +0 . 012 ( BDI ) −0 . 006 ( BAI ) . Paranoia was a significant predictor of κ ( β = 0 . 343 , t = 2 . 470 , p=0 . 016 , CI=[0 . 012 , 0 . 113] ) but depression and anxiety were not ( BDI: β = 0 . 086 , t = 0 . 423 , p=0 . 674 , CI=[−0 . 043 , 0 . 066]; BAI: β = −0 . 043 , t = −0 . 218 , p=0 . 828 , CI=[−0 . 063 , 0 . 050] ) . Examination of correlation plots for κ ( Figure 4 ) revealed a much stronger relationship when analyses were restricted to individuals with paranoia scores greater than 0 ( i . e . , endorsement of at least one item ) ; among participants who denied all questionnaire items , a minority ( seven out of 32 ) exhibited elevated κ . To account for the possibility that some individuals with severe paranoia may avoid disclosing sensitive information , we performed additional analyses of participants who endorsed one or more paranoia item . The correlation between paranoia and κ in the first block of the task increases from r = 0 . 3 , p=0 . 011 , CI=[0 . 074 , 0 . 497] ( all participants , n = 72 ) to r = 0 . 588 , p=8 . 0E-5 , CI=[0 . 335 , 0 . 762] ( participants with paranoia >0 , n = 39 ) . In this subset , a significant regression equation was also found ( F ( 3 , 35 ) =6 . 322 , p=0 . 002 ) , with an R2of 0 . 351 ( Figure 4 ) . Participants’ predicted κ was equal to 0 . 432 + 0 . 150 ( PARANOIA ) +0 . 013 ( BDI ) −0 . 004 ( BAI ) . Paranoia was a significant predictor of κ ( β = 0 . 538 , t = 2 . 983 , p=0 . 005 , CI=[0 . 048 , 0 . 252] ) but depression and anxiety were not ( BDI: β = 0 . 111 , t = 0 . 494 , p=0 . 624 , CI=[−0 . 041 , 0 . 067]; BAI: β = −0 . 035 , t = −0 . 163 , p=0 . 872 , CI=[−0 . 057 , 0 . 049] ) . Thus , paranoia predicts kappa across participants . Anxiety and depression do not predict kappa . Win-switching was the prominent behavioral feature of both paranoid participants and rats exposed to methamphetamine ( Table 1 , Table 2; Groman et al . , 2018 ) . Collapsed across blocks and task versions , our Experiment 2 data demonstrated a main effect of paranoia group ( Figure 3b; F ( 1 , 299 ) =9 . 207 , p=0 . 003 , ηp2=0 . 030 , MD = 0 . 059 , CI=[0 . 021 , 0 . 097]; version trend: F ( 3299 ) =2 . 263 p=0 . 081 , ηp2=0 . 022; low paranoia: m = 0 . 06 [0 . 01] , high paranoia: m = 0 . 12 [0 . 02] ) . To elucidate whether this behavior was stochastic or predictable ( e . g . , switching back to a previously rewarding option ) , we calculated U-values ( Kong et al . , 2017 ) , a metric of behavioral variability employed by behavioral ecologists ( increasingly an inspiration for human behavioral analysis [Fung et al . , 2019] ) , particularly with regards to predator-prey relationships ( Humphries and Driver , 1970 ) . When a predator is approaching a prey animal , the prey’s best course of action is to behave randomly , or in a protean fashion , in order to evade capture ( Humphries and Driver , 1970 ) . The more protean or stochastic the behavior , the closer to the U-value is to 1 . Across task blocks , paranoid participants exhibited elevated choice stochasticity ( paranoia by version interaction , F ( 3 , 298 ) =3 . 438 , p=0 . 017 , ηp2=0 . 033; Table 2 ) . Post-hoc tests indicate that this stochasticity was specific to versions with contingency transition , suggesting a relationship to unexpected uncertainty ( Figure 3b; version 3 , F ( 1 , 298 ) =17 . 585 , p=3 . 6E-5 , ηp2=0 . 056 , MD = 0 . 071 , CI=[0 . 038 , 0 . 104]; version 4 , F ( 1 , 298 ) =6 . 397 , p=0 . 012 , ηp2=0 . 021 , MD = 0 . 039 , CI=[0 . 009 , 0 . 07] ) . Our task manipulation , increasing unexpected volatility , increases win-switching behavior and stochastic choice more in more paranoid participants . To test the propriety of our model , we simulated data for each subject in online version 3 and determined whether or not key behavioral effects ( Figure 7a , Table 1 , Table 8 ) were present . Using individually estimated HGF parameters to generate ten simulations per participant , we recapitulated both elevated win-switch behavior ( paranoia effect , F ( 1 , 70 ) =15 . 394 , p=2 . 01E-4 , ηp2=0 . 180 , MD = 0 . 186 , CI=[0 . 091 , 0 . 28] ) and choice stochasticity ( U-value; paranoia effect , F ( 1 , 70 ) =13 . 362 , p=0 . 0005 , ηp2=0 . 160 , MD = 0 . 065 , CI=[0 . 030 , 0 . 101] ) in simulated paranoid participants ( Figure 7b; simulated win-switch rate , low paranoia: m = 0 . 24 [0 . 02] , high paranoia: m = 0 . 43 [0 . 04]; simulated U-value , low paranoia: m = 0 . 851 [0 . 008] , high paranoia: m = 0 . 916 [0 . 016] ) . Neither real nor simulated data showed any significant relationship between lose-stay behavior and paranoia ( Table 1 , Table 2 , Table 8 ) . To demonstrate the effects of parameters on task performance , we performed additional simulations in which we doubled or halved a single parameter at a time from the baseline average of low paranoia participants . These results confirmed the impact of κ , ω2 , and ω3 on win-shift behavior ( Figure 4 ) . Parameter recovery revealed significant correlations for κ and ω2 between original subject parameters and those estimated from simulations ( Figure 6; ω: r = 0 . 702 , p=2 . 52E-11 , CI=[0 . 557 , 0 . 805]; κ: r = 0 . 305 , p=0 . 011 , CI=[0 . 072 , 0 . 506] ) . Higher level parameters ( ω3 , μ30 ) were less consistently recovered , as noted in previous publications ( Bröker et al . , 2018 ) . Thus , the model we chose , with meta-volatility and three coupled layers of belief , successfully simulates the key features of paranoid behavior: higher win-switching and stochastic choice . Our model is complex and other simpler reinforcement learning models might explain behavior on this task . Given the win-switching behavior we sought to understand , we fit a model from Lefebvre and colleagues that instantiated biased belief updating via differential weighting of positive and negative prediction errors ( Lefebvre et al . , 2018 ) . Fitting this model to online version 3 , we saw no significant paranoia group differences in learning rates for positive or negative prediction errors in parameters derived from all 180 trials ( independent samples t-test: α+ , t ( 70 ) =-0 . 532 , p=0 . 597; α- , t ( 70 ) =0 . 963 , p=0 . 339 ) , nor did we see any significant block*paranoia or paranoia group effects by repeated measures ANOVA ( block*paranoia: α+ , F ( 1 , 70 ) =0 . 188 , p=0 . 732 , α- , F ( 1 , 70 ) =0 . 378 , p=0 . 540; paranoia group: α+ , F ( 1 , 70 ) =0 . 243 , p=0 . 623 , α- , F ( 1 , 70 ) =1 . 292 , p=0 . 260 ) . See Table 9 . We can also simplify within our hierarchical Gaussian Filter framework . The model we chose had three layers of beliefs and the highest level seemed to capture most of the task and paranoia effects of interest ( Figure 8 ) . To confirm this suspicion , we removed the third layer , fitting an HGF model that had beliefs about outcomes and deck values but no beliefs about volatility , no unexpected volatility learning rate , nor meta-volatility . This model failed to capture the task effects or group differences in its parameters ( see Table 9 ) . Therefore , a more complicated model , one that captures higher-level beliefs about contingency transitions or learning when to learn , seems most appropriate , and indeed , that type of model was able to simulate the key features of our data ( Palminteri et al . , 2017 ) . Future work will compare and contrast different potential computational models included , but not limited to Bayesian Hidden State Markov Models ( Hampton et al . , 2006 ) , as well as switching ( Gershman et al . , 2014 ) and volatile Kalman Filters ( Piray and Daw , 2020 ) . Given the apparent similarity in effects of paranoia and methamphetamine in humans and rats , respectively ( Figure 2b ) , we searched for latent structure in our data using two-step cluster analysis ( Tkaczynski , 2017 ) . This approach sorts subjects into groups ( clusters ) on the basis of some experimenter-selected variables such as estimated model parameters . The goal is to find distinct subsets in the data such that each cluster exhibits a cohesive pattern of relationships between the variables . Whereas some clustering approaches require the experimenter to predefine the expected number of clusters , two step-clustering determines both the optimal number of clusters and the composition of each cluster . The greater the similarity ( or homogeneity ) within a group and the greater the difference between groups , the better the clustering . Considering that paranoia and methamphetamine exposure share a pattern of elevated μ30 and κ accompanied by decreased ω2 and ω3 ( Table 10 ) , we hypothesized that these four variables would yield a distinct cluster: a ‘paranoid style’ across species . We analyzed μ30 , κ , ω2 , and ω3 estimates derived from the first block of experiment one and online version 3 ( pre-context change data , because rats do not experience a context shift ) with post-chronic exposure rat data ( methamphetamine and saline ) . We identified two clusters with good cohesion and separation , meaning that subjects sorted into two groups ( each containing rodents and humans ) whose parameters travelled in such a way that their values were close to the centroid or mean of the cluster they were in and as far as possible from the centroid of the other cluster ( average silhouette coefficient = 0 . 7; cluster size ratio = 2 . 46; Figure 9a ) . All parameters contributed to clustering; κ contributed most strongly ( Figure 9b ) . Importantly , the cluster solution did not separate rats from humans ( despite the differences in task structure , incentives , manipulanda , and phylogeny ) . Relative to the overall distribution , Cluster one was characterized by high κ and μ30 , and decreased ω2 and ω3 . Cluster one membership was significantly associated with high paranoia and methamphetamine exposure , χ2 ( 1 , n = 121 ) =29 . 447 , p=5 . 75E-8 , Cramer’s V = 0 . 493 ( Figure 9c ) . Notably , no participants in the low paranoia group with paranoia scores above zero were ascribed Cluster one membership . The cluster solution was robust to validation by split-half analysis ( removing half of the participants and repeating the clustering ) , removal of the rat subjects , and removal of human participants . In each case , we identified two clusters with good cohesion and separation ( Split-half 1 , n = 19 cluster 1 , 42 cluster 2: silhouette coefficient = 0 . 6; Split-half 2 , n = 17 cluster 1 , 43 cluster 2: silhouette coefficient = 0 . 7; No Rat , n = 26 cluster 1 , 78 cluster 2: silhouette coefficient = 0 . 7; Rat Only , n = 6 cluster 1 , 11 cluster 2: silhouette coefficient = 0 . 7 ) . In summary , paranoid participants and methamphetamine-exposed rats cluster together ( high μ30 , high κ , low ω2 , and low ω3 ) , suggesting that these parameters share an underlying generative process and that paranoia and methamphetamine have similar effects on reversal-learning .
During non-social probabilistic reversal-learning , paranoid individuals and rats chronically exposed to methamphetamine have higher initial expectations of task volatility ( μ30 ) . In other words , they start the task anticipating more changes in stimulus-outcome associations , and they switch choices readily and excessively in anticipation of reversal events . By relying more on their expectations of volatility than on actual experience ( exemplified by switching even after positive feedback ) , they are slower to learn about changes in task volatility . This manifests as decreased meta-volatility learning ( ω3 ) and failure to significantly adjust μ30 after contingency transitions . More paranoid individuals are similarly slower to adjust expected deck values ( lower ω2 ) but faster to attribute volatility to reversal events ( elevated κ ) , perceiving change ( unexpected uncertainty ) instead of normal statistical variation ( expected uncertainty ) . They sit at Hofstadter’s ‘turning point’ , constantly expecting change but never learning appropriately from it . In the reversal learning literature , choice switching after positive feedback has garnered less attention than perseverative behavior and sensitivity to negative feedback ( Izquierdo et al . , 2017; Waltz , 2017 ) . Individuals with depression and schizophrenia seemingly perseverate less than healthy controls , but this has formerly been attributed to increased sensitivity to negative feedback ( Waltz , 2017; Robinson et al . , 2012 ) . However , elevated win-switch tendencies have been reported in youths with bipolar disorder , major depressive disorder , and anxiety disorder ( Dickstein et al . , 2010 ) . A prior study in people with schizophrenia described excessive win-switch behavior that correlated with the severity of delusional beliefs and hallucinations ( Waltz , 2017 ) . Likewise , an elevated prior on environmental volatility ( μ30 ) and higher sensitivity to this volatility ( κ ) have been observed in HGF analyses of 2-choice probabilistic reversal-learning in medicated and unmedicated patients with schizophrenia ( Deserno , 2018 ) . These authors did not explore paranoia specifically . We assessed paranoia across the continuum of health and mental illness , provided three choice options , and explicitly manipulated unexpected volatility across task versions . The version that shifted from an easier to a more difficult contingency context ( version 3 ) was associated with paranoia group effects on μ30 , κ , and ω2 , and a meta-analytic effect on ω3 . Furthermore , this contingency transition – an exposure to truly unexpected volatility – rendered low paranoia controls more similar to their paranoid counterparts by decreasing their meta-volatility learning ( ω3 ) . Paranoid participants responded to contingency transitions in version 3 and version four by switching stochastically . These findings suggest a continuum of behavioral responses to volatility , moving from optimal learning to diminished feedback sensitivity ( i . e , decreased ω3 in low paranoia participants ) and from diminished feedback sensitivity ( lower ω3 and increased win-switching in high paranoia participants ) toward complete dissociation from experienced feedback ( stochastic switching ) . Unexpected uncertainty , the perception of change in the probabilities of the environment — particularly ‘unsignaled context switches” ( Yu and Dayan , 2005 ) which increase unexpected volatility — is thought to promote abandonment of old associations and new learning . However , our results suggest that this response might vary according to a hierarchy of belief . Paranoid participants were quick to abandon ‘best deck’ associations and explore alternative options ( i . e . , x2 beliefs ) , but in turn they relied more on their higher-level beliefs about the task volatility ( x3 beliefs ) and less on sensory feedback ( lower metavolatility learning ) . Our analysis of covariates warrants specific focus on κ , the sensitivity to unexpected volatility . Other parameter-paranoia associations did not endure after controlling for demographic factors ( age , gender , ethnicity , and race ) , although we see their derangement in our rodent study as well as in the significant meta-analytic effects across our experiments . Furthermore , these demographic factors are themselves strong predictors of paranoia ( Holt and Albert , 2006; Iacovino et al . , 2014; Mahoney et al . , 2010 ) . It is notable too that κ was the most powerful discriminator of the two clusters of human and animal participants . We conclude that elevated κ - belief updating tethered to unexpected volatility - is the parameter change most robustly associated with paranoia . Doubling κ in our simulations induced significantly more win-switching . Multiple neurobiological manipulations may induce such win-switching behavior . Lesions of the mediodorsal thalamus in non-human primates ( Chakraborty et al . , 2016 ) or neurons projecting from the amygdala to orbitofrontal cortex in rats ( Groman et al . , 2019 ) engender win-switching . Unexpected uncertainty , and the κ parameter of the HGF in particular ( Marshall et al . , 2016 ) , are thought to be signaled via the locus coeruleus and noradrenaline ( Yu and Dayan , 2005; Payzan-LeNestour and Bossaerts , 2011; Payzan-LeNestour et al . , 2013; Tervo et al . , 2014 ) . This mechanism is thought to modulate switching versus staying behaviors ( Kane et al . , 2017; Aston-Jones and Cohen , 2005; Aston-Jones et al . , 1999; Eldar et al . , 2013 ) , as well as responses to stress ( Borodovitsyna et al . , 2018; McCall et al . , 2015; Atzori et al . , 2016 ) and subliminal fear cues ( Liddell et al . , 2005 ) to coordinate fight-or-flight responses ( Atzori et al . , 2016 ) . The dual role of the locus coeruleus in recognizing and responding to threats as well as unexpected uncertainty suggests that dysfunction could produce both paranoia and the inferential abnormalities we observed . Methamphetamine may induce similar dysfunction ( Ferrucci et al . , 2019; Ferrucci et al . , 2013; Ferrucci et al . , 2008 ) . Acute moderate doses increase pre-synaptic catecholamine release , particularly noradrenaline ( Rothman et al . , 2001 ) , and induce exploratory locomotive effects modulated through adrenoceptors on dopamine neurons ( Ferrucci et al . , 2013 ) . Excessive release of noradrenaline from the locus coeruleus into the anterior cingulate cortex drives feedback insensitivity and stochastic switching behavior in rats completing a three-option counter prediction task ( Tervo et al . , 2014 ) . Evolutionarily , departure from predictable , rational actions might offer an adaptive mechanism for escape from intractable threat . As a protean defense mechanism , behavioral stochasticity impedes predators’ abilities to create accurate , actionable countermeasures ( Humphries and Driver , 1970; Richardson et al . , 2018; Humphries and Driver , 1967 ) . If driven by excessive unexpected uncertainty , underwritten by noradrenaline , protean defense may represent a heavily conserved , continuous common mechanism underlying vigilance and false alarms ( Aston-Jones et al . , 1994; Rajkowski et al . , 1994; Usher et al . , 1999 ) , arousal-linked attentional biases ( Eldar et al . , 2013 ) and selective processing of social threats . However , protean behaviors are not necessarily adaptive . Pathological insensitivity to feedback and reliance on internal beliefs over evidence constitute a ‘break from reality’ – in other words , psychosis . Efference copy models of motor control Wolpert and Ghahramani , 2000 have been evoked to explain psychotic symptoms ( Blakemore et al . , 2000; Blakemore et al . , 1998; Blakemore et al . , 1999; Blakemore et al . , 2002; Frith et al . , 2000a; Frith et al . , 2000b; Shergill et al . , 2005; Shergill et al . , 2014 ) . Aberrant mismatches between expected and experienced sensory consequences of actions , weighted by their uncertainty ( Wolpert and Ghahramani , 2000 ) , can lead to the misattribution of one’s movements to an external agent ( Blakemore et al . , 2000; Blakemore et al . , 1998; Blakemore et al . , 1999; Blakemore et al . , 2002; Frith et al . , 2000a; Frith et al . , 2000b; Shergill et al . , 2005; Shergill et al . , 2014 ) . Since we model others’ intentions with reference to our model of ourselves ( Friston and Frith , 2015 ) , volatile experiences of ones’ body and actions will lead to uncertain and ultimately more threatening inferences about others ( Friston and Frith , 2015 ) . This would be entirely consistent with the present observations . When confronted with intractable unexpected uncertainty our participants rely on higher-level beliefs about the task environment . When humans experience non-social volatility , ( For example through threats to their sense of control [Whitson and Galinsky , 2008] or exposure to surprising non-social stimuli [Proulx et al . , 2012; Heine et al . , 2006] ) , they appeal to the influence of powerful enemies , even when those enemies’ influence is not obviously linked to the volatility ( Sullivan et al . , 2010 ) . Our account places the locus of paranoia at the level of the individual . Here , our account departs from evolutionary accounts of paranoia grounded in coalitional threat ( Raihani and Bell , 2019; persecutors are not scapegoats that increase group cohesion . Rather , when paranoid , we have a ready explanation for hazards . With a well-defined persecutor in mind , a volatile world may be perceived to have less randomly distributed risk ( Sullivan et al . , 2010 ) . However , paranoia might become a self-fulfilling prophecy , engendering more volatility and negative social interactions . This aspect may be captured in our task through win-switch behavior . By failing to incorporate positive feedback from the best option , paranoid individuals sample sub-optimal options which delivers misleading positive feedback . There are some important limitations to our conclusions . Compared with humans , rats are relatively asocial . But they are not completely asocial . In our experiment they were housed in pairs , and , more broadly , they evince social affiliative interactions with other rats ( Donaldson et al . , 2018; Kondrakiewicz et al . , 2019; Urbach et al . , 2010 ) . A further limitation centers on the comparability of our experimental designs . In humans our comparisons were both within ( contingency transition ) and between groups ( low versus high paranoia ) . In rats , the model was also mixed with some between ( saline versus methamphetamine ) and some within-subject ( pre versus post chronic treatment ) comparisons . We should be clear that there was no contingency context transition in the rat study . However , just as that transition made low paranoia humans behave like high paranoia , chronic methamphetamine exposure made rats behave on a stable contingency much like high paranoia humans - even in the absence of contingency transition . The comparable results across species , despite these differences , warrant the inference that our basic , relatively asocial , approach provides a robust tool for computational dissection of learning mechanisms . Social interactions play a rich and undeniable role in paranoia , but translational , domain-general approaches may ultimately facilitate biological insights into paranoia , psychosis and delusions ( Corlett and Fletcher , 2014; Feeney et al . , 2017 ) . Whilst we contend that our task is relatively free of social features ( certainly compared to others [Raihani and Bell , 2017] ) , the possibility remains that the elevated U-values in our participants are reflective of attempts ( and perhaps failures ) to predict our intentions as experimenters . Indeed , this is a possibility raised previously with regards to simple conditioned behaviors in experimental animals . Even during Pavlovian conditioning , animals may attempt to infer a generative model of the task environment , which might , ultimately , include the experimenter arranging the contingencies ( Gershman and Niv , 2012; Gershman and Niv , 2010 ) . It is possible that all instances of human cognitive testing involve an element of inference by the participant with regards to the intentions of the experimenter , whether or not the task at hand is explicitly social , and indeed , all cognitive functions may be aimed at or modulated by such inferences ( Turner et al . , 1994 ) . In summary , a strong belief in the volatility of the world necessitates hypervigilance and a facility with change . However , in paranoia , that belief in the volatility of the world is itself resistant to change , making it difficult to reassure , teach , or change the minds of people who are paranoid . They remain ‘on guard , ’ adhering to expectations over evidence . By using a non-social task , we have shown that this paranoid style is not restricted to the social domain , and that it can be modeled in relatively asocial animals . Additionally , our domain-general approach reaffirms the merit of establishing expectations of a stable , predictable environment to promote recovery from paranoia-associated illness ( Powers et al . , 2018 ) . We note with interest the apparent relationship between conspiratorial ideation and societal crisis situations ( terrorist attacks , plane crashes , natural disasters or war ) throughout history , with peaks around the great fire of Rome ( AD 64 ) , the industrial revolution , the beginning of the cold war , 9/11 , and contemporary financial crises ( van Prooijen and Douglas , 2017 ) . In today’s world of escalating uncertainty and volatilty – particularly environmental climate change and viral pandemics – our findings suggest that the paranoid style of inference may prove particularly maladaptive for coordinating collaborative solutions .
English-speaking participants aged 18 to 65 ( n = 34 ) were recruited from the greater New Haven area through public fliers and mental health provider referrals . Exclusion criteria included history of cognitive or neurologic disorder ( e . g . , dementia ) , intellectual impairment , or epilepsy; current substance dependence or intoxication; cognition-impairing medications or doses ( e . g . opiates , high dose benzodiazepines ) ; history of special education; and color blindness . Participants were classified as healthy controls ( n = 18 ) , schizophrenia spectrum patients ( schizophrenia or schizoaffective disorder; n = 8 ) , and mood disorder patients ( depression , bipolar disorder , generalized anxiety disorder , post-traumatic stress disorder; n = 8 ) on the basis of clinician referrals and/or self-report . Participants were compensated $10 for enrolment with an additional $10 upon completion . Two healthy controls were excluded from analyses due to failure to complete the questionnaires and suspected substance use , respectively . 332 participants were recruited online via Amazon Mechanical Turk ( MTurk ) . The study advertisement was accessible to MTurk workers with a 90% or higher HIT approval rate located within the United States . To discourage bot submissions and verify human participation , we required participants to answer open-ended free response questions; submit unique , separate completion codes for the behavioral task and questionnaires; and enter MTurk IDs into specific boxes within the questionnaires . All submissions were reviewed for completion code accuracy , completeness of responses ( i . e . , declining no more than 30% of questionnaire items ) , quality of free response items ( e . g . , length , appropriate grammar and content ) , and use of virtual private servers ( VPS ) to submit multiple responses and/or conceal non-US locations ( Dennis VPS paper , 2018 ) . Upon approval , workers were compensated $6 . Those who scored in the top 25% on the card game ( reversal-learning task ) earned a $2 bonus . We rejected or excluded 19 submissions that geolocation services ( https://www . iplocation . net/ ) identified as originating outside of the United States or from suspected server farms , four submissions for failure to manually enter MTurk ID codes , and two submissions for insufficient questionnaire completion . Submissions with grossly incorrect completion codes were rejected without further review . Subject information , behavioral data acquisition , and behavioral analyses were described previously ( Groman et al . , 2018 ) . Long Evans rats ( Charles River; n = 20 ) ranged from 7 to 9 weeks of age . Rats were exposed to escalating doses and frequency of saline ( n = 10 ) or methamphetamine ( n = 10 , three withdrawn during dosing ) , imitating patterns of human methamphetamine users ( Segal et al . , 2003; Han et al . , 2011 ) . Prior to dosing ( Pre-Rx ) , rats completed 26 within-session reversal sessions , including up to eight reversals per session . Post-dosing ( Post-Rx ) , rats completed one test session per week for four weeks . Computational model parameters were estimated from each session and averaged across treatment conditions to yield one Pre-Rx and Post-Rx set of parameters per rat . Participants completed a 3-option probabilistic reversal-learning paradigm . Three decks of cards were displayed on a computer monitor for 160 trials . Participants selected a deck on each trial by pressing the predesignated key . We advised participants that each deck contained winning and losing cards ( +100 and −50 points ) , but in different amounts . We also stated that the best deck may change . Participants were instructed to find the best deck and earn as many points as possible . Probabilities switched between decks when the highest probability deck was selected in 9 out of 10 consecutive trials ( performance-dependent reversal ) . Every 40 trials the participant was provided a break , following which probabilities automatically reassigned ( performance-independent reversal ) . In Experiment 1 , the task was presented via Eprime 2 . 0 software ( Psychology Software Tools , Sharpsburg , PA ) . Participants were limited to a 3 s response window , after which the trial would time out and record a null response . A fixation cross appeared during variable inter-trial intervals ( jittering ) . Task pacing remained independent of response time . In block 1 ( trials 1–80 ) the reward probabilities ( contingency ) of the three decks were 90% , 50% , and 10% ( 90-50–10% ) . Without cue or warning ( i . e . unsignaled to the participants ) the contingency transitioned to 80% , 40% , and 20% ( 80-40–20% ) upon initiation of block 2 ( trials 81–160 ) . In Experiment 2 , the task was administered via web browser link from the MTurk marketplace . We changed the task timing to self-paced and eliminated null trials and inter-trial jittering . A progress tracker was provided every 40 trials . Workers were randomly assigned to one of four task versions , using restricted block randomization to ensure comparable numbers of high paranoia participants across task versions . Version one had a constant contingency of 90-50–10% . Version 4 maintained a constant contingency of 80-40–20% . Version 3 replicated the 90-50–10% ( block 1 ) to 80-40–20% ( block 2 ) context transition of Experiment 1 . Version 4 presented the reversed contingency transition , 80-40–20% ( block 1 ) to 90-50–10% ( block 2 ) . We analyzed attrition rates across the four versions . Following task completion , questionnaires were administered via the Qualtrics survey platform ( Qualtrics Labs , Inc , Provo , UT ) . Items included demographic information ( age , gender , educational attainment , ethnicity , and race ) and mental health questions ( past or present diagnosis , medication use , Structured Clinical Interview for DSM-IV Axis II Personality Disorders ( SCID-II ) ( Ryder et al . , 2007 ) , Beck’s Anxiety Inventory ( BAI ) ( Beck et al . , 1988 ) , Beck’s Depression Inventory ( BDI ) ( Beck et al . , 1961 ) . We removed the single suicidality question from the BDI for Experiment 2 . Experiment 2 included additional items: income , three cognitive reflection questions ( Table 7 ) , and three free response items ( ‘What do you think the card game was testing ? ’ , ‘Did you use any particular strategy or strategies ? If yes , please describe’ , and ‘Did you find yourself switching strategies over the course of the game ? ’ ) . We quantified trait-level paranoia using the paranoid personality subscale of the SCID-II , and we included an ideas of reference item from the schizotypy subscale ( ‘When you are out in public and see people talking , do you often feel that they are talking about you ? ’ ) This item , along with other SCID-II items , has previously been included as a metric of paranoia in the general population ( Bebbington et al . , 2013; Bell and O'Driscoll , 2018 ) . Participants who endorsed four or more paranoid personality items ( i . e . , the cut-off for the top third identified in Experiment 1 ) were classified as ‘high paranoia . ’ Each participant’s SCID-II , BAI , and BDI scores were normalized by total scale items answered . Response rates were higher than 90% for all questionnaire items and scales ( Table 11 ) . We analyzed tendencies to choose alternative decks after positive feedback ( win-switch ) and select the same deck after negative feedback ( lose-stay ) . Win-switch rates were calculated as the number of trials in which the participant switched after positive feedback divided by the number of trials in which they received positive feedback . Lose-stay rates were calculated as number of trials in which a participant persisted after negative feedback divided by total negative feedback trials . In Experiment 1 , we excluded post-null trials from these analyses . To further characterize switching behavior , we calculated U-values , a measure of choice stochasticity: ( 1 ) U−value=−Σi=1βlog ( αi ) x αilog ( β ) where β is the number of possible choice options ( i . e . , card decks or noseports ) and α equals the relative frequency of choice option i ( Kong et al . , 2017 ) . To avoid any choice counterbalancing effects across reversals , choice frequencies were determined by the underlying probabilities of the decks rather than their physical attributes ( e . g . , deck position or color ) . Additional behavioral analyses included trials to first reversal , trials to post-reversal recovery , and trials to post-reversal switch . The latter two were restricted to the first reversal in the first block . Trials post-reversal were counted from the first-negative feedback trial following the true reversal event . Recovery was defined as switching to the best deck and staying for at least one additional trial . Data are available on ModelDB ( McDougal et al . , 2017; http://modeldb . yale . edu/258631 ) with accession code p2c8q74m . | Everyone has had fleeting concerns that others might be against them at some point in their lives . Sometimes these concerns can escalate into paranoia and become debilitating . Paranoia is a common symptom in serious mental illnesses like schizophrenia . It can cause extreme distress and is linked with an increased risk of violence towards oneself or others . Understanding what happens in the brains of people experiencing paranoia might lead to better ways to treat or manage it . Some experts argue that paranoia is caused by errors in the way people assess social situations . An alternative idea is that paranoia stems from the way the brain forms and updates beliefs about the world . Now , Reed et al . show that both people with paranoia and rats exposed to a paranoia-inducing substance expect the world will change frequently , change their minds often , and have a harder time learning in response to changing circumstances . In the experiments , human volunteers with and without psychiatric disorders played a game where the best choices change . Then , the participants completed a survey to assess their level of paranoia . People with higher levels of paranoia predicted more changes would occur and made less predictable choices . In a second set of experiments , rats were put in a cage with three holes where they sometimes received sugar rewards . Some of the rats received methamphetamine , a drug that causes paranoia in humans . Rats given the drug also expected the location of the sugar reward would change often . The drugged animals had harder time learning and adapting to changing circumstances . The experiments suggest that brain processes found in both rats , which are less social than humans , and humans contribute to paranoia . This suggests paranoia may make it harder to update beliefs . This may help scientists understand what causes paranoia and develop therapies or drugs that can reduce paranoia . This information may also help scientists understand why during societal crises like wars or natural disasters humans are prone to believing conspiracies . This is particularly important now as the world grapples with climate change and a global pandemic . Reed et al . note paranoia may impede the coordination of collaborative solutions to these challenging situations . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2020 | Paranoia as a deficit in non-social belief updating |
Viruses interact with hundreds to thousands of proteins in mammals , yet adaptation against viruses has only been studied in a few proteins specialized in antiviral defense . Whether adaptation to viruses typically involves only specialized antiviral proteins or affects a broad array of virus-interacting proteins is unknown . Here , we analyze adaptation in ~1300 virus-interacting proteins manually curated from a set of 9900 proteins conserved in all sequenced mammalian genomes . We show that viruses ( i ) use the more evolutionarily constrained proteins within the cellular functions they interact with and that ( ii ) despite this high constraint , virus-interacting proteins account for a high proportion of all protein adaptation in humans and other mammals . Adaptation is elevated in virus-interacting proteins across all functional categories , including both immune and non-immune functions . We conservatively estimate that viruses have driven close to 30% of all adaptive amino acid changes in the part of the human proteome conserved within mammals . Our results suggest that viruses are one of the most dominant drivers of evolutionary change across mammalian and human proteomes .
A number of proteins with a specialized role in antiviral defense have been shown to have exceptionally high rates of adaptation ( Cagliani et al . , 2011; Cagliani et al . , 2012; Elde et al . , 2009; Fumagalli et al . , 2010; Kerns et al . , 2008; Liu et al . , 2005; Sawyer et al . , 2004; Sawyer et al . , 2005; Sawyer et al . , 2007; Sironi et al . , 2012; Vasseur et al . , 2011 ) . One example is protein kinase R ( PKR ) , which recognizes viral double-stranded RNA upon infection , halts translation , and as a result blocks viral replication ( Elde et al . , 2009 ) . PKR is one of the fastest adaptively evolving proteins in mammals . Specific amino acid changes in PKR have been shown to be associated with an arms race against viral decoys for the control of translation ( Elde et al . , 2009 ) . However , PKR and other fast-evolving antiviral defense proteins may not be representative of the hundreds or even thousands of other proteins that interact physically with viruses ( virus-interacting proteins or VIPs in the rest of this manuscript ) . Most VIPs are not specialized in antiviral defense and do not have known roles in immunity . Many of these VIPs play instead key functions in basic cellular processes , some of which might be essential for viral replication . In principle some VIPs without specific antiretroviral functions might nonetheless evolve to limit viral replication or alleviate deleterious effects of viruses despite the need to balance this evolutionary response with the maintenance of the key cellular functions they play . There are reasons to believe that such an evolutionary response to viruses might be limited , however . First , most VIPs evolve unusually slowly rather than unusually fast both in animals ( Davis et al . , 2015; Jäger et al . , 2011 ) and in plants ( Mukhtar et al . , 2011; Weßling et al . , 2014 ) . Second , VIPs tend to interact with proteins that are functionally important hubs in the protein-protein interaction network of the host possibly limiting their ability to adapt ( Dyer et al . , 2008; Halehalli and Nagarajaram , 2015 ) . Finally , very few cases of adaptation to viruses are known outside of fast evolving , specialized antiviral proteins ( Demogines et al . , 2012; Meyerson et al . , 2014; Meyerson and Sawyer , 2011; Meyerson et al . , 2015; Ng et al . , 2015; Ortiz et al . , 2009; Schaller et al . , 2011 ) . Transferrin receptor or TFRC is the most notable exception , and serves as a striking example of a non-immune , housekeeping protein used by viruses ( Demogines et al . , 2013; Kaelber et al . , 2012 ) . TFRC is responsible for iron uptake in many different cell types and is used as a cell surface receptor by diverse viruses in rodents and carnivores . TFRC has repeatedly evaded binding by viruses through recurrent adaptive amino acid changes . As such , TFRC is the only clear-cut example of a host protein not involved in antiviral response that is known to adapt in response to viruses . Here we analyze patterns of evolutionary constraint and adaptation in a high quality set of ~1300 VIPs that we manually curated from virology literature . These 1300 VIPs come from a set of ~10000 proteins conserved across 24 well-sequenced mammalian genomes ( Materials and methods ) . As expected , the vast majority of these VIPs ( ~80% ) have no known antiviral or any other more broadly defined immune activity . We confirm that VIPs do tend to evolve slowly and demonstrate that this is because VIPs experience much stronger evolutionary constraint than other proteins within the same functional categories . However , despite this greater evolutionary constraint , VIPs display higher rates of adaptation compared to other proteins . This excess of adaptation is visible in VIPs across biological functions , on multiple time scales , in multiple taxa , and across multiple studied viruses . Finally , we showcase the power of our global scan for adaptation in VIPs by studying the case of aminopeptidase N , a well-known multifunctional enzyme ( Mina-Osorio , 2008 ) used by coronaviruses as a receptor ( Delmas et al . , 1992; Yeager et al . , 1992 ) . Using our approach we reach an amino-acid level understanding of parallel adaptive evolution in aminopeptidase N in response to coronaviruses in a wide range of mammals .
We curated a set of 1256 VIPs from the low-throughput virology literature ( Materials and methods and Supplementary file 1A ) . VIPs were defined as proteins that interact physically with viral proteins , viral RNA , and/or viral DNA ( Supplementary file 1A ) . We excluded interactions identified by high-throughput experiments because we were concerned about a high rate of false positives ( Mellacheruvu et al . , 2013 ) . The 1256 VIPs were annotated from an initial set of 9861 proteins with clear orthologs in all 24 analyzed mammalian high quality genomes ( Figure 1 , Supplementary file 1B and Materials and methods ) ( Enard et al . , 2016 ) . 10 . 7554/eLife . 12469 . 003Figure 1 . Tree of 24 mammals used in the analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 12469 . 003 Most of the VIPs ( 95% ) correspond to an interaction between a human protein and a virus infecting humans ( Supplementary file 1A ) . Human Immunodeficiency Virus type 1 ( HIV-1 ) is the best-represented virus with 240 VIPs , with nine other viruses ( HPV , HCV , EBV , HBV , HSV , Influenza Virus , ADV , HTLV and KSHV ) having at least 50 VIPs ( Supplementary file 1A ) . This dataset represents the largest , most up-to-date set of VIPs backed by individual low-throughput publications . Nonetheless , given that many VIPs were discovered only recently , with half of all publications reporting VIPs published in the past 7 years ( Figure 2 ) , it is likely that many additional VIPs remain to be discovered . 10 . 7554/eLife . 12469 . 004Figure 2 . Number of VIPs discovered per year until 2014 . DOI: http://dx . doi . org/10 . 7554/eLife . 12469 . 004 The identified 1256 VIPs are involved in diverse cellular and supracellular processes with 162 overlapping GO cellular and supracellular processes having more than 50 VIPs ( Gene Ontology ( GO ) classes ( October 2013 version ) ( Ashburner et al . , 2000; The Gene Ontology Consortium , 2015 ) ; Supplementary file 1C ) . These cellular processes include transcription ( 354 VIPs ) , post-translational protein modification ( 224 VIPs ) , signal transduction ( 396 VIPs ) , apoptosis ( 185 VIPs ) , and transport ( 264 VIPs ) . The supracellular processes notably include defense response ( 103 VIPs ) and developmental processes ( 327 VIPs ) . Only 57 VIPs or 5% of VIPs have known antiviral activity ( Supplementary file 1D ) . These 57 antiviral VIPs are part of a larger group of 241 VIPs ( 20% of VIPs ) with known immune functions , defined here as any activity that modulates the immune response or involved in the development of the immune response ( Materials and methods and Supplementary file 1D ) . Most - more than 80% - of the VIPs have no known immune activity . We analyze both purifying selection and positive selection in VIPs versus non-VIPs at two distinct evolutionary time scales: ( i ) in the great apes in general and in the human branch specifically and ( ii ) across the entire mammalian phylogeny . We use the ratio of nonsynonymous to synonymous polymorphisms ( abbreviated as pN/pS ) within humans and great apes as a measure of purifying selection . We use McDonald-Kreitman ( MK ) and the branch-site tests of positive selection using the BS-REL ( Kosakovsky Pond et al . , 2011 ) and BUSTED ( Murrell et al . , 2015 ) tests from the HYPHY package ( Pond et al . , 2005 ) to assess the prevalence of positive selection in VIPs compared to non-VIPs in the human lineage and in mammals in general ( Material and methods ) . We confirm that VIPs tend to evolve slowly ( Jäger et al . , 2011; Davis et al . , 2015 ) . On average , the VIPs have ~15% lower mammal-wide dN/dS ratio compared to non-VIPs ( 0 . 124 versus 0 . 145 , 95% CI [0 . 136 , 0 . 148]; Materials and methods ) . The difference in dN/dS is highly significant ( permutation test P=0 after 109 iterations; Supplementary file 1B ) . In order to disentangle whether this slower evolution of VIPs is due to stronger purifying selection or to a lower rate of adaptation , we first assess the strength of purifying selection in the VIPs using the pN/pS ratio . Genome-wide polymorphism data required to measure pN/pS are available in humans ( Abecasis et al . , 2012 ) ( 1000 Genomes Project ) ( Supplementary file 1E ) , and other great apes: chimpanzee , gorilla , and orangutans ( Prado-Martinez et al . , 2013 ) ( Great Apes Genome Project ) ( Supplementary file 1F ) . The 1000 Genomes Project and the Great Apes Genome Project are complementary for this analysis . On the one hand , the 1000 Genomes Project provides high quality variants with frequencies estimated from a large number of individuals . On the other hand while the Great Ape Genome project includes fewer individuals and provides coarser frequency data , it provides substantially higher pN and pS counts than the 1000 Genomes data because non-human great apes tend to be more polymorphic overall ( Prado-Martinez et al . , 2013 ) . In the human African populations from the 1000 Genomes project ( Materials and methods ) , the average pN/pS is 21% lower in VIPs compared to non-VIPs ( 0 . 759 versus 0 . 966 , 95% CI [0 . 92 , 1 . 01] , simple permutation test P=0 after 109 iterations ) . VIPs also show an excess of low frequency ( ≤10% ) deleterious non-synonymous variants compared to non-VIPs ( Figure 3—figure supplement 1; simple permutation test P=0 after 109 iterations ) . In great apes , the average pN/pS ratio is 25% lower in VIPs compared to non-VIPs ( 0 . 526 versus 0 . 697 , 95% CI [0 . 66 , 0 . 72] , simple permutation test P=0 after 109 iterations; Figure 3A ) . Finally , stronger purifying selection acting on VIPs is widespread and is not limited to VIPs interacting with any one particular virus ( Figure 3B ) . 10 . 7554/eLife . 12469 . 005Figure 3 . Patterns of purifying selection in VIPs . ( A ) Distribution of pN/pS in VIPs ( blue ) and non-VIPs ( pink ) . The blue curve is the density curve of pN/ ( pS+1 ) for 1256 VIPs . We use pN/ ( pS+1 ) instead of pN/pS to account for those coding sequences where pS=0 . pN and pS are measured using great ape genomes from the Great Ape Genome Project ( Materials and methods ) . The pink area represents the superimposition of the density curves for each of 5000 sets of randomly sampled non-VIPs . ( B ) Average pN/pS in VIPs ( blue dot ) versus average pN/pS in non-VIPs ( red dot and red 95% confidence interval ) within ten viruses with more than 50 VIPs The number between parentheses is the number of VIPs for each virus . KSHV: Kaposi’s Sarcoma Herpesvirus . HIV-1: Human Immunodeficiency Virus type 1 . HBV: Hepatitis B Virus . ADV: Adenovirus . HPV: Human Papillomavirus . HSV: Herpes Simplex Virus . EBV: Epstein-Barr Virus . Influenza: Influenza Virus . HTLV: Human T-lymphotropic Virus . HCV: Hepatitis C virus . ( C ) Same as B ) , but for the 20 most high level GO processes with the highest number of VIPs . The full GO process name for “protein modification” as written in the figure is “post-translational protein modification” . DOI: http://dx . doi . org/10 . 7554/eLife . 12469 . 00510 . 7554/eLife . 12469 . 006Figure 3—figure supplement 1 . Site Frequency Spectrum of non-synonymous variants in VIPs and non-VIPs in African populations Red: VIPs . Blue: non-VIPs . The number pN for non-VIPs is rescaled to the actual number pN multiplied by the number of VIPs divided by the number of non-VIPs so that VIPs and non-VIPs can be compared . The x-axis gives the upper threshold for each bin . For example for the second bin , the upper frequency threshold is 0 . 002 and the lower frequency is 0 . 001 , which is also the upper threshold of the first bin . DOI: http://dx . doi . org/10 . 7554/eLife . 12469 . 006 VIPs and non-VIPs have slightly different coding sequence GC content ( 0 . 516 versus 0 . 523 on average , P=6x10-4 ) , coding sequence lengths ( 668 versus 606 amino acids on average , P=0 ) and recombination rates ( Kong et al . , 2010 ) ( 1 . 145 cM/Mb versus 1 . 175 cM/Mb on average , P=0 . 21 ) . To ensure that the difference in pN/pS between VIPs and non-VIPs is robust to these differences , we compare VIPs with non-VIPs with similar values for each potential confounding factor using permutations with a target average ( Materials and methods ) . The difference in pN/pS in great apes between VIPs and non-VIPs persists when comparing VIPs and non-VIPs with similar GC content ( 0 . 526 versus 0 . 655 , P=0 after 109 iterations ) , similar coding sequence length ( 0 . 526 versus 0 . 654 , P=0 ) , or similar recombination ( 0 . 526 versus 0 . 702 , P=0 ) . The difference in pN/pS between VIPs and non-VIPs is therefore a genuine difference in the strength of purifying selection and not due to confounding factors biasing the pN/pS ratio . VIPs have been shown before to be broadly expressed genes and to serve as hubs in the human protein-protein interactions network ( Dyer et al . , 2008 , Halehalli and Nagarajaram , 2015 ) . These differences in gene expression and the number of protein-protein interactions may explain the stronger purifying selection experienced by VIPs . We confirm that VIPs are indeed expressed in more tissues than non-VIPs both at the RNA level ( GTEx Consortium , 2015 ) ( GTEx V4 RNA-seq expression RPKM≥10 in 25 . 5 tissues on average in VIPs versus 11 . 9 tissues in non-VIPs , simple permutation test P=0 ) and at the protein level ( Kim et al . , 2014 ) ( Human Proteome Map spectral count≥5 in 15 . 1 tissues on average for VIPs versus 6 . 1 for non-VIPs , simple permutation test P=0 ) . VIPs also have many more protein-protein interaction partners than non-VIPs based on a dataset of human protein-protein interactions curated by ( Luisi et al . , 2015 ) from the Biogrid database ( Stark et al . , 2011 ) ( 18 . 4 on average versus 3 . 2 , simple permutation test P=0 ) . The magnitude of the difference in pN/pS between VIPs and non-VIPs expressed in a similar number of tissues at the RNA level ( GTEx ) ( 0 . 526 versus 0 . 647 , P=0 ) or in a similar number of tissues at the protein level ( Human protein Map ) ( 0 . 526 versus 0 . 662 , P=0 ) remains largely unchanged . In contrast , the difference in pN/pS is strongly affected when comparing VIPs and non-VIPs with a similar number of protein-protein interactions . Indeed , non-VIPs with the same number of interacting partners as VIPs have a pN/pS ratio of 0 . 605 versus 0 . 697 for all non-VIPs , and the difference in the pN/pS ratios between VIPs and non-VIPs is reduced from 25% to 13% . These results show that VIPs do experience stronger purifying selection than non-VIPs , and that the difference in purifying selection is driven at least partly by the fact that VIPs tend to be hubs with many interacting partners in the human protein-protein interactions network . The higher level of purifying selection in VIPs might be due to the fact that VIPs participate in the more constrained host functions , or , alternatively , because within each specific host function , viruses tend to interact with the more constrained proteins . In order to test these two non-mutually exclusive scenarios we generated 104 control sets of non-VIPs chosen to be in the same 162 Gene Ontology processes as VIPs ( GO processes with more than 50 VIPs; Supplementary file 1C and Materials and methods ) . In great apes , GO-matched non-VIPs still have a much higher pN/pS ratio compared to VIPs , suggesting that VIPs tend to be more conserved than non-VIPs from the same GO category . On average , pN/pS in the GO-matched non-VIPs is 0 . 647 ( 95% CI [0 . 621 , 0 . 674] ) . This is only slightly lower than the average ratio in non-VIPs in general ( pN/pS=0 . 697 , P=2x10-3 ) , but much higher than the average ratio in VIPs ( 0 . 526 , permutation test P=0 after 104 iterations ) . Moreover , the stronger purifying selection acting on VIPs is apparent within most functions . Figure 3C shows stronger purifying selection in the 20 high level GO categories with the most VIPs . In all the 20 GO categories pN/pS is lower in VIPs than in non-VIPs , and the difference is significant for 17 of these categories ( Supplementary file 1C ) . This shows that within a wide range of host functions , viruses tend to interact with the most conserved proteins . Interestingly , even immune VIPs ( Supplementary file 1D ) have a significantly reduced pN/pS ratio compared to immune non-VIPs ( Figure 3C ) , which suggests that immune proteins in direct physical contact with viruses are more constrained . The reduction in pN/pS in non-immune VIPs is very similar to the reduction observed in the entire set of VIPs ( Figure 3C ) . The table at Supplementary file 1C further shows stronger purifying selection in 124 of the 162 GO categories ( 77% ) with more than 50 VIPs . The increased rate of adaptation in VIPs in the human lineage strongly suggests that VIPs in our dataset , 95% of which interact with modern viruses affecting humans ( Supplementary file 1A ) , were also VIPs during the last 7 million years of human evolution since the split with chimpanzees . It is also plausible that a substantial proportion of the VIPs we study are also VIPs in multiple mammalian lineages . Indeed , viruses infecting humans ( including the ten viruses with the most VIPs ) are known to have close viral relatives in many other mammals , with the exception of Hepatitis C Virus ( HCV ) for which only distant relatives are known and primarily in bats ( Quan et al . , 2013 ) . There is also growing evidence that distantly related viruses tend to interact with overlapping sets of host proteins ( Jäger et al . , 2011 , Davis et al . , 2015 ) . We thus hypothesize that VIPs , while identified primarily in humans , may have also experienced frequent adaptation in mammals in general , with the possible exception of the VIPs interacting with HCV . To test this hypothesis we use the Branch-Site Random Effect Likelihood test ( BS-REL test ) ( Kosakovsky Pond et al . , 2011 ) and the BUSTED test ( Murrell et al . , 2015 ) both available in the HYPHY package ( Pond et al . , 2005 ) in order to detect episodes of adaptive evolution in each of the 44 branches of the mammalian tree used for the analysis ( Materials and methods ) . For a specific coding sequence , the BS-REL and BUSTED tests estimate the proportion of codons where the rate of non-synonymous substitutions is higher than the rate of synonymous substitutions ( dN/dS>1 ) , which is a hallmark of adaptive evolution . The BS-REL test estimates proportions of selected codons specifically for each branch , whereas BUSTED estimates an overall proportion of selected codons across the entire tree . Both tests then compare two competing models of evolution , one with adaptive substitutions and one without adaptive substitutions , and decide whether the model with adaptation is a significantly better fit to the data . The BUSTED P-value is a good measure of whether a specific protein experienced adaptation in the history of mammalian evolution . In addition to presence/absence of adaptation , we assess the amount of adaptation experienced by a particular protein by estimating the average proportion of selected codons from the BS-REL test along all mammalian branches . We compare the proportion of selected codons detected by the BS-REL test between VIPs and non-VIPs . The statistical power of BUSTED and the BS-REL test has been shown to depend strongly on the amount of constraint in a coding sequence , with higher constraint/purifying selection decreasing the ability to detect adaptation ( Kosakovsky Pond et al . , 2011 ) . We confirm this in our dataset by observing a strong positive correlation between the pN/pS ratio in great apes and the proportion of selected codons across mammals estimated by the BS-REL test ( Spearman’s rank correlation ρ =0 . 34 , p<2x10-16 , n=9861 ) . We therefore use a permutation test with a target average ( Materials and methods ) that matches VIPs and non-VIPs with similar pN/pS ratios in order to compare VIPs and non-VIPs that experience similar levels of purifying selection and providing us with similar power to detect adaptation ( Materials and methods and Figure 5—figure supplements 1 and 2 ) . The permutation test shows that adaptation has been much more common in VIPs than in non-VIPs across mammals ( Figure 5 ) . We estimate that all VIPs have experienced twice as many adaptive amino acid changes on average compared to non-VIPs ( Figure 5A , permutation test P=0 after 109 iterations ) . We further use an increasingly strict level of evidence for the presence of adaptation , by including only proteins with increasingly low BUSTED P-values; that is , increasingly high probability that adaptation occurred somewhere on the tree ( Figure 5A ) . Figure 5A shows that VIPs with the strongest evidence of adaptation ( BUSTED P-values lower than 10-5 ) have a six-fold excess of strong signals of adaptation ( permutation test P=0 after 109 permutations ) . In Figure 5—figure supplement 3 we further show that this excess of adaptation in VIPs is due to i ) more VIPs with signals of adaptation than non-VIPs , ii ) more branches of the tree per VIP showing adaptation , and iii ) a greater proportion of codons evolving adaptively per branch . In line with the MK test , we find that the excess of adaptation in mammals is robust to the potential confounding factors of expression at the RNA and protein levels , and to the number of host protein-protein interactions ( Supplementary file 1H ) . Indeed , adaptation in mammals remains at least twice more frequent in VIPs compared to non-VIPs expressed at the RNA level in many tissues ( permutation test p<10-5for VIPs and non-VIPs expressed in at least 10 , 20 , 30 or 40 GTEx tissues ) , VIPs and non-VIPs expressed at the protein level in many tissues ( permutation test p<for VIPs and non-VIPs expressed in at least 10 or 20 Human Proteome Map tissues ) , and VIPs and non-VIPs with a high number of protein-interacting partners ( permutation test P<10–5for VIPs and non-VIPs with at least 5 or 10 protein-interacting partners ) . 10 . 7554/eLife . 12469 . 008Figure 5 . Excess of adaptation across mammals in VIPs The excess of adaptation is measured as the extra percentage of adaptation in VIPs compared to non-VIPs . For example , if VIPs have 1 . 5 times or 50% more adaptation , then the adaptation excess is 50% . ( A ) Thick black curve: average excess of adaptation in all VIPs . Dotted black curves: 95% confidence interval for the excess of adaptation in all VIPs . Thick grey curve: excess of adaptation in non-immune VIPs . Dotted grey curves: 95% confidence interval for the excess of adaptation in non-immune VIPs . ( B ) Virus-by-virus excess of adaptation in VIPs . Black dot is the average excess and the represented interval is the 95% confidence interval . Excess is shown for BUSTED p≤0 . 5 . ( C ) Excess of adaptation within the top 20 high-level GO processes with the most VIPs . Excess is shown for BUSTED p≤0 . 5 . ( D ) Proportions of selected codons in VIPs ( blue dot ) and non-VIPs ( red dot and 95% confidence interval ) in the mammalian clades represented by more than one species in the tree . All: entire tree . Primata: primates . Glires: rodents and rabbit . Cetartyodactyla: sheep , cow , pig . Zooamata: carnivores and horse . Excess is shown for BUSTED p≤0 . 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 12469 . 00810 . 7554/eLife . 12469 . 009Figure 5—figure supplement 1 . How to compare VIPs and non-VIPs across mammals Red: part of dN/dS explained by adaptive evolution . Grey: part of dN/dS explained by neutral evolution . Non-VIP I has the same amount of purifying selection as the VIP . Non-VIPs number II , III and IV have the same dN/dS as the VIP . Non-VIP II has the same amount of purifying selection as the VIP , non-VIP III has less purifying selection ( more neutral evolution ) and non-VIP IV has more purifying selection . In all cases , matching by dN/dS would be overly conservative . Upper arrow: observed dN/dS . Lower arrow: expected dN/dS if there was no adaptive evolution and only neutral evolution . DOI: http://dx . doi . org/10 . 7554/eLife . 12469 . 00910 . 7554/eLife . 12469 . 010Figure 5—figure supplement 2 . Scheme for the permutation test with a target average using the example of purifying selection . A full explanation of the permutation scheme is provided in Materials and methods . In brief , we sample non-VIPs that maintain the cumulated average of all sampled non-VIPs within the target interval [dN ( inf ) ;dN ( sup ) ] ( blue dots on the scheme ) . Every fixed number of sampled non-VIPs , we authorize one non-VIPs to drive the cumulated average outside of the target interval ( red dots ) . When it does happen that the cumulated average is driven outside of the interval , we sample as many non-VIPs as necessary that decrease or increase the cumulated average back to the target interval based on whether the cumulated average is above or below the target interval ( grey dots ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12469 . 01010 . 7554/eLife . 12469 . 011Figure 5—figure supplement 3 . Contributions of the number of genes , number of branches and proportion of selected codons to the excess of adaptation in VIPs . The excess of adaptation could be due to more genes with evidence of adaptation , more branches per gene with adaptation , and/or a greater proportion of selected codons per branch . Upper plain line: excess of adaptation in VIPs measured using the proportion of selected codons . Middle dotted line: excess of adaptation in VIPs measured using the number of branches with evidence of adaptation . Lower dotted line: excess of adaptation measured by counting the number of genes with evidence of adaptation . DOI: http://dx . doi . org/10 . 7554/eLife . 12469 . 011 We further quantify the excess of adaptation specifically for each of the ten viruses with more than 50 VIPs ( HIV-1 , HPV , HCV , EBV , HBV , HSV , Influenza Virus , ADV , HTLV and KSHV ) ( Figure 5B ) . Nine out of the ten viruses show a strong excess of adaptation in their respective VIPs . HIV-1 and HBV VIPs have the strongest excess , with three times as many selected codons as non-VIPs . As mentioned above , HCV stands out among the ten viruses with the largest number of VIPs in humans in that it has no known close viral relatives despite extensive screening of diverse mammalian species ( Quan et al . , 2013 ) . If this reflects a true lack of close viral relatives of HCV , then we predict a limited excess of adaptation in HCV VIPs . In line with this prediction , the 109 VIPs of HCV are the only ones where we do not detect any excess of adaptation along the mammalian tree ( Figure 5B ) despite being one of the largest groups of VIPs . VIPs that interact with all other viruses all show substantial elevation of adaptation ( Figure 5B ) . VIPs are represented in a wide range of GO functions with 162 GO categories having more than 50 VIPs ( Supplementary file 1C ) . Of these 162 GO categories , 118 ( 73% ) have a 50% or greater excess of adaptation in VIPs ( Supplementary file 1C , permutation test p<0 . 05 in all cases ) . The excess of adaptation in VIPs is therefore widespread across host functions . GO processes with a strong excess of adaptation include cellular processes such as transcription , signal transduction , apoptosis , or post-translational protein modification , but also supracellular processes related to development ( Figure 5C and Supplementary file 5 ) . As expected , immune VIPs have a very strong excess of adaptation compared to immune non-VIPs ( Figure 5C ) . Importantly , VIPs with no known immune function ( Supplementary file 1D ) show a very similar excess of adaptation compared to all VIPs ( Figure 5A , black versus grey lines; permutation test P=0 after 109 iterations ) . Overall these results suggest that the arms race with viruses has strongly increased the rate of adaptation in a wide range of VIPs . Since 95% of the VIPs were discovered for viruses infecting humans , it is possible that the observed excess of adaptation in VIPs in mammals is due to higher rates of adaptation exclusively in the primate branches of the mammalian tree ( Figure 1 ) . However , all mammalian clades in the tree show a similar excess of adaptation in VIPs ( Figure 5D ) . Primates stand out due to their low overall proportions of positively selected codons compared to the other mammalian clades in the tree ( Figure 5D ) . This is most likely due to a lower statistical power of the BS-REL test in the short primate branches ( Kosakovsky Pond et al . , 2011 ) . In line with this , VIPs with strong signals of adaptation show such signals in all the mammalian clades represented ( Figure 6 ) . This includes well-known antiviral VIPs ( Figure 6A ) , antiviral VIPs where adaptation was previously unknown across mammals ( Figure 6B ) , and non-antiviral VIPs with diverse , well-studied functions in the mammalian hosts ( Figure 6C ) . This phylogenetically widespread excess of recurrent adaptation suggests that many of the VIPs annotated in humans were also VIPs for a substantial evolutionary time in a wide range of mammals . 10 . 7554/eLife . 12469 . 012Figure 6 . Examples of mammalian orthologs with adaptation spread across clades . ( A ) Signals of adaptation in eight antiviral proteins with well-known adaptation across mammals . Red: BS-REL p≤0 . 001 . Orange: BS-REL p≤0 . 05 . Yellow: BS-REL p≤0 . 1 . ( B ) Top eight antiviral proteins with the highest number of branches under selection , and no previously know adaption spread across mammals . Note that adaptation was previously found for TRIM21 in primates but no other mammalian clade ( Malfavon-Borja et al . , 2013 ) . ( C ) Top eight non-antiviral proteins with well-known functions and the highest number of branches under selection across mammals . Proteins are ordered according to the number of branches with signals of adaptation . DOI: http://dx . doi . org/10 . 7554/eLife . 12469 . 012 We have shown that rates of adaptation are globally elevated in VIPs in humans and mammals in general , suggesting the existence of tens of thousands of isolated events of adaptations to a diverse range of viruses . Here , we test if our global approach has enough power to isolate new specific cases of adaptation to viruses by looking for instances where viruses are the plausible cause of adaptation in a VIP with no known antiviral activity . This is particularly relevant because , to our knowledge , the transferrin receptor ( TFRC ) is one of the only well understood case of a non-antiviral protein adapting in response to viruses ( Demogines et al . , 2013 ) . To identify a non-antiviral VIP for in-depth investigation we first excluded all VIPs with a well-known antiviral activity ( Supplementary file 6; here as in the rest of the manuscript antiviral means a protein activity that restricts viral infection ) and then selected all remaining VIPs with strong overall evidence of adaptation ( Supplementary file 1I ) and at least 10 branches with signals of adaptation . We also selected proteins with i ) at least one available tertiary structure , ii ) amino acid level resolution of the interaction with one or more viruses , and iii ) host tropism . The most positively selected non-antiviral VIP that fulfills all these requirements is aminopeptidase N , abbreviated ANPEP , APN or CD13 ( Mina-Osorio , 2008 ) . The analysis of a phylogenetic tree including 84 mammals ( Supplementary file 1J ) confirms pervasive adaptation of ANPEP across mammals , with 76 out of 165 branches in the tree showing signals of adaptation ( Figure 7A ) . Note that adaptation of ANPEP has previously been detected in the context of oxidative stress in Cetaceans ( Yim et al . , 2014 ) . ANPEP is a cell-surface enzyme well known for its surprisingly wide range and diversity of functions ( Mina-Osorio , 2008 ) . In particular , it is used by group I coronaviruses as a receptor , including the Human Coronavirus 229E ( HCoV-229E ) ( Yeager et al . , 1992 ) , Transmissible Gastroenteritis Virus ( TGEV ) ( Delmas et al . , 1992 ) , Feline Coronavirus ( FCoV ) ( Tresnan and Holmes , 1998 ) , Canine Coronavirus ( CCoV ) ( Tusell et al . , 2007 ) , Porcine Respiratory Coronavirus ( PRCV ) ( Delmas et al . , 1993 ) and Porcine Epidemic Diarrhea Virus ( PEDV ) ( Oh et al . , 2003 ) . Reguera et al . ( Reguera et al . , 2012 ) solved the structure of porcine ANPEP bound together with TGEV and PRCV . The authors identified in the extracellular domain of ANPEP 22 amino acids that form a surface of contact with TGEV and PRCV ( Figure 7B ) ( Pettersen et al . , 2004 ) . The most important component of this contact surface for host tropism is a N-glycosylation site at position 736 in porcine ANPEP ( orthologous position 739 in human ANPEP ) that forms hydrogen bonds with TGEV and PRCV ( Tusell et al . , 2007; Reguera et al . , 2012 ) . Deleting this site abolishes the ability of TGEV and PRCV to bind porcine ANPEP ( Reguera et al . , 2012 ) . Adding the glycosylation site in human ANPEP that natively lacks it transforms it into a receptor for TGEV and PRCV ( Reguera et al . , 2012 ) . 10 . 7554/eLife . 12469 . 013Figure 7 . Patterns of adaptation to coronaviruses in aminopeptidase N . ( A ) BS-REL test results for ANPEP in a tree of 84 mammalian species . Legend is on the figure . ( B ) Contact surface with PRCV and TGEV on ANPEP structure ( PDB 4FYQ ) . The figure includes visualizations of all the six different faces of ANPEP . Legend is in the figure . ( C ) Excess of adaptation in and near the contact interface with PRCV and TGEV . Within the contact interface plus a given number of neighboring amino acids ( one , five , ten or 20 in the figure ) , adaptation excess ( y axis ) is defined as the number of observed codons with a MEME P-value lower than the P-value threshold on the x axis , divided by the average number of codons under the same P-value threshold obtained after randomizing the location of adaptation signals over the entire ANPEP coding sequence 5000 times . Dark red curve: adaptation excess within the contact interface with TGEV and PRCV plus one neighboring amino acid . Red curve: plus five neighboring amino acids . Orange: plus ten neighboring amino acids . Light orange: plus 20 neighboring amino acids . Numbers in the figure represent the number of adapting codons , and the stars give the significance of the excess . One star: excess p≤0 . 05 . Two stars: p≤0 . 01 . ( D ) Losses and gains of the N-glycosylation across the mammalian phylogeny . DOI: http://dx . doi . org/10 . 7554/eLife . 12469 . 013 We use the MEME test from HYPHY ( Murrell et al . , 2012 ) to identify codons in ANPEP that were under episodic adaptive evolution in mammals . MEME detects significant adaptation ( MEME p≤0 . 05 ) in 85 of the 931 aligned codons . Interestingly , several of these adaptively evolving codons are within , or right next to the surface of contact with TGEV and PRCV ( Figure 7B and C ) . The codons in contact with TGEV and PRCV and their neighbors are strongly enriched in adaptation compared to ANPEP codons as a whole ( Figure 7C ) . This enrichment fades very rapidly as one gets further from the surface of contact with TGEV and PRCV , consistent with detected adaptation being related to interaction with coronaviruses , and not to a more diffuse , less specific enrichment within a wider segment of ANPEP ( Figure 7C ) . Adaptively evolving codons in the contact surface with TGEV and PRCV most notably include two codons within the consensus motif for the N-glycosylation site responsible for host tropism ( Figure 7B ) . N-glycosylation is governed by a three amino acids consensus , NX[ST] , where X can be anything except proline ( Bause , 1983 ) . The first and third positions in the consensus evolved adaptively in mammals ( MEME P=0 . 005 for both ) . The ancestral states of the two positions shows that the mammalian ancestor had a fully functional consensus for N-glycosylation , and that the consensus was lost independently 11 times in mammals , either by modification of the first or third position ( Figure 7D ) . The consensus was regained only two times after loss ( Figure 7D ) . This suggests that the signals of adaptation detected at the first and third positions in the consensus mainly reflect parallel , adaptive losses of the N-glycosylation site in multiple mammalian lineages . Given the crucial role of this N-glycosylation site in the binding of TGEV , PRCV , FCoV and CCoV to ANPEP , it is probable that these parallel adaptive losses were due to selective pressure exerted by ancient coronaviruses .
Here , we have shown that viruses have been a major selective pressure in the evolution of the mammalian proteome . Indeed viruses appear to drive ~30% of all adaptive amino acid changes in the conserved part of the human proteome , as evidenced by the MK test . Furthermore , the footprints of the arms race with viruses are visible in a large number of VIPs , and in a broad range of mammals . Importantly , we find a substantial enrichment in strong signals of adaptation in VIPs with no known antiviral or other immune functions ( Figures 4B and 5A ) . Instead adaptation to viruses is visible in VIPs with a very diverse range of functions including such core functions as transcription or signal transduction ( Figures 4B , 5C and Supplementary file 1C ) . This very diverse range of functions strongly argues in favor of a pervasive , external selective pressure –in our case viruses– as the cause for the observed signals . Our results thus draw a broader picture where adaptation against viruses involves not only the specialized antiviral response , but also the entire population of host proteins that come into contact with viruses . The best-known case of a housekeeping protein having adapted in response to viruses , the transferrin receptor , may thus represent the rule more than an exception ( Demogines et al . , 2013 ) . In line with this , a new non-antiviral protein , NPC1 , has very recently been shown to adapt against its use as a receptor by filoviruses in bats ( Ng et al . , 2015 ) . Although we find a strong signal of increased adaptation , the amount of adaptive evolution that can be attributed to viruses is probably underestimated by our analysis . First , there likely to be many undiscovered VIPs . There is no sign that the pace of discovery of new VIPs is slowing down ( Figure 2 ) . This means that a substantial number of proteins classified as non-VIPs in this analysis are in fact VIPs , making non-VIPs a conservative control . Second , adaptation in response to viruses is most likely not restricted to proteins that physically interact with viruses . For example , adaptation to viruses might happen in proteins that act downstream of VIPs in signaling cascades , or in non-coding sequences that regulate the expression of VIPs . Third , not all of the 1256 VIPs we use here have been consistently interacting with viruses during evolution . Most VIPs in the dataset ( 95% ) were discovered in humans , and how frequently these VIPs have also been interacting with viruses in other mammals is currently unknown . Some VIPs like PKR have probably been in very frequent contact with diverse viruses . Conversely , other VIPs may have been in contact with viruses for a very limited evolutionary time in mammals , and only in a limited range of lineages . This would apply to VIPs that interact with viruses with a limited host range and few other phylogenetically closely related viruses , as is the case with HCV ( Figure 5B ) . In addition , we could only work with VIPs active in current human populations reflecting the set of viruses infecting humans at present . This means that a potentially large number of proteins classified as non-VIPs in our study were actually VIPs during past human evolution or during the evolution of other mammalian lineages . Altogether , the bias of our sample towards present human VIPs thus makes our results conservative . We speculate that our results might explain a puzzling observation that rates of protein adaptation appear relatively invariant across different biological functions assessed using GO analysis ( Bierne and Eyre-Walker , 2004 ) . Because viruses ( and likely other pathogens ) interact with diverse proteins across most GO functions , they elevate the rate of adaptation across the whole proteome in a way that appears independent of specific functions in the GO analysis . We argue that grouping genes together based on the way they interact with diverse pathogens or other environmental stimuli might be a profitable way for discerning the nature of selective pressures that have molded animal genomes . In conclusion , our analysis suggests that viruses have exerted a very powerful selective pressure across the breadth of the mammalian proteome , and suggests the possibility that pathogens in general are the key driver of protein adaptation in mammals and likely other lineages and might have driven many pleiotropic effects on diverse biological functions .
We identified 1256 proteins that physically interact with viruses out of a total of 9 , 861 proteins with orthologs in the genomes of the 24 mammals included in the analysis ( Figure 1 and Supplementary files 1A and 1B ) . Annotation of the 1256 interacting proteins was performed by querying PUBMED ( http://www . ncbi . nlm . nih . gov/pubmed ) . We started by downloading all the known human gene identifiers gathered by the HUGO Gene Nomenclature Committee ( http://www . genenames . org/ ) ( Gray et al . , 2015 ) for each of the 9861 orthologs ( on the 24th of June 2013 ) . Each gene identifier was then used to automatically query PUBMED together with the term virus . The first automatic query was performed on the 24th of June 2013 , and the last update was performed by automatically querying PUBMED again on the 10th of December 2014 . We then manually went through all the matches by first looking at the titles of the publications retrieved . Titles of publications unlikely to report an interaction , such as publications about antiviral medications or publications about the prevalence of a virus in a given population , were discarded from the annotation process . Titles of publications directly mentioning an interaction , or of publications reporting insights at the cellular/molecular biology level were retained for further inspection of their abstracts . For the vast majority of cases , interactions between a host protein and a viral protein , RNA or DNA are clearly reported in the abstracts , as it usually represents an important finding of many publications in the virology literature . For more ambiguous cases , we went through the full text of publications . We did not consider interactions identified only through high throughput methods such as two-hybrid or mass spectrometry screens to limit the number of false positives in our dataset ( Mellacheruvu et al . , 2013 ) . In total , we could identify 982 proteins with at least one known interaction with a viral protein , RNA or DNA . We completed our own annotations with 274 additional proteins identified with low throughput methods listed in the VirHostNet ( http://virhostnet . prabi . fr/ ) ( Guirimand et al . , 2015 ) and HPIDB ( http://www . agbase . msstate . edu/hpi/main . html ) ( Kumar and Nanduri , 2010 ) databases as of February 4th 2015 . In total , 1256 of the 9861 orthologous proteins ( 13% ) were found to interact with viral proteins , RNA or DNA according to low throughput methods . These interactions are available online as a supplemental table ( Supplementary file 1A ) . We identified and aligned orthologous coding sequences in the genomes of 24 mammals ( Figure 1 and Supplementary file 1B ) . Those 24 mammals were those with high sequencing depth genomes as of December 2012 . They include the assemblies hg19 for human , chimpanzee panTro4 , gorilla gorGor3 , orangutan ponAbe2 , gibbon nomLeu3 , macaque rheMac3 , baboon papAnu2 , marmoset calJac3 , bushbaby otoGar3 , mouse mm10 , rat rn5 , guinea pig cavPor3 , squirrel speTri2 , rabbit oryCun2 , sheep oviAri3 , cow bosTau7 , pig susScr3 , microbat myoLuc2 , panda ailMel1 , ferret musFur1 , dog canFam3 , cat felCat5 , horse equCab2 , and elephant loxAfr3 . We first used Blat ( Kent , 2002 ) to find all the homologous matches of 22 , 074 human coding sequences ( CDS ) from Ensembl v69 ( Flicek et al . , 2012 ) in all the assemblies listed above . These CDS are the longest CDS for their respective human genes . Blat was parameterized for high sensitivity , using translated genomes ( options -q=dnax and -t=dnax ) , translated queries and setting the minimum identity to 50% , in addition to using the -fine option . Best Blat matches with the highest number of matching positions were then blatted back on the human genome to identify best reciprocal hits . This way we could find a total of 9861 human CDS with best reciprocal hits in all the other 23 species , no in-frame stop codon and at least 30% of the length of the human CDS ( Supplementary file 1B ) . Of these , 9338 human CDS were found to have clear conserved synteny in at least 18 of the 23 non-human species , with at least five conserved neighboring protein coding genes within 500 kb upstream their 5’start and 500 kb downstream their 3’ end ( Supplementary file 1B ) . The CDS of these 9338 orthologs were then aligned using PRANK under the codon evolution model ( Loytynoja and Goldman , 2008 ) . Any codon present in less than eight species was discarded . The combination of Blat to extract the homologous sequences and PRANK to realign them ensures we work with high quality alignments . Indeed , the first step of local alignment with Blat excludes segments of the CDS that are too diverged between mammals to be properly aligned during the subsequent step of global alignment with PRANK , which has been shown to be the best coding sequence aligner available ( Fletcher and Yang , 2010; Jordan and Goldman , 2012 ) . We consider proteins as antiviral if they restrict viral replication in any way , for example by directly engaging viruses for recognition and/or degradation of viral molecules , or if they have been shown to be required for the proper unfolding of the antiviral response . Antiviral proteins were annotated at the same time as VIPs and are listed in Supplementary file 1D . We also annotated VIPs with immune functions while identifying VIPs . The final set of immune VIPs is made of the 203 VIPs annotated with the GO categories 'immune system process' ( GO:0002376 ) , 'defense response' ( GO:0006952 ) or 'immune response' ( GO:0006955 ) ( 2015 ) , in addition to 38 immune VIPs we annotated based on publications reporting roles in various parts of the immune response , from innate to adaptive immune response , from regulators to effectors of the immune response , to VIPs involved in the development of the immune system . Supplementary file 1D lists all the immune VIPs identified . We estimate and compare the proportion of adaptive amino acid changes , noted α , in VIPs and non-VIPs using either the classic McDonald-Kreitman test ( MK test ) ( McDonald and Kreitman , 1991 ) or an asymptotic MK test ( Messer and Petrov , 2013 ) . The MK test measures α as follows:α=1− ( DS∗pN ) / ( DN∗pS ) where DN and DS are the number of fixed amino acid substitutions and synonymous substitutions in the lineage studied since divergence with a closely related species , and pN and pS are the number of polymorphic non-synonymous and synonymous sites , respectively . For our study , DN and DS are the number of substitutions in the human lineage since divergence with chimpanzee . Theses substitutions are identified in human-chimpanzee-orangutan alignments of the mammalian orthologous CDS ( Supplementary file 1E ) . These alignments are prepared the same way as the alignments of 24 mammals ( see Materials and methods , 'Multiple alignments of mammalian orthologs' ) , except for the fact that we include all isoforms of every ortholog in the analysis . The substitutions are those positions where chimpanzee and orangutan have the same nucleotide but human has a different nucleotide . We further require the substitution to be fixed in human , as shown by the fact that the position is not polymorphic in the 1000 Genomes Project data ( Abecasis et al . , 2012 ) . The measures of pN and pS are obtained as the numbers of polymorphic positions in African populations from the final phase I 1000 Genomes Project . For the classic MK test , we only use variants with a derived allele frequency between 0 . 1 and 0 . 9 in African populations in order to limit the effect of deleterious mutations . Note however that the classic MK test has been shown to underestimate α even when excluding low frequency variants . The classic MK test is thus well suited to measure the relative difference in adaptation between VIPs and non-VIPs , but it is not well suited to actually measure the true absolute α in VIPs and non-VIPs . For this we use the asymptotic MK test since it does not underestimate α ( Messer and Petrov , 2013 ) . It is nevertheless limited to using large numbers of genes ( >1000 ) . The asymptotic MK test is robust to the presence of deleterious mutations and to demography . The asymptotic MK test works by estimating α in bins of derived allele frequencies . For example , α can be calculated in the bin of frequencies from 0 . 1 to 0 . 2 by counting only variants with a derived allele frequency between 0 . 1 and 0 . 2 to measure pN and pS . An exponential curve is then fitted to the estimates of α across bins of frequency . The value taken by the fitted curve for a derived allele frequency of 100% provides the estimate for α ( Messer and Petrov , 2013 ) . Using the asymptotic MK test , Messer and Petrov ( Messer and Petrov , 2013 ) estimated that α is 57% in Drosophila melanogaster , and 13% in human . For both species , these estimates were obtained based on polymorphism and divergence data for most of the proteome ( more than 104 protein coding sequences in both cases ) . Here , we need to estimate α in 1256 VIPs , and in the same number of randomly sampled non-VIPs . This is an order of magnitude less than the number of coding sequences used by Messer and Petrov ( Messer and Petrov , 2013 ) , which makes curve fitting challenging . Indeed , the low number of high frequency variants means that estimates of α in the high frequency bins are very noisy . Using the stable release 1000 Genomes Project final phase I variants from African populations , we count that VIPs only have 143 non-synonymous and 343 synonymous variants with a derived allele frequency above 0 . 5 , respectively . The [0 . 8 , 0 . 9] bin of frequency only has 18 non-synonymous and 51 synonymous variants . In comparison , the [0 . 1 , 0 . 2] bin has 149 non-synonymous and 370 synonymous variants . The [0 . 4 , 0 . 5] bin still has twice more variants ( 36 non-synonymous , 101 synonymous ) than the [0 . 8 , 0 . 9] bin . The low number of high frequency variants is however not the only issue . A second potential issue when trying to fit a curve to predict α in the asymptotic McDonald-Kreitman test is the mispolarization of alleles as ancestral or derived . Mispolarization is a common problem that distorts the unfolded Site Frequency Spectrum ( SFS ) ( Hernandez et al . , 2007 ) . The most severe distortion is usually within the high frequency part of the SFS ( Hernandez et al . , 2007 ) . Indeed , abundant low-frequency derived variants are often misidentified as high frequency derived variants . This can result in substantial overestimations of the number of high frequency variants . The number of non-synonymous variants pN might be more severely overestimated than pS , since less high frequency and more low frequency non-synonymous variants are expected in the first place . This could hypothetically result in underestimates of α within high frequency bins . Here we modify the asymptotic MK test to circumvent the mispolarization and high frequency , high noise issues . We do so by estimating α based only on derived allele frequencies lower than 0 . 5 , where the distortion of the SFS due to mispolarized alleles is negligible . This also makes the asymptotic McDonald-Kreitman less reliant on bins of high frequencies with very noisy estimates of α , due to small values of pN and pS . We use either a logarithmic fit of the form y=a+b ( ln ( x+c ) ) over the range of frequencies 0 to 0 . 5 ( Figure 4C ) , or an exponential fit of the form y=a+b∗exp ( −x/c ) . Both the logarithmic fit and the exponential fits provide accurate estimates of α for a wide range of evolutionary scenarios , as shown by forward population simulations using SLIM ( Messer , 2013 ) . We use the forward population simulator SLIM ( Messer , 2013 ) to simulate a typical , 400 codons , six exons coding sequence . Each exon is separated by 4000 bp long introns . One in four coding sites is synonymous and only experiences neutral mutations . Non-synonymous sites experience neutral , advantageous , strongly deleterious and slightly deleterious mutations . The coding sequence evolves for 20 , 000 generations in a population of 1000 individuals , at a uniform mutation rate of 2 . 5 x 10−7 and with a uniform recombination rate of 10 cM/Mb . These parameters are equivalent to 200 , 000 generations of evolution of a 10 , 000 individuals population with a mutation rate of 2 . 5 x 10−8 and a recombination rate of 1 cM/Mb . This results roughly in the amount of divergence observed in the human lineage since divergence with chimpanzee . The rescaling by a factor of ten greatly speeds up the simulations . Roughly matching the observed DN , DS , pN and pS in VIPs requires simulating 1000 coding sequences . The true α obtained from simply counting adaptive fixations in the simulations can then be compared with the α estimated from DN , DS , pN and pS . By repeating the simulation of sets of 1000 coding sequences many times , we can get the variance of the estimation of α both by the modified asymptotic MK test . By repeating the simulations 100 times , we show that the modified asymptotic MK test gives accurate estimates of α for all evolutionary scenarios tested ( Supplementary file 1G ) . In practice , the logarithmic fit is easier to use than the exponential fit . Indeed , fitting algorithms such as the ones implemented in the LM ( ) function or the nlsLM ( ) function from the minpack . lm package in R often fail to converge for the exponential fit . We therefore use the logarithmic fit . We use the multiple alignments of the coding sequences from the 24 mammals listed above to quantify adaptation across mammals . There are three different types of tests aimed at detecting and quantifying adaptation in a multi-species coding sequence alignment: branch tests , site tests , and branch–site tests . The so-called branch tests look for branches in a tree where the ratio of non-synonymous to synonymous substitutions dN/dS exceeds one for the entire coding sequence . In order to happen this requires an extreme amount of adaptation in a specific branch . Branch tests thus detect only the most extreme bursts of adaptation , and have very low statistical power to detect the vast majority of more moderate bursts of adaptation in a phylogeny ( Nielsen et al . , 2005 ) . This makes them a very poor choice to quantify adaptation within an entire phylogeny . Site tests look for specific codons of a coding sequence where dN/dS significantly exceeds one across the entire phylogeny . Codons with dN/dS >> 1 are codons that have accumulated many adaptive non-synonymous substitutions across the tested phylogeny . This means site tests ignore the case where specific codons have evolved adaptively on a specific branch , probably the most common case in coding sequence evolution ( Murrell et al . , 2012 ) . Although site tests are well suited for cases where there is a strong a priori expectation about which sites should evolve adaptively , as is for example the case of TFRC , here we have no a priori knowledge about the sites that are expected to evolve adaptively in VIPs in response to viruses . Instead we use branch-site tests which are designed to detect adaptation at specific codons in specific branches . There are currently two main implementations of the branch-site test , one available in PAML ( Zhang et al . , 2005 , Yang , 2007 ) and one available in the HYPHY package ( Kosakovsky Pond et al . , 2011 ) . The two tests are both likelihood ratio tests that compare a model integrating positive selection with a neutral model without positive selection . The PAML branch-site test and the HYPHY BS-REL branch-site test differ mainly in the assumptions of their evolutionary models . The PAML branch-site test defines two kinds of branches in the phylogenetic tree used , the foreground and background branches . The foreground branch is the branch where the presence of positive selection is tested . The evolutionary model of the branch-site test authorizes positive selection in the foreground branch , but not in the background branch . Unlike the PAML branch-site test , the HYPHY BS-REL test uses a model that has no limitation regarding the occurrence of adaptation across the tree . This difference in the models used has very profound consequences for the ability of the two tests to detect and quantify recurrent adaptation ( Kosakovsky Pond et al . , 2011 ) . Indeed , the HYPHY test has good power to detect recurrent adaptation . Because it does not allow adaptation in the background branches , the PAML tests suffers a severe loss of statistical power when recurrent adaptation does occur in the background branches . As an example , the HYPHY BS-REL test detects significant ( BS-REL test p≤0 . 05 ) signals of adaptation in 18 branches of the mammalian tree used in this study for PKR ( Figure 6 ) . In comparison , the PAML test detects only nine branches ( PAML test p≤0 . 05 ) . This is a crucial difference between the two tests in our case given that the arms race with viruses is likely to trigger recurrent bursts of adaptation across mammals . For this reason we use HYPHY BS-REL to quantify adaptation in mammals . More specifically , we use the proportion of selected codons estimated by the BS-REL test to quantify adaptation . To estimate the strength of the evidence in favor of adaptation across the entire mammalian tree , we use the P-value of the BUSTED test in HYPHY that uses the same codon evolution model as the BS-REL test . In this study , we compare VIPs and non-VIPs for weak ( BUSTED P-values ≤0 . 9 ) to increasingly strong ( BUSTED P-values ≤10–5 ) evidence of adaptation . We start by computing the average proportion of codons under adaptive evolution for VIPs and the same average proportion for the sets of randomly matched non-VIPs ( see the description of the permutation test ) . For each coding sequence , we retrieve the proportion of positively selected codons on each branch , and compute the average of this proportion across branches . More specifically , we only count branches of the tree with conserved synteny ( Supplementary file 1B ) . That is , if 40 of the 44 branches of the tree have conserved synteny ( see above ) , we compute the average proportion of selected codons only from these 40 branches . In practice we tolerate only up to five branches in the tree with no conserved synteny ( at least 39 branches with conserved synteny; Supplementary file 3 ) . This reduces the dataset of orthologs that can be used in the analysis only slightly , from 9861 to 9338 total , and among those the number of VIPs from 1256 to 1193 . Then adaptation is simply quantified as the average proportions of selected codons in valid branches across VIPs ( or the same number of matched non-VIPs ) . If the threshold for BUSTED P-value is set to 10−x , we only include in the quantification the average proportions of selected codons from coding sequences with BUSTED P-value≤10−x . For a low x , we compare how much selection occurred in VIPs and non-VIPs counting both weak and strong signals of adaptation . For a high x , we compare how much strong , highly significant signals of adaptation occurred in VIPs compared to non-VIPs . We designed a permutation test that makes it possible to compare adaptation in VIPs with adaptation in non-VIPs with the same amount of purifying selection . The amount of purifying selection in a protein corresponds to the proportion of amino acids that cannot change , or very infrequently during evolution . On average VIPs experience much more purifying selection than non-VIPs . This means that mechanically , a smaller proportion of amino acids can possibly be targeted by adaptive evolution in VIPs . A naïve comparison of VIPs and non-VIPs would therefore tell more about the difference in purifying selection than about the difference in the amount of adaptation . Instead , the idea is to compare adaptation in the 1256 VIPs with adaptation in 1256 non-VIPs with the same overall average and variance in levels of purifying selection . The sampling of non-VIPs with similar purifying selection VIPs is however challenging . The first question is which measure of purifying selection to use ? The ratio of non-synonymous to synonymous substitution rates dN/dS is often used as a measure of purifying selection . How much smaller dN is compared to dS can indeed tell how evolutionarily constrained a protein is . However the problem with dN/dS is that dN not only reflects purifying selection , but also reflects adaptive amino acid substitutions . This means that comparing VIPs with non-VIPs with similar dN/dS ratios would underestimate an excess of adaptation in VIPs . This is because more adaptation would increase dN/dS in VIPs more than it does in non-VIPs ( Figure 5—figure supplement 1; compare VIP with non-VIP I ) . As a result , VIPs would be matched with either non-VIPs that have the same dN/dS because they have the same levels of both purifying selection and adaptation ( Figure 5—figure supplement 1; compare VIP with non-VIP II ) , or non-VIPs that have the same dN/dS because they have experienced less purifying selection , and thus have also had more codons available for adaptation to happen ( Figure 5—figure supplement 1; compare VIP with non-VIP III ) , or non-VIPs that have the same dN/dS because they are under both stronger purifying selection and adaptation ( Figure 5—figure supplement 1; compare VIP with non-VIP IV ) . In all three cases , non-VIPs selected as controls would experience more adaptation than non-VIPs that would have been matched based on a more accurate measure of purifying selection ( Figure 5—figure supplement 1; compare non-VIP I with non-VIPs II , III and IV ) . This ultimately results in underestimating any excess of adaptation in VIPs compared to non-VIPs . Unlike the dN/dS ratio , the ratio of non-synonymous to synonymous polymorphism pN/pS only reflects purifying selection . Indeed , pN is decreased by purifying selection , but is not affected by adaptive mutations that segregate for very short times in populations . This makes pN/pS a much better measure of purifying selection than dN/dS that can be used to match VIPs with similarly constrained non-VIPs . There are however two problems with the pN/pS ratio . The first is that proteome-wide estimates of pN/pS are not available for all the mammals included in this analysis . Good estimates of pN/pS require sequencing the genomes of a sufficient number of non-inbred individuals , ideally more than ten , within a given species . The pN/pS ratio is publicly available , based on the genome sequences of a sufficient number of individuals , in human ( Abecasis et al . , 2012 ) and the non-human primate species represented in the Great Ape Genome Project , namely chimpanzee , gorilla and orangutan ( Prado-Martinez et al . , 2013 ) . The limited number of species of the mammalian tree with pN/pS information can still be used as a control of purifying selection in the permutation test for all mammals . It is true that the pN/pS ratio within a species or a subset of species does not represent the absolute , overall level of purifying selection in the entire mammalian tree . It is known for instance that primates experience weaker purifying selection than rodents . What matters however for the permutation test is not the absolute level of purifying selection , but the relative difference in pN/pS between VIPs and non-VIPs . Indeed , VIPs and matched non-VIPs with similar pN/pS experience similar purifying selection , even if pN/pS is from a subset of species in the mammalian tree . Whether pN/pS is overall skewed towards higher or towards lower values in the subset of species used , then the skew is still the same for both VIPs and non-VIPs . This means that the relative difference in pN/pS is still a good measure of the general difference in purifying selection across mammals . A different skew in VIPs and non-VIPs requires invoking unlikely scenarios where VIPs would experience a global relaxation or intensification of constraint specifically in the primate species where pN/pS is available . Given the high number of VIPs and their high functional diversity ( Supplementary file 1C ) , such a global trend towards relaxation or higher constraint in primates is extremely unlikely . The pN/pS ratio from primate species can therefore be used as a control for purifying selection . More specifically , we use the pN/pS ratio from populations of chimpanzees ( Nigeria-Cameroon , Eastern and Central populations ) , gorillas ( Western lowland population ) and orangutans ( Sumatran and Bornean populations ) ( Supplementary file 1F ) . These populations are the populations included in the Great Apes Genome Project with the highest effective population sizes . Indeed , the pN/pS ratio is less noisy and available for more proteins in populations with higher population sizes and higher genetic diversity . For each VIP and non-VIP , the value of pN/pS used in the permutation test is simply the sum of pN across all the primate populations divided by the sum of pS across the same populations ( Supplementary file 1F ) . In each primate population , pN and pS are measured excluding singletons to limit the influence of potential erroneous variant calls . The second potential problem with pN/pS is that it is a noisy measure of purifying selection . At any time in a population of primates , only few positions are polymorphic within a typical ( ~300 codons ) coding sequence . As a consequence , a highly constrained coding sequence may by chance have more non-synonymous variants than synonymous variants , and a high pN/pS ratio . Conversely , a weakly constrained coding sequence may by chance have less non-synonymous variants , and a low pN/pS ratio . This is problematic if we want to use pN/pS as a control for purifying selection . One can consider the case of VIPs where pN/pS is substantially lower than in non-VIPs . Matching VIPs with non-VIPs with a similarly lower pN/pS , we would end up selecting non-VIPs with a lower pN/pS not because of purifying selection but merely because of noise . This makes controlling for purifying selection less straightforward than directly matching each individual VIP with non-VIPs with similar pN/pS ratios . Instead we use an indirect matching strategy . As described above , the pN/pS ratio is a noisy measure of purifying selection . This means that we cannot use a direct matching strategy between VIPs and non-VIPs for the permutation test . An indirect matching strategy can however still be used , that uses the mammals-wide rate of non-synonymous substitutions dN as an intermediate . In particular , we use PAML to estimate dN and dS under the M8 evolution model ( Yang , 2007 ) . The dN/dS ratio for the whole mammalian tree ( Supplementary file 1B ) integrates hundreds of millions of years of evolution . In the absence of adaptation , it would therefore be a much less noisy measure of purifying selection than pN/pS . The issue is however that dN is influenced by both purifying selection and by adaptation . If VIPs experience more adaptation than non-VIPs , then purifying selection being equal , we expect dN/dS to be higher in VIPs than in non-VIPs . If VIPs experience less adaptation than non-VIPs , then purifying selection being equal , we expect dN/dS to be lower in VIPs than in non-VIPs . VIPs have a 25% lower pN/pS than non-VIPs in great apes , but only a 15% lower dN/dS than non-VIPs . The smaller difference in dN/dS than pN/pS is an indication that adaptation has increased dN more strongly in VIPs than in non-VIPs . Purifying selection being equal , dN/dS is therefore higher in VIPs than in non-VIPs . This means that non-VIPs with the same pN/pS ( purifying selection ) as VIPs have a lower dN/dS . We can therefore match VIPs and non-VIPs with the same pN/pS by selecting non-VIPs with a dN/dS ratio that is only a fraction of the dN/dS in VIPs . This fraction can be adjusted through trial and error until finding the one that matches VIPs with non-VIPs with the same overall average pN/pS . This indirect matching strategy makes it possible to compare VIPs and non-VIPs with the same level of purifying selection while avoiding the pitfall of noise in pN/pS . The random sets of non-VIPs must fulfill two criteria to be comparable to VIPs . First , non-VIPs should have the same overall average pN/pS as VIPs . Second , non-VIPs should have the same variance in pN/pS , i . e . pN/pS values in non-VIPs are spread as much as they are in VIPs . This can all be achieved by using a permutation scheme where samples of non-VIPs must satisfy a pre-fixed , target average ( Figure 5—figure supplement 2 ) . Note that although here we detail the case of purifying selection , permutations with a target average can be used to get samples of non-VIPs similar to VIPs for any possible factor . For the case of purifying selection , the permutations with an average target work as follows . We first measure the average dN in VIPs , noted dN ( vip ) . Then we define the target average dN we wish the chosen non-VIPs to exhibit at the end of the sampling . In specific , we ask the target dN to be a fraction a of dN ( vip ) , plus or minus 5% . The target dNfor non-VIPs can therefore take values between dN ( inf ) and dN ( sup ) , where dN ( inf ) = 0 . 95 ( adN ( vip ) ) and dN ( sup ) = 1 . 05 ( adN ( vip ) ) . The fraction a is set manually through trial and error so that the sampled non-VIPs have the same average pN/pS as VIPs . Note that we use dN and not dN/dS to avoid giving too much weight to dS , as it tends to saturate and take much greater values than dN ( Supplementary file 1B ) and thus bears much more heavily on the dN/dS ratio . Non-VIPs are sampled using a simple algorithm described in Figure 5—figure supplement 2 . We first randomly sample a set of five non-VIPs . This initial sampling of five non-VIPs is repeated until their average dN falls within the target interval [dNinf , dNsup] . We then add randomly sampled non-VIPs one at a time until their number matches the number of VIPs . The average of all the sampled non-VIPs has to remain within [dN ( inf ) , dN ( sup ) ] ( blue dots in Figure 5—figure supplement 2 ) , except for every X non-VIP that is sampled completely randomly ( red dots in Figure 5—figure supplement 2 ) . This means that in the latter case the average dN of the sampled non-VIPs can fall out of [dN ( inf ) , dN ( sup ) ] . When this happens we sample non-VIPs with dN values that bring the average dN back within [dN ( inf ) , dN ( sup ) ] ( grey dots in Figure 5—figure supplement 2 ) . That is , if the average dN is above dN ( sup ) , we sample as many non-VIPs as necessary that each lower the average dN until it falls back within [dN ( inf ) , dN ( sup ) ] . If the average dN is below dN ( inf ) , we sample non-VIPs that each increase the average dN until it falls back within [dN ( inf ) , dN ( sup ) ] . The parameter X is the parameter of the test that makes it possible to match the variance in pN/pS of the sampled non-VIPs with the variance observed for VIPs . A low X gives samples of non-VIPs with a higher variance . A high X gives samples of non-VIPs with a lower variance . To define the fraction a and X , we get 104 random samples of non-VIPs . We then test whether the average and variance of pN/pS in VIPs are significantly different or not from the distributions of averages and variances of pN/pS given by the 104 random samples of non-VIPs . We find that a=0 . 7 and X=3 give samples of non-VIPs with slightly significantly higher average pN/pS than VIPs’ pN/pS ( 0 . 56 in non-VIPs versus 0 . 526 in VIPs , P=0 . 03 ) and a very similar variance ( 0 . 53 in non-VIPs vs 0 . 552 in VIPs , P=0 . 35 ) . The pN/pS ratio in human african populations is also slightly higher in non-VIPs compared to non-VIPs ( 0 . 81 in non-VIPs compared to 0 . 76 in VIPs , P=0 . 04 ) , which shows that our calibration is robust to the species used to measure pN/pS . There is no combination of a and X where both the average and variance of pN/pS are identical in VIPs and the sampled non-VIPs . The combination of a=0 . 7 and X=3 gives the closest matching variances and a slightly higher pN/pS ( lower purifying selection , see above for numbers ) in non-VIPs than in VIPs . Other combinations give closer averages of pN/pS , but more distant variances . To be conservative , we thus choose to use a=0 . 7 and X=3 . The fact that the sampled non-VIPs experience slightly less purifying selection than VIPs makes the comparison conservative ( the less purifying selection in non-VIPs compared to non-VIPs , the more opportunities there were for adaptation to happen at positions of coding sequences that can change ) . Finally , using a=0 . 97 and X=2 , we can compare VIPs and non-VIPs with similar dN/dS ratios ( VIPs’ and non-VIPs’ average dN/dS=0 . 124 , P=0 . 51 ) . As expected ( Figure 5—figure supplement 1 ) , using matching dN/dS instead of pN/pS strongly underestimates , but yet still reveals a substantial excess of adaptation in VIPs compared to non-VIPs ( 39% adaptation excess , P=0 versus 117% excess , P=0 after 109 iterations , when matching pN/pS; see Figure 5A ) . Throughout this analysis we distinguish between the effects due to viruses and the effects due to the functional roles that VIPs play in the host . This is done by comparing VIPs with matching control sets of non-VIPs with similar Gene Ontology ( GO ) processes . There are 162 GO processes with 50 or more VIPs ( Supplementary file 1C ) . The matching procedure is conducted using only these 162 processes . For each VIP , we find all the non-VIPs that have at least 60% of GO processes in common , and where the total number of processes does not exceed 140% of the number in the VIP to be matched with . We then randomly choose one non-VIPs among all those that fulfill these requirements . With the parameters used , we find each VIP always has more than 5 non-VIPs to choose from , and many more for most VIPs . Furthermore , these parameters give control sets of non-VIPs with representations of GO processes very similar to their representation in VIPs . On average the representation of each GO process is only 18% lower or higher in the matching controls , versus 60% lower or higher in non-matching , randomly sampled sets of non-VIPs . Note that perfect matching is impossible to achieve given that different proteins can have very different and specific combinations of associated GO processes . | When an environmental change occurs , species are able to adapt in response due to mutations in their DNA . Although these mutations occur randomly , by chance some of them make the organism better suited to their new environment . These are known as adaptive mutations . In the past ten years , evolutionary biologists have discovered a large number of adaptive mutations in a wide variety of locations in the genome – the complete set of DNA – of humans and other mammals . The fact that adaptive mutations are so pervasive is puzzling . What kind of environmental pressure could possibly drive so much adaptation in so many parts of the genome ? Viruses are ideal suspects since they are always present , ever-changing and interact with many different locations of the genome . However , only a few mammalian genes had been studied to see whether they adapt to the presence of viruses . By studying thousands of proteins whose genetic sequence is conserved in all mammalian species , Enard et al . now suggest that viruses explain a substantial part of the total adaptation observed in the genomes of humans and other mammals . For instance , as much as one third of the adaptive mutations that affect human proteins seem to have occurred in response to viruses . So far , Enard et al . have only studied old adaptations that occurred millions of years ago in humans and other mammals . Further studies will investigate how much of the recent adaptation in the human genome can also be explained by the arms race against viruses . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"Mmethods"
] | [
"evolutionary",
"biology",
"computational",
"and",
"systems",
"biology"
] | 2016 | Viruses are a dominant driver of protein adaptation in mammals |
Formation of the Drosophila adult abdomen involves a process of tissue replacement in which larval epidermal cells are replaced by adult cells . The progenitors of the adult epidermis are specified during embryogenesis and , unlike the imaginal discs that make up the thoracic and head segments , they remain quiescent during larval development . During pupal development , the abdominal histoblast cells proliferate and migrate to replace the larval epidermis . Here , we provide evidence that the microRNA , miR-965 , acts via string and wingless to control histoblast proliferation and migration . Ecdysone signaling downregulates miR-965 at the onset of pupariation , linking activation of the histoblast nests to the hormonal control of metamorphosis . Replacement of the larval epidermis by adult epidermal progenitors involves regulation of both cell-intrinsic events and cell communication . By regulating both cell proliferation and cell migration , miR-965 contributes to the robustness of this morphogenetic system .
Tissue morphogenesis is a complex process , through which the organism coordinates cell proliferation and cell death with cell migration and rearrangements to achieve final organ shape and size . Mechanisms controlling these processes play important role in morphogenesis , tissue repair and regeneration , and in cancer ( Friedl and Gilmour , 2009; Rorth , 2009 ) . The abdominal epithelium of Drosophila provides a useful model system in which to study the dynamics of tissue morphogenesis in vivo and to explore the genetic and cellular mechanisms that control these complex morphogenetic processes . During metamorphosis , larval epidermal tissues undergo cellular restructuring and rearrangement to give rise to adult abdominal epithelium . The adult abdominal epithelium is produced from progenitor cells , known as histoblast cells . Histoblasts are small diploid cells , easily distinguishable from the large polyploid larval epidermal cells ( LEC ) that surround them . Histoblasts are specified in the embryo and lie quiescent throughout larval development ( Guerra et al . , 1973; Simcox et al . , 1991 ) . There are four pairs of histoblast nests in each segment , which merge to assemble the adult abdominal epidermis ( Madhavan and Schneiderman , 1977 ) . The anterior and posterior dorsal histoblast nests give rise to the external dorsal cuticle of the abdominal segments ( tergites ) , while the ventral pair give rise to the ventral cuticle ( sternites ) . During larval stages , histoblasts are arrested in the G2 phase of the cell cycle . At pupariation , an ecdysone pulse triggers the expression of string ( cdc25 ) , which activates cyclin/CDK and pushes the histoblasts into rapid proliferation ( Edgar and O'Farrell , 1990; Gautier et al . , 1991 ) . The proliferative phase is divided into two stages with distinct features ( Madhavan and Madhavan , 1980; Ninov et al . , 2009 ) . The early ‘division phase’ is characterized by rapid and synchronous cell division without intervening G2 phases to allow for cell growth . Cells double in number and decrease in size with each division , with little change in the size of the nests . This is followed by a phase of slower division , from 15–40 hr of pupal development , in which proliferation is accompanied by longer intervening gap phases to allow cell growth . During this phase , epidermal growth factor receptor ( EGFR ) and Insulin receptor/PI3K- signaling coordinate the growth of cells with proliferation ( Ninov et al . , 2009 ) . During the growth phase , the histoblast nests begin to spread to replace the LEC . The expansion of the histoblast nests by cell migration is accompanied by programmed cell death , so that larval cells are replaced by the expanding histoblast population to maintain integrity of the epithelium ( Bischoff and Cseresnyes , 2009; Nakajima et al . , 2011 ) . Patterning of the adult abdominal segments makes use of many of the signaling pathways that pattern the embryonic and larval segments , including hedgehog , wingless , decapentaplegic and EGFR . Interactions between morphogen gradients produced by some of these proteins determine anterior-posterior and dorsal-ventral patterning of segments ( Sanicola et al . , 1995; Shirras and Couso , 1996; Struhl et al . , 1997; Kopp et al . , 1999; Ninov et al . , 2010 ) . Although abdominal segmental patterning has been extensively studied , the molecular mechanisms regulating cell division , migration , cell replacement and their interactions during formation of segments remain less well understood . Here , we provide evidence that the microRNA , miR-965 , is required in the histoblast nests , where it acts via regulation of string and wingless to control histoblast proliferation and migration during pupal morphogenesis .
Mutants lacking miR-965 had only minor effects on survival during development . Mutant adults appeared to be morphologically normal with the exception of defects in abdominal segmentation ( Figure 1E ) . Among the affected individuals , the predominant defect was a dorsal gap in one or more abdominal segments , in some cases leading to segment fusion ( Figure 1—figure supplement 2 ) . In addition , formation of ectopic bristles associated with a polarity defect was observed in ∼15% of affected individuals ( Figure 1—figure supplement 3 ) . Polarity reversal was always accompanied by a gap or segment fusion phenotype . These defects were rescued by restoring miRNA expression using the miR-965 RMCE rescue allele ( Figure 1E ) . The abdominal segmentation defects observed in miR-965 mutants suggested that the miRNA might be required in the histoblasts . To visualize miR-965 activity , we made use of a sensor transgene consisting of a ubiquitously expressed GFP reporter with a perfect target site for miR-965 in the 3′ UTR . Sensors of this design allow miRNA activity to be visualized by downregulation of GFP ( Brennecke et al . , 2003 ) . The control sensor , lacking the miRNA-target site , was expressed at comparable levels in the large polyploid LEC and in the smaller histoblast cells ( hb , Figure 2 ) . miR-965 sensor GFP levels were lower in the histoblast nests , compared to the adjacent larval cells , particularly in the histoblast cells near the edge of the nests ( Figure 2 ) . This difference was lost when the sensor was introduced into the miR-965 mutant background , providing evidence that the reduced GFP level in the histoblasts is due to miR-965-mediated repression ( Figure 2 ) . 10 . 7554/eLife . 07389 . 007Figure 2 . miR-965 expression in histoblasts . Top: design of the control and miR-965 sensor transgenes . EGFP was under control of the tubulin promoter . For the miR-965 sensor , 1 copy of a perfect miR-965 target sequence was placed into the SV40 UTR . Images showing GFP expression from the control sensor ( left ) and miR-965 sensor ( middle ) transgenes at 21 hr APF . Histoblast nests consist of small diploid histoblast cells ( hb ) surrounded by large polyploid larval epidermal cells ( LEC ) . Nuclei were labeled with histone-RFP ( red ) . Downregulation of GFP was lost when the transgene was placed in the KO1/KO2 miR-965 mutant background ( right ) . Anterior ( A ) , posterior ( P ) . Scale bar: 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 007 The escargot ( esg ) gene is expressed in histoblasts from the time of their specification in the embryo until the late pupal stage , and provides a marker to visualize histoblast development ( Hayashi et al . , 1993 ) . Figure 3A presents still images taken from time-lapse videos of esg-GAL4 , UAS-GFP pupae to visualize the first 3 mitotic divisions of the histoblasts . In the controls , cells became smaller and the number of cells doubled after each division ( Video 1 ) . This pattern of synchronous division was perturbed in the miR-965 mutant . Asynchronous division of miR-965 mutant histoblasts resulted in the presence of cells of different sizes ( Figure 3A , Video 2 ) . In the mutant , a subset of histoblast nuclei became pyknotic , fragmented and disappeared , indicating cell death ( Video 3 ) . Cell death was rare during this phase in the control samples . Synchronous division was restored and cell death was suppressed in the miR-965-rescue allele ( Figure 3A , Video 4 ) . 10 . 7554/eLife . 07389 . 008Figure 3 . Abnormal histoblast proliferation and migration in the miR-965 mutant . ( A ) Still images taken from time-lapse videos of control , miR-965 mutant ( KO1/KO2 ) and rescued mutant showing the reduction divisions of the early histoblast proliferation phase . M1 , M2 and M3 indicate images taken after mitosis 1 , 2 or 3 . Imaging was started 0–1 hr APF . Histoblasts were labeled by esg-Gal4 directed expression of UAS-nuclear GFP . ADHN and PDHN represent anterior dorsal histoblast nests and posterior dorsal histoblast nests . Scale bars: 50 µm . Note the different cell sizes in the miR-965 mutant histoblast nests . ( B ) Still images taken from time-lapse videos at 24 , 33 and 42 hr APF from control , miR-965 mutant and rescued mutant to illustrate expansion of the histoblast nests to replace LECs . Histoblasts were labeled by esg-Gal4 directed expression of cytoplasmic GFP . esg-GAL4 and UAS-GFP were recombined onto the miR-965 mutant and onto the miR-965 Rescue chromosome . Nuclei were labeled red with H2-RFP A and P indicate anterior and posterior orientation . Scale bars: 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 00810 . 7554/eLife . 07389 . 009Figure 3—figure supplement 1 . Rate of histoblast nest expansion measured from time-lapse videos . Each data point corresponds to one histoblast nest . The leading edge of each histoblast nest was tracked using imageJ . Speed was calculated measuring total distance covered ( micrometer/hour ) . Genotypes: Control was esg-GAL4 , UAS-GFP . esg-GAL4 and UAS-GFP were recombined onto the KO1 and KO2 chromosomes and onto the miR-965-Rescue chromosome . Data include examples with both recombinant mutant chromosomes . No difference between these two recombinants was apparent . n = 18 for the miR-965 ( KO1/KO2 ) mutant combination . n = 15 for control and rescue . Left panel: p < 0 . 0001 comparing KO1/KO2 with control , p < 0 . 01 comparing KO1/KO2 with rescue using one-way ANOVA . Right panel: p < 0 . 001 comparing KO1/KO2 with control , p < 0 . 01 comparing KO1/KO2 with rescue using one-way ANOVA . Refers to Figure 3B and Videos 5–7 . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 00910 . 7554/eLife . 07389 . 010Figure 3—figure supplement 2 . Large polyploid cells in miR-965 mutant histoblast nests . Histoblast nests were labeled with esg-Gal4-directed expression of UAS-GFP at 24 hr APF . Note the presence of large polyploid cells in the histoblast nest in the miR-965 mutant ( arrows ) . At the start of the imaging period , the large polyploid cells marked by the arrows did not express GFP , but began to express GFP after making contact with the expanding histoblast nests . Possible explanations for the appearance of GFP in large polyploidy cells include ( 1 ) induction of esg-Gal4 activity in the larval cells that cannot be eliminated by the expanding histoblast nests , perhaps by signals from the histoblasts; ( 2 ) fusion of polyploidy LEC with esg-Gal4-expressing histoblasts . Scale bar: 50 µm . Refers to Figure 3B . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 01010 . 7554/eLife . 07389 . 011Figure 3—figure supplement 3 . Pupal survival assays . Pupal survival was assayed for flies of the indicated genotypes . 6 batches of pupae were sampled/genotype . The data present the total number of surviving adults ( live ) and the total number of dead pupae ( dead ) . There was no significant difference between the mutant and control genotypes used to make the videos: p = 0 . 67 comparing KO2 esgG4>GFP/+ vs KO2 esgG4>GFP/KO1 ( Mann–Whitney test ) . p = 1 comparing KO1 esgG4>GFP/+ vs KO1 esgG4>GFP/KO2 ( Mann–Whitney test ) . Refers to Figure 3B . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 01110 . 7554/eLife . 07389 . 012Video 1 . Control , division phase . Early synchronous divisions of histoblast nests in Esg-GAL4>UAS-nuclear GFP controls . GFP was used to track anterior and posterior dorsal histoblast nests during the early division phase . Animals were collected for imaging at 0 hr APF , the white pre-pupal stage . ADHN: Anterior dorsal histoblast nest . PDHN: posterior dorsal histoblast nest . A and P indicate anterior and posterior orientation . Scale bar: 50 µM . Refers to Figure 3A . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 01210 . 7554/eLife . 07389 . 013Video 2 . miR-965 mutant division phase . Early asynchronous divisions in the miR-965 mutant ( KO2 , esg-GAL4>UAS-nuclear GFP/KO1 ) . Esg-GAL4>UAS-nuclear GFP was recombined onto miR-965 RMCE mutant ( KO2 ) chromosome . ADHN and PDHN indicate anterior and posterior dorsal histoblast nests . Animals were collected for imaging at 0 hr APF . Scale bar: 50 µM . Refers to Figure 3A . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 01310 . 7554/eLife . 07389 . 014Video 3 . miR-965 mutant apoptosis . Apoptotic cells are seen during the early histoblast division phase in the miR-965 mutant . Esg-GAL4>UAS-nuclear GFP was recombined onto miR-965 RMCE mutant ( KO2 ) chromosome . ADHN and PDHN indicate anterior and posterior dorsal histoblast nests . Animals were collected for imaging at 0 hr APF . Scale bar: 50 µM . Refers to Figure 3A . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 01410 . 7554/eLife . 07389 . 015Video 4 . miR-965-Rescue division phase . Early synchronous divisions of histoblasts in miR-965 RMCE rescue . Esg-GAL4>UAS-nuclear GFP was recombined onto miR-965 RMCE rescue chromosome . ADHN and PDHN indicate anterior and posterior dorsal histoblast nests . Animals were collected for imaging at 0 hr APF . Scale bar: 50 µM . Refers to Figure 3A . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 015 Subsequently , the rate of histoblast nest expansion was slower in the miR-965 mutant , compared to the controls ( Figure 3B ) . Expansion of the histoblast nests was quantified in segments 3 and 4 by monitoring the speed of migration ( Figure 3—figure supplement 1 , Videos 5 , 6 ) . The average speed of migration of third and fourth histoblast nests in the control samples was 15 µm/hr , compare with ∼6 . 5 µm/hr in the mutant . The rate of nest expansion was increased to 12–14 µm/hr by restoring miR-965 expression with the rescue allele the miR-965 mutant background ( Video 7 ) . 10 . 7554/eLife . 07389 . 016Video 5 . Control , growth phase . Migration of histoblast nests during the growth phase in an Esg-GAL4>UAS-cytoplasmic GFP control pupa . GFP ( green ) is used to monitor growth and migration of histoblast nests . H2-RFP ( red ) marks the nuclei . Big nuclei are LECs and small nuclei are histoblast cells . ADHN and PDHN indicate anterior and posterior dorsal histoblast nests . Scale bar: 100 µM . Refers to Figure 3—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 01610 . 7554/eLife . 07389 . 017Video 6 . miR-965 mutant growth phase . Delayed migration of histoblast nests during the growth phase in the miR-965 mutant . Esg-GAL4>UAS-cytoplasmic GFP was recombined onto both KO1 and KO2 mutant chromosomes . This video shows a KO2 , esg-GAL4>UAS-cytoplasmic GFP/KO1 pupa . H2-RFP ( red ) marks the nuclei . ADHN and PDHN indicate anterior and posterior dorsal histoblast nests . Scale bar: 100 µM . Refers to Figure 3—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 01710 . 7554/eLife . 07389 . 018Video 7 . miR-965 Rescue growth phase . Migration of histoblast nests during the growth phase in the miR-965 RMCE rescue genotype . Esg-GAL4>UAS-cytoplasmic GFP was recombined onto the miR-965 RMCE rescue chromosome . H2-RFP ( red ) marks the nuclei . ADHN and PDHN indicate anterior and posterior dorsal histoblast nests . Scale bar: 100 µM . Refers to Figure 3—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 018 LEC generally undergo programmed cell death as the histoblasts nests start to grow ( Nakajima et al . , 2011 ) . In miR-965 mutants , we observed that some of these cells were still present in the nests , suggesting a failure to eliminate LECs in the mutant ( Figure 3—figure supplement 2 ) . Delayed expansion of the histoblast nests , combined with the persistence of LECs , is likely to be responsible for the gaps observed in adult abdominal segments . It was noted that the migration defects observed in the videos seem to be more severe than the segmentation defects observed in the adult flies . However , we did not observe significant pre-eclosion lethality ( Figure 3—figure supplement 3 ) . Given that there was no evidence for loss of a class of more severely affected animals , we suggest that the apparent difference reflects a delay in tissue replacement in the mutant , so that most animals end up with milder defects by the end of pupariation than were apparent during the early pupal time-window in the videos . miR-965 is predicted to target 69 genes ( www . targetscan . org ) . polycomb , string , wingless , homothorax , Tor , Hsp83 and jumeau were selected for further analysis based on the quality of the predicted target site and on Flybase annotation suggesting roles in segmentation . Of these , only string and wingless ( wg ) mRNAs were upregulated in RNA isolated from miR-965 mutant pupae , and were also restored to near normal levels in the rescued mutant ( Figure 4A , B ) , as would be expected for functional miRNA targets . 10 . 7554/eLife . 07389 . 019Figure 4 . miR-965 regulates string and wingless . ( A , B ) string ( stg ) and wingless ( wg ) transcript levels measured by quantitative real time RT-PCR in RNA isolated from w1118 control , KO1/KO2 and 965-rescue pupae at 21 hr APF . Data represent the average of three independent RNA collections ± SD . ANOVA: p < 0 . 01 comparing KO1/KO2 with control or with rescue for stg and wg . ( C ) Top: diagram of the predicted miR-965 target site in the wg 3′ UTR , showing pairing to the miRNA seed sequence . Residues shown in red were mutated in the mutant version of the wg 3′ UTR luciferase reporter . Below: luciferase activity in S2 cells transfected to express a tubulin-promoter miR-965 transgene , Renilla luciferase and the indicated firefly luciferase reporters . Control indicates the luciferase reporter with the SV40 3′ UTR , which lacks miRNA binding sites . wg UTR indicates the intact full-length wg 3′ UTR . Mut indicates the wg 3′ UTR with the miRNA seed site mutated as indicated in red . Data represent the average of 3 independent experiments ± SD . ANOVA: p < 0 . 001 comparing control to the intact 3′ UTR . p = 0 . 001 comparing the intact and site mutant versions of the 3′ UTR . ( D ) Top: diagram of the predicted miR-965 target site in the stg 3′ UTR , showing pairing to the miRNA seed . Residues shown in red were mutated in the seed mutant version of the reporter . The changes made in the extended target site mutant reporter are shown in Figure 3 . Below: luciferase activity as in panel C . Data represent the average of 3 independent experiments ± SD . ANOVA: p < 0 . 0001 comparing control to the intact 3′ UTR and comparing intact to seed mutant and multiple mutant UTR reporters . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 01910 . 7554/eLife . 07389 . 020Figure 4—figure supplement 1 . ( A ) Predicted miR-965 sites in the string 3′UTR . Based on the potential for strong 3′ pairing in the Seed 1 mutant ( shown in Figure 4D ) , as well as the presence of a second nearby non-canonical seed match ( seed 2 ) , a more extensively mutated UTR was made to eliminate pairing to both potential sites . Nucleotides mutated are shown in red . Refers to Figure 4D . ( B ) Structure of the miR-965 site in the string 3′ UTR , as predicted by RNAHybrid ( http://bibiserv . techfak . uni---bielefeld . de/ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 020 The wg 3′UTR was used to make a luciferase reporter transgene and tested for regulation by miR-965 in S2 cells . miR-965 expression reduced wg 3′UTR reporter activity , and this regulation was lost in the wg UTR reporter mutated to disrupt pairing with the miRNA seed sequence ( Figure 4C , mutated residues shown in red ) . miR-965 expression also reduced expression of the string 3′UTR luciferase reporter ( Figure 4D ) . Mutation of the predicted target site to disrupt seed pairing partially offset regulation of the string 3′UTR luciferase reporter ( Figure 4D , mutated residues in red ) . We noted the presence of a second potential target site nearby ( Figure 4—figure supplement 1 ) . More extensive mutation to disrupt both sites further compromised regulation by miR-965 ( Figure 4D ) . Other non-canonical target sites might be responsible for the remaining regulation by miR-965 , however , we do not exclude the possibility that there could also be indirect effects of miR-965 on expression of this reporter . These findings provide evidence that miR-965 can regulate expression of string and wg . To ask whether the increases in string and wg expression might be responsible for the miR-965 mutant phenotype , we first asked if overexpressing them in an otherwise normal genetic background could phenocopy the mutant . esg-Gal4 driven expression of a UAS-string transgene produced abdominal segment gaps , segment fusion and polarity reversal phenotypes , similar to those observed in the miR-965 mutant ( Figure 5A Figure 5—figure supplement 1 ) . The proportion of flies with defects caused by string overexpression was similar to that in the miR-965 mutant ( Figure 5B ) . string overexpression caused asynchronous histoblast division and apoptosis during the early division phase ( Figure 5—figure supplement 2 , Video 8 ) . Expansion of the string-overexpressing histoblast nests was slowed , and the histoblasts were unable to fully replace the LECs ( Figure 5—figure supplements 2 , 3 , Video 9 ) , similar to what was observed in miR-965 mutants . Thus , string overexpression was sufficient to reproduce the miR-965 mutant phenotype . 10 . 7554/eLife . 07389 . 021Figure 5 . Overexpression of string and wg contributes to the miR-965 mutant phenotype . ( A ) Dorsal views of abdomens from adult female esg-Gal4 UAS-string flies illustrating the segment gap , segment fusion and polarity reversal phenotypes . ( B ) Penetrance of abdominal defects of all classes in esg-Gal4 UAS-string vs mutant . esg-Gal4 UAS-string: n = 97/469; KO1/KO2 n = 110/446 . p = 0 . 16 Fishers exact test . ( C ) Penetrance of abdominal defects in esg-Gal4 UAS-wgts flies reared at 18° and 25°C vs KO1/KO2 . esg-Gal4 UAS-wgts reared at 18°C: n = 9/129; esg-Gal4 UAS-wgts at 25°C n = 1/254; KO1/KO2 n = 110/446 . p = 0 . 014 comparing wgts at 18 vs 25°C , Fishers exact test . ( D ) Penetrance of abdominal defects comparing KO1/KO2 mutants with KO1/KO2 mutants carrying one copy of stringEY12388 or string4 alleles . p < 0 . 001 comparing KO1/KO2 to KO1/KO2; stgEY/+ or stg4/+ using Fisher's exact test . ( E ) Confocal micrographs showing dorsal histoblast nests of wild-type ( WT ) and miR-965 mutant ( KO ) at ∼24 hr APF labeled with anti-Wg ( red ) . Nuclei were labeled with DAPI ( blue ) . Scale bar: 20 µm . Anterior and dorsal histoblast nests in the miR-965 mutants were not yet fused at 24 hr APF , due to delayed migration . Images were captured using identical microscope settings . ( F ) Penetrance of abdominal segmentation defects comparing KO1/KO2 mutants with KO1/KO2 mutants carrying one copy of wgSP-1 or wgl-12 temperature sensitive alleles or carrying one copy of wgSP-1 and stg4 together . p < 0 . 05 comparing KO1/KO2 to KO1 , wgSP-1/KO2 using Fisher's exact test . KO1/KO2 was not significantly different from KO1 , wgI-12/KO2 , perhaps because wgI-12 is a weaker , temperature sensitive allele . p < 0 . 001 comparing KO1/KO2 with KO1 , wgSP-1/KO2; stg4/+ using Fisher's exact test . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 02110 . 7554/eLife . 07389 . 022Figure 5—figure supplement 1 . The proportion of flies with defects caused by string overexpression . Penetrance of the three types of abdominal defect in flies overexpressing UAS-String under esg-Gal4 control . n = 102 flies . Refers to Figure 5A . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 02210 . 7554/eLife . 07389 . 023Figure 5—figure supplement 2 . Still images from a time-lapse video of esg-Gal4>UAS-string histoblasts . Left: rapid proliferation phase . Note the presence of cells of different sizes , indicative of asynchronous division . Images represent 0 hr and mitosis M1 , M2 and M3 . Genotype: esg-Gal4 , UAS-string , UAS-GFP . Right: growth and migration phase . Note the delayed spreading and incomplete replacement of the LEC , compared to controls at the equivalent time points ( Figure 3C ) . Genotype: esg-Gal4 , UAS-string , UAS-GFP . Refers to Figure 5B and Videos 8 , 9 . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 02310 . 7554/eLife . 07389 . 024Figure 5—figure supplement 3 . Speed of histoblast nest migration . n = 15 for control , n = 18 for KO1/KO2 and n = 11 for esg-GAL4>UAS-stg . p < 0 . 001 comparing control and esg-GAL4>UAS-stg using one-way ANOVA . esg-GAL4>UAS-stg is not significantly different from the miR-965 mutant ( KO1/KO2 ) . Control and miR-965 mutant samples are same as in Figure 3—figure supplement 1 . Refers to Figure 5B and Video 9 . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 02410 . 7554/eLife . 07389 . 025Figure 5—figure supplement 4 . Rescue of the migration defect of miR-965 mutants with reduced levels of string . Left: still images taken from a time-lapse video showing histoblast divisions in the miR-965 mutant ( KO1/KO2 ) with reduced levels of string transcript using the stgEY12388 allele . M1 and M2 indicate mitosis 1 and 2 . Cell membranes were labeled using Atpα-GFP ( green ) . Nuclei were labeled using Histone2-RFP ( red ) . ADHN: anterior dorsal histoblast nest . PDHN: posterior dorsal histoblast nest . Scale bars: 50 µM . Right: still images taken from a time-lapse video showing histoblast nest migration in animals of the same genotypes . Scale bars: 100 µM . Refers to Figure 5D and Video 11 . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 02510 . 7554/eLife . 07389 . 026Figure 5—figure supplement 5 . Speed of histoblast migration restored by reduced string activity . Left: speed of histoblast nest migration in the third abdominal segment . p < 0 . 05 comparing KO1/KO2 with KO1/KO2; stgEY12388 . Control ( n = 15 ) , KO1/KO2 ( n = 18 ) and KO1/KO2; stgEY12388 ( n = 14 ) . The Control and KO1/KO2 samples are the same as those in Figure 3—figure supplement 1 . The two experiments were done together . Right: speed of histoblast migration in the fourth abdominal segment . p < 0 . 05 comparing KO1/KO2 with KO1/KO2; stgEY12388 . Refers to Figure 5D and Video 11 . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 02610 . 7554/eLife . 07389 . 027Video 8 . Esg-Gal4 UAS-Stg division phase . Early divisions in a pupa overexpressing String under esg-GAL4 control ( genotype: esg-GAL4 , UAS-nuclear GFP/UAS-stg ) . ADHN and PDHN indicate anterior and posterior dorsal histoblast nests . Animals were collected for imaging at 0 hr APF . Scale bar: 50 µM . Refers to Figure 5—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 02710 . 7554/eLife . 07389 . 028Video 9 . Esg-Gal4 UAS-Stg growth phase . Growth and migration phase in a pupa overexpressing String under esg-GAL4 control , as in Video 8 . ADHN and PDHN indicate anterior and posterior dorsal histoblast nests . Scale bar: 100 µM . Refers to Figure 5—figure supplements 2 , 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 028 To assess the contribution of wg overexpression , we made use of a UAS transgene directing expression of a temperature sensitive form of Wg protein ( Wilder and Perrimon , 1995 ) . Use of Wgts was required to allow stage specific activation of Wg in the histoblasts . Continuous expression of wild-type Wg under esg-Gal4 control was lethal . Wgts protein is inactive at 25°C . Flies carrying esg-Gal4 and UAS-wgts were reared at 25°C or shifted to 18°C in the third larval instar to allow expression of active Wg as the histoblasts began proliferation . This resulted in a phenocopy of the segment gap phenotype in 7% of animals ( Figure 5C ) . Phenocopy was rare in the animals raised continuously at 25°C to maintain low Wg activity ( 0 . 4% affected , Figure 5C; p = 0 . 014 comparing UAS-wgts at 18° vs 25° , Fisher's exact test ) . Next we asked whether limiting target gene overexpression could suppress the mutant phenotype . Lowering string activity by introducing string mutant alleles reduced the penetrance of the segment gap phenotype in the miR-965 ( KO1/KO2 ) mutant background from 25% to 6–7% ( Figure 5D , Video 10 , p < 0 . 001 comparing KO1/KO2 vs KO1/KO2; stgEY or stg4/+ , Fisher's exact test ) . Lowering string activity also suppressed the asynchronous division phenotype and the slow cell migration phenotype of the miR-965 mutant ( Figure 5—figure supplements 4 , 5 , Video 11 ) . 10 . 7554/eLife . 07389 . 029Video 10 . miR-965 mutant with reduced string levels division phase . The string mutant allele , stgEY12388 was used to reduce string levels in the miR-965 mutant ( KO1/KO2 ) background . Atpα-GFP ( green ) was used to mark cell membranes . H2-RFP ( red ) marks the nuclei . ADHN and PDHN indicate anterior and posterior dorsal histoblast nests . Scale bar: 50 µM . Refers to Figure 5D . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 02910 . 7554/eLife . 07389 . 030Video 11 . miR-965 mutant with reduced string levels with reduced string levels growth phase . Normal growth and migration of histoblast nests in the miR-965 mutant with reduced string levels . The string mutant allele , stgEY12388 was used to reduce string levels in the miR-965 mutant ( KO1/KO2 ) background . Atpα-GFP ( green ) was used to mark cell membranes . H2-RFP ( red ) marks the nuclei . ADHN and PDHN indicate anterior and posterior dorsal histoblast nests . Scale bar: 50 µM . Refers to Figure 5—figure supplements 4 , 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 030 Wingless protein ( Wg ) is expressed in the anterior dorsal histoblast nests , during their growth and migration ( Kopp et al . , 1999 ) . The level of Wg protein was higher in the miR-965 mutant ( Figure 5E ) . Wg is a secreted protein , and its distribution appears to be broader , reaching some of the LECs in the mutant , perhaps reflecting the higher level of protein produced ( Figure 5E ) . To ask whether this elevated Wg expression contributes to the defects in the mutant , we introduced wg mutant alleles into the miR-965 ( KO1/KO2 ) mutant background . The segment gap phenotype was reduced comparing KO1/KO2 vs KO1 wgSP-1/KO2 ( Figure 5F , p < 0 . 05 , Fisher's exact test; suppression by the wgI-12 allele was not statistically significant ) . Next , we tested the effect of removing one copy each of stg and wg . The segment gap phenotype was reduced comparing KO1/KO2 vs KO1 wgSP-1/KO2; stg4/+ ( Figure 5F , p < 0 . 001 , Fisher's exact test ) . Taking out one copy each of stg and wg lowered penetrance of the phenotype to ∼4% , compared with 6–7% for stg/+ alone or 12% for wg/+ alone . These experiments provide evidence that overexpression of string and wg each contribute to the miR-965 mutant histoblast phenotypes . The effect of limiting Wg expression in the mutant background may appear to be smaller than that of limiting String . However , to make a meaningful comparison of the relative contribution of these two targets , it would be necessary to restore each of them to normal levels . The genetic method used to reduce target activity in the mutant background does not allow precise control over the final target level , so this question cannot be addressed . We also do not exclude the possibility that there could be other functionally significant targets in addition to Wg and String . Ecdysone pulses at the beginning of pupariation have been shown to induce string expression in order to reactive histoblast proliferation ( Ninov et al . , 2009 ) . In light of the relationship between miR-965 and string , we asked whether miR-965 expression might be under Ecdysone control at this stage . We made use of a UAS-EcRRNAi transgene expressed under esg-Gal4 to reduce Ecdysone receptor levels in the histoblasts . RNAi-mediated depletion of EcR mRNA led to an increase in the level of the miR-965 primary transcript and mature miRNA ( Figure 6A , Figure 6—figure supplement 1 ) , and to reduced string mRNA levels in RNA samples isolated from early pupae ( Figure 6A ) . EcR binding sites have been identified near the host gene , kismet ( Gauhar et al . , 2009 ) , consistent with the possibility that ecdysone regulates expression of both kismet and miR-965 . When miR-965 was overexpressed in the histoblast cells using esg-GAL4 , string transcript levels were reduced ( Figure 6B ) and histoblast cell divisions were arrested ( Figure 6C , Video 12 ) . This phenotype resembles EcR-B mutants , in which histoblast division is compromised ( Bender et al . , 1997 ) . Pupae overexpressing miR-965 in histoblast cells did not survive beyond ∼12 APF , so it was not possible to monitor the later stages of histoblast migration in this genotype . 10 . 7554/eLife . 07389 . 031Figure 6 . Regulation of miR-965 by ecdysone at the beginning of pupariation . ( A ) Quantitative RT-PCR showing levels of miR-965 primary transcript , EcR , and string mRNAs in RNA isolated from pupae expressing esg-GAL4 ( control ) and esg-GAL4 driving UAS-EcR-RNAi to deplete EcR mRNA . Samples were collected at 0 hr APF . Data were normalized to rp49 and to the esg-GAL4 control . Data represents average of three independent samples ± SD . ( B ) Quantitative RT-PCR showing string mRNA in 0 hr pupae overexpressing miR-965 in histoblast cells . For quantitative microRNA PCR , data were normalized to U14 , U27 , SnoR422 . Data were normalized to rp49 for string mRNA qPCR . Data represent the average of four independent samples ± SD . ( C ) Images from time-lapse videos showing the effects of miR-965 overexpression in histoblast cells during the synchronous division phase . M1 , M2 and M3 indicate three consecutive mitotic divisions in dorsal histoblast nests . Scale bar: 50 µm . ( D ) Quantitative RT-PCR showing levels of string , EcR primary transcript ( EcR-PT ) and mature mRNA in RNA isolated from pupae expressing esg-GAL4 ( control ) , esg-GAL4 in the miR-965 mutant with and without UAS-EcR-RNAi to deplete EcR mRNA . esg-GAL4 was recombined onto the KO2 mutant chromosome . Samples were collected at 0 hr APF . Data were normalized to rp49 and to the esg-GAL4 control . Data represents average of six independent samples ± SD . p = 0 . 37 for stg levels between KO2 , esg-GAL4/KO1 and KO2 , esg-GAL4/KO1>EcR-RNAi . p ≤ 0 . 01 comparing primary and mature EcR transcripts between esg-GAL4 control and KO2 , esg-GAL4/KO1 mutant samples . ( E ) Diagram of the regulatory relationships between EcR , miR-965 and the miR-965 targets string and wg . The symbols represent repression of gene expression . miR-965 and EcR repress each other at the primary transcript level . The effect of miR-965 on EcR primary transcript is most likely indirect . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 03110 . 7554/eLife . 07389 . 032Figure 6—figure supplement 1 . Mature miR-965 miRNA regulation by EcR . miRNA quantitative RT-PCR showing the levels of mature miR-965 in RNA extracted from 0 hr pupae . esg-GAL4 was used to direct UAS-EcRRNAi expression in histoblasts . Data were normalized to U27 and snoR422 and to the esg-GAL4/+ control sample . Data represent the average of three independent biological samples ± SD . Refers to Figure 6A . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 03210 . 7554/eLife . 07389 . 033Figure 6—figure supplement 2 . EcR 3′ UTR reporter expression in the miR-965 mutant . A reporter transgene containing the n EcR 3′ UTR linked to GFP was introduced into the miR-965 KO1/KO2 mutant background . GFP expression ( green ) did not increase in the histoblast nests in the miRNA mutant compared to the control . Thus , there was no indication that the miRNA acts directly on EcR transcript . There were no good quality miR-965 sites predicted in the various EcR 3′ UTR isoforms . Nuclei were labeled with H2-RFP . Scale bar—100 µM . Refers to Figure 6D . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 03310 . 7554/eLife . 07389 . 034Video 12 . miR-965 overexpression . Early division arrest in a pupa overexpressing miR-965 under esg-GAL4 control ( genotype: esg-GAL4 , UAS-nuclear GFP/UAS-miR-965 ) . Animals were collected for imaging at 0 hr APF . Scale bar: 50 µM . Refers to Figure 6C . DOI: http://dx . doi . org/10 . 7554/eLife . 07389 . 034 In light of the finding that EcR limits miR-965 expression , we asked whether Ecdysone signaling might act via miR-965 to regulate string . We introduced a UAS-EcRRNAi into the miR-965 mutant background and measured the levels of string mRNAs . Depletion of EcR did not reduce string mRNA in animals lacking miR-965 ( Figure 6D ) . This provides evidence that Ecdysone signaling is mediated through regulation of the miR-965 miRNA to regulate histoblast proliferation . In the course of this analysis , we observed that EcR transcript levels increased in the miR-965 mutant ( Figure 6D ) . Both mature and primary transcript levels increased , suggesting an indirect effect of the miRNA on EcR transcription . To ask if there might also be a post-transcriptional component to the regulation of EcR by miR-965 , we used an EcR 3′ UTR reporter transgene linked to GFP ( Varghese and Cohen , 2007 ) . GFP expression in the histoblast nests did not increase in the miRNA mutant background , indicating indirect regulation of EcR by miR-965 ( Figure 6—figure supplement 2 ) . These experiments provide evidence for a regulatory feedback relationship between miR-965 and the Ecdysone receptor ( Figure 6E ) . EcR activity limits miR-965 expression . miR-965 activity limits EcR primary transcript levels , suggesting an effect on transcription .
Our findings link regulation of the miR-965 microRNA to the onset of histoblast proliferation at the larval to pupal transition . Previous reports have provided evidence that Ecdysone signaling activates string expression to trigger the onset of histoblast proliferation at the beginning of pupal development ( Ninov et al . , 2009 ) . Our findings provide evidence that Ecdysone signaling works though regulation of miR-965 , which in turn regulates string . Interestingly , we also find evidence for negative feedback regulation of miR-965 on EcR . Mutual repression circuitry of this type can contribute a switch-like function: EcR activity lowers miR-965 activity , which allows greater EcR expression/activity by alleviating miR-965 mediated repression . In a circuit of this design , there will be a delay between reduced transcription of the miRNA primary transcript and the decay of the mature miRNA product . Hence sustained EcR activity is needed to throw the switch . We have previously reported that EcR shows positive transcriptional autoregulation and that this is buffered by miR-14 in a mutual repression circuit ( Varghese and Cohen , 2007 ) . Positive feedback allows for a sharp switch-like response , but also makes the system very sensitive to stochastic fluctuation in EcR activity . Coupling EcR positive auto-feedback to miRNA-mediated repression allows a robust switch function upon Ecdysone stimulation , while protecting the system from the effects of biological noise . This study provides evidence that miR-965 plays an analogous role in regulating EcR response and suggests that miR-965 confers robustness to the EcR response in the histoblasts . Upregulation of string in the miR-965 mutant contributes to the defects in histoblast proliferation . How misregulation of string might contribute to the migration defects is less immediately obvious . Previous work has shown that cell cycle progression in the histoblast population is required to trigger programmed cell death in the surrounding LEC ( Nakajima et al . , 2011 ) . Those authors provided evidence that cell growth and the expansion of the histoblast nests may be required to elicit LEC apoptosis . Although the mechanism by which expansion of the histoblasts triggers LEC death is not clear , elevated string expression in the miR-965 mutant is likely to be responsible for the cell cycle progression defects during this phase , hindering normal LEC removal and histoblast migration . Persistence of the LECs might also be a consequence of the increased expression of Wg protein in the mutant histoblast nests . Wg acts in combination with EGFR and Dpp signals to control abdominal segment patterning ( Shirras and Couso , 1996; Kopp et al . , 1999; Ninov et al . , 2009 , 2010 ) . These signals are thought to control differential cell adhesion , which may be important for elimination of the LECs as well as for proper segmental fusion of the histoblast nests . Elevated expression of Wg protein may lead to an expanded range of action , perhaps resulting in ectopic Wg activity in the LECs . Each adult abdominal segment has a well-defined anterior-posterior polarity . Wg is required from 15–20 hr APF for bristle formation and from 18–28 hr APF for tergite differentiation and pigmentation . Overexpression of wg has been shown to cause ectopic bristle formation , and shaggy mutant clones , which constitutively activate wg signaling , can cause polarity reversal in abdominal bristles , while EGFR , FGF , dpp and Notch signaling have no effect on the polarity of bristles in adult epidermis ( Lawrence et al . , 2002 ) . Wg levels are normally higher in the posterior region of the anterior histoblast nests and lower more anteriorly . Our finding that Wg levels were elevated and that the distribution of Wg was broader than normal suggests ectopic Wg activity throughout the histoblast nest , including cells that normally experience low Wg levels . Ectopic spread of Wg could be responsible for the formation of ectopic bristles and for the occasional instances of polarity reversal observed in the anterior part of tergites in the miR-965 mutants . Replacement of the larval epidermis during metamorphosis involves regulation of both cell-intrinsic events in the abdominal histoblasts and communication between histoblasts and the larval cells they will replace . miR-965 acts on at least two separate processes required during histoblast morphogenesis . A miRNA with multiple targets can add a layer of regulation , acting across different pathways to integrate their activities ( Herranz and Cohen , 2010 ) . In doing so , the miR-965 miRNA appears to contribute to the robustness of this complex morphogenetic system .
w1118 was used as the control genotype unless otherwise indicated . esg-GAL4 , UAS-GFP was obtained from Shigeo Hayashi . UAS-stg . N4 , stg4 , stgEY12388 , wgSp-1 , wgI-12 , Kis1 , Kisk13416 , Kisk11324 , Kisk10237 , KisBG01657 , KisKG08532 , KisEY12846 , Df ( 2L ) Exe7702 , Df ( 2L ) ED19 are from Bloomington stock center . EcR-RNAi ( v37059 ) was from Vienna Drosophila Research Center and P{GawB}esgNP7011 and ZCL2207 ( Atpα-GFP ) was from DGRC , Kyoto . miR-965 ( KO1 ) and miR-965RMCE ( KO2 ) deletion mutants were generated by targeted homologous recombination as described ( Chen et al . , 2011 , 2014 ) . Left and right homology arms were amplified by PCR from genomic DNA . The 4055 bp left homology arm was amplified with primers: LF- 5′ TTAGAGCTATTGCAACGAAAAGTG 3′ , LR- 5′ GTGTAACGGGGATAATAGGATCTG 3′; the 4435 bp right arm was amplified with: RF- 5′ AACACACACAGATGCAGATACAGA 3′ and RR- 5′ AAATAAACGGTTCACTTCTTCTGC 3′ . Following recombination , 153 bp spanning the miRNA-965 hairpin was deleted and replaced with a mini-white cassette . miR-965 mutants were crossed to heat shock-CRE flies and given heat shock treatment to excise the mini-white cassette . Deletion of miR-965 in both mutants was confirmed by genomic DNA PCR and by microRNA quantitative-PCR . All genetic tests were done using flies carrying two independently generated alleles or an allele in trans to a chromosomal deletion , Df ( 2L ) ED19 , which removes the miR-965 locus . For generation of Rescue allele , 158 bp of genomic region containing miR-965 hairpin was amplified with primers F- 5′ GCGGGCATGTCGAGGTCGACAAGTAAAATAGCGGAATCAAAATAAT 3′ , R- 5′ GCTCTAGAACTAGTGGATCCAACACTTTTCGTTGCAATAGCTC 3′ and replaced mini-white gene in miR-965RMCE ( KO2 ) mutant allele . miR-965 sensor: for microRNA-GFP sensors , mature miR-965 sequence with primers F- 5′ CTAGAAAGGGGAAAAGCTATACGCTTAC 3′ and R- 5′ TCGAGTAAGCGTATAGCTTTTCCCCTTT 3′ was cloned into 3′UTR of EGFP driven by tubulin promoter in pCasper4 . For UAS constructs , the miR-965 hairpin was amplified using F- 5′ TATAGCGGCCGCAAGTAAAATAGCGGAATCAAAATAAT 3′ and R- 5′ TATATCTAGAAACACTTTTCGTTGCAATAGCTC 3′ and cloned into pUAST-DsRed . miR-965 expression plasmids were generated by cloning the miR-965 hairpin into pCasper-tub-SV40 with primers F- 5′ TCTAGACTTTCATTTTAAGTAAAATAGCGG 3′ and R- 5′ CCTCGAGAACACTTTTCGTTGCAATAGCTCT 3′ . For luciferase reporter constructs , 3′UTR of target genes were cloned into pCasper4-tub-Fluc-SV40 firefly luciferase vector . The following primers were used for cloning 3′UTRs into luciferase vector . Stg 3′UTR F- 5′ GATCGCCGTGTAATTCTAGAGATGATCGTGCAGTTCGTTATC 3′ . Stg 3′UTR R- 5′ GGCTGCAGGTCGACCTCGAGTTCTTTTTCGTCGTGTATTAATGT 3′ . Wg 3′UTR F- 5′ GATCGCCGTGTAATTCTAGACCGCCCTCTTCGTTCTTTGT 3′ . Wg 3′UTR R- 5′ GGCTGCAGGTCGACCTCGAGACTCATTGTCGTTTTGTGTTTTT 3′ . Mutation in the miR-965 seed region in stg 3′UTR was done using following primers: Stg mut UTR up F- 5′ GGGCGGAAAGATCGCCGTGTAATTCTAGAGATGATCGTGCAGTTCG 3′ . Stg mut UTR up R- 5′ CAAATAATGATCATAAATTGTACCTAGCAGAAGTT 3′ . Stg mut UTR down F- 5′ TTATGATCATTATTTGTTTATTTTTATGTAATCCG 3′ . Stg mut UTR down F- 5′ ATAAACAAATAAAATTGTACCTAGCAGAAGTT 3′ . Stg extensive mut 1F- 5′ GATCGCCGTGTAATTCTAGAGATGATCGTGCAGTTCGTTATC 3′ . Stg extensive mut 1R- 5′ CATCACTTAGGCGTAATGTCGGATAAATAAAGTTTTATGG 3′ . Stg extensive mut 2F- 5′ACGCCTAAGTGATGCCAGATGTACCCTACTGCTAGGTACAATTTA 3′ . Stg extensive mut 2R- 5′ GGCTGCAGGTCGACCTCGAGTTCTTTTTCGTCGTGTATTAATGT 3′ . Mutation in miR-965 seed region in wg 3′UTR was done using primers: Wg mut up F- 5′ GAACTGCCTGCGTGAGATTCTCGCATGCCAGAGATCCTA 3′ . Wg mut up R- 5′ CTAATAACAAAGGCTGAGTGGAGACAAAATACATAACACA 3′ . Wg mut down F- 5′ TGTGTTATGTATTTTGTCTCCACTCAGCCTTTGTTATTAG 3′ . Wg mut down R- 5′GGCTGCAGGTCGACCTCGAGACTCATTGTCGTTTTGTGTTTTT 3′ . Drosophila Schneider cells ( S2 ) were grown at 25°C in the absence of CO2 , with serum free medium ( SFM , Gibco ) supplemented with L-Glutamine . 2 × 106 S2 cells were transfected in 24-well plates with 25 ng of the firefly luciferase reporter and Renilla luciferase control plasmids , and 250 ng of the miRNA-965 expression plasmid or empty vector . Transfection was done using Cellfectin II ( Invitrogen ) . Transfections were performed in triplicate and each experiment was performed in at least three independent replicates . Cells were lysed 2 . 5 days after transfection in 100 μl passive lysis buffer , shaken at room temperature for 20 min and dual-luciferase assays were performed according to the manufacturers protocol ( Promega ) . Total RNA was extracted from 0 hr or 21 hr pupae and used for cDNA synthesis . Mature miR-965 transcript level was measured by using TaqMan miRNA assays and normalized to U14 , U27 or snoR422 control primers . For target mRNA qRT-PCR , total RNA was treated with RNAse-free DNAse . First strand cDNA was synthesized using oligo-dT primers and SuperScript RT-III ( Invitrogen ) . qRT-PCR was performed using SYBR green ( Applied Biosystems ) . Measurements were normalized to Ribosomal Protein 49 . Primers: rp49 F- GCTAAGCTGTCGCACAAA and rp49 R- TCCGGTGGGCAGCATGTG . kis F- TTCACGGAAATCATCAAGGA and kis R- CTGTTGCTGTAGCGGATGTG stg F- ATTCTCCCATTTTCCCAGTTTT and stg R- CTTCCCATCCTATCCTTTCCTT . wg F- GTCAGGGACGCAAGCATAAT and wg R- GCGAAGGCTCCAGATAGACA . EcR F- TAACGGCCAACTGATTGTACG and EcR R- GCGGCCAAGACTTTGTTAAGA . Pri-965 F- AAATCACAAAGCAGAAGAAGTGAA and Pri-965 R- ACAGAAGGGCACATATAACGTACA . For video of the division phase ( 0–8 hr ) , 0 hr white pupa were washed with PBS , dried and stuck to imaging dishes ( MatTek ) with a drop of mineral oil . For growth phase videos , white pupae and allowed to age until 15–20 hr at 25°C . The outer cuticle was removed without damaging internal tissues . Pupae were mounted into imaging dishes with a drop of mineral oil . Zeiss LSM700 and Leica SP5 microscopes were used for imaging and videos were processed by imageJ and photoshop . videos were taken at 5 frames/s for division phase and 10 frames/s for growth phase . For measurement of speed of migration of histoblast nests , the distance moved by the leading edges of the histoblast nests from segment 3 and 4 was measured and migration speed was calculated in micrometer/hour . Adult flies were soaked in Ethanol:Glycerol ( 3:1 ) , then boiled with 10% KOH at 90°C for 2–5 min . Cuticles were rinsed once with PBT ( PBS with 0 . 1% TritonX-100 ) , then 3 times with PBS , 15 min each and mounted with glycerol . For immunostaining , 0 hr white pupae were transferred to fresh vials and raised at 25°C for staging . Pupae ∼24 hr APF were attached to double sided tape and bisected with a blade . Gut and fat body were removed without disturbing the inner epithelial layer . Cuticle with attached epithelium was washed with PBS and fixed with 4% paraformaldehyde for one and half hour at 4°C . Samples were rinsed 4 × 15 min with PBT ( PBS with 0 . 1% Triton X-100 ) and blocked for 2 hr with 1% BSA in PBT . Samples were incubated with mouse anti-Wg antibody ( 1:20 , DSHB ) overnight at 4°C . Samples were washed with PBT ( 4 × 15 min ) and incubated with AlexaFluor secondary antibody ( 1:500 , Invitrogen ) for 2 hr at room temperature . Samples were washed again 4 × 15 min with PBT and stained with DAPI ( 1:1000 ) for 10 min at room temperature before mounting . | Tissues in living organisms are shaped via complex processes that are collectively called ‘morphogenesis’ . Many researchers have used the fruit fly Drosophila as a model to understand morphogenesis , which occurs both during the development of a Drosophila embryo and during metamorphosis ( when the pupa changes to become an adult fly ) . Like other insects , adult fruit flies have three main body sections ( a head , a thorax and an abdomen ) , which are further divided into segments . The adult's abdomen forms inside the pupa from precursor cells called histoblasts . These cells are unusual in that they develop in the embryo but remain inactive during the larval stages of life . During pupation , these cells are reactivated by a hormone called ecdysone , and gradually replace the larva's tissues . However , it was not clear how this process was coordinated . Verma and Cohen have now demonstrated that a small RNA molecule—a microRNA called miR-965—acts in histoblast cells and controls how much these cells divide as well as how they migrate during morphogenesis to replace the larval cells . MicroRNAs regulate other RNAs , called messenger RNAs , typically by targeting them for destruction . This prevents the messenger RNA molecules from being used to make proteins . When flies develop without miR-965 , they are mostly normal but have defects in the formation of segments in the abdomen . Verma and Cohen revealed that miR-965 acts by targeting two important messenger RNAs for destruction . These messenger RNAs encode a protein called String , which regulates histoblast proliferation , and another protein called Wingless . Once the pupa starts to form , the ecdysone hormone reduces the production of miR-965 to increase histoblast proliferation and migration . The miR-965 microRNA in turn reduces the level of ecdysone receptor . The ecdysone hormone acts as an all-or-nothing switch to make an irreversible change from the larval to the pupal stage . The hormone boosts its own activity by increasing expression of its own receptor . This ‘positive feedback loop’ acts like a switch and is very sensitive to small changes in the amount of hormone present . Verma and Cohen propose that by reducing the levels of the hormone receptor , miR-965 makes the system more stable . This is because the hormone must first overcome the action of miR-965 before it can kick off the positive feedback loop . This takes time , and means that any change in the amount of hormone must be around for a while to have an effect . This mechanism buffers against short-lived , small changes in hormone levels that might throw the switch at the wrong time—a feature known as ‘robustness’ . This seems to be a complicated process to go from one state to another ( i . e . , from a larva to a pupa ) . But , the existence of the many distinct checks and balances makes sure the switch is thrown only when it is needed . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology"
] | 2015 | miR-965 controls cell proliferation and migration during tissue morphogenesis in the Drosophila abdomen |
Without transposon-silencing Piwi-interacting RNAs ( piRNAs ) , transposition causes an ovarian atrophy syndrome in Drosophila called gonadal dysgenesis ( GD ) . Harwich ( Har ) strains with P-elements cause severe GD in F1 daughters when Har fathers mate with mothers lacking P-element-piRNAs ( i . e . ISO1 strain ) . To address the mystery of why Har induces severe GD , we bred hybrid Drosophila with Har genomic fragments into the ISO1 background to create HISR-D or HISR-N lines that still cause Dysgenesis or are Non-dysgenic , respectively . In these lines , we discovered a highly truncated P-element variant we named ‘Har-P’ as the most frequent de novo insertion . Although HISR-D lines still contain full-length P-elements , HISR-N lines lost functional P-transposase but retained Har-P’s that when crossed back to P-transposase restores GD induction . Finally , we uncovered P-element-piRNA-directed repression on Har-P’s transmitted paternally to suppress somatic transposition . The Drosophila short Har-P’s and full-length P-elements relationship parallels the MITEs/DNA-transposase in plants and SINEs/LINEs in mammals .
The sterility syndrome of ‘P-M’ hybrid dysgenesis in Drosophila melanogaster ( Engels and Preston , 1979; Kidwell et al . , 1977 ) is due to uncontrolled P-element transposition that damages ovarian development and induces female sterility ( Bingham et al . , 1982 and reviewed in Kelleher , 2016 ) . This gonadal dysgenesis ( GD ) phenotype occurs in hybrid F1 daughters whose paternal genome comes from a father possessing active P-elements ( a ‘P’ strain ) and a maternal genome unable to express P-element piRNAs ( an ‘M’ strain ) ( Brennecke et al . , 2008; Khurana et al . , 2011 ) . The fascinating nature of this genetic syndrome is complete fertility in daughters from the reciprocal cross because the mother possessing active P-elements contribute P-element-derived Piwi-interacting RNAs ( piRNAs ) to silence these transposons in daughters ( Brennecke et al . , 2008; Khurana et al . , 2011 ) . Thus , despite identical genetic makeup in daughters between the reciprocal crosses , the epigenetic maternal transmission of transposon repression by piRNAs starkly defines female fertility ( illustrated in Figure 1A ) . Between fertility and complete sterility lies a spectrum of GD induction variation amongst different strain crosses that may be attributed to differential P-element copy numbers in different strain genomes ( Anxolabéhère et al . , 1985; Bergman et al . , 2017; Biémont et al . , 1990; Bingham et al . , 1982; Boussy et al . , 1988; Kidwell et al . , 1981; Ronsseray et al . , 1989; Srivastav and Kelleher , 2017; Yoshitake et al . , 2018 ) , and capacity to generate piRNAs ( Wakisaka et al . , 2017 ) . In addition , there are many non-autonomous P-element variants that can be mobilized by P-transposases , including very short elements from the pi[2] strain ( Bingham et al . , 1982; O'Hare and Rubin , 1983 ) that actually assemble in vitro with the P-transposase tetramer complex >100X more efficiently than the full-length P-element ( Tang et al . , 2007 ) . However , many earlier studies perceived truncated variants such as the ‘KP2’ variants as inhibitors of transposition by acting to titrate P-transposase since P-element piRNAs were unknown at the time ( Black et al . , 1987; Gloor et al . , 1993; Jackson et al . , 1988; Robertson and Engels , 1989; Simmons et al . , 2002a ) . Most studies of GD were typically calibrated with a strong paternal inducer ‘P’-strain like Harwich ( Har ) or pi[2] when mated with ‘M’ strain females lacking P-elements ( Bingham et al . , 1982; Brennecke et al . , 2008; Kidwell et al . , 1977; Rubin et al . , 1982 ) . Despite over 40 years of study , what defines a strong paternal inducer of GD has remained a mystery . Although P-element copy numbers in Har are significant ( 120–140 copies; Khurana et al . , 2011 ) , strains with even more copies like OreR-MOD do not induce GD whereas other strong inducer strains that have >75% fewer P-element copies than Har can also trigger complete GD ( Figure 1B and C ) . Thus , there is a lack of correlation between P-element copy number and GD induction ( Figure 1D ) that we and others have previously observed ( Bergman et al . , 2017; Ronsseray et al . , 1989; Srivastav and Kelleher , 2017 ) . Since P-element copy numbers do not explain GD severity , we hypothesized that a special P-element variant or insertion locus might underlie the strong GD phenotype in certain strong ‘P’ strains like Har . To discover this P-element variant , we undertook a reductionist approach to find specific P-element variant ( s ) required for GD induction that revealed unexpectedly a short variant from the Har strain that may act together with the full-length P-transposase to drive strong GD .
To genetically isolate the causative transposon strongly inducing GD and facilitate discovery by whole genome sequencing ( WGS ) , we generated hybrid lines where only a minor fraction of the Har genome is within the background of the ISO1 reference genome sequence . We first conducted several fertility-permissive backcrosses between female Har and male ISO1 , selecting hybrid progeny that propagated a red-eye phenotype which we attributed to the ‘red’ eyes due to Har alleles replacing the cn , bw , sp , alleles on Chromosome 2R ( Chr2R ) of ISO1 ( Figure 2A-abridged scheme , Figure 2—figure supplement 1-detailed scheme ) . We then performed an initial GD validation screen with many vials of individual hybrid males crossed to ISO1 females and selecting for lines that caused 100% GD from this cross . Lines were propagated with additional self-crosses and further in-bred with single-sibling pairs . We then subjected multiple independent Har-ISO1-Selfed-Red ( HISR ) lines to a second GD assay . Finally , we conducted qPCR to identify the lines with the greatest reduction of P-element copy numbers ( Figure 2B ) and settled on four lines each that either retained severe paternally-induced GD ( HISR-D ) or had lost this capacity ( HISR-N ) ( Figure 2C ) . Genomic PCR genotyping of deletion loci of Har compared to ISO1 in HISR lines indicated that these lines carried mostly ISO1 genomes ( Figure 2—figure supplement 2 ) . Therefore , we performed WGS of the parental Har and ISO1 strains , the 8 HISR lines , and the pi[2] and Birmingham ( Birm ) strains , two classic strains with similar numbers of P-elements but diametric capacity to induce GD ( Engels et al . , 1987; Simmons et al . , 1987 ) . Single-nucleotide polymorphism ( SNP ) profiles of HISR lines confirmed that only a small percentage of the Har genome was retained in mostly an ISO1 background ( Figure 2D ) . Quantification of P-element copies from WGS with the TIDAL program ( Transposon Insertion and Depletion AnaLyzer , Figure 2E ) ( Rahman et al . , 2015 ) was also consistent with qPCR measurements ( Figure 2F ) . To determine how substantial reduction in P-element copy numbers in HISR lines affected P-element-directed piRNA production , we generated and sequenced highly-consistent ovarian small RNA libraries ( Figure 3A ) and confirmed the expected presence and absence of P-element piRNAs in Har and ISO1 ovaries , respectively ( Figure 3B ) . Surprisingly , there were similar-to-increased levels of P-element piRNAs between HISR-D and Har strains , whereas amongst the HISR-N lines , only HISR-N10 retained P-element piRNAs ( Figure 3C ) . Our own mapping analysis indicated a common 3' end antisense bias of P-element piRNAs that we also confirmed with an independent piRNA analysis pipeline ( Han et al . , 2015 ) . These mapping patterns are consistent with piRNAs silencing transposons and suppressing hybrid GD ( Brennecke et al . , 2008; Erwin et al . , 2015; Khurana et al . , 2011 ) as well as correlating with all the HISR-D’s and HISR-N10’s immunity to strong GD induction when these females are mated to Har males ( Figure 3D ) . Additional piRNAs broadly cover the full length of P-element in Har and HISR-D lines ( Figure 3B-top and right ) , but the notable depletion of internal P-element piRNAs in HISR-N10 ( Figure 3B-middle left ) prompted us to conjecture which of its 23 TIDAL-mapped P-elements might be stimulating this novel piRNA pattern . We only found one euchromatic P-element insertion in HISR-N10 that specifically coincided with an increase of local piRNAs ( Figure 3—figure supplement 1A ) . This P-element inserted into the 5' UTR of DIP1 , adjacent to the enhancer and promoter region of Flamenco , the major piRNA cluster located in a pericentromeric region of the X-chromosome ( Brennecke et al . , 2007 ) . However , when we selected just the HISR-N10 X-chromosome balanced with the FM7a balancer chromosome ( Figure 3—figure supplement 1B ) , this X-chromosome locus did not generate enough P-element piRNAs to provide full GD immunity . It is possible for additional P-elements to have inserted into major piRNA cluster loci like 42AB , Flamenco and TAS-regions as part of the endogenizing process ( Khurana et al . , 2011; Moon et al . , 2018 ) , but the intractable repetitiveness of piRNA cluster regions prevents bioinformatic programs from pinpointing P-element insertions in these regions . Interestingly , all the P-element piRNAs detected in HISR-N10 , -D29 , and parental Har appeared to be expressed in the germline due to the detection of clear ping-pong piRNA biogenesis signatures ( Figure 3—figure supplement 2 ) . Finally , the P-element piRNA patterns in HISR-N10 can be explained by the abundant P-element variant that will be discussed below . The selection for ‘red’ eyes of Har alleles in HISR lines should have replaced the cn , bw , sp , alleles on Chromosome 2R ( Chr2R ) of ISO1 , therefore we had hoped that WGS of HISR line genomes might to point to a specific set of P-elements responsible for inducing strong GD . Unexpectedly , the P-element insertions were not confined to Chr2R , but rather were dispersed across the entire genomes of all HISR lines ( Figure 4A ) , seemingly defying the genomic PCR genotyping and WGS-SNP profiling that indicated sufficient backcrossing to favor mostly the ISO1 genetic background ( Figure 4—figure supplement 1 ) . To explore this conundrum , we examined how many of the original P-elements in the Har genome were conserved in the HISR lines’ genomes ( Figure 4B ) . As expected for HISR-D29 whose P-element copy numbers was closest to Har , this line conserved the highest share of parental Har P-elements compared to other HISR lines . However , there were also 35 novel P-element insertions ( ~45% ) in HISR-D29 absent from Har . Surprisingly , the vast majority of the P-element insertions across all HISR lines were also de novo P-element insertions ( Figure 4C ) , with each line clearing out nearly all parental Har P-element insertions and developing unique landscapes of P-element insertions . These data suggest that during the course of stabilizing the HISR lines , there were bursts of new P-element transpositions resulting in novel transposon landscapes that are completely distinct from the parental Har genome . Although this dispersion of de novo P-elements in HISR lines’ genomes stymied our goal to pinpoint a particular Har locus strongly inducing GD , we next cloned and sequenced genomic PCR amplicons of all P-elements from the various P-element-containing strains . By using a single oligonucleotide that primes from both the 5' and 3' Terminal Inverted Repeats ( TIRs ) , we amplified full-length P-elements as well as several additional truncation variants ( Figure 5A ) that have been missed in other genomic PCR assays using internal primers ( Wakisaka et al . , 2017 ) . The most abundant variant accounting for more P-element copies in OreR-MOD and OreR-TK strains compared to Har were the ‘KP’ variant shown to encode a dominant negative protein that inhibits full-length P-transposase activity ( Jackson et al . , 1988; Simmons et al . , 1990 ) ( Figure 5A–5C ) , thus explaining the innocuous accumulation of these P-element variants in these OreR strains . Full-length P-elements were also sequenced from Har , pi[2] and Lerik-P strains , but there were no sequence differences in these clones from the original full-length P-element sequence in GenBank that might suggest a superlative quality to the full-length P-element in these strong GD-inducing strains . Interestingly , we sequenced short ~630 bp P-element variants that were all very similar in configuration in Har , pi[2] , and Birm strains , which only retains ~130 bp of the 5' end and ~500 bp of the 3' end of the P-element ( Figure 5A and B ) . By retaining functional TIRs , these short elements can still be detected by TIDAL in WGS , can mobilize during crosses with the pi[2] strain ( Bingham et al . , 1982; Mullins et al . , 1989; O'Hare and Rubin , 1983 ) ; and were previously shown to be able to assemble in vitro with the P-transposase tetramer complex >100X more efficiently than the full-length P-element ( Tang et al . , 2007 ) . In addition , these short P-element variants seemed unlikely to translate into a protein due to multiple premature stop codons introduced by the massive internal deletion . In all HISR-D lines that retain strong GD induction , we detected this short P-element variant and the full-length P-element encoding P-transposase , whereas the HISR-N lines retained the short variant but appeared to have lost the full-length P-element ( Figure 5A , right panel ) . With the smaller number of TIDAL-predicted P-element insertions in HISR-N lines , we confirmed by locus-specific PCR the absence of full-length P-elements and that the majority of P-element insertions ( ~55–95% ) were these de novo short P-element insertions ( Figure 5D and Figure 5—figure supplement 1 ) . We name this short variant ‘Har-Ps’ ( Harwich P’s ) in homage to Harpies , highly mobile hybrid bird-human creatures from the Greek mythological stories of the Argonauts . To further support the conclusion that Har-P-like elements are the ammunition to drive severe GD , we examined additional fly lines from the DGRP collection ( Mackay et al . , 2012 ) which are of completely independent origin from Har , pi[2] , and Birm strains . Indeed , the strains RAL-42 and RAL-377 that cause severe GD despite having a fraction of the load of P-elements as Har also possessed Har-P-like short variants ( Figure 5—figure supplement 2 ) . Meanwhile , two other P-element-containing strains RAL-508 and RAL-855 did not induce GD because only the longer KP-like elements were present ( Figure 5—figure supplement 2 ) . We hypothesized that Har-Ps combined with P-transposase from full-length P-elements could be the drivers of strong GD induction from pi[2] , Har , and HISR-D strains . To test this hypothesis , we used negative-control yw-background females that lack P-transposase and transgenic H{CP}3 females that only express P-transposase in the germline ( Simmons et al . , 2002b ) in crosses with males that either lack Har-P copies ( ISO1 , Lerik-P , OreR-MOD ) or contain many Har-P copies ( Har , HISR-N’s , Birm ) ( Figure 6A ) . GD induction was only restored in the F1 daughters of this cross in strains with many Har-Ps ( Figure 6B ) . To avoid silencing of P-transposase by maternal P-element piRNAs in these strains , these crosses specifically used males that should only contribute paternal chromatin without contributing piRNAs ( Figure 6A ) . Notably , the KP-length and full-length P-elements in OreR-MOD and Lerik-P , respectively , did not restore GD ( Figure 6B , right most bars of left graph ) . These results suggest P-transposase act upon Har-P loci rather than longer P-element variants to induce GD and support the observation for Har-P loci making up the majority of the de novo P-element insertions in HISR-N lines ( Figure 6C ) . Our data now genetically explain a previously described biochemical result showing that P-transposase assembles much more efficiently in vitro on short P-elements compared to the full-length P-element ( Tang et al . , 2007 ) . We noticed that GD severity in crossing HISR-N with the H{CP}3 transgenic line was not completely penetrant like GD assays with the parental Har ( Figure 6B versus Figure 1B ) because Har contributes both multiple copies of full-length P-elements and Har-P loci versus the single copy of the natural P-element transgene in H{CP}3 ( Simmons et al . , 2002b ) . In addition , natural P-element translation is inhibited by strong somatic splicing inhibition of the native P-element’s third intron ( IVS3 ) containing a premature stop codon and only inefficient splicing in the Drosophila germline that is further suppressed by piRNAs ( Siebel et al . , 1994; Teixeira et al . , 2017 ) . We also confirmed that IVS3 intron splicing was the main alteration that increased P-element expression in ovaries from a dysgenic cross between Har and ISO1 , whereas Open Reading Frame ( ORF ) parts of the P-element transcript are only modestly increased ( Figure 6—figure supplement 1A ) . We believe this sufficient expression of P-transposase promotes the preferred mobilization of Har-P short variants in dysgenic cross ovaries , but the cut-and-paste transposition mechanism of P-transposase should theoretically conserve the total copy number of P-elements . By using digital droplet PCR to precisely quantity total P-element copy numbers , we confirmed that total P-element copy numbers were stable across ovaries of daughters from two sets of dysgenic and non-dysgenic crosses ( Figure 6—figure supplement 1B ) . To test whether a stronger expressing P-transposase transgene could induce the complete GD in crosses with HISR-N lines , we turned to the delta[2-3] P-transposase transgenes that lack the IVS3 intron to enable strong somatic and germline P-transposase activity ( Robertson et al . , 1988 ) . When we crossed two different delta[2-3] female strains to males of HISR-N17 , -N25 , and -N31 which lack P-element piRNA expression but have Har-Ps , we were unable to assay GD because of extensive pupal lethality ( Figure 7A ) . We also confirmed extensive pupal lethality in crosses between delta[2-3] and the Birm strain ( Figure 7A ) as previously described ( Engels et al . , 1987; Simmons et al . , 1987 ) . Since we also detected very short P variants in Birm that are similar to Har-P ( Figure 5A and B ) we conclude that somatically expressed P-transposase acting only on the Har-Ps in Birm , HISR-N17 , -N25 , and -N31 is sufficient to disrupt pupal development . Unexpectedly , the pupal lethality was suppressed when delta[2-3] females were crossed with Har-P-containing males that also expressed P-element piRNAs , such as Har , pi[2] , the four HISR-D lines , and HISR-N10 ( Figure 7A ) . These hybrid F1 progeny developed into adults , but the adult females of Har and HISR-N10 hybrids with delta[2-3] still exhibited severe GD ( Figure 7B ) . In addition , we also observed severe pupal lethality when delta[2-3] females were crossed to RAL-42 but not RAL-377 , although strong GD was still observed with RAL-377 ( Figure 7—figure supplement 1 ) . These RAL strains of independent origin from Har , pi[2] , and HISR lines provide convincing support for the conclusion that P-element piRNAs impart a paternally-transmitted imprint on Har-P loci that resists mobilization with somatically-expressed P-transposase and enables development to adulthood . However , this imprint is either erased in ovaries or insufficient to prevent ovarian GD . Finally , the notable P-element piRNA pattern of HISR-N10 perfectly matches the Har-P structure since many internal piRNAs are absent ( Figure 3B ) , but overall P-element piRNAs in HISR-N10 are equivalent to Har and HISR-D lines ( Figure 3C ) , and therefore are sufficient to repress Har-P’s epigenetically from being mobilized in the soma by the delta[2-3] P-transposase .
After a Drosophila strain has silenced an invading transposon through the Piwi/piRNA pathway , the neutered transposon will naturally decay into various truncations that are presumed to be neutral or even beneficial to host fitness ( Kelleher , 2016 ) , such as natural KP2 truncation variants that inhibit P-element transposition ( Jackson et al . , 1988; Simmons et al . , 1990 ) . However , we discovered one such truncation we call Har-P via our unbiased genetic and molecular approach that can actually be detrimental to the host . Our findings resonate with the previous finding that P-transposase assembles in vitro much more efficiently on very short natural P-element variants ( Tang et al . , 2007 ) , therefore we propose a new model for catastrophic P-element transposition in strong GD inducer strains like pi[2] and Har ( Figure 6C ) . When a P-element truncates to a ~630 bp Har-P variant , this non-autonomous variant dominates as the main mobilizing P-element during a dysgenic cross to induce strong GD . Thus , previous studies examining GD variability across other Drosophila strains and isolates may now be explained by whether these genomes contain both full length and very short P-elements ( Bergman et al . , 2017; Kozeretska et al . , 2018; Ronsseray et al . , 1989; Srivastav and Kelleher , 2017; Wakisaka et al . , 2017; Yoshitake et al . , 2018 ) . Moreover , this particular deletion size of a ~630 bp P-element variant that arose in at least four completely independent lines has persisted without detriment to these animals either because of piRNA silencing ( i . e . Har , pi[2] , RAL-377 ) or because the P-transposase was separated ( i . e . HISR-N and Birm ) . The short configuration must be special because additional sequence lengths such as P-element-based transgenes that were mobilized by P-transposase into transgenic strains are not strong triggers of GD like the Har-P elements ( Figure 7—figure supplement 1C ) . Although our future goal will be to determine which specific epigenetic marks are deposited at full length P-elements and Har-P’s by piRNAs , we believe a chromatin mark resisting P-transposase activity is more likely than somatic piRNAs or siRNAs ( Chung et al . , 2008; Ghildiyal et al . , 2008; Kawamura et al . , 2008 ) silencing the delta[2-3] P-transposase in our pupal lethality crosses because we confirmed robust P-transposase mRNA expression regardless of the expression of P-element piRNAs ( Figure 7—figure supplement 2A ) . A second future goal will be to generate transgenic flies with single or multiple synthetic Har-P copies to determine the precise dosage of Har-P’s that would trigger GD or pupal lethality . However , in addition to copy number , genomic location may also influence host tolerance of Har-P’s , because we observed a significant rescue of viable pupae in crosses between delta[2-3] P-transposase and a derivative strain of HISR-N17 with Har-P’s only on Chromosome 3 with six P-elements , while no pupae survived with delta[2-3] P-transposase and Har-P’s on Chromosome 2 with nine P-elements ( Figure 7—figure supplement 2B ) . The Drosophila P-element system of hybrid GD mainly affects female sterility and requires maternally contributed P-element piRNAs to propagate transgenerational P-element silencing in daughters via trimethylation of histone H3 lysine 9 ( H3K9me3 ) ( Josse et al . , 2007; Le Thomas et al . , 2014 ) . Although previous studies of dysgenic crosses focused on complete GD in females ( Bingham et al . , 1982; Brennecke et al . , 2008; Khurana et al . , 2011; Rubin et al . , 1982 ) , sons respond differently because they are fertile despite presumed somatic P-element excision ( Wei et al . , 1991 ) . Since previous studies of P-M hybrid dysgenesis above never considered a paternal imprint on P-elements that our findings now suggest is being propagated ( Figure 7C ) , future studies with HISR-N strains will enable us to dissect a paternally-transmitted small RNA-directed silencing effect in Drosophila that harkens to also a similar paternally-transmitted RNAi effects observed earlier in nematodes ( Grishok et al . , 2000 ) . Mouse piRNAs bound by MIWI2 direct the re-establishment of DNA methylation marks on transposons like L1 and IAP ( Aravin et al . , 2008 ) , which may propagate in sperm , but Drosophila DNA methylation is not prominent ( Krauss and Reuter , 2011 ) . Similar to other metazoans , Drosophila sperm also undergoes histone exchange with protamines ( ProtB ) , with little contribution of paternal cytoplasm ( Rathke et al . , 2007 ) . However , recent data do support the retention of some H3K9me3 in sperm ( Yamaguchi et al . , 2018 ) , which might underlie the paternal imprint of piRNA-silencing of P-elements that will be investigated further in future studies . This interplay between the truncated Har-Ps and full-length P-element DNA transposon resembles other examples in nature , such as the extreme proliferation of MITEs ( miniature inverted repeat transposable element ) in rice scavenging other transposases to mobilize ( Yang et al . , 2009 ) , and short mammalian SINEs retrotransposons taking advantage of the transposition machinery of longer LINEs , since SINEs persist in much greater numbers than their longer LINE counterparts ( Hancks and Kazazian , 2012 ) . However , while the full impact of MITEs and SINEs on organism development is still obscure , our study indicates that Har-Ps combined with the P-transposase to trigger transposition events so efficiently to be detrimental to ovarian and pupal development . Notwithstanding , the high efficiency of Har-P mobilization by P-transposase may also be engineered into a new generation of transposon-based mutagenesis approaches .
All strains were maintained on standard cornmeal medium at 22°C . Because the ISO1 ( BDSC#2057 ) stock had accumulated >180 new transposon insertions relative to the original stock sequenced in the Berkeley Drosophila genome project ( Adams et al . , 2000; Rahman et al . , 2015 ) , we obtained the ISO1 strain from Susan Celniker’s lab ( ISO1-SC ) . The Har strain was obtained from ( Har-WET ) was obtained from the William Theurkauf’s lab ( Khurana et al . , 2011 ) . Three Oregon-R strains were obtained from Terry Orr-Weaver’s lab , OreR-TOW , OreR-TK ( Kaufman , BDSC#2376 ) and OreR-MOD ( BDSC#25211 ) . The Lerik-P strain was obtained from Stephane Ronserray’s lab ( Josse et al . , 2007; Marin et al . , 2000 ) . All the following strains were also directly obtained from the BDSC – RAL-42 ( #28127 ) , RAL-377 ( #28186 ) , RAL-508 ( #28205 ) , RAL-855 ( #28251 ) , pi[2] ( #2384 ) , y[1] w[67c23]; H{w[+mC]=hsp/CP}3 ( #64160 ) , Birmingham; Sb[1]/TM6 ( #2539 ) , w[*]; wg[Sp-1]/CyO; ry[506] Sb[1] P{ry[+t7 . 2]=Delta2-3}99B/TM6B , Tb[+] ( #3629 ) , y[1] w[67c23]; H{w[+mC]=w[+] . Delta2-3 . M}6 ( #64161 ) , H{w[+mC]=KP} ( #64175 ) , P{TubGal80}Chr2 ( #7108 ) , P{TubGal80}Chr3 ( #7017 ) , and P{Elav-Gal4}Chr2 ( #8765 ) . Sp/CyO;TM6b/Sb was obtained from Michael Rosbash’s lab . All crosses were set up with 3–5 virgin females and 2–4 young males per replicate on standard cornmeal medium at 25°C and parents were purged after 5 days of egg laying ( Srivastav and Kelleher , 2017 ) . For GD assays , F1 females aged to 4–5 days at 25°C were examined for GD using food dye and GD % shown is average of 3 replicate crosses with total minimum of 100 F1 females assayed ( Simmons et al . , 2007 ) . Somatic pupal lethality was recorded by counting dead ( uneclosed ) and empty pupae ( eclosed ) 6 days after first eclosion was observed in respective control cross ( P{delta[2-3]}99B x ISO1 or H{P delta[2-3]} x ISO1 ) ( Engels et al . , 1987 ) . Pupal lethality percentage shown is average of two or more replicate crosses that obtained at least total of 50 F1 pupae each . The detailed crossing scheme is illustrated in Figure 2—figure supplement 1 . After a first cross between virgin Har females and ISO1 males , three more backcrosses of virgin Har/ISO1 hybrid progeny females mated to ISO1 males were performed and following the progeny with red eyes to select for the Har segment segregating with the cn , bw , sp , alleles on Chromosome 2R . We hoped that a particular set of P-elements that drive strong GD induction would co-segregate with red eye color . We then performed a ‘Validation Cross’ with the F4 hybrid males individually mated to ISO1 females . We screened >100 individual groups of F4 males for their GD induction , where the early-hatching 3 day old daughters were screened via the squash assay for 100% GD . Only the F5 vials showing 100% GD from F4 males crossed to ISO1 females were kept , and then were allowed to age and self-crossed and propagated in 11 more generations to attempt to create recombining-inbred-lines ( RILs ) . Selecting only flies with red eyes required purging any flies emerging with the ‘white’ eyes of ISO1 and discarding many vials that failed to generate progeny due to genotoxic collapse from inability to silence P-element transposition . At the F16 stage , Har/ISO1 Selfed Red ( HISR ) lines males were rescreened in a Validation Cross with ISO1 females , this time keeping lines that still caused 100% GD and designated as HISR-D ( Dysgenic ) lines . We also selected additional lines that had now lost GD and allowing for >50% of females to generate egg chambers , and these were designated HISR-N ( Non-dysgenic ) lines . We performed 2 rounds of single-sibling pair mating to further inbreed these lines in an attempt to stabilize the genotypes , and we maintained 4 lines of each HISR-D and HISR-N for true propagation of just the red or cinnabar eyes and speck phenotype . Genomic DNA was prepared from 10 young female flies by homogenizing tissues with plastic pestle in 300 µL Lysis buffer ( 10 mM Tris pH-8 . 0 , 100 mM NaCl , 10 mM EDTA , 0 . 5% SDS , and Proteinase K at 50 µg/ml ) and incubated at 65°C overnight followed by treatment of RNase A at 100 µg/ml at 37°C for 30 mins . 200 µL of 0 . 5M NaCl was added followed by one volume of Phenol:CHCl3:IAA ( at 25:24:1 ) and spun at 14 , 000 rpm for 10 min to isolate DNA in aqueous phase . Aqueous phase was extracted again with one volume of CHCl3:IAA ( at 24:1 ) and supplemented with one volume of 5M LiCl and incubated at −20°C and then spun at 15 , 000 rpm for 15 mins to precipitate RNA . Supernatant was isolated and supplemented with 2 volumes of 100% ethanol and incubated in −20°C for 2 hr and then spun at 15 , 000 rpm for 20 mins . DNA pellet was washed with chilled 70% ethanol and dissolved in nuclease free water . DNA integrity checked ( >10 kb ) by running 1 µg on 1% agarose gel with EtBr . Genomic PCR reactions to characterize P-element structural variation were set up in 30 µL reactions of 1X NEB GC buffer , 300 µM dNTPs , 0 . 5M Betaine , 2 . 5 mM MgCl2 , 0 . 25 µM of IR primer ( Rasmusson et al . , 1993 ) , 1 µL of Phusion polymerase and 50 ng of genomic DNA and cycled at 94°C for 1 min , 62°C for two mins , 72°C for 4 mins for 27 cycles and followed by 72°C for 15 min . Genomic PCR reactions to characterize P-element structural variation in HISR lines , predicted by TIDAL were also set up similarly using P-element insertion locus specific primers . Genomic PCR reactions for genotyping of HISR-N lines were set up similarly but cycled at 94°C for 30 s , 60°C for 15 s , 72°C for 30 s for 27 cycles and followed by 72°C for 5 min . Genomic qPCR experiments were performed in three biological replicates with two 20 µL technical reactions replicates each , using Luna Mastermix ( NEB ) , primers at 0 . 5 µM and 20 ng of genomic DNA per reaction in real time quantitative PCR . P-element load was calculated from 2^ ( -∆∆Ct ) normalized to Har at 100% and ∆Ct from RP49 . All primers used for are listed in Supplementary file 1 . For the Droplet Digital PCR ( ddPCR ) , we utilized the Evagreen Mastermix ( Biorad ) and conducted on a QX500 ddPCR machine with manual setting of droplet signal thresholds . 10–15 pairs of ovaries and corresponding carcass from 4 to 5 day old F1 females were dissected from dysgenic and non-dysgenic crosses of Har and HISR-D51 with ISO1 strain at 18 °C . DNA was extracted from the ovaries and carcass and quantified using Qubit 2 . 0 Fluorometer . Digital PCR probe assays were conducted in 40 µL droplet reactions , generated from 25 µL digital PCR reaction and 70 µL droplet oil each . 25 µL digital PCR reactions were set up with BioRad ddPCR probe supermix , P-element7a ( FAM ) and rp49 ( HEX ) probes each at 250 nM and 200 pg of DNA . Reactions were cycled at 95 °C for 10 mins followed by 95 °C for 30 s and 58 °C for 1 min for 40 cycles , and 98 °C for 10 mins . Copies/µL values were extracted from QuantaSoft ( BioRad ) software and P-element copies per genome were calculated normalized to rp49 . P-elements amplified from IR PCR were purified from 1% agarose gel using QIAquick Gel extraction kit and cloned into pCR4-TOPO vector using Zero Blunt TOPO PCR Cloning Kit at RT , followed by transformation of chemically competent DH10β cells , which were then grown on LB plates with 0 . 05 mg/ml Kanamycin overnight . 5–10 colonies were screened by PCR and two colonies positive for P-element cloned were chosen for plasmid mini-prep and sequenced using M13 forward and reverse primers for all variants in addition to internal primers to complete the sequencing of full-length P-elements . Genomic DNA libraries were prepared using NEB Ultra II FS kit E7805 . Briefly , 500 ng of genomic DNA ( >10 kb ) was fragmented at 37°C for 12 min , followed by adaptor ligation and loop excision according to kit manual protocol . Size selection was performed with two rounds of AmpureXP beads addition to select for insert size 150–250 bp as per kit manual . Library PCR amplification was also carried out as per manual instructions for six cycles and purified using one round of AmpureXP beads addition at 0 . 9X volume . Individual barcoded libraries were quantified on NanoDrop and each diluted to 2 nM and then pooled to produce equimolar concentration . Whole genome sequencing was performed on an Illumina NextSeq 500 with paired-end reads of 75 bp x 75 bp in the Rosbash lab at Brandeis University . Reads were demultiplexed and trimmed by Trimmomatic to remove low quality bases , and then reads were analyzed by the TIDAL program ( Rahman et al . , 2015 ) . TIDAL outputs were sorted for P-element insertions and the insertion coordinates were compared across the HISR lines using SQL queries in MS-Access . To calculate the Single Nucleotide Polymorphism ( SNP ) profiles , paired-end reads were mapped to the Dm6 ISO1 genome with ‘BWA MEM’ ( Li and Durbin , 2010 ) using default parameters . PCR duplicates are removed with Picard and SNPs are called with GATK HaplotypeCaller ( Danecek et al . , 2011; DePristo et al . , 2011; McKenna et al . , 2010 ) . We then generated the nucleotide distribution for each SNP to ensure that there are at least 20 reads supporting each SNP . Then , we created a unified SNP list by using the union of SNPs from all libraries and carefully noted if each SNP is present in each library . The SNP counts were binned by 5 kb segments and converted into a graphical representation as differences between the reference genome and strain/line in Figure 4—figure supplement 1 . To remove the 2S rRNA from Drosophila ovaries , we adapted a protocol from our previous Q-sepharose beads matrix technique ( Lau et al . , 2009 ) . About 50 ovaries per parental Har and ISO1 strains and HISR lines were dissected from young adult females . Ovaries were then lysed in ice cold 500 ul Elution Buffer ( 20 mM Hepes pH 7 . 9 ( with KOH ) , 10% glycerol , 400 mM KOAc , 0 . 2 mM EDTA , 1 . 5 mM MgCl2 , 1 . 0 mM DTT , 1X Roche Complete EDTA-free Protease Inhibitor Cocktail ) using one freeze-thaw cycle and pulverizing with a blue plastic pestle . A 1 . 5 ml aliquot of Q-Sepharose FF matrix suspension was washed 1X in water , then 3X in Elution buffer , then incubated for 10 min with the ovaries lysate with occasional agitation in cold room . Ribosomal RNA gets bound by the Q-sepharose , while small RNA RNPs remains in the elution buffer . Elution buffer was removed and then subjected to small RNA extraction with the Tri-reagent protocol . The precipitated small RNAs where then converted into Illumina libraries using the NEBNext Small RNA Library Construction kit . One modification we employed during the overnight linker ligation is to supplement the reactions to 12 . 5% PEG 8000 to reduce the potential sequence biases from T4 RNA ligase activity . Small RNA libraries were sequenced as 75 bp single end reads on the NextSeq550 . Adapters for the small RNA libraries were removed with CutAdapt and then mapped to the Drosophila transposon consensus sequences from RepBase and Flybase using Bowtie v1 with up to two mismatches and R plotting scripts as applied in our previous published studies on Drosophila piRNAs ( Clark et al . , 2017; Sytnikova et al . , 2014 ) . For this assay , 5–10 pairs of ovaries were dissected from 3 to 5 day old F1 females of dysgenic and non-dysgenic cross with Har and ISO1 , as well as with Har and HISR-D46 . RNA was extracted from such ovaries and integrity checked by running 1 µg RNA at 2% Agarose II gel ( Fischer BioReagents ) . 3 µg was reverse transcribed using Protoscript RT enzyme ( NEB ) as per manufacturer’s protocol and negative RT control was carried out similarly without RT enzyme . 50 ng of cDNA was used for setting up rp49 PCR reactions ( as described above ) from RT and corresponding negative RT reactions to evaluate DNA contamination . qPCR reactions for P-element ORF2 , ORF3 , IVS3 were also carried out as genomic qPCR reactions with 20 ng cDNA input and ΔCt were calculated similarly using rp49 RNA levels . In the RT-qPCR analysis of H{P delta[2-3]} gene expression in gonadal dysgenesis and pupal lethality , 5–10 pairs of ovaries and corresponding carcass were dissected from F1 females of pupal lethality crosses conducted at 18 °C . RNA extraction , reverse transcription , PCR and qPCR reactions were carried out similarly as above . ΔCt were calculated similarly using rp49 RNA levels . Fold change values were obtained from normalizing F1 carcass P-element RNA levels to H{P delta[2-3]} carcass and F1 ovary P-element RNA levels were normalized to H{P delta[2-3]} carcass in Figure 7—figure supplement 2 . HISR-N17 autosomes were isolated first by crossing virgin HISR-N17 females with Sp/CyO;TM6b/Sb stock males and using virgin F1 females with CyO and TM6 to cross again with Sp/CyO;TM6b/Sb . F2 males with either HISR-N17 Chr2 or HISR-N17 Chr3 were crossed to virgin H{P delta[2-3]} . All crosses were performed in triplicates at 25 °C . F3 pupal lethality was recorded on 16th day of the H{P delta[2-3]} crosses . | DNA provides the instructions needed for life , a role that relies on it being a very stable and organized molecule . However , some sections of DNA are able to move from one place in the genome to another . When these “mobile genetic elements” move they may disrupt other genes and cause disease . For example , a mobile section of DNA known as the P-element causes a condition called gonadal dysgenesis in female fruit flies , leading to infertility . Only certain strains of fruit flies carry P-elements and the severity of gonadal dysgenesis in their daughters varies . For example , when male fruit flies of a strain known as Harwich ( or Har for short ) is crossed with female fruit flies that do not contain P-elements , all of their daughters develop severe gonadal dysgenesis and are infertile . However , if the cross is done the other way around , and female Har flies mate with males that do not contain P-elements , the daughters are fertile because the Har mothers provide their daughters with protective molecules that silence the P-elements . But it was a mystery as to why the P-elements from the Har fathers always caused such severe gonadal dysgenesis in all the daughters . Here , Srivastav et al . bred fruit flies to create offspring that had different pieces of Har DNA in a genetic background that was normally free from P-elements; they then analyzed the ‘hybrid’ offspring to identify which pieces of the Har genome caused gonadal dysgenesis in the daughter flies . These experiments showed that Har flies possess a very short variant of the P-element ( named “Har-P” ) that is more mobile than other variants . However , the Har-P variants still depended on an enzyme known as P-transposase encoded by the full-length P-elements to move around the genome . Further experiments showed that other strains of fruit flies that cause severe gonadal dysgenesis also had very short P-element variants that were almost identical to Har-P . These findings may explain why Har and some other strains of fruit flies drive severe gonadal dysgenesis . In the future , it may be possible to transfer P-transposase and Har-P into mosquitoes , ticks and other biting insects to make them infertile and help reduce the spread of certain diseases in humans . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"genetics",
"and",
"genomics"
] | 2019 | Har-P, a short P-element variant, weaponizes P-transposase to severely impair Drosophila development |
The ability to detect and/or manipulate specific cell populations based upon the presence of intracellular protein epitopes would enable many types of studies and applications . Protein binders such as nanobodies ( Nbs ) can target untagged proteins ( antigens ) in the intracellular environment . However , genetically expressed protein binders are stable regardless of antigen expression , complicating their use for applications that require cell-specificity . Here , we created a conditional system in which the stability of an Nb depends upon an antigen of interest . We identified Nb framework mutations that can be used to rapidly create destabilized Nbs . Fusion of destabilized Nbs to various proteins enabled applications in living cells , such as optogenetic control of neural activity in specific cell types in the mouse brain , and detection of HIV-infected human cells by flow cytometry . These approaches are generalizable to other protein binders , and enable the rapid generation of single-polypeptide sensors and effectors active in cells expressing specific intracellular epitopes .
Many applications in biology and medicine require the ability to target a subset of cells in a population based upon specific cellular characteristics . Although this can be achieved by exploiting transcriptional elements that are selectively active in a subset of cells , specific elements are often not available , and can be difficult to generate . Alternatively , other features that distinguish cells , such as expression of a specific RNA or protein , may be exploited . Recently , it has become possible to utilize specific intracellular proteins to drive desired molecular events , using RNA-based binders in cells ( Auslander et al . , 2014; Culler et al . , 2010; Kennedy et al . , 2014; Saito et al . , 2011 ) , and protein-based binders in cells and animals ( Tang et al . , 2013 , 2015 ) . While current methods are promising , protein-responsive systems are continually evolving . There remains a need for generalizable strategies that enable rapid conversion of diverse classes of binders into protein-responsive tools . Antibodies are widely adopted reagents used for protein detection and manipulation . Their popularity derives from their superior specificity and high affinity , achieved in large part by the stringent selection in an immunized animal . Nbs , the antigen recognition portions of single chain antibodies found in camelids ( Hamers-Casterman et al . , 1993 ) and cartilaginous fishes ( Greenberg et al . , 1995 ) , bind their cognate antigens with high affinity and specificity , and have the added advantage over heterotetrameric antibodies in that they are very stable in the intracellular environment . Fusions between Nbs and proteins with desirable activities have enabled a number of applications in living cells ( Caussinus et al . , 2012; Irannejad et al . , 2013; Kirchhofer et al . , 2010; Rothbauer et al . , 2006; Tang et al . , 2013 , 2015 ) . Despite these successes , it has been difficult to take advantage of Nbs for live cell applications requiring cell-specificity , as genetically expressed Nb-fusion proteins are stable and active even in cells that do not express the cognate antigens . This is a general problem that applies to any class of protein-based binder . To address this issue , we reasoned that the Nb portion of a single Nb-fusion protein could be modified to be conditionally stable in living cells , with stability conferred by the presence of antigen ( Figure 1A ) . A similar approach has been used to create small molecule-dependent domains , for temporal control or tuning of protein activity ( Banaszynski et al . , 2006 ) . 10 . 7554/eLife . 15312 . 003Figure 1 . Isolation of a destabilized Nb whose protein level depends upon antigen co-expression . ( A ) Concept of antigen-controlled protein stabilization . FP , fusion protein; An , antigen . ( B ) Strategy for isolating dNbs . LTR , long terminal repeat . ( C ) GFP-dependent stabilization of dGBP1 tagged with TagBFP . ( D ) Western blot of transfected 293T cell lysate for TagBFP and βgal ( a transfection control ) shows that the level of dGBP1-TagBFP is dependent upon YFP co-expression . YFP is a derivative of GFP . ( E ) YFP ( green ) -dependent dGBP1-TagBFP fluorescence in transfected 293T cells . t-HcRed ( red ) is a transfection marker for cells with TagBFP ( gray ) fusion constructs . Scale bar , 100 μm . All results from ( D , E ) are representative of three independent experiments , for 3 biological replicates . ( F ) Schematic of electroporation experiment . ( G ) GFP ( green ) , but not DsRed ( red ) , promotes dGBP1-TagBFP fluorescence ( magenta ) in the mouse retina . ONL , outer nuclear layer; INL , inner nuclear layer . ( H ) The expression pattern of TagBFP protein , as detected by Anti-TagBFP ( magenta ) , can be altered by changing the GFP expression pattern with broadly active ( CAG ) or rod photoreceptor-specific ( Rho ) promoters . TagBFP immunodetection was not carried out in ( B ) as the TagBFP antibody cross-reacts with DsRed . Scale bar is 20 μm . Biological replicates ( retinas ) : n = 3 for all conditions . ( I ) Dependence of dGBP1-TagBFP protein level on YFP dose in transfected 293T cells . ( J ) The ubiquitin proteasome pathway is involved in degradation of dGBP1 . Transfected 293T cells were treated with drugs for 20 hr . BTZ , Bortezomib . All results from ( I , J ) are representative of three independent experiments , for 3 biological replicates . Additional data related to characterizations of dGBP1-TagBFP in vitro and in vivo are shown in Figure 1—figure supplements 1–2 . DOI: http://dx . doi . org/10 . 7554/eLife . 15312 . 00310 . 7554/eLife . 15312 . 004Figure 1—figure supplement 1 . YFP-dependent dGBP1-TagBFP stabilization in cells . ( A ) Comparison of background fluorescence between two GBP1 variants . Constructs were expressed using MMLV-based vectors , which were transfected into 293T cells . Images were acquired 2 days post-transfection . TagBFP fused to dGBP1 had no detectable background fluorescence in the absence of YFP , whereas TagBFP fused to GBP1 variant S19L , K67N had occasional fluorescent aggregates in the absence of YFP . Images are representative of at least 2 independent experiments . Scale bar is 50 μm . ( B ) Quantification of results from Figure 1E . Transfected cells were identified by expression of either YFP or t-HcRed . Plot shows median and maximum-to-minimum range . Data are from 3 independent experiments , for 3 biological replicates . ( C ) YFP stabilizes wildtype GBP1 protein in cells . 293T cells transfected with MMLV-GBP1-TagBFP , CAG-nlacZ and filler plasmid or pCAG-YFP were harvested 2 days post-transfection and whole cell lysates were blotted for anti-TagBFP as well as anti-βgal . Densitometry measurements of western blot bands were plotted as the relative density of anti-TagBFP bands compared to that of anti-βgal bands in the same lane . a . u . is arbitrary unit . n = 3 . Comparison of the two conditions gave P value 0 . 026 . *indicates p<0 . 05 , Student’s t-test . Plot shows median and range . Results are representative of at least 3 independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 15312 . 00410 . 7554/eLife . 15312 . 005Figure 1—figure supplement 2 . Detection of antigen-expressing cells with dNb in vivo . Quantification of results from Figure 1G , H . ( A ) Tight coupling of GFP expression ( green ) and Anti-TagBFP staining ( red ) from ONL cells in the +CAG-GFP condition . Scale bar is 20 μm . ( B–E ) . Quantification of electroporation results . ( B ) GFP-dependency of TagBFP expression . Counted cells from ONL . Plotted % TagBFP+ cells given GFP+ ( from +CAG-GFP ) or DsRed+ cells ( from +CAG-DsRed ) . ( C ) Efficiency of GFP-dependent protein stabilization . Efficiency is % Anti-TagBFP+ cells given GFP+ cells . ( D ) GFP-specificity of system , as determined by% GFP+ cells given Anti-TagBFP+ cells . ( E ) dGBP1-TagBFP expression pattern closely matches that of GFP . All electroporated cells , as defined by GFP or TagBFP expression , were quantified across a 20 μm retinal section and represented as % of total number of cells counted . Graphs and values shown are as mean ± standard deviation . Biological replicates ( retinas ) : n = 3 for all conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 15312 . 005 Here , we report the isolation of destabilized Nbs ( dNbs ) using a strategy that should be generalizable to other types of protein-based binders . We isolated a dNb whose destabilizing mutations fell within the structurally conserved framework region of Nbs . These destabilizing mutations could simply be transferred to other Nbs to rapidly generate antigen-dependent stability . dNbs were able to destabilize fusion partners having a variety of activities , including fluorescent proteins , site-specific recombinases and genome editing enzymes . We used these reagents to optogenetically control neural activities in specific cell types , as well as detect and isolate Human immunodeficiency virus ( HIV ) infected cells based upon the expression of the HIV-1 capsid protein . Thus , this work offers a generalizable strategy to label and manipulate specific cell populations in cellular and animal systems , with specificity endowed by protein expression and/or specific cellular features .
To test whether it is possible to modify an Nb such that its intracellular protein level is strongly dependent upon antigen co-expression , we used the GFP-binding Nb , GBP1 , for proof-of-concept experiments ( Kirchhofer et al . , 2010; Rothbauer et al . , 2006 ) ( Figure 1B , C ) . We generated a Moloney murine leukemia virus ( MMLV ) library encoding randomly mutagenized variants of GBP1 fused to the blue fluorescent protein , TagBFP ( Subach et al . , 2008 ) . t-HcRed ( Gurskaya et al . , 2001 ) was co-expressed via an IRES to report infection . TagBFP and t-HcRed bear little amino acid similarity to Aequorea-derived GFP and its derivatives . We infected 293T cells with this library , and combined FACS with super-infection by a GFP-encoding recombinant adeno-associated virus ( rAAV ) to isolate GBP1-TagBFP variants whose blue fluorescence depended upon GFP expression ( Figure 1B; Materials and methods ) . One hundred GBP1 variants were then individually screened for enhanced TagBFP expression in the presence of yellow fluorescent protein ( YFP ) , a GFP derivative known to also interact with GBP1 ( Rothbauer et al . , 2008; Tang et al . , 2013 ) . Some variants showed fusion TagBFP aggregates within well-transfected cells when YFP was absent , but became soluble in the cytoplasm when YFP was present ( Figure 1—figure supplement 1A ) . Notably , a GBP1 variant carrying 6 amino acid changes ( A25V , E63V , S73R , S98Y , Q109H , S117F ) gave little to no TagBFP fluorescence , with no signs of aggregation in the absence of YFP . We focused our efforts on this variant , which will hereafter be referred to as destabilized GBP1 ( dGBP1 ) . dGBP1-TagBFP showed strong fluorescence and protein level when co-expressed with GFP or YFP , but became weakly detectable or undetectable when antigen was absent ( Figure 1D , E and Figure 1—figure supplement 1 ) . In contrast , unmodified GBP1-TagBFP showed strong fluorescence and protein level regardless of antigen co-expression ( Figure 1D , E ) . Interestingly , we detected an increase in the level of wildtype GBP1-TagBFP protein in the presence of YFP ( Figure 1—figure supplement 1C ) . In an electroporation experiment using the mouse retina , dGBP1-TagBFP fluorescence and protein level were detected only upon GFP co-expression in vivo ( Figure 1F–H , Figure 1—figure supplement 2 ) . Strikingly , the efficiency of TagBFP stabilization by GFP expression was nearly 100% , i . e . almost every GFP+ cell was TagBFP+ ( Figure 1—figure supplement 2 ) . The efficiency of the TDDOG and CRE-DOG systems was , at the highest , ~60% in similarly designed electroporation experiments ( Tang et al . , 2013 , 2015 ) . This difference likely reflects the requirement for delivery of a greater number of components for T-DDOG and CRE-DOG experiments . Taken together , these data show that one can create a highly destabilized Nb whose protein level is dependent upon co-expression with its cognate antigen in vitro and in vivo . We previously created Nb-based , antigen-dependent systems that use the antigen as a scaffold for the assembly of split protein domains or fragments ( Tang et al . , 2013 , 2015 ) . Complex assembly can be inhibited when excessive antigen levels saturate antigen-binding sites in Nb-fusion proteins ( Tang et al . , 2013 , 2015 ) . In contrast , a single polypeptide , dNb-fusion protein should not suffer the same limitation . Indeed , YFP promoted dGBP1-TagBFP stability in a dose-dependent manner , with no adverse effects even when YFP plasmid was transfected at ten-fold excess of dGBP1-TagBFP plasmid ( Figure 1I ) . To investigate the mechanism of dNb destabilization , dGBP1-TagBFP-transfected 293T cells were treated with the ubiquitin proteasome inhibitors , MG132 or Bortezomib ( BTZ ) ( Kisselev et al . , 2012 ) . dGBP1-TagBFP protein was evident following addition of either inhibitor and was absent without inhibitors , indicating that it was degraded by the ubiquitin proteasome system ( UPS ) ( Figure 1J ) . Discosoma-derived mCherry fused to dGBP1 ( dGBP1-mCherry ) also showed antigen-dependent stabilization . Unlike dGBP1-TagBFP ( Figure 1E ) , some aggregation of dGBP1-mCherry occurred inside cells when antigen was absent ( Figure 2—figure supplement 1 ) . We use dGBP1-mCherry as a sensitized reporter to map the key residues involved in GBP1 stability , by comparing the level of fluorescence and aggregation of the fusion proteins in cells . ( Figure 2—figure supplements 1 , 2 ) . C/S98Y and S117F showed strong destabilizing effects , as seen in both sufficiency and necessity experiments . S73R and Q109H also had destabilizing effects in single mutant analyses . GFP rescued the destabilization phenotype of all mutants . Thus , specific single dGBP1 mutations had clear destabilizing effects , which could be enhanced by combination with other destabilizing mutations . The dGBP1 mutations mapped onto the structurally conserved framework regions of Nbs ( Muyldermans , 2013 ) , and 99–100% of Nbs ( n=76 ) shared the same residue as GBP1 at each of the 3 most destabilizing positions ( 3maj: S73R , C/S98Y , S117F ) ( Figure 2A; Figure 2—figure supplement 3A ) . Further , a survey across 76 unique Nb-antigen interfaces , gathered from a total of 102 crystal structures , indicated that Nb positions corresponding to those of dGBP1 A25V , S73R , S98Y and S117F were universally located outside of all Nb-antigen interfaces ( Figure 2—figure supplement 3B ) . Nb positions corresponding to dGBP1 Q109H fell outside of 99% , or 75 of 76 unique Nb-antigen interfaces . Positions equivalent to dGBP1 E63V were found in 22% , or 17 of 76 unique Nb-antigen interfaces , and in close proximity to the interface in 9% , or 7 of 76 of the cases . Given these results , we hypothesized that the destabilizing framework mutations could be transferred across Nbs to rapidly create antigen-dependent stability . We transferred all dGBP1 mutations ( 6mut ) and the 3maj mutations to Nbs targeting the HIV-1 capsid protein ( αCA ) and Escherichia coli dihydrofolate reductase ( αDHFR ) , respectively . dNbs created by mutation transfer behaved similarly as dGBP1 in that TagBFP fusion fluorescence and protein level both depended upon expression of the cognate antigen ( Figure 2B–E ) . Destabilization also depended on degradation by the UPS ( Figure 2F ) . We then explored whether dNb-TagBFP expression had an adverse effect on antigen expression . Using western blots to quantify protein levels , we found that dGBP1-TagBFP did not have an obvious effect on YFP protein level , when compared to the negative control condition whereby αCA6mut-TagBFP replaced dGBP1-TagBFP ( Figure 2—figure supplement 3C ) . 10 . 7554/eLife . 15312 . 006Figure 2 . dGBP1 destabilizing mutations can be transferred to Nbs derived from different species to create antigen-dependent stability . ( A ) Protein alignment of Nbs against GFP ( GBP1 ) , HIV-1 CA ( αCA ) and E . coli DHFR ( αDHFR ) . Amino acid positions numbered according to the ImMunoGeneTics information system ( IMGT ) . FR , framework; CDR , complementarity determining region . Purple- and green-highlighted residues indicate dGBP1 mutation position . Green-highlighted residues ( 3maj ) are most destabilizing when mutated . The underlined serine was a cysteine in the original GBP1 . ( B–E ) . Transfer of dGBP1 mutations to other Nbs . Destabilized , but not unmodified αCA and αDHFR showed antigen-dependent fluorescence ( B , D ) as well as protein level ( C , E ) . DsRed ( red ) indicates transfected cells in ( B , D ) and is only shown when TagBFP ( gray ) results are negative . Scale bar , 50 μm . ( F ) A dNb generated by mutation transfer was degraded by the UPS . αCA-dNb6mut-TagBFP showed an increase in protein level when transfected 293T cells were treated with MG132 for 6 hr . Results representative of 3 independent experiments . ( G ) Heat map showing median TagBFP fluorescence intensity of Nb-TagBFP fusions . All Nbs shown recognize epitopes of intracellular origin . Fluorescent readings were normalized to that of unmodified Nb ( no antigen ) condition , which was set to 100 . n = 3 biological replicates pooled from 3 independent experiments per Nb . Each biological replicate result is shown as a horizontal bar in the heat map . Bar graphs indicate median and maximum-to-minimum range . Additional data related to mutation transfer of destabilizing mutations are shown in Figure 2—figure supplements 1–3 . Source data for % Nb-TagBFP fluorescence values are shown in Figure 2—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 15312 . 00610 . 7554/eLife . 15312 . 007Figure 2—source data 1 . Source data for fluorescence quantifications of Nb-TagBFP tests . This excel file contains numerical values for the % Nb-TagBFP fluorescence parameter shown as a colored heat map in Figure 2G . DOI: http://dx . doi . org/10 . 7554/eLife . 15312 . 00710 . 7554/eLife . 15312 . 008Figure 2—figure supplement 1 . Mapping of mutations necessary for dGBP1 destabilization . ( A ) Representative images showing expression of GBP1 variants tagged with mCherry in 293T cells . Images taken 17 hr post-transfection . Scale bar , 50 μm . ( B ) Semi-quantitative summary of mCherry fluorescence intensity as well as cellular solubility phenotype . Sol , soluble . Agg , aggregate . Plot is mean ± standard deviation . Asterisk indicates mutations that showed clear increase in fluorescence compared to dGBP1-mCherry . n = 4–5 per condition . Consistent results were obtained in at least 3 biological replicates ( transfected wells ) in at least 3 independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 15312 . 00810 . 7554/eLife . 15312 . 009Figure 2—figure supplement 2 . Mapping mutations sufficient for dGBP1 destabilization . ( A ) Representative images showing expression of GBP1 variants tagged with mCherry in 293T cells . Images taken 17 hr post-transfection . Scale bar , 50 μm . ( B ) Semi-quantitative summary of mCherry fluorescence intensity as well as cellular solubility phenotype . Sol , soluble . Agg , aggregate . Plot is mean ± standard deviation . Asterisk indicates clearly destabilized mutations compared to GBP1-mCherry . n = 6 per condition except GBP1 control ( n = 3 ) . Consistent results were obtained in at least 3 biological replicates ( transfected wells ) in at least 3 independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 15312 . 00910 . 7554/eLife . 15312 . 010Figure 2—figure supplement 3 . Generation of dNbs by mutation transfer . ( A ) Conservation of dGBP1 mutations across 76 Nbs derived from Camelus dromedarius , Llama glama and Vicugna pacos . ( B ) Mapping of Nb destabilizing positions in relation to binding interfaces across Nb-Antigen complexes . ( C ) dGBP1-TagBFP expression does not have a clear effect on YFP protein level . Transfected 293T cells were harvested 1 day post-transfection . Whole cell lysates were blotted for anti-YFP as well as anti-βgal . Densitometry measurements of western blot bands were plotted as the relative density of anti-TagBFP bands compared to that of anti-βgal bands in the same lane . a . u . is arbitrary unit . n = 3 . Plot shows median and range . Results are representative of at least 3 independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 15312 . 010 To further investigate the generality of the mutation transfer approach , we transferred the 3maj mutations to 9 Nbs that recognize epitopes of intracellular origin ( Figure 2G; Materials and methods ) . All dNb-TagBFPs showed strongly reduced fluorescence relative to their unmodified Nb counterparts and occasionally formed faint fluorescent punctae in cells over-expressing the fusion proteins ( Figure 2B , D and 2G ) . Importantly , whereas no unmodified Nb showed >2 fold increase in TagBFP fluorescence in response to antigen co-expression , 8 of 9 , or 89% of dNbs , passed this threshold ( Figure 2G ) . Notably , mutations that destabilized a Camelus dromedarius Nb ( GBP1 ) had very similar effects on Nbs derived from Vicugna pacos and Llama glama , indicating that destabilizing mutations can be transferred across Nbs from different camelid species to create antigen-dependence ( Figure 2A and G; Table 1 ) . 10 . 7554/eLife . 15312 . 011Table 1 . List of tested nanobodies and their associated antigens . DOI: http://dx . doi . org/10 . 7554/eLife . 15312 . 011Nb PDB codeSpecies of originTested antigenAntigen species/pathogenEndogenous location of epitope3K1KC . droGFPAequorea victoriaIntracellularYFPAequorea victoriaIntracellularYFP-FLAGAequorea victoriaIntracellular2XV6V . pacosCapsid protein p24 C-terminal domain , residues 278-352 of gag polyproteinHIV-1IntracellularCapsid protein p24HIV-1Intracellular4EIGL . glamaDihydrofolate reductaseEscherichia coliIntracellular4TVSV . pacosTorsin-1A-interacting protein 1 , UNP residues 356-583Homo sapiensIntracellular4EIZL . glamaDihydrofolate reductaseEscherichia coliIntracellular4FHBL . glamaDihydrofolate reductaseEscherichia coliIntracellular4QO1L . glamaCellular tumor antigen p53 DBD , UNP residues 92-312Homo sapiensIntracellular3K7UL . glamaMP18 RNA editing proteinTrypanosoma bruceiIntracellular4GFTL . glamaMyosin A tail domain interacting protein C-terminal domain , UNP residues 137-204Plasmodium falciparumIntracellular4C57L . glamaCyclin-G associated kinase , kinase domain residues 14-351Homo sapiensIntracellularC . dro: Camelus dromedarius; L . glama: Lama glama; V . pacos: Vicugna pacos . To explore if additional fusion proteins could be rendered antigen-dependent , we engineered dNbs onto two popular site-specific recombinases , Cre and codon-optimized Flp ( Flpo ) ( Luo et al . , 2008; Raymond and Soriano , 2007 ) . Both enzymes were rendered GFP-dependent after fusion to dGBP1 , but not GBP1 ( Figure 3A–B ) . By increasing the number of dGBP1 domains fused to either enzyme , residual GFP-independent recombination of the initial fusions was decreased without affecting GFP-dependent recombination ( Figure 3A–B and Figure 3—figure supplement 1A–B ) . Notably , Flpo fused to tandemly repeated dGBP1 ( dGBP1x2-Flpo , or Flp dependent on GFP [Flp-DOG] ) had insignificant background signal and over 600 fold induction by GFP ( Figure 3B ) . We also rapidly constructed an Flpo fusion protein dependent upon the C-terminal portion of HIV-1 CA ( C-CA ) using mutation transfer . This construct was functional in vitro and in vivo ( Figure 3C and Figure 3—figure supplement 1D ) . Further , both GFP- and C-CA-dependent Flpo responded to antigen in a dose-dependent manner ( Figure 3—figure supplement 1C–D ) . Thus , dNbs can confer antigen-dependent control over different types of effector proteins , and can adequately reduce the activity of a highly sensitive enzyme when antigen is absent . 10 . 7554/eLife . 15312 . 012Figure 3 . Generation and optimization of antigen-specific effectors based on dNbs . ( A ) Reporter assay of transfected 293T cells testing dGBP1-Cre fusion constructs for activation of a Cre-dependent luciferase construct . ( B ) Reporter assay of transfected 293T cells testing dGBP1-Flpo fusion constructs for activation of a Flpo-dependent luciferase construct . Cells were harvested at 15 hr ( A ) or 36 hr ( B ) post-transfection . All results are representative of 3 independent experiments . Sample size per condition , n = 12 ( A ) and n = 18 ( B ) . ( C ) Rapid generation of a C-CA-dependent Flpo by transfer of dGBP1 mutations to αCA Nb . Schematic of C-CA-dependent Flpo ( left ) . C-CA can promote Flpo recombination in the mouse retina ( right ) . n-βgal was an electroporation marker . Scale bar , 20 μm . ( D ) Quantification of C-CA-dependent Flpo activity . In ( C , D ) , 4 and 3 electroporated retinas were analyzed for +C-CA and +GFP conditions , respectively . ( E , F ) Reporter assay in 293T cells using Flp-dependent luc2 construct , CAFNF-luc2 . Results show dependence of Flpo activity on the presence of both GFP and C-CA ( E ) or GFP and DHFR ( F ) . n = 6 per condition in ( E ) and n = 9 per condition in ( F ) . Consistent results were obtained in 3 independent experiments . All plots indicate mean and standard deviation . Additional characterization of dNb-based effectors is shown in Figure 3—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 15312 . 01210 . 7554/eLife . 15312 . 013Figure 3—figure supplement 1 . Additional information on antigen-specific effectors . ( A–B ) Plasmids encoding unmodified or destabilized GBP1 fusion to Cre or Flpo were transfected into 293T cells along with CAG-driven , loxP-Neo-loxP- ( LNL- ) or FRT-Neo-FRT- ( FNF- ) DsRed . Cre and Flpo were each individually fused to either GBP1 , dGBP1-GBP1 , or dGBP1x2 , respectively ( see cartoon ) . Compared to dGBP1-GBP1 fusion , dGBP1x2 gave increased suppression of background recombinase activity . O/E: overexposed . Cells imaged at 22 hr ( A ) or 50 hr ( B ) post-transfection . Images representative of three independent experiments . Scale bar , 100 μm . ( C–D ) . Antigen-specific effectors depend on antigen dose . Transfected 293T cells were assayed for Flpo recombination by the CAG-driven FNF-luc2 reporter . ( C ) GFP dose dependency of GFP-dependent Flpo . n = 6 . Data pooled from 3 independent experiments . ( D ) C-CA dose dependency of C-CA-dependent Flpo . n = 6 . Data pooled from 3 independent experiments . In ( C–D ) , Boxplots indicate maximum-to-minimum range . Box boundaries range from 25th to 75th percentile . The line in box indicates median . ( E ) Schematic of double dNb-fusion system for coincidence detection of antigen . FP , fusion protein . DOI: http://dx . doi . org/10 . 7554/eLife . 15312 . 013 The ability to control Flpo activity with tandem dNb-fusions raised the possibility of imposing dual regulation on effector protein activity using two dNbs , each targeting a different antigen ( Figure 3D–E and Figure 3—figure supplement 1E ) . We tested this by generating Flpo fused to dGBP1 and αCA dNb6mut ( dGC-Flpo ) or dGBP1 and αDHFR dNb3maj ( dGD-Flpo ) . Strong Flpo recombination was triggered only when both antigens were present ( Figure 3D–E ) . This shows the feasibility of using dNb-fusion proteins to create synthetic circuits whereby dual inputs are integrated entirely at the protein level . We further tested whether it was possible to perform genome targeting and editing under the control of specific antigen ( s ) ( Figure 4A ) . We created a fusion between two αCA dNb6mut and Cas9 ( dCC-Cas9 ) and delivered the construct to an engineered human cell line that expresses β-galactosidase upon removal of a loxP-stop-loxP cassette . We also delivered a guide RNA that can specifically target the loxP sites , leading to Cas9-mediated deletion of the stop cassette ( dCC-Cas9-loxPgRNA ) ( Figure 4B–C ) . Co-expression of C-CA with dCC-Cas9 and loxPgRNA triggered genome-editing events , while little to no β-galactosidase expression was detected when C-CA was absent ( Figure 4D ) . The efficiency of C-CA-dependent genome editing approached that of control Cas9 ( Figure 4D ) . This result demonstrated the feasibility of using intracellular epitopes to initiate genome editing or targeting . 10 . 7554/eLife . 15312 . 014Figure 4 . Intracellular antigen can trigger genome editing via dNbs . ( A ) Schematic of Cas9 fusion protein inducible by C-CA binding . ( B ) Fusion configuration of tandemly repeated C-CA Nb to Cas9 , giving dCC-Cas9 . ( C ) dCC-Cas9 activity was assayed for βgal expression in a human TE671 cell line engineered to contain a lacZ reporter inactive in expression due to a loxP-STOP-loxP transcriptional termination cassette . gRNA targets both loxP sequences for Cas9 cleavage and STOP removal . ( D ) dCC-Cas9 shows C-CA-dependent activity . Cas9 activity is represented as number of βgal+ cells induced as a percentage of unfused Cas9 activity ( 100% ) . AU1 is used as a negative control peptide , expressed in place of C-CA . Plots were mean ± standard deviation . n = 3 or 5 biological replicates ( transfected wells ) per condition . Consistent results were obtained in 3 independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 15312 . 014 To evaluate the usefulness of dNbs , we applied dNb-fusion proteins in situations where we could not control antigen level . GFP and its derivatives ( Tsien , 1998 ) are widely used to label cell types , with specificity dependent upon cellular features such as gene transcription ( Chalfie et al . , 1994 ) or neuronal connectivity ( Beier et al . , 2011; DeFalco et al . , 2001; Ekstrand et al . , 2014; Lo and Anderson , 2011; Wickersham et al . , 2007 ) . Genetic manipulation of GFP-labeled cells can reveal their functions , but current approaches require delivery of 2 or more αGFP Nb-fusion proteins ( Tang et al . , 2015 , 2013 ) . Here , we used electroporation and rAAV to deliver the one-component Flp-DOG along with Flp-dependent constructs to the retinas of Tg ( CRX-GFP ) ( Samson et al . , 2009 ) and cerebella of Tg ( GAD67-GFP ) ( Tamamaki et al . , 2003 ) lines , respectively . In both instances , robust Flpo recombination was detected in GFP+ tissues , but not in GFP-negative tissues labeled with electroporation , infection or injection markers ( Figure 5A–C; Figure 5—figure supplement 1 , and Figure 5—figure supplement 2A–E ) . We used rAAV-delivered Flp-DOG to induce ChR2-mCherry expression in GABAergic Purkinje cells ( PCs ) of Tg ( GAD67-GFP ) cerebella ( Figure 5C ) . Under conditions in which infection did not alter spontaneous firing frequency or input resistance , we evoked excitatory photocurrents and synaptic inhibitory currents in ChR2-mCherry+ PCs , with inhibitory inputs from neighboring ChR2-mCherry+ neurons that contacted the recorded PCs ( Figure 5; Figure 5—figure supplement 2F–G ) . GFP+ neurons that did not express ChR2-mCherry , as well as control ZsGreen+ neurons in GFP-negative animals , never showed light-evoked photocurrents , indicating antigen-specificity of the system ( Figure 5D ) . Thus , Flp-DOG provides a much simpler approach to manipulate GFP-defined cell types over pre-existing methods . Overall , these results demonstrate that dNb-fusion proteins enable functional manipulation of antigen-expressing cells in vivo . 10 . 7554/eLife . 15312 . 015Figure 5 . Applying Flp-DOG for optogenetic manipulation of transgenic GFP-labeled cell types in the mouse cerebellum . ( A ) rAAV reagents for ( B–D ) . ( B ) Schematic showing delivery of rAAVs to the mouse cerebellum for cell type-specific manipulation in Tg ( GAD67-GFP ) animals . ( C ) Representative image showing that rAAV-encoded , GFP-dependent Flpo activates ChR2-mCherry expression selectively in the cerebellar cortex of Tg ( GAD67-GFP ) ( n = 4 ) , but not wildtype ( n = 2 ) animals . ZsGreen is unrelated to GFP and was used as an infection marker for GFP-negative animals . Scale bar , 50 μm . ( D ) Optogenetic manipulation of GFP+ , ChR2-mCherry+ cells . A pulse of blue light ( blue bar ) evoked a photocurrent at -60 mV holding potential , and an inhibitory synaptic current at 0 mV that was blocked by 5 µM of the GABAA receptor antagonist SR 95531 , indicating activation of mCherry-ChR2+ cells synapsing onto the recorded PC . No photocurrents or synaptic currents were detected in GFP+/mCherry- PCs ( n = 12 cells ) from Tg ( GAD67-GFP ) animals as well as ZsGreen+/mCherry- PCs ( n = 9 cells ) from wildtype animals . Two-photon images show cells identified live for optogenetic manipulation and physiology . Bright ZsGreen aggregates were sometimes detected as faint signals in the mCherry channel . Scale bar , 10 μm . ZsG , ZsGreen; mC , mCherry . Additional data related to evaluation of Flp-DOG specificity for GFP expression are shown in Figure 5—figure supplements 1 and 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 15312 . 01510 . 7554/eLife . 15312 . 016Figure 5—figure supplement 1 . Retrofitting transgenic GFP line for cell-specific gene manipulation in the mouse retina . ( A ) Schematic of electroporation experiment . CAG-nlacZ was an electroporation marker . ( B ) Electroporation of CAG-driven Flp-DOG into Tg ( CRX-GFP ) retinas resulted in strong activation of FNF-DsRed reporter . DsRed was not detected in electroporated wildtype retinas . Scale bar , 20 μm . ( C ) Quantification of GFP-dependent Flpo recombination in the outer nuclear layer ( ONL ) of Tg ( CRX-GFP ) or wildtype retinas . Sample size , n = 4 retinas for GFP+ condition , n = 3 retinas for GFP-negative condition . Plots were mean ± standard deviation . Consistent results were obtained in 2 independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 15312 . 01610 . 7554/eLife . 15312 . 017Figure 5—figure supplement 2 . Characterization of mouse cerebella infected with AAV-Flp-DOG . ( A ) Reagents for ( B–D ) . ( B ) Schematic showing delivery of rAAVs and green fluorescent beads to the mouse cerebellum for cell type-specific manipulation in Tg ( GAD67-GFP ) animals . ( C ) rAAV-delivered , Flp-DOG activated ChR2-mCherry expression selectively in the cerebellar cortex of GFP+ , but not GFP-negative animals . Green fluorescent beads mark the site of infection . Pictures representative of 2 injected animals per condition . Scale bar , 50 μm . ( D ) Quantification of Flpo activity with regards to ( B–C ) . Plots show number of mCherry+ ( mC+ ) cells counted in whole brain slices labeled with beads . All mCherry+ cells were counted in 9 and 7 whole brain sections with highest bead density , for Tg ( GAD67-GFP ) and GFP-negative brains , respectively . n = 2 animals per condition . ( E ) Quantification of Flpo activity in Tg ( GAD67-GFP ) animals injected with rAAV-EF1a-Flp-DOG and rAAV-FLEXFRT-ChR2-mCherry , as well as in wildtype animals injected with rAAV-EF1a-ZsGreen , rAAV-EF1a-Flp-DOG and rAAV-FLEXFRT-ChR2-mCherry . Plot shows percentage of GFP+ PCs expressing ChR2-mCherry ( 335 cells counted , 15 sections , 4 Gad67-GFP+ animals ) , and percentage of mCherry+ PCs in ZsGreen+ PCs ( 108 cells counted , 8 sections , 2 wildtype animals ) . ( F ) Input resistance and spontaneous firing frequency measured in GFP+ , mCherry+ PCs ( 39 ± 6 MΩ , n = 21; 62 ± 11 Hz , n = 10 ) and GFP+ , mCherry- PCs ( 50 ± 10 MΩ , n = 12; 65 ± 12 Hz , n = 7 ) of rAAV-injected Tg ( GAD67-GFP ) animals , as well as in ZsGreen+/mCherry- PCs of rAAV-injected wildtype animals ( 51 ± 5 MΩ , n = 9; 81 ± 13 Hz , n = 8 ) , were not significantly different ( p = 0 . 433 for comparisons of input resistance , p = 0 . 522 for comparisons of spontaneous frequency , one-way ANOVA ) . Values are listed as mean ± SEM . All plots show median and interquartile range . DOI: http://dx . doi . org/10 . 7554/eLife . 15312 . 017 Next , we tested whether dNbs could be used to detect and isolate live , antigen-expressing cells . ACH-2 ( Folks et al . , 1989 ) , a human T-cell line chronically infected with HIV-1 , is widely used to study HIV-1 persistence ( Clouse et al . , 1989 ) . Destabilized αCA fused to either TagBFP or TagRFP ( Matz et al . , 1999 ) were expressed in ACH-2 or the uninfected parental cell line ( CEM ) , under conditions in which HIV-1 was reactivated with phorbol 12-myristate 13-acetate ( PMA ) ( Poli et al . , 1990 ) ( Figure 6A and Figure 6—figure supplement 1 ) . Using flow cytometry , we detected fluorescence of the destabilized fusions selectively in ACH-2 , but not CEM cells ( Figure 6B–C and Figure 6—figure supplement 1B ) . ACH-2-specific fluorescence was dependent upon CA recognition , as the effect was not observed with dGBP1-TagBFP . Importantly , unmodified αCA Nb-TagBFP fluoresced strongly in both cell lines and could not be used to distinguish between the two ( Figure 6B–C ) . As positive controls , we confirmed that αCA dNb6mut-TagBFP could be stabilized by C-CA co-expression in CEM cells ( data not shown ) and that the HIV-1 CA antigen was specifically detected by immunofluorescence in ACH-2 , but not CEM cells ( Figure 6—figure supplement 1C ) . Thus , dNbs make possible detection of intracellular viral epitopes without the need for cell fixation or membrane permeabilization , enabling live monitoring of intracellular viral protein expression and specific isolation of infected live cells with a choice of spectrally distinct fluorescent proteins . 10 . 7554/eLife . 15312 . 018Figure 6 . Detection of HIV-1 reactivated cells with a CA-specific sensor . ( A ) Schematic showing isolation of HIV-1 cells via flow cytometry using αCA-specific , dNb sensor . Both ACH-2 ( HIV+ ) and CEM ( HIV- ) cells were treated with 10 nM PMA prior to transfection of sensors . CAG-DsRed was a transfection marker . ( B ) Destabilized , but not unmodified αCA Nb enabled selective isolation of reactivated HIV-1 cells using flow cytometry ( P = 0 . 0009 for comparison between destabilized αCA , ACH-2 vs . CEM ) . Plot shows median and maximum-to-minimum range . The number of biological replicates ( equal to number of independent experiments ) for each condition is shown in parentheses . **p<10–3 , Mann-Whitney test . ( C ) Example of flow cytometry gating to isolate HIV-1 cells based on expression of CA . Cell populations are represented as log contour maps . Percentages of DsRed+ cells that were TagBFP+ are indicated for each condition . All cell populations were gated for DsRed expression . Results shown are representative of the number of biological replicates indicated in ( B ) . Additional data related to HIV-1 sensors are shown in Figure 6—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 15312 . 01810 . 7554/eLife . 15312 . 019Figure 6—figure supplement 1 . dNb sensor against HIV-1 CA enables detection and isolation of reactivated HIV-1+ cells with flow cytometry . ( A ) Schematic showing isolation of HIV-1 cells via flow cytometry using HIV-1 CA-specific , red fluorescent sensor . ( B ) Destabilized , but not unmodified αCA Nb enabled selective isolation of reactivated HIV-1 cells using red fluorescence in flow cytometry . Plot shows median and maximum-to-minimum range . 3 independent experiments were performed for each condition , yielding 3 biological replicate values . ( C ) Confirmation of CA immunoreactivity in PMA-stimulated ACH-2 , but not CEM cells . KC57-RD1 is a RD1 dye-conjugated , mouse monoclonal antibody that recognizes CA protein . Cell populations are represented as log contour maps . Unlike αCA-Nb6mut TagBFP , immunostaining for KC57-RD1 requires cell fixation and membrane permeabilization . DOI: http://dx . doi . org/10 . 7554/eLife . 15312 . 019
A key aspect of this study is the demonstration that these Nb-based , conditionally stable reagents were effective even when the experimenter did not choose the antigen levels . This was true for experiments conducted in human cell culture and in mice . Importantly , the detection of HIV-1 CA+ cells was achieved using a dNb that was rapidly generated by mutation transfer rather than by isolation from a screen . In future work , one may create fusion proteins that specifically manipulate or kill infected cells , e . g . by fusing a conditional Nb to a cellular toxin . Studies of model organisms often take advantage of transgenic lines that express an exogenous protein in specific cell types ( Luo et al . , 2008 ) . Driver molecules , such as transcription factors and site-specific recombinases , can respond to the introduction of DNA cassettes to enable the manipulation of gene expression in a cell type-specific manner . Here , we used GFP , which has no naturally known regulatory abilities , as a novel driver molecule . The GFP/dGBP1 binary system is thus analogous to the popular GAL4/UAS , Cre/loxP and Flp/FRT systems . Flp-DOG is immediately useful for studies in model organisms such as the mouse , by making use of existing transgenic GFP reporter lines ( >1000 lines in the mouse ) ( Chalfie , 2009; Gong et al . , 2003; Heintz , 2004; Siegert et al . , 2009; Tang et al . , 2015 , 2013 ) or virally labeled neural circuits ( Beier et al . , 2011; DeFalco et al . , 2001; Ekstrand et al . , 2014; Lo and Anderson , 2011; Schwarz et al . , 2015; Wickersham et al . , 2007 ) for cell-specific manipulation studies . In addition , one can combine GFP and the popular Cre recombinase for intersectional Cre + Flp cell targeting studies ( Dymecki et al . , 2010; Fenno et al . , 2014 ) . Destabilized nanobodies were originally developed with the desire to simplify the delivery of GFP-dependent reagents as well as to improve their performance . Indeed , we observed that the most direct output of the GFP/dGBP1 system – a dGBP1 fusion protein ( dGBP1-TagBFP ) , could be stabilized by YFP with close to 100% efficiency amongst electroporated cells of the retina ( Figure 1—figure supplement 2B and C ) . In comparison , the efficiency of direct reporter output using the T-DDOG and CRE-DOG systems was ~60% in similarly designed experiments , likely due to the need to co-deliver multiple reagents to the same cell ( Tang et al . , 2015 , 2013 ) . Also , whereas the previous dimerizer-based systems were inhibited by excessive amounts of GFP , the activity of Flp-DOG continued to rise along with increasing GFP levels ( Figure 3—figure supplement 1 ) ( Tang et al . , 2015 , 2013 ) . These results highlight the improvements in the GFP/dGBP1 approach compared to previous approaches . However , there are several potential caveats of the GFP/dGBP1 system . First , although Flp-DOG is relatively easier to deliver to tissues as a single-component coding sequence , enhanced construct delivery may lead to enhanced background activity . Indeed , we found that it was necessary to remove the woodchuck hepatitis virus post-transcriptional regulatory element ( WPRE ) sequence from the rAAV expression cassette in order to avoid background Flp-DOG activity with rAAV infection . Second , tandemly repeated dGBP1s were fused to Flpo in order obtain tight GFP-dependent recombination . The requirement for two GFP molecules to stabilize a single Flp-DOG construct is predicted to reduce the sensitivity of Flp-DOG for lower GFP expression levels . However , in practice , Flp-DOG had adequate activity for the targeting of GFP cell types of the two different transgenic GFP lines tested . Nevertheless , these considerations suggest that one should establish an appropriate level of construct delivery , such as the amount of DNA plasmid or virus to deliver for optimal Flp-DOG performance . Beyond fluorescent proteins , endogenous proteins should be usable as driver molecules to trigger sensor or effector activity . Endogenous proteins would enable one to selectively target specific cell types in wildtype animals for experimentation , without requiring any knowledge of cell type-specific promoters or creation of knock-in alleles . Such an approach would especially benefit studies of non-model organisms , with the only demand being a method to introduce genetic constructs , e . g . via viral vectors . As a possible caveat , the range of biological activities that could be driven by an endogenous protein may be limited by the protein’s natural functions and/or sub-cellular localization . For example , it may be problematic to use a membrane-localized protein to trigger DNA recombination events in the nucleus . One may overcome this by choosing a dNb-fusion protein that , when stabilized , is biologically active in the same sub-cellular localization as that of the antigen . Another possibility is that an antigen-stabilized complex may travel from the site where stability is conferred to another sub-cellular locations to exert its function . These possibilities will likely be idiosyncratic to the fusion protein and antigen . An additional caveat is the effect that a dNb fusion protein might have on the targeted endogenous protein . There could be cases where antigen-Nb interactions lead to less antigen activity . Reduction in activity will depend on several variables . The ratio of Nb fusion to antigen , the binding site of the Nb fusion , and perhaps the particular Nb fusion structure , all have the potential to create changes in antigen activity . Whether or not a reduction in cellular function will be effected will depend upon the sensitivity of the cell to the level of antigen activity . In most heterozygous loss of function mutations in mice and humans , a phenotype is not noted . Although we cannot predict the frequency with which Nb fusions will result in a phenotype , we believe that the method is strong enough to encourage its continued development and application . Although dGBP1-TagBFP showed virtually no background signal , fusions of dGBP1 to some fusion partners gave significant background signals . Background signals could be addressed by simply increasing the number of dNbs fused to the protein partner . Additional engineering efforts could further reduce background activity of particular fusion constructs . For example , background activity of dGBP1-Cre could be further controlled by fusion to an ERT2 domain to create small-molecule dependency ( Feil et al . , 1997 ) . Lastly , one could perform additional screens to isolate novel destabilizing mutant combinations that enhance the antigen-specificity of a wider variety of sensor/effector fusion partners . Such an approach could help eliminate the background fluorescent aggregates seen with some dNb-fluorescent protein fusions . Additional improvements may be made to enhance the response of dNb fusion constructs to antigen co-expression . First , the activity of a protein might be affected by fusion to the Nb . For example , Cre activity was reduced upon binding of GFP to GBP1-Cre , possibly as a result of steric hindrance . Second , the sensitivity of the dNb fusion construct might be sub-optimal . Although the proportion of CA+ cells detected by the dNb sensor was approximately 1/3 to 1/2 of that detected by a mouse monoclonal antibody , the efficiency of antigen detection may be improved by optimization of the dNb fusion construct or of the gene delivery protocol . Possible optimization steps may include exploring different fusion orientations , linker lengths or linker compositions . The fusion of protein binders to fluorescent proteins enables visualization of antigen localization in living cells . Optimal signal-to-noise detection requires that the fluorescent fusion proteins be strictly co-localized with the antigen . This may not occur if the number of fluorescent fusion proteins exceeds the number of target antigens . One could address this by designing a transcriptional feedback mechanism to control the level of a fluorescent fusion protein ( Gross et al . , 2013 ) . This method requires that the antigen be localized outside of the nucleus . The use of dNb fluorescent protein fusions is not limited by this requirement , as the mechanism for background reduction involves protein degradation rather than transcriptional feedback . Indeed , we found that a dNb-TagBFP construct became strictly localized to the nucleus upon co-expression with its NLS-tagged antigen ( data not shown ) . The finding that one can fuse an effector protein to two dNbs led to the development of a strategy wherein two different antigens bound to distinct dNbs could stabilize the effector protein , Flpo . With further development , dual antigen dependence may enable one to precisely target specific cell populations in ways similar to established intersectional strategies , but using proteins that do not necessarily have any defined regulatory abilities ( Dymecki et al . , 2010; Luo et al . , 2008 ) . Here , we demonstrated the ability to integrate dNb with CRISPR/Cas technology to perform genome editing selectively in antigen-expressing cells . A concern with the expression of Cas9 and gRNA in cells is that there is non-specific genome editing , and several methods are being developed to address this problem ( Hsu et al . , 2014; Sander and Joung , 2014 ) . The strategy developed here , wherein dNb-Cas9 activity is suppressed until antigen can stabilize the fusion protein , offers a novel strategy to restrict the cell populations that might suffer from off-targeting events . The screening strategy described here should enable the creation of additional protein-responsive reagents useful for control of fusion protein activity in specific cell populations . A key feature of our screen is the use of rAAVs to deliver the antigen to cells . Virtually all MMLV-infected cells in culture can be super-infected by rAAV , and the cells remain viable for subsequent culture expansion and FACS . In principle , this screening strategy can be extended to generate a diversity of protein-responsive , destabilized binders based on the Nb scaffold or other protein scaffolds ( Wurch et al . , 2012; Helma et al . , 2015 ) . Over the past 20 years , ~100 crystal structures featuring Nb-antigen complexes have been solved . We leveraged this resource to establish a phylogenetic and structural basis for transferring destabilizing mutations across Nbs . All successfully modified Nbs were derived from camelid species different from that of GBP1 , demonstrating the broad transferability of the mutations discovered here . As one would expect , dNb-TagBFPs generated by mutation transfer showed a spectrum of fluorescence fold-change in response to antigen co-expression . This is likely due to multiple factors , including variable Nb affinity for antigen , variable antigen stability , and variable Nb stability even before destabilization . In addition , Nbs might have variable tolerance to the destabilization mutations tested . Thus , although the high percentage of successful mutation transfers indicate that the strategy is generally applicable , it would be beneficial to derive novel combinations of destabilizing mutations that are even better tolerated across Nbs . Lastly , although dNb generation may be limited by the availability of Nbs isolated from immunized animals , additional Nbs and dNbs may be isolated from in vitro screening technologies that are constantly being improved upon . Beyond Nbs , multiple classes of artificially derived binding proteins that are amenable to expression in living cells are being developed for antigen-recognition ( Helma et al . , 2015; Wurch et al . , 2012 ) . As epitope-specific binders are typically generated by varying loops or surfaces on a common structural scaffold , it should be possible to generate epitope-responsive properties by incorporating a common set of mutations onto conserved and non-epitope binding regions of the scaffold . Future developments building upon this work should expand our ability to rapidly generate sensors and effectors against a diversity of intracellular epitopes , for cell- or antigen-specific applications in biology and medicine .
The Institutional Animal Care and Use Committee at Harvard University approved all animal experiments . Timed pregnant CD1 ( Charles River Breeding Laboratories , Boston , MA ) were used for electroporation experiments . Tg ( CRX-GFP ) ( Samson et al . , 2009 ) and Tg ( GAD67-GFP ) ( Tamamaki et al . , 2003 ) were kept on a C57/BL6J background . pCAG-GFP ( Addgene plasmid 11150 ) ( Matsuda and Cepko , 2004 ) . pCAG-YFP ( Addgene plasmid 11180 ) ( Matsuda and Cepko , 2004 ) . pCAG-DsRed ( Addgene plasmid 11151 ) ( Matsuda and Cepko , 2004 ) . pRho-GFP-IRES-AP ( referred to as Rho-GFP ) ( Emerson and Cepko , 2011 ) . pCAG-nlacZ ( Cepko lab , Harvard Medical School ) pCAGEN ( Addgene plasmid 11160 ) ( Matsuda and Cepko , 2004 ) . pCALNL-DsRed ( Addgene plasmid 13769 ) ( Matsuda and Cepko , 2004 ) . pCAFNF-DsRed ( Addgene plasmid 13771 ) . ( Matsuda and Cepko , 2004 ) . pCALNL-luc2 ( Tang et al . , 2015 ) . pRL-TK ( #E2241; Promega , Madison , WI ) . Antibodies used were rabbit anti-TagRFP ( also targets TagBFP; 1:5 , 000 dilution for immunoblot , 1:1 , 000 for immunohistochemistry ) ( AB233; Evrogen , Moscow , Russia ) , mouse anti-βgal ( 1:50 for immunoblot ) ( 40-1a supernatant; Developmental Studies Hybridoma Bank , University of Iowa ) , chicken anti-βgal ( 1:1 , 000 for immunohistochemistry ) ( ab9361;Abcam , Cambridge , MA ) , rabbit-anti-GFP ( 1:500 for immunohistochemistry ) ( A-6455; Invitrogen , Carlsbad , CA ) , rabbit anti-GAPDH ( 1:10 , 000 for immunoblot ) ( A300-641A; Bethyl Laboratories , Inc . , Montgomery , TX ) , mouse anti-FLAG M2 ( 1:1 , 000 for immunoblot ) ( F1804; Sigma-Aldrich , St . Louis , MO ) , mouse anti-KC57-RD1 ( 5ul per 1 million cells ) ( 6604667; Beckman Coulter , Danvers , MA ) . Secondary antibodies used were goat anti-chicken Alexa Fluor 647 ( 1:500 of 50% glycerol stock ) ( 102371; Jackson ImmunoResearch Laboratories Inc . , West Grove , PA ) , goat anti-rabbit DyLight 649 ( 1:500 of 50% glycerol stock ) ( 111-495-144; Jackson ImmunoResearch Laboratories Inc . ) , anti-rabbit or anti-mouse IgG-Horseradish Peroxidase ( GE Healthcare , Little Chalfront , UK ) . 293T cells ( Cepko lab stock ) . ACH-2 ( NIH AIDS Reagent Program , Germantown , MD ) . CEM ( NIH AIDS Reagent Program ) . HIV-1 integration sites in ACH-2 cell line were authenticated by integration site analysis of HIV-1 genome , confirming the major integration site is on chromosome 7 . Using immunostaining followed by flow cytometry analysis , HIV-1 CA protein was confirmed to be present in ACH-2 cells , and to be absent in CEM cells . All antigen and Nb protein sequences , except YFP , were acquired from Protein Data Bank ( PDB ) . Protein sequences were backtranslated into DNA sequences , using codons optimized for Mus musculus . The list of tested Nbs and their antigens are listed in Table 1 . In general , an antigen sequence was synthesized as gBlock fragments , which were inserted into an EcoRI/NotI digested pCAG vector via Gibson Assembly , giving pCAG-antigen plasmids used for co-expression of antigen in cells . In general , Nb sequences were synthesized as gBlock fragments , and individually inserted into an EcoRI/SphI digested pCAG-TagBFP vector via Gibson Assembly , giving pCAG-Nb-TagBFP plasmids . To destabilize Nbs , mutations were introduced into residue positions that aligned with the dGBP1 mutation positions . Equivalent residues were easy to identify since surrounding amino acid sequences were highly conserved . For 6mut combo , the dGBP1 mutations were A25V , E63V , S73R , C/S98Y , Q109H and S117F . For 3maj combo , the dGBP1 mutations were S73R , S98Y and S117F . Note that C/S98Y in GBP1 was originally a cysteine , but was mutated to serine in earlier studies to avoid complications with disulfide bond formation . pBMN-GBP1-TagBFP - A GBP1-TagBFP construct was inserted into a BamHI/NotI digested pBMN DHFR ( DD ) -YFP ( a gift from Thomas Wandless; Addgene plasmid # 29325 ) ( Iwamoto et al . , 2010 ) , replacing the DHFR ( DD ) -YFP insert and generating pBMN-GBP1-TagBFP . This became the host vector for mutagenized GBP1 inserts . pBMN-dGBP1-Cre and pBMN-dGBP1-Flpo –pBMN-dGBP1-TagBFP were digested with SphI/SalI , liberating TagBFP as well as the IRES-t-HcRed element . PCR-amplified Cre and Flpo fragments were then inserted into the digested vector via Gibson Assembly . pBMN-GBP1-Cre and pBMN-GBP1-Flpo – PCR fragments of GBP1 were inserted into BspEI/SphI digested pBMN-dGBP1-Cre and pBMN-dGBP1-Flpo by Gibson Assembly , resulting in pBMN-GBP1-Cre and pBMN-GBP1-Flpo , respectively . dGBP1 sequence was removed by BspEI/SphI digest . pBMN-dGBP1x2-Cre , pBMN-dGBP1-GBP1-Cre , pBMN-dGBP1x2-Flpo , pBMN-dGBP1-GBP1-Flpo – pBMN-dGBP1-Cre or –Flpo plasmids were digested with SphI . A gBlock fragment encoding a codon modified dGBP1 was inserted into this site via Gibson Assembly , generating pBMN-dGBP1x2-Cre or –Flpo . Using a GBP1 gBlock fragment instead of dGBP1 gave pBMN-dGBP1-GBP1-Cre or –Flpo . pCAFNF-luc2 – An EcoRI-Kozak-luc2-NotI DNA fragment liberated from pCALNL-luc2 ( Tang et al . , 2015 ) was sub-cloned into EcoRI/NotI digested pCAFNF-DsRed vector , giving pCAFNF-luc2 . pCAG-dGBP1-TagBFP – Using PCR , an AgeI-Kozak-dGBP1-TagBFP-NotI was generated from pBMN-dGBP1-TagBFP . This fragment was sub-cloned into AgeI/NotI digested pCAG-GFP , giving pCAG-dGBP1-TagBFP and removing GFP from the construct . pCAG-dGBP1-TagBFP-FLAG – A gBlock fragment encoding Kozak-TagBFP-FLAG was inserted into SphI/NotI digested pCAG-dGBP1-TagBFP via Gibson Assembly , giving pCAG-dGBP1-TagBFP-FLAG and removing untagged TagBFP from the construct . pCAG-YFP-FLAG – A gBlock fragment encoding Kozak-YFP-FLAG was inserted into EcoRI/NotI digested pCAG-αCA-dNb6mut-TagBFP , giving pCAG-YFP-FLAG and removing αCA dNb6mut-TagBFP from the construct . pCAG-dGBP1-mCherry – PCR amplified mCherry was inserted into a SphI/NotI digested pCAG-dGBP1-TagBFP vector , resulting in replacement of TagBFP with mCherry . The vector became pCAG-dGBP1-mCherry . pCAG-GBP1-mCherry – A gBlock fragment encoding GBP1 was inserted into a EcoRI/SphI digested pCAG-dGBP1-mCherry vector , resulting in replacement of dGBP1 with GBP1 . The vector became pCAG-GBP1-mCherry . pCAG-αCA-Nb-TagBFP , pCAG-αDHFR-Nb-TagBFP , pCAG-αCA-dNb6mut -TagBFP and pCAG-αCA-dNb3maj-TagBFP –A gBlock fragment carrying either the αCA Nb or αDHFR Nb coding sequence was inserted into an EcoRI/SphI digested pCAG-TagBFP vector via Gibson Assembly , resulting in pCAG-αCA-Nb-TagBFP or pCAG-αDHFR-Nb-TagBFP . gBlocks carrying these mutations in the respective Nbs were introduced into the EcoRI/SphI digested pCAG-TagBFP vector via Gibson Assembly , giving either pCAG-αCA-dNb6mut-TagBFP or pCAG-αDHFR-dNb3maj-TagBFP . pCAG-dGC-Flpo and pCAG-dGD-Flpo – A gBlock fragment carrying either αCA-dNb6mut or αDHFR-dNb3maj coding sequence were inserted into SphI digested pCAG-dGBP1-Flpo vector ( Tang , J . C . Y . , Cepko lab ) via Gibson Assembly , giving either pCAG-dGC-Flpo or pCAG-dGD-Flpo , respectively . pCAG-dGBP1x2-Flpo – An AgeI-Kozak-dGBP1x2-Flpo-NotI fragment was generated by PCR using pBMN-dGBP1x2-Flpo as a template . This fragment was sub-cloned into AgeI/NotI-digested pCAG vector , giving pCAG-dGBP1x2-Flpo . pCAG-αCA-dNb6mutx2-Flpo – Two gBlock fragments , together encoding αCA-dNb6mutx2 , was inserted into EcoRI/SphI-digested pCAG-dGBP1x2-Flpo , giving pCAG-αCA-dNb6mutx2-Flpo and replacing dGBP1x2 from the construct . pCAG-αCA-dNb6mut-TagRFP – A gBlock fragment carrying the TagRFP coding sequence was inserted into SphI/NotI digested pCAG-αCA-dNb6mut-TagBFP via Gibson Assembly , giving pCAG-αCA-dNb6mut-TagRFP and removing TagBFP from the construct . pAAV-EF1α-dGBP1x2-Flpo-NW- A BamHI-Kozak-dGBP1x2-Flpo-EcoRI PCR fragment was inserted into BamHI/EcoRI digested pAAV-EF1α-N-CretrcintG ( Tang et al . , 2015 ) , giving pAAV-EF1α-dGBP1x2-Flpo . The WPRE element was subsequently removed from this plasmid via EcoRV/AfeI digest and re-ligation , giving pAAV-EF1α-dGBP1x2-Flpo-NW . pAAV-CAG-FLEXFRT-ChR2 ( H134R ) -mCherry- A Chr2 ( H134R ) -mCherry DNA fragment was digested with NheI and inserted into NheI digested pAAV-CAG-FRTed-SynGFPreverse-WPRE ( Pivetta et al . , 2014 ) ( a gift from Sylvia Arber ) ( Pivetta et al . , 2014 ) . A clone with ChR2 ( H134R ) -mCherry inserted in the reverse direction relative to CAG promoter were chosen , giving pAAV-FLEXFRT-ChR2 ( H134R ) -mCherry . pCAG-C-CA and pCAG-DHFR– A gBlock fragment carrying either the HIV-1 C-CA coding sequence ( encoding residue 278–352 of HIV-1 gag polyprotein ) or E . coli DHFR coding sequence was inserted into EcoRI/NotI digested pCAG-GFP vector via Gibson Assembly; C-CA or DHFR replaced GFP in the cassette . For sample size , we chose to reproduce all our results in at least 3 independent experiments ( equal to at least 3 biological replicates , or transfected wells ) . We consider this to be a sufficient sample size for demonstrating reproducibility of our findings . For statistical analysis , we chose to increase our independent experiments ( equal to biological replicates ) to 8 . 293T cells were seeded onto 24 or 96 well plates and used for transfection when the cells reached between 60–95% confluency , usually 1–2 days later . Transfections were achieved with polyethyleneimine ( PEI ) at a 1:4 DNA mass:PEI volume ratio . PEI stock was 1 mg/ml . A total of between 100 and 400 ng of DNA were transfected into single wells of 96 well plates for fluorescence analysis of destabilized mutants . Approximately 70 ng total DNA were transfected into single wells of 96 well plates for luciferase analysis . Approximately 400 to 520 ng of total DNA were transfected into single wells of 24 well plates for fluorescence imaging and western blot analysis . We exhaustively searched the Protein Data Bank ( PDB ) and , at the time of analysis , identified 77 unique camelid single-chain antibody fragments ( VHH or VH , here collectively referred to as nanobodies ( Nbs ) ) that have been co-crystallized with their respective antigens . We removed one structure ( PDB ID 3J6A ) from the analysis because it was a low-resolution structure produced by cryo-electron microscopy . We used the PDBePISA online server tool ( http://www . ebi . ac . uk/pdbe/pisa/ ) to evaluate whether residue positions equivalent to those of dGBP1 mutations are located outside of antigen-Nb interfaces across different crystallized complexes . PDBePISA produces an analysis of buried surface area ( BSA ) , defined as the solvent-accessible surface area of the corresponding residue that is buried upon interface formation , in Å2 . We considered an Nb residue to be in the interface with the antigen if its BSA value is above 0 Å2 . We confirmed that this metric is a reliable indicator of an Nb residue’s proximity to the antigen by examining all structures by eye using tools such as PyMol . In a few cases the Nb bound to more than one antigen . We took this into consideration by analyzing the interfaces formed between an Nb and each of the two antigens . We used protein alignment to determine the residue positions corresponding to the mutations located in dGBP1 . The same 76 Nbs were used to determine the extent of GBP1 residue conservation across Nbs . Analysis across 76 unique Nb-antigen interfaces , and a total of 102 uniquely crystallized interfaces , indicated that Nb positions corresponding to those of dGBP1 A25V , S73R , S98Y and S117F were universally located outside of all Nb-antigen interfaces ( Figure 2—figure supplement 3B ) . Nb positions corresponding to dGBP1 Q109H fell outside of 99% , or 75 of 76 unique Nb-antigen interfaces . Positions equivalent to dGBP1 E63V were directly found in 22% , or 17 of 76 unique Nb-antigen interfaces , and in close proximity to the interface in 9% , or 7 of 76 of the cases . 16 identical or highly similar Nb-antigen complexes had been crystallized under similar or differing conditions , allowing us to validate the mutation mapping results by comparing across identical or related crystal structures . There was agreement in mutation mapping results between redundant or similar crystal structures for 94% , or 16 of 17 unique Nb-antigen interfaces . The lone exception concerned a unique Nb ( PDB ID 4KRM and 4KRL ) . The Q109H equivalent position in the Nb of structure 4KRL was identified to be at the interface . This discrepancy may be explained by the fact that the two structures were crystallized under very different pH conditions . For the mutation transfer experiment , we tested 18 Nbs that , when fused to TagBFP , showed strong blue fluorescence and soluble , diffuse localization in 293T cells . Some available nanobodies were too problematic to be included for analysis because they recognized problematic antigens such as the Ricin toxin . Nevertheless , Nbs that bound to proteins originating from both intracellular and extracellular locations were selected . We used a pCAG expression vector to express antigen deposited in PDB for each crystal structure . During analysis , we noticed a strong correlation between dNbs that failed to be stabilized by antigen ( <2-fold TagBFP fluorescence induction by antigen ) and the use of extracellular epitopes . dNbs targeting extracellular epitopes were thus excluded from evaluation of mutation transfer generality . To determine whether the antigen used for TagBFP stabilization assays derived from intracellular or extracellular proteins , we studied the annotation and literature reports of each antigen’s cellular localization . In all in vivo experiments , biological replicates are defined in terms of cells , retinas or animals . Technical replicates are defined in terms of whole brain sections . We consider 3 biological replicates to be a sufficient sample size for demonstrating reproducibility of our findings . As an exception , we had 2 biological replicate for injection of green fluorescent beads/rAAV mix into GFP and wildtype mouse brains ( Figure 5—figure supplement 2 ) . However , we deemed this sufficient as the results basically replicated our findings in an equivalent experiment using a slightly different injection mix ( Figure 6 ) . For statistical analysis , we used data that consisted of 7–21 cells . rAAV ( 2/1 ) virus preparations were made from pAAV-EF1α-dGBP1x2-Flpo-NW and pAAV-CAG-FLEXFRT-ChR2-mCherry . All rAAVs were injected in the range of 1013–1014 genome copies/ml , assayed by PCR of rAAV vectors at Boston Children’s Hospital ( Zhigang He lab ) . Primers targeted the ITR region of AAV vectors , and were: Forward - 5’-GACCTTTGGTCGCCCGGCCT-3’ , Reverse - 5’-GAGTTGGCCACTCCCTCTCTGC-3’ . Note that we found this titering method gave about 10-100 fold higher numerical value in titer than other titering methods . The human LoxP-LacZ cell line was obtained from Allele Biotech ( San Diego , CA ) ( SKU: ABP-RP-CLACLOXE ) , and cultured as instructed in the product manual . Cas9 activity was assessed by detecting βgal-expressing cells in wells transfected with pX330-dCC-Cas9 and either pCAG-C-CA or pCAG-GAPDH-AU1 control construct ( simply called AU1 in the main text ) . In addition , pCAG-mCherry is included as a transfection marker . For X-gal staining , cells were fixed on ice with 0 . 5% Glutaraldehyde for 5min . X-gal staining was performed as previously described . Cells were left at room temperature overnight for color development . Images were acquired by Keyence BZ9000 microscope . The number of mCherry+ and X-gal+ cells was quantified by Fiji software . The normalized Cas9 activity is calculated by dividing individual replicate values of specific conditions by the average number of X-gal+ cells induced by pX330-loxPgRNA alone . | Biologists often wish to study the role of a particular cell type within an organism , but such studies are often not possible due to the lack of reagents that allow one to gain control of the cell type of interest . One method that can be used to detect and manipulate the cells that express specific proteins uses molecules called antibodies . An antibody can strongly bind to a specific part of a protein , and a diversity of antibodies that bind to different proteins can be isolated by animal immunization , or by using molecular or cell-based methods . Antibodies from camelid species ( which include camels and llamas ) are increasingly being used to detect and manipulate proteins in living cells . The variable region of these antibodies – also known as the nanobody – recognises the proteins that the antibody binds to , and often just this fragment of the antibody is used in protein detection experiments . However , nanobodies are stable even in cells that do not contain their target proteins , which makes it difficult to use nanobodies to study just a specific cell type within an organism . Tang , Drokhlyansky et al . have now developed a way of engineering the sequence of a nanobody so that it is broken down in living cells unless it is bound to its protein target inside the cell . Any protein that is tethered to the engineered nanobody is also broken down . For example , some tethered proteins with useful biological activities are fluorescent proteins and enzymes that can modify DNA . When one of these engineered nanobodies binds to a protein target of interest , the activity of the nanobody-tethered protein can be turned on in just those cells that produce the targeted protein . Thus , this strategy of engineering allows “conditionally stable” tools to be generated . A core set of sequence alterations can be used to modify different nanobodies that target different proteins . Tang , Drokhlyansky et al . have demonstrated the uses of several of the resulting conditionally stable nanobodies . In one application , the nanobodies were used to target specific cell types in the mouse brain in a way that allowed the activity of these cells to be controlled by light . Another application of the technique enables live human cells that have been infected with HIV to be detected and isolated . The conditionally stable nanobody tools can be used to detect and manipulate cells that express any protein for which a camelid antibody exists . Tang , Drokhlyansky et al . therefore hope that biologists who work in a wide range of fields will find the tools useful for studying many different types of organisms and biological processes . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"neuroscience"
] | 2016 | Detection and manipulation of live antigen-expressing cells using conditionally stable nanobodies |
Cell migration and polarization is controlled by signals in the environment . Migrating cells typically form filopodia that extend from the cell surface , but the precise function of these structures in cell polarization and guided migration is poorly understood . Using the in vivo model of zebrafish primordial germ cells for studying chemokine-directed single cell migration , we show that filopodia distribution and their dynamics are dictated by the gradient of the chemokine Cxcl12a . By specifically interfering with filopodia formation , we demonstrate for the first time that these protrusions play an important role in cell polarization by Cxcl12a , as manifested by elevation of intracellular pH and Rac1 activity at the cell front . The establishment of this polarity is at the basis of effective cell migration towards the target . Together , we show that filopodia allow the interpretation of the chemotactic gradient in vivo by directing single-cell polarization in response to the guidance cue .
Cell migration is essential for tissue and organ development , for tissue homeostasis and function . The process of migration is also at the basis of pathological conditions such as inflammation and cancer metastasis ( Friedl and Gilmour , 2009 ) . Migrating cells generate finger-like cellular extensions that contain parallel actin-bundles ( Mallavarapu and Mitchison , 1999; Yang et al . , 2009 ) , termed filopodia , that constitute a characteristic feature of metastasizing cells ( Machesky , 2008; Arjonen et al . , 2011; Pan et al . , 2011 ) . It has been proposed that filopodia transport signalling molecules to neighbouring cells ( Sanders et al . , 2013; Roy et al . , 2014 ) , promote adhesion and are therefore important for generation of traction forces ( Albuschies and Vogel , 2013; Fierro-González et al . , 2013 ) and serve as a sensing organelle ( reviewed in [Wood and Martin , 2002] ) . The latter notion arose from functional studies that were carried out in vitro and involved manipulations that could have affected different processes in addition to filopodia formation in chemotrophic growth cones ( Zheng et al . , 1996; Rajnicek et al . , 2006 ) . Other studies demonstrated an EGF retrograde transport on protrusions of epidermoid carcinoma cells in culture ( Lidke et al . , 2005 ) , although the functional significance of the findings was not explored . Consistent with a role in sensing , filopodia formation was observed in response to VEGF that directs angiogenic sprouting ( Gerhardt et al . , 2003 ) , to FGF that instructs branching morphogenesis in the tracheal system ( Ribeiro et al . , 2002 ) , as well as in response to morphogen signals in Drosophila ( Roy et al . , 2011 ) . In the context of group cell migration , inhibiting filopodia formation decreased the migration velocity , yet the cellular basis for this effect has not been further investigated ( Phng et al . , 2013 ) . Similarly , it was suggested that the migration of neural crest cells as streams require filopodia function , since a neuronal crest subset failed to migrate properly in zebrafish mutants that lacked the fascin1a gene whose actin bundling function is required for filopodia formation ( Boer et al . , 2015 ) . Nevertheless , the precise consequence of impaired filopodia formation in migrating single cells in vivo and the mechanism underlying their action during normal migration in the context of the intact tissue have thus far not been reported . As a useful in vivo model for exploring the regulation and function of filopodia in cell migration , we employed zebrafish Primordial germ cells ( PGCs ) . These cells perform long-range migration as single cells within a complex environment from the position where they are specified towards their target ( Richardson and Lehmann , 2010; Tarbashevich and Raz , 2010 ) . PGC migration is guided by the chemokine Cxcl12a that binds Cxcr4b , which is expressed on the surface of these cells ( Doitsidou et al . , 2002; Knaut et al . , 2003 ) . This specific receptor-ligand pair has been shown to control among other processes , stem-cell homing ( Chute , 2006 ) , cancer metastasis ( Zlotnik , 2008 ) and inflammation ( Werner et al . , 2013 ) . Interestingly , similar to other migrating cells types in normal and disease contexts , zebrafish PGCs form filopodia , protrusions whose precise function in guided migration has thus far not been characterized . We show here that in response to Cxcl12a gradients in the environment , filopodia exhibit polar distribution around the cell perimeter and alter their structural and dynamic characteristics . We demonstrate that PGCs guided by Cxcl12a form more filopodia at the cell front , filopodia that exhibit higher dynamics and play a critical role in receiving and transmitting the polarized signal . Specifically , we show that the short-lived actin-rich filopodia formed at the front of cells migrating within a Cxcl12a gradient are essential for conferring polar pH distribution and Rac1 activity in response to the guidance cue , thus facilitating effective cell polarization and advance in the correct direction . Together , these results provide novel insights into the role of filopodia in chemokine-guided single cell migration , underlining their function in orienting cell migration .
Guided towards their target by the chemokine Cxcl12a , zebrafish PGCs generate blebs primarily at the cell aspect facing the migration direction ( Reichman-Fried et al . , 2004 ) . To define the mechanisms that could contribute to the apparent polarity of migrating PGCs , we first measured the distribution of Cxcr4b on the cell membrane around the cell perimeter . Similar to findings in Dictyostelium discoideum cells , in which the guidance receptor is evenly distributed around the cell membrane ( Ueda et al . , 2001 ) and consistent with our previous results ( Minina et al . , 2007 ) , the level of a GFP-tagged Cxcr4b ( expressed at low amounts that do not affect the migration ) measured at the cell front and its back was similar ( Figure 1A ) . Furthermore , the receptor turnover on the plasma membrane , as visualized by a Cxcr4b tandem fluorescent timer ( tft ) ( Khmelinskii et al . , 2012 ) expressed in PGCs ( Figure 1—figure supplement 1A–E ) , which are directed by the endogenous Cxcl12a gradient ( Figure 1B ) , did not reveal a significant difference between the front and the back of the cell . Together , employing the tools described above , we could not detect an asymmetric receptor distribution or differential turnover around the cell perimeter of PGCs in the wild type situation . These findings prompted us to explore qualitative and quantitative differences between the cell front and back , specifically by examining cellular structures that could contribute to the polarity of Cxcr4b signalling . 10 . 7554/eLife . 05279 . 003Figure 1 . In wild type embryos the Cxcr4b receptor is uniformly distributed on the migrating PGC membrane , and its turnover is uniform around the cell circumference . ( A ) A graph showing the Cxcr4b-GFP protein level measured at the front and the back ( normalized to the mCherry-F' ) of individual migrating PGCs under conditions of endogenous Cxcl12a distribution in the embryo ( A , n = 18 ) . A representative cell is shown , with the areas defined as front and back indicated by lines and the arrows designate the migration direction . Scale bars signify 10 µm . ( B ) A graph showing the protein age ( lifetime ratio , see Figure 1—figure supplement 1 ) measured at the front and the back of individual migrating PGCs under conditions of endogenous Cxcl12a distribution in the embryo ( n = 20 ) . A representative cell is shown , with the areas defined as front and back indicated by lines and the arrows designate the migration direction . Scale bars signify 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 00310 . 7554/eLife . 05279 . 004Figure 1—figure supplement 1 . Functionality of the Cxcr4b tandem fluorescent timer ( tft ) in the context of PGC migration . ( A ) Schematic representation of the cxcr4b tft RNA , showing the cxcr4b open reading frame cloned upstream to the slow maturing mCherry and to the fast maturing sfGFP , followed by nanos3′UTR that drives preferential expression of the protein within the PGCs . The Red to Green intensity ratio reflects the protein age such that the higher the ratio , the more mature the protein is . ( B ) The Cxcr4b tft protein can guide PGCs to their target in embryos lacking functional Cxcr4b ( ody−/− ) . Arrowheads point at PGCs found in ectopic positions in 24 hr post fertilisation ( hpf ) ody−/− mutants ( left panel ) and at the target position following introduction of cxcr4b tft RNA ( right panel ) . ( C ) The Cxcr4b tft protein expressed in PGCs is internalized upon exposure to Cxcl12a ( right panel ) . The functionality of the Cxcr4b tandem fluorescent timer ( tft ) is evidenced by the fact that despite the uniform distribution of Cxcr4b on the cell membrane ( D , n = 14 ) , cells located within an artificially-generated steep Cxcl12a gradient exhibit reduced age at their cell front ( E , n = 18 ) . Scale bars signify 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 004 As zebrafish PGCs migrate , they extend protrusions on their circumference that appear like filopodia ( Ahmed et al . , 2010 ) . To determine the role these protrusions play in the process of bleb-dependent single cell migration in vivo , we first studied their characteristics in the course of PGC migration within the three dimensional ( 3D ) cellular environment of the embryo . Since filopodia are defined as F-actin containing cellular extensions ( Mallavarapu and Mitchison , 1999 ) , we followed the distribution of F-actin using an Lifeact-EGFP fusion protein . While F-actin was detected in all of the filopodia formed at the cell front , about 50% of those localized to the back of the cell showed no detectable actin signal ( Figure 2—figure supplement 1A ) , pointing at differences in actin content between the two populations of filopodia . In the filopodia formed by PGCs migrating in embryos lacking Cxcl12 ( medNY054 , [Valentin et al . , 2007] ) , we observed a higher proportion of filopodia in which actin is not detected ( Figure 2—figure supplement 1B ) . Filopodia-like structures devoid of F-actin were previously described ( Phng et al . , 2013; Yang et al . , 2009 ) . To characterize the distribution of filopodia in the migrating cells , we imaged PGCs expressing farnesylated EGFP ( EGFP-F' ) . A characteristic wild type PGC is presented in Figure 2A and Video 1 , showing polarized formation of an array of filopodia pointing in the direction of migration ( arrow in Figure 2A ) . For a quantitative measure of this finding , we categorized the position of filopodia around the circumference of the PGCs as front , back or side filopodia as illustrated in Figure 2—figure supplement 2A–A‴ . We found that at any given time point , migrating PGCs show on average 5 . 6 ± 0 . 5 filopodia , with a strong bias for filopodia generation at the cell front ( wild type in Figure 2C , D ) . Whereas filopodia formed preferentially at the cell front , their distribution was independent of that of the forming blebs . Specifically , we found that blebs can form in parts of the cells where filopodia are not present ( Figure 2—figure supplement 3A , B , red asterisk in 40″ , Video 2 ) and conversely , filopodia were maintained close to a forming bleb ( Figure 2—figure supplement 3B , arrowheads in 20″ , Video 2 ) . 10 . 7554/eLife . 05279 . 005Figure 2 . The polar positioning and number of filopodia are determined by the Cxcl12a gradient . ( A ) A PGC extending filopodia during migration in wild type embryos ( Video 1 ) , and ( B ) in medusa ( medNY054 ) mutant embryos lacking Cxcl12a ( Video 3 ) . ( C ) Filopodia distribution and number in PGCs migrating within wild type and medNY054 homozygous embryos ( for segmentation see Figure 2—figure supplement 2C for more details ) . ‘n’ indicates the number of cells analysed . ( D ) Examples of a wild type ( left ) and a medNY054 PGC in which the characteristic distribution of the filopodia is indicated by arrowheads . ( E ) Filopodia number and distribution in PGCs migrating within in medNY054 homozygous embryos knocked down for Cxcr7b and that express either uniform levels of control RNA ( light bars ) or Cxcl12a-encoding RNA ( dark bars , see also Video 4 ) . ( F ) Examples of PGCs migrating within an environment lacking ( left ) , or containing uniform Cxcl12a ( right ) in which the characteristic distribution of the filopodia is indicated by arrowheads . In A , B , D and F the membrane of the PGCs is labelled with EGFP-F' and arrows indicate direction of movement . Scale bar is 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 00510 . 7554/eLife . 05279 . 006Figure 2—figure supplement 1 . F-actin content in filopodia extended by PGCs . ( A ) A graph showing the frequency of F-actin-containing filopodia in PGCs of transgenic embryos co-expressing Lifeact-EGFP and mCherry-F' . In the cell presented on the right F-actin can be detected in the filopodia formed at the front ( filled arrowheads ) , while the empty arrows point at filopodia at the back of the cell where F-actin is not detected . ‘n’ represents the number of cells analysed . ( B ) A graph showing the frequency of F-actin-containing filopodia ( determined as presented in A ) that are formed by PGCs in medNY054 embryos . ‘n’ represents the number of cells analysed . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 00610 . 7554/eLife . 05279 . 007Figure 2—figure supplement 2 . The procedure for cell segmentation , defining the front , back and sides of the cell . ( A ) The segmentation process is performed on a 3D reconstructed image of a polarized PGC labelled with EGFP-F' and imaged by a spinning-disk microscope . ( A′ ) Two orthogonal lines are drawn , with their cross positioned at the centre of the cell . The orientation of the cross relates to the direction of migration and to the morphological axis of the cell , such that one right angle points at the front and the other to the back . ( A″ ) The positions where the lines of the cross meet the cell membrane are connected by the white segmentation lines at the front and the back . ( A‴ ) 3D analysis of filopodia distribution and number in the three areas ( front , side and back ) is conducted while turning the cell in all dimensions . Arrows indicate the direction of movement . Scale bars signify 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 00710 . 7554/eLife . 05279 . 008Figure 2—figure supplement 3 . Filopodia do not appear to play an essential direct role in the generation of blebs . ( A ) A graph showing the percentage of blebs forming relative to filopodia position ( without filopodia = no filopodia on top or immediately next to bleb; with filopodia = filopodia on top or immediately next to bleb ) . ( B ) Panels presenting individual snapshots from Video 2 , where a first bleb ( asterisk ) is formed in a close proximity to filopodia ( arrowheads , at 10–20″ ) and a second is formed in a region devoid of filopodia ( 40″ ) . Filopodia persist as a bleb inflates in the immediate vicinity ( arrowheads , 10–20″ ) . Scale bar is 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 00810 . 7554/eLife . 05279 . 009Figure 2—figure supplement 4 . Filopodia distribution and number in PGCs knocked down for Cxcr4b . ( A ) PGCs in embryos knocked down for Cxcr4b ( dark bars ) show apolar distribution of filopodia and an overall increased formation of these cellular protrusions as compared with control cells ( light bars ) . ( B ) Representative examples of a control ( left ) and Cxcr4b-knocked down PGCs . Arrowheads mark filopodia and ‘n’ indicates the number of cells analysed . Arrows show the direction of movement . Scale bars signify 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 00910 . 7554/eLife . 05279 . 010Video 1 . Migrating PGCs in wild type embryos show enhanced formation of filopodia in the direction of migration . A 10 min time-lapse video of a PGC in kop-egfp-f'nos3′UTR embryo was captured using a 63× objective on a Zeiss AxioImager . M2 microscope equipped with a Photometrics camera ( Cascade II ) and VS-Laser Control . Z-stacks include 29 planes per time point at focal planes 1 µm apart , 10 s interval , 300 ms exposure time with binning one . Time in minutes and seconds . Scale bar represents 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 01010 . 7554/eLife . 05279 . 011Video 2 . Filopodia are not directly controlling bleb formation . Bleb-formation ( asterisk ) can be observed first in close proximity to filopodia ( arrowheads , 10–20″ ) and then in a region devoid of filopodia ( 40″ ) . A 40-s time-lapse video of a membrane-labelled PGC ( in a kop-egfp-f'nos3′UTR embryo ) using a 63× objective on a Zeiss AxioImager . M2 microscope equipped with a dual view filter ( MAG Biosystems ) , Photometrics camera ( Cascade II ) and VS-Laser Control . Z-stacks were captured with 29 planes at focal planes per time point , 1 µm apart at 10 s interval , 300 ms exposure time and with binning one . Scale bar represents 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 011 By following filopodia dynamics we determined that the extensions persisted for an average of 33 ± 2 . 5 s ( wild type in Figure 3A , whole cell ) , from the time of appearance to complete retraction and that on average , they extended to a maximum length of 3 . 4 ± 0 . 1 µm ( wild type in Figure 3B , whole cell ) . Interestingly , the dynamic parameters ( persistence and length ) were dramatically different between filopodia at the front and back of the cell . That is , while filopodia in the front persisted for 28 ± 3 . 5 s and were 3 . 0 ± 0 . 1 µm long , filopodia at the back of the cell persisted for 51 ± 8 . 2 s and were 4 . 3 ± 0 . 4 µm long ( wild type in Figure 3A , B ) , thus revealing that filopodia located at the cell front are significantly more dynamic . Importantly , the enhanced dynamics of filopodia at the cell front was not a result of the protrusions being engulfed by blebs that preferentially form at this aspect of the cell . Only 30% of the filopodia at the cell front were engulfed by forming blebs and those that were not engulfed showed a reduced persistence as compared with filopodia formed at the cell back ( Figure 3—figure supplement 1A ) . 10 . 7554/eLife . 05279 . 012Figure 3 . The dynamics of filopodia at the cell front are determined by the distribution of Cxcl12a . ( A ) Persistence and ( B ) maximum length of filopodia in PGCs migrating within wild type ( light bars ) and medNY054 homozygous embryos ( dark bars ) . ‘n’ indicates the number of filopodia analysed in 10 wild type cells and 8 medNY054cells over 10 min . ( C ) Persistence and ( D ) maximum length of filopodia in PGCs migrating within medNY054 homozygous embryos knocked down for Cxcr7b and injected with 2 pg control mRNA ( light bars ) or 2 pg cxcl12a mRNA ( dark bars ) . ‘n’ indicates the number of filopodia in 10 cells over 10 min . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 01210 . 7554/eLife . 05279 . 013Figure 3—figure supplement 1 . The enhanced dynamics of filopodia at the cell front is independent of bleb formation at this aspect of the cell . ( A ) Persistence of filopodia at the cell front , which are not engulfed by blebs , compared to that of filopodia at the back of PGCs migrating within wild type embryos ( from dataset used in Figure 3 ) . ‘n’ represents number of filopodia analysed . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 01310 . 7554/eLife . 05279 . 014Figure 3—figure supplement 2 . Properties of PGCs migrating in embryos expressing a low concentration of uniform Cxcl12a . ( A ) Injection of 2 pg of Cxcl12a-encoding RNA into embryos results in an increase in the proportion of the ectopic PGCs per embryo at 24 hpf . Representative embryos are shown on the right . An asterisk labels the site where the gonad develops; ectopic PGCs are labelled with arrowheads . 39 control and 40 embryos expressing 2 pg Cxcl12a were analysed . ( B ) Migration tracks of PGCs in medNY054 homozygous embryos knocked down for Cxcr7b and injected with 2 pg control mRNA ( upper panel ) or 2 pg cxcl12a mRNA ( lower panel ) . Tracks represent 58 min of PGC migration in 7 hpf embryos acquired at a 2 min interval ( Video 4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 014 Together , the formation of dynamic filopodia on migrating PGCs is biased towards the leading edge of the cell , with filopodia formed at the front and back exhibiting significant differences in number , length , persistence and actin content . Since PGCs migrate towards regions expressing higher levels of Cxcl12a and filopodia formation is biased in the same direction , we sought to determine whether Cxcl12a in the environment influences the generation of filopodia . To this end , we analysed filopodia distribution and their dynamics in medNY05 embryos lacking functional Cxcl12a ( Figure 2B and Video 3 ) . PGCs in medNY054 embryos show a dramatic increase in the number of filopodia ( 11 . 4 ± 0 . 8 , Figure 2C , whole cell ) relative to wild type ( 5 . 6 ± 0 . 5 , Figure 2C , whole cell ) . PGCs in medNY054 embryos also show an even distribution of filopodia around the cell circumference , which is in sharp contrast to the polar distribution observed in the wild type embryos ( Figure 2C , D ) . Similar results were obtained upon disruption of the Cxcl12a/Cxcr4b signalling pathway by morpholino-mediated Cxcr4b knockdown ( Figure 2—figure supplement 4A , B ) . Strikingly , irrespective of their position around the cell perimeter , filopodia of PGCs in medNY054 embryos revealed average persistence ( 51 ± 3 . 1 s , Figure 3A , whole cell ) and maximum lengths ( 3 . 9 ± 0 . 1 µm Figure 3B , whole cell ) similar to filopodia located at the back of wild type PGCs ( Figure 3A , B , wild type ) . Filopodia of PGCs in medNY054 embryos also included a higher frequency of those with no detectable actin , thus exhibiting further similarities to filopodia in the back of wild type cells ( Figure 2—figure supplement 1B ) . Next , we examined the effect of a uniform Cxcl12a distribution on filopodia formation . Here , we injected cxcl12a RNA into medNY054 embryos at a concentration , which is sufficient to interfere with PGC arrival at their target region in wild type embryos , but does not immobilize them ( Figure 3—figure supplement 2A , B and Video 4 ) . In these experiments we co-injected morpholino against cxcr7b , the somatic decoy receptor for Cxcl12a ( Boldajipour et al . , 2008 ) , to avoid any local decrease in Cxcl12 levels that would result in gradient formation . Similar to the results observed in medNY054 embryos ( Figure 2C , medNYO54 and Figure 2E , control ) , numerous , evenly distributed filopodia appeared on the cell surface of PGCs ( Figure 2E , F , cxcl12a ) and their persistence ( 58 ± 3 . 0 s ) , as well as their maximum length ( 4 ± 0 . 1 µm ) was significantly elevated to the same extent as filopodia in medNYO54 embryos ( Figure 3C , D and Video 5 ) . We conclude therefore , that the graded distribution of Cxcl12 determines the number , the distribution and the dynamic behaviour of filopodia around the PGC circumference . 10 . 7554/eLife . 05279 . 015Video 3 . Migrating PGCs in medusa ( medNY054 ) mutant embryos that lack Cxcl12a show enhanced apolar filopodia formation . A 10 min time-lapse video of a PGC in medNY054 homozygous embryo was captured using a 63× objective on a Zeiss AxioImager . M2 microscope equipped with a Photometrics camera ( Cascade II ) and VS-Laser Control . Z-stacks include 29 planes per time point at focal planes 1 µm apart , 10 s interval , 300 ms exposure time with binning one . Time in minutes and seconds . Scale bar represents 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 01510 . 7554/eLife . 05279 . 016Video 4 . Low concentration of uniformly expressed Cxcl12a affects cell migration , but does not immobilize the cells . Migration tracks of PGCs in medNY054 homozygous embryos knocked down for Cxcr7b , injected with mcherry_h2b_globin3′UTR mRNA ( mCherry labelling nuclei of all cells ) and either with 2 pg control mRNA ( first video ) or 2 pg cxcl12a mRNA ( second video ) . Tracks represent 58 min of development in 7 hpf embryos . Snapshots captured at 2-min intervals with 300 ms exposure time at three focal planes ( 15 µm apart ) using a 10× objective to generate the Z-stacks on a Zeiss AxioImager . M1 microscope equipped with a Photometrics camera ( CoolSNAP ES2 ) . Time in minutes and scale bar represents 50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 01610 . 7554/eLife . 05279 . 017Video 5 . Uniform Cxcl12a in the environment induces apolar formation of numerous long and persisting filopodia on the PGC surface . A 10 min time-lapse video of a PGC in medNY054 homozygous embryo knocked down for Cxcr7b and that expresses uniform levels of Cxcl12a-encoding RNA was captured using a 63× objective on a Zeiss AxioImager . M2 microscope equipped with a Photometrics camera ( Cascade II ) and VS-Laser Control . Z-stacks include 29 planes per time point at focal planes 1 µm apart , 10 s interval , 300 ms exposure time with binning one . Time in minutes and seconds . Scale bar represents 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 017 Considering that filopodia formation is influenced by the distribution of Cxcl12a , we further monitored the immediate response of PGCs to an artificial Cxcl12a gradient generated in Cxcl12a mutant embryos ( medNY054 ) . We focused on the formation of filopodia as the PGCs switch from random to directed migration upon exposure to the chemokine gradient . A steep Cxcl12a gradient was generated in medNYO54 embryos by transplanting mCherry-labelled cells that were treated with cxcr7b morpholino and expressed either cxcl12a or control RNA ( Figure 4—figure supplement 1A ) . PGCs exposed to the control transplant generated randomly distributed filopodia ( Figure 4A , C and Figure 4—figure supplement 1B ) and were not attracted to the transplanted cells ( see Video 6 control transplant , with snapshots presented in Figure 4A , 2/23 cells migrated towards the transplant ) . In contrast , measuring the angle of filopodia to a Cxcl12a-expressing transplant ( Figure 4D and Figure 4—figure supplement 1B ) we observed that PGCs exhibited gradual polarization of filopodia directed towards the attractant source . The biased filopodia orientation was followed by cell polarization ( blue arrow in Figure 4D ) as manifested by the elongation of the cell along the axis of migration , positioning of the golgi to the back of the cell ( Figure 4—figure supplement 1A ) and migration towards the chemokine source ( Video 6 , Cxcl12a transplant , with snapshots presented in Figure 4B of a single cell representing 15/17 cells that showed this response ) . These experiments show that PGCs respond to the Cxcl12a gradient first by formation of filopodia oriented towards the Cxcl12a source and then by cell polarization and migration in the direction of the chemokine . 10 . 7554/eLife . 05279 . 018Figure 4 . PGCs extend filopodia towards the chemokine source prior to cell polarization and directed migration towards the attractant . ( A , B ) The cellular behaviour of PGCs ( green ) in response to transplanted control cells ( magenta in A ) or to Cxcl12a-expressing cells ( magenta in B ) . Upper panels show the cells immediately after transplantation and lower panels show snapshots from Video 6 presenting the behaviour of the cells in the following 28 min . In B , additional images present the polar position of the golgi ( red asterisk , labelled by EGFP-F' , as defined in Figure 4—figure supplement 1A ) at the back of the cell . Green arrowheads mark filopodia , blue dots indicate no migration and blue arrows show the direction of PGC movement . Scale bar is 10 µm . ( C , D ) The angle of filopodia orientation relative to the position of transplanted cells ( located at 0° , see also Figure 4—figure supplement 1B ) in the case of a control transplant ( C ) and with respect to Cxcl12a expressing transplant ( D , three examples ) over 25 min . Blue arrows signify the time of morphological cell polarization and movement . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 01810 . 7554/eLife . 05279 . 019Figure 4—figure supplement 1 . Polarization of PGCs encountering an artificially generated Cxcl12a gradient . ( A ) Schematic experimental setup . Cells from 4 hpf medNY054 homozygous embryos expressing Cxcl12a and mCherry-F' , in which Cxcr7b expression was inhibited , were transplanted into 6 hpf medNY054 homozygous embryos to examine the response of PGCs to the chemokine gradient . Control cells were similarly labelled but lacked Cxcl12a expression . The image to the right shows a polarized PGC with F-actin at the cell front labelled by Lifeact-EGFP ( magenta ) and the golgi apparatus at the cell back labelled with ECFP- tagged human beta1 , -4-galactosyltransferase ( yellow ) . The mCherry-F' labels the cell membrane , as well as the area of the golgi at the back of the cell . ( B ) A schematic representation of the angle measurements presented in Figure 4C , D for filopodia orientation relative to the position of the transplanted cells ( located at 0° ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 01910 . 7554/eLife . 05279 . 020Figure 4—figure supplement 2 . PGCs extend filopodia in the direction of migration prior to polarization and actual onset of migration . ( A ) Snapshots from Video 7 presenting the behaviour of a migrating cell , which makes a 90° turn ( upper panel ) and of a cell , which depolarizes and then migrates in the opposite direction ( lower panel ) . Green arrowheads mark filopodia; blue dot indicates no migration and blue arrows show the direction of PGC movement . Scale bar signifies 10 µm . ( B ) Filopodia orientation relative to the forthcoming direction of migration ( located at 0° ) in the two cells presented in Video 7 . An additional example is presented on the right . Blue arrows signify the time of morphological cell polarization and migration . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 02010 . 7554/eLife . 05279 . 021Video 6 . PGCs extend filopodia towards the chemokine source , prior to their morphological polarization and directed migration towards cells producing the attractant . Cells from 4 hpf medNY054 homozygous embryos , in which Cxcr7b expression was inhibited and which express mCherry-F' ( red ) and either control RNA ( part control transplant ) or express cxcl12a RNA ( part Cxcl12a transplant ) were transplanted into 6 hpf host medNY054 homozygous embryos . The PGCs in the host embryos were labelled with EGFP-F' and their reaction to the transplant was monitored on a Zeiss AxioImager . M1 microscope equipped with a dual view filter ( MAG Biosystems ) , Photometrics camera ( Cascade II ) and VS-Laser Control . Acquisition of the 28 min time-lapse video started immediately following the transplantation using a 40× objective . Z-stacks were captured with 24 planes per time point , at focal planes 4 µm apart , 15 s interval , 300 ms exposure time with binning one . Green arrowheads mark the position of filopodia , red asterisks mark the position of the golgi , blue dots indicate no migration and blue arrows indicate morphological PGC polarization and movement . Time in minutes and scale bar represents 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 021 To examine the relevance of these findings for cells migrating within a non-manipulated environment , we monitored the morphology and the behaviour of EGFP-F' labelled cells in wild type embryos . Interestingly , cells migrating within an unmodified environment behaved similarly to those responding to artificial gradients . Specifically , in cases where migrating PGCs performed sharp turns , polarized filopodia formation preceded the change in migration direction ( Figure 4—figure supplement 2 and Video 7 ) . Taken together , the observed PGC response to an artificial and endogenous chemokine gradient is consistent with the idea that filopodia formation constitutes an early response to distribution of Cxcl12 that is followed by cell polarization and directed migration . 10 . 7554/eLife . 05279 . 022Video 7 . PGCs extend filopodia in the direction of migration , prior to polarization and actual onset of migration . 10 min time-lapse videos of PGCs in wild type embryos were captured using a 63× objective on a Zeiss AxioImager . M2 microscope equipped with a Photometrics camera ( Cascade II ) and VS-Laser Control . Z-stacks include 29 planes per time point at focal planes 1 µm apart , 11 s interval , 300 ms exposure time with binning one . Time in minutes and seconds . Scale bar represents 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 022 Cells respond to graded Cxcl12a in the environment by polarized formation of filopodia ( Figure 2 ) , which could be important for the orientation of the cell and migration in the direction of the chemokine source . This observation may imply that filopodia could play a role in the transmission of the Cxcl12a signal into the leading edge of the cell by locally increasing the surface area for ligand binding . To address this notion , we overexpressed fluorescently labelled Cxcl12a in cell clones within the embryo ( Figure 5A ) and followed its distribution relative to the cells and particularly , to the filopodia . As shown in Figure 5B and Video 8 , Cxcl12a-Venus foci that were initially detected on the cell surface were subsequently internalized into the cell . Strikingly , a similar interaction between Cxcl12 foci and PGCs was observed on the surface of filopodia ( Figure 5C and Video 9 ) . Binding of Cxcl12 was observed at the tip of a filopodium ( red arrowhead in Figure 5C ) that extended about 12 µm away from the cell body , as well as closer to the cell body ( yellow arrowhead in Figure 5C ) . Retraction of the filopodium resulted in dragging of the distant Cxcl12 focus to the cell body and a bleb forming at this location engulfs the closer Cxcl12 focus ( Figure 5C ) . The internalization of Cxcl12a-Venus ( yellow arrowhead in Figure 5D ) is consistent with the idea that the ligand interacts with its receptor Cxcr4b on the filopodia membrane and elicits a signal directing cell polarization relevant for cell migration in the direction dictated by the chemotactic gradient . Whereas the filopodia at the cell front are more dynamic and higher in number , consistent with the uniform distribution of Cxcr4b on the cell membrane ( Figure 1A ) , we could detect interaction of overexpressed Venus-labelled Cxcl12 with filopodia at the back of the cell as well ( Figure 5—figure supplement 1A–A″ ) . 10 . 7554/eLife . 05279 . 023Figure 5 . Cxcl12a internalization and interaction with filopodia . ( A ) Schematic experimental setup . Cxcr7b function was knocked down in embryos , in which PGCs express mCherry on their membrane . At 16-cell stage , these embryos were injected with cxcl12a-venus RNA directed into a corner cell for a mosaic expression of the chemokine . ( B ) Cxcl12a ( green ) is bound to the PGC membrane ( magenta ) and internalizes into the cell . Snapshots from Video 8 , showing an optical section of a PGC ( a Z-projection of two 1-µm-slices ) . An arrowhead points at an internalizing Cxcl12a spot . Scale bar is 5 µm . ( C ) Snapshots from Video 9 , showing an optical section of a PGC ( a Z-projection of 4 1-µm-slices ) with Cxcl12a ( green ) interaction seen on the filopodium ( magenta ) . The red arrowhead indicates a Cxcl12a spot bound to the tip of the retracting filopodium and the yellow arrowhead points at Cxcl12a , which is bound to the filopodium closer to the cell body and is then engulfed by the cell ( 1:30–3:30 min ) . Scale bar is 5 µm . ( D ) PGC from panel C at 3:30 min ( D′ ) where the area magnified in D″ and D‴ is delineated in a red box . ( D″ and D‴ ) A 3D wire presentation of the magnified cell surface area ( magenta ) and the relevant Cxcl12 foci ( green ) . ( D″ ) A 3D image orientated as the original panel in C and ( D‴ ) is horizontally rotated ( 90° clockwise ) to visualize internalization of Cxcl12a . Scale bar is 2 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 02310 . 7554/eLife . 05279 . 024Figure 5—figure supplement 1 . Cxcl12a interaction with filopodia at the front and back of the cell . ( A–A″ ) Cxcl12a ( Venus-tagged , green ) interacts with filopodia of a PGC ( mCherry-F , magenta ) . ( A′ ) An optical section of a PGC ( a Z-projection of 4 1-µm-slices ) with Cxcl12a ( green ) interaction with filopodia ( magenta ) observed over 1:15 min . ( A and A″ ) Magnified insets marked in A′ as red squares . The yellow arrowheads point at Cxcl12a spots along filopodia . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 02410 . 7554/eLife . 05279 . 025Video 8 . Cxcl12a internalization . Cxcl12a ( green ) is bound to the PGC membrane ( magenta ) and internalizes into the cell . The video was captured using a 63× objective on a Zeiss AxioImager . M2 microscope equipped with a Photometrics camera ( Cascade II ) and VS-Laser Control at 15 s intervals for 2:15 min and shows an optical section of a PGC ( a Z-projection of two 1-µm-slices ) . Arrows indicate a bound and an internalizing Cxcl12a spot . Scale bar is 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 02510 . 7554/eLife . 05279 . 026Video 9 . Cxcl12a interaction with filopodia . Cxcl12a ( green ) is bound to a filopodium and internalizes into the cell . An optical section of a PGC ( a Z-projection of 4 1-µm-slices ) is shown with Cxcl12a ( green ) interaction observed on the filopodium ( magenta ) . The video was captured with a Lightsheet Z1 microscope ( Zeiss , Germany ) at 15 s intervals for 3:30 min The red arrowhead indicates a Cxcl12a spot bound to the tip of the retracting filopodium and the yellow arrowhead points at Cxcl12a , which is bound to the filopodium more proximally and is then engulfed by the cell . Wire presentation visualizes Cxcl12a inside the cell . Scale bar is 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 026 To determine the functional significance of the Cxcl12a-dependent polarized generation of filopodia , we specifically interfered with their formation and examined the effect on cell polarization . For that purpose , we initially employed the insulin receptor substrate protein 53 ( Irsp53 ) , a key protein in the process of filopodia formation ( Ahmed et al . , 2010; Krugmann et al . , 2001 ) . The Irsp53 protein contains an Inverse-Bin-Amphiphysin-Rvs ( I-BAR ) domain ( Mattila et al . , 2007 ) that promotes membrane deformation and includes actin binding sites ( Yamagishi et al . , 2004 ) . The protein also contains a partial CRIB motif for Cdc42-GTP mediated activation ( Govind et al . , 2001 ) and an SH3 domain for interaction with additional filopodia regulators such as Mena ( Krugmann et al . , 2001 ) , Eps8 ( Vaggi et al . , 2011 ) , mDia1 ( Goh et al . , 2012 ) , WAVE2 ( Goh et al . , 2012 ) , N-WASP ( Oh et al . , 2013 ) , Lin7 ( Crespi et al . , 2012 ) and WIRE ( Misra et al . , 2010 ) ( Figure 6A ) . We found that the expression pattern of irsp53 mRNA in zebrafish embryos spans the relevant early stages during which PGCs migrate . Specifically , irsp53 RNA is maternally provided and ubiquitously present in 1-hr post fertilization ( hpf ) embryos , as well as at later stages ( 6 and 10 hpf , Figure 6B ) . We also found that an Irsp53-mCherry fusion protein ( Figure 6A ) was expressed in the cytoplasm and importantly , in filopodia ( Figure 6C , upper panel and Figure 6—figure supplement 1A ) , particularly in those located at the cell front . The expression pattern of irsp53 RNA and the subcellular localization of the Irsp53 fusion protein to filopodia are therefore consistent with the possibility that Irsp53 plays a role in filopodia formation in migrating PGCs . 10 . 7554/eLife . 05279 . 027Figure 6 . irsp53 RNA expression , Irsp53 protein localization and the role of the protein in filopodia formation . ( A ) Schematic structure of the Irsp53 protein domains . The position of the mCherry fluorophore fusion at the C-terminus of the protein is presented , proteins interacting with the SH3 and CRIB domains are indicated and the mutations introduced into the I-BAR domain to generate the dominant-negative ( DN ) Irsp53 version are marked . ( B ) Ubiquitous expression of the irsp53 RNA in 1 , 6 and 10-hpf embryos . ( C ) A single plane of PGCs expressing EGFP-F' and an Irsp53-mCherry protein fusion showing localization of Irsp53 to filopodia ( upper panel ) , while the dominant-negative Irsp53 protein is not found in the filopodia ( lower panel ) . Rectangles delineate the area of magnification shown in the right panels . ( D ) Reduction of filopodia number at the cell front in PGCs in wild type embryos expressing the DN Irsp53 protein ( dark bars ) relative to control PGCs ( light bars ) . ‘n’ indicates the number of cells analysed . ( E ) Examples of a control ( left ) and a dn irsp53-expressing ( right ) PGCs in wild type embryos . ( F ) Expression of dn irsp53 in PGCs of medNY054 homozygous embryos shows no effect on filopodia number and distribution around the cell perimeter ( dark bars ) as compared with control PGCs ( light bars ) . ( G ) Examples of a control ( left ) and a dn irsp53-expressing ( right ) PGCs in medNY054 homozygous embryos . ‘n’ indicates the number of cells analysed , arrows show direction of cell migration and arrowheads mark filopodia . Scale bar is 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 02710 . 7554/eLife . 05279 . 028Figure 6—figure supplement 1 . Irsp53 localization within filopodia . ( A ) Irsp53 can be detected in a higher proportion of filopodia at the cell front than within filopodia at the side and back of the cell . The dominant-negative version of Irsp53-mCherry cannot be detected within front filopodia and can be observed in a small proportion of side and back filopodia . ‘n’ signifies the number of cells analysed . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 02810 . 7554/eLife . 05279 . 029Figure 6—figure supplement 2 . Inhibition of Irsp53 does not affect bleb formation at the cell front . ( A ) Number of Cxcl12a-independent blebs formed by PGCs expressing a dominant-negative version of irsp53 , as compared with the number of blebs formed by control cells in medNY054 homozygous embryos . ‘n’ signifies the number of cells analysed . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 029 To assess filopodia function in migrating PGCs , we set out to disrupt their formation by expressing a dominant-negative ( DN ) form of the Irsp53 protein . The dominant-negative form of the protein harbours mutations in the four actin-bundling sites ( Millard et al . , 2005; Lim et al . , 2008 ) within the I-BAR domain ( Figure 6A ) . A DN Irsp53-mCherry protein version expressed in PGCs was present primarily in the cell body , rather than in filopodia ( Figure 6C , lower panel , Figure 6—figure supplement 1 ) . Moreover , expression of the DN Irsp53 version in PGCs resulted in a significant decrease in filopodia number at the cell front ( Figure 6D , E ) . However , inhibition of Irsp53 function had no effect on filopodia formation in medNY054 embryos ( Figure 6F , G ) , in which the filopodia exhibit posterior characteristics ( Figure 3A ) . The inhibition of Irsp53 function specifically affected filopodia formation rather than influencing cell migration in general , as this treatment had no effect on the formation of blebs at the PGC front in embryos lacking the guidance cue ( Figure 6—figure supplement 2A ) . These results provide evidence that Irsp53 is required for filopodia formation in response to Cxcl12a gradient and can serve as a specific reagent for interfering with their formation . In a second line of experiments attempting to affect filopodia formation we employed the actin filament associated protein 1L1 ( AFAP1L1 ) . AFAP1L1 was shown to play a role in tumour metastasis and to constitute a cancer prognostic marker ( Furu et al . , 2011 ) , to be localised to invadosomes and along with cortactin , to podosomes ( Snyder et al . , 2011 ) . AFAP1L1 contains two PH domains that serve as a link to the membrane , a serine-threonin-rich substrate domain ( SD ) in between , a leucine zipper ( Lzip ) , an actin-binding domain ( ABD ) at the C- terminus , as well as one SH3 and two SH2 domains ( Figure 7A ) ( Snyder et al . , 2011 ) . While AFAP1L1 had not been previously implicated in filopodia formation , its subcellular localization in PGCs was intriguing and prompted our following investigation . Namely , when an mCherry-tagged version of Afap1L1a ( Figure 7A ) was expressed in PGCs , it was found to be specifically enriched at the base of most filopodia ( 84 ± 9 . 7% ) at the level of the cell cortex ( Figure 7C ) , suggesting a function for the protein in filopodia formation or maintenance . We subsequently ascertained that afap1l1a RNA is expressed at the time at which PGCs migrate towards their target and showed that afap1l1a RNA is maternally provided and is uniformly expressed during early embryonic development ( Figure 7B ) . 10 . 7554/eLife . 05279 . 030Figure 7 . afap1l1a RNA expression , Afap1L1a protein localization and the role of the protein in filopodia formation . ( A ) Schematic structure of the Afap1L1a protein domains including a serine-threonine-rich substrate domain ( SD ) flanked by two PH domains , a leucine zipper ( Lzip ) , an actin-binding domain ( ABD ) , two SH2 and one SH3 domains . mCherry fluorophore was fused to the N-terminus of Afap1L1a to determine the subcellular localization of the protein . ( B ) Ubiquitous expression of the afap1l1a RNA in 1 , 6 and 10-hpf embryos . ( C ) A single plane of PGCs expressing EGFP-F' and the mCherry-Afap1L1a fusion protein reveals weak expression of the protein around the cell perimeter ( C′ ) and a predominant strong association with the base of filopodia ( C′–C‴ ) . In C′ , one marks the site of a retracted filopodium , two the sites of extended filopodia and arrow indicates the direction of migration . ( D ) The number and distribution of filopodia in migrating PGCs overexpressing afap1l1a ( dark bars ) as compared with control PGCs ( light bars ) . ‘n’ indicates the number of cells analysed . ( E ) Representative images of control ( left ) and of afap1l1a-overexpressing PGCs . Arrows indicate the direction of movement and arrowheads point at filopodia . Scale bars signify 10 µm except for C″ and C‴ , where the scale is 3 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 03010 . 7554/eLife . 05279 . 031Figure 7—figure supplement 1 . The effect of Afap1L1a overexpression on filopodia dynamics in PGCs migrating within a wild type environment . ( A ) Persistence and ( B ) maximum length of filopodia in PGCs that express control RNA ( light bars ) or overexpress afap1l1a ( dark bars ) . ‘n’ signifies the number of filopodia analysed in six cells during 10 min . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 03110 . 7554/eLife . 05279 . 032Figure 7—figure supplement 2 . Afap1L1a overexpression does not affect bleb formation at the cell front . ( A ) Number of Cxcl12a-independent bleb formation by PGCs overexpressing afap1L1a , as compared with control cells in medNY054 homozygous embryos . ‘n’ signifies the number of cells analysed . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 032 To test whether Afap1L1a could serve as a tool for altering filopodia formation , we overexpressed the protein in PGCs . Remarkably , overexpression of the protein resulted in a significant increase in filopodia number as compared to control cells ( Figure 7D , whole cell ) . Furthermore , this treatment changed the distribution of the filopodia significantly , resulting in a decrease in filopodia number at the cell front and an increase in number at the sides and back of the cells ( Figure 7D , E ) , effectively reversing the normal distribution of filopodia in the cell . Based on their increased persistence and length ( Figure 7—figure supplement 1A , B ) , the filopodia formed by cells overexpressing Afap1L1a were of the type generated at the back of PGCs . Similar to the Irsp53 effect , overexpression of Afap1L1a did not significantly alter bleb formation at the PGC front in embryos lacking the guidance cue ( Figure 7—figure supplement 2A ) . Taken together , these findings suggest that Afap1L1 could serve as a specific tool for manipulating the formation of filopodia in PGCs . Since manipulating Irsp53 and Afap1L1a function in PGCs altered filopodia distribution and number , we were in a position to probe into the functional significance of the polar formation of filopodia as potential sensors of the guidance cue . To this end , we made use of our recent findings showing that in response to a graded Cxcl12a concentration in the environment , zebrafish PGCs exhibit increased pH at the cell front ( 7% higher on the logarithmic pH scale , reflecting difference of 15% at the level of H+ concentration ) . This polar pH elevation could control the activity of various proteins important for cell motility ( Stock and Schwab , 2009 ) and in the case of PGCs , was shown to be essential for a local increase in the activity of the small GTPase Rac1 and thus , for directing actin polymerization to the cell front ( Tarbashevich et al . , 2015 ) . The loss of the Cxcl12a gradient leads the loss of the difference in pH between the cell front and its rear ( Tarbashevich et al . , 2015 ) . We reasoned that if filopodia were indeed important for Cxcl12 sensing , they should be functionally linked to pH elevation and Rac1 activity at the front of PGCs . To test this possibility , we first performed FRET-based pH measurements in cells in which filopodia formation was manipulated as described above . Strikingly , the reduction of filopodia formation at the cell front reduced ( in the case of Irsp53 ) , or abolished ( in the case of the Afap1L1a ) the increased pH at the leading edge ( Figure 8A ) . Conversely , inhibiting the relative elevation of pH at the cell front by inhibiting the translation of the Carbonic anhydrase 15b ( Tarbashevich et al . , 2015 ) had no effect on filopodia formation ( Figure 8—figure supplement 1A , B ) . To determine whether the reduction of filopodia formation at the cell front had any effect on Rac1 activity , we expressed a FRET-based Rac1 activity reporter in cells overexpressing Afap1L1a . Importantly , as compared with control cells , Rac1 activity was significantly reduced in these cells ( Figure 8B ) . A milder effect on the difference between the front and rear pH , as induced by DN Irsp53 ( Figure 8A ) could affect different proteins important for cell motility ( Stock and Schwab , 2009 ) , but was not associated with a significant reduction in Rac1 activity as judged by the FRET level . Consistent with the idea that filopodia function upstream of the Cxcl12a signalling cascade , expression of a constitutively activated version of Rac1 ( Rac1V12 ) in germ cells ( that stop migrating as a result ) had no effect on filopodia formation beyond the increase in their numbers observed in cells rendered immotile ( Figure 8—figure supplement 1C , D ) . These results suggest that when filopodia function at the cell front is hindered , the ability of the PGCs to effectively sense the chemokine gradient is impaired , as reflected by the lack of elevated pH and by the reduced Rac1 activity at the leading edge . These results are also consistent with the idea that as a cellular structure , filopodia function in controlling intracellular pH at the cell front and are formed independent of it . 10 . 7554/eLife . 05279 . 033Figure 8 . Filopodia are required for cellular response to polarized Cxcl12a distribution . ( A ) 6–7 hpf control PGCs exhibit polarized distribution of intracellular pH as determined by the FRET efficiency of the pH sensor pH-lameleon5 protein in the cells ( left ) . Expression of the dominant negative form of Irsp53 or overexpression of the Afap1L1a abrogates the formation of high pH in the front . The graph represents average pH-FRET ratios between the front and the rear ( as indicated by the circles ) . The values for each cell are averages of 20 time points . ( B ) 8–9 hpf PGCs overexpressing Afap1L1a exhibit a decrease in Rac1 activity , as determined by differences in FRET generated by a Rac1-FRET activity reporter . Arrows indicate the direction of movement . ‘n’ indicates the number of cells analysed . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 03310 . 7554/eLife . 05279 . 034Figure 8—figure supplement 1 . Filopodia formation is independent of the elevated pH at the cell front and of Rac1 activity in migrating PGCs . ( A ) PGCs in embryos knocked down for ca15b ( dark bars ) show similar distribution and number of filopodia to that of control cells ( light bars ) . ( B ) Examples of a control ( left ) and ca15b -knocked down PGCs . Arrows indicate the direction of movement . ( C ) The number of filopodia in control ( light bar ) and in PGCs expressing constitutive active ( ca ) rac1 ( rac1V12 ) ( dark bar ) is similar . Since cells expressing the activated version of Rac1 are immotile , we immobilized both control and experimental cells by DN Rock expression , so the cells can be compared with respect to filopodia formation . ( D ) Representative images of a control ( left ) and a rac1V12 expressing PGCs . Arrowheads indicate filopodia . ‘n’ indicates the number of cells analysed . Scale bars signify 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 034 Together , these experiments show that the filopodia at the cell front are important for the downstream signalling by Cxcr4b and thus , for cell polarization and establishment of the cell front in response to chemokine cues . To determine whether inhibition of front filopodia formation affects the directed migration of PGCs , we examined manipulated cells for their migration tracks and arrival at their target . Interestingly , PGCs expressing the DN Irsp53 protein exhibited a significant reduction in displacement ( the shortest distance between the start and end points ) as a result of reduced track straightness ( the degree of changes in path direction ) ( Figure 9A , B ) . The inhibition of front filopodia mediated by DN Irsp53 appears therefore to impair sensing of the Cxcl12 guidance cue as reflected by the altered characteristics of the migration route . The specific role of front filopodia in the response to the chemokine gradient is further highlighted by the finding , that inhibition of Irsp53 function in medNY054 embryos lacking functional chemokine had no effect on the migration tracks of PGCs ( Figure 9—figure supplement 1A , B ) . These results also emphasize the fact that Irsp53 , which specifically affects filopodia formation at the cell front , is not essential for the migratory behaviour per se , but is rather important in the context of chemotaxis . 10 . 7554/eLife . 05279 . 035Figure 9 . Manipulations of filopodia formation lead to PGC migration defects . ( A ) Representative migration tracks of control PGCs ( upper panel ) and PGCs expressing the dominant-negative ( dn ) irsp53 version ( lower panel ) . ( B ) Analysis of PGC migration tracks assessing displacement , straightness and migration speed , comparing control cells ( light bars ) with dn Irsp53-expressing cells ( dark bars ) . ‘n’ indicates the number of migration tracks analysed . ( C ) Representative migration tracks of control PGCs ( upper panel ) and PGCs overexpressing ( oex ) afap1L1a ( lower panel ) . ( D ) Analysis of PGC displacement , track straightness and migration speed , comparing control cells ( light bars ) with afap1l1a overexpressing cells ( dark bars ) . ‘n’ indicates the number of migration tracks analysed . ( E–H ) Expression of the dn irsp53 ( E , F ) or overexpressing afap1l1a ( G , H ) in PGCs results in an increase of ectopic cells at 24 hpf as compared with embryos whose PGCs express a control RNA . ‘n’ indicates the number of embryos analysed . Arrowheads point at ectopic PGCs and asterisks mark the site of the developing gonad . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 03510 . 7554/eLife . 05279 . 036Figure 9—figure supplement 1 . Reducing the activity of Irsp53 has no effect on cell migration in the absence of Cxcl12a . ( A ) Representative migration tracks of control PGCs ( upper panel ) and PGCs expressing the dominant-negative Irsp53 version ( lower panel ) in medNY054 homozygous embryos . ( B ) Analysis of PGC migration tracks assessing displacement , track straightness and speed . Control cells ( light bars ) are compared with dn irsp53-expressing cells ( dark bars ) migrating in medNY054 homozygous embryos . ‘n’ indicates the number of migration tracks analysed . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 036 Alterations in the migration track properties were observed yet to a greater extent , upon overexpression of the Afap1L1a in PGCs . Specifically , overexpression of the Afap1L1a that led to a reversal of filopodia distribution with predominance of filopodia in the back , also led to a reduction in cell displacement , track straightness and speed ( Figure 9C , D ) . We conclude therefore , that tight regulation of filopodia formation and distribution around the cell surface is essential for effective guided PGC migration . The defects in the dynamic characteristics of cell migration manifested in the cell tracks of manipulated PGCs resulted in an increased proportion of chemokine-guided cells that failed to reach their target at 24 hpf ( Figure 9E–H ) .
In this work we provide evidence that filopodia are asymmetrically formed in response to a chemokine gradient and in turn , play a critical role in polarization of cells and their response to directional cues . We found that filopodia properties differ in the front from the back of the cell , such that front filopodia are larger in number , turn over more rapidly , are shorter and contain more F-actin . These differences in filopodia distribution and dynamics are governed by the graded distribution of the chemokine in the environment . The basis for this notion is the finding that filopodia formation in PGCs migrating within embryos lacking Cxcl12a was altered in a similar manner to that observed under conditions where the chemokine is uniformly expressed . The precise function of the front filopodia emerges when PGCs are experimentally exposed to a focused Cxcl12a source: there , filopodia formation appears to constitute an early response of the cells to the graded distribution of the chemokine . The formation of filopodia precedes further polarization such as directed bleb formation and onset of migration towards the chemoattractant source . A similar phenomenon was observed in neuronal growth cones , which respond to a glutamate gradient by an asymmetric distribution of filopodia prior to turning , in an in vitro setting ( Zheng et al . , 1996 ) . Similarly , endothelial tip cells of the early postnatal retina in mice respond to a source of VEGF by filopodia formation ( Gerhardt et al . , 2003 ) and RhoD-activated fibroblasts extend cytoneme-like protrusions towards an FGF source in vitro ( Koizumi et al . , 2012 ) . Whereas these observations were consistent with the idea that polarized filopodia formation constitutes a response to a guidance cue , the relevance for directed cell migration was not known . Our finding that the chemokine appears to bind and be transported on the filopodium , followed by internalization , is consistent with the idea that filopodia function in Cxcl12a uptake and in increasing the surface relevant for this action at the cell front . In this study , we carried out specific manipulations affecting filopodia formation and show an adverse effect on the ability of migrating single cells to respond to Cxcl12a by directional migration . The effect of specifically decreasing filopodia formation at the cell front ( by inhibiting Irsp53 function ) resulted in inefficient propagation of the external signal into the cell as evidenced by the strong reduction in the difference between the pH at the cell front and its back . In the same direction , overexpression of the Afap1L1 protein erased the normal front-back polarity with respect to filopodia distribution . In the absence of polar distribution of filopodia in the direction of the attractive cue , the cells could not effectively detect the graded chemokine cue , establish a stable front and thus , did not elevate the pH and Rac1 activity at the leading edge . These effects on the characteristic front-back cell polarity impaired the directional migration in response to a guidance cue , culminating in ectopic localization of the cells at the end of the migration process . Taken together , this work presents important insights into the mechanism by which dynamic filopodia located at the front of migrating cells participate in the transmission of the guidance cue to allow directional migration ( Figure 10 ) . We suggest that following the activation of Cxcr4b by Cxcl12a , the scaffold protein Irsp53 is activated to promote the formation of dynamic filopodia at the cell front . These filopodia extend in the direction of higher chemokine concentration and expand the surface for Cxcl12a binding , thereby further enhancing Cxcr4b signalling at the cell front and effectively enhancing the absolute differences in ligand levels detected along the front-back axis of the cell . We suggest that the positive feedback loop presented in Figure 10 allows the PGCs to interpret shallow gradients by forming a front that is more sensitive to the guidance cue . The uneven receptor activation along the back–front axis of the cell that includes the filopodia is a prerequisite for the establishment of the intracellular elevated pH at the cell front that facilitates actin polymerization at this location ( Tarbashevich et al . , 2015 ) leading to functional polarization of the cell . Under conditions of no , or uniform activation of Cxcr4b a cell front is not established . In such cases , the positive feedback loop that normally stabilizes the front in one aspect of the migrating cell is not present , affecting cell polarity and the features of migration in the direction of the chemokine source . 10 . 7554/eLife . 05279 . 037Figure 10 . Regulation of dynamic filopodia at the cell front and their role in cell polarization and directed cell migration . Following the activation of Cxcr4b by Cxcl12a ( 1 ) , the scaffold protein Irsp53 is activated ( 2 ) and promotes the formation of dynamic filopodia at the cell front ( 3 ) . The dynamic filopodia that extend in the direction of higher Cxcl12a concentration increase the surface for chemokine binding ( 4 ) , thereby enhancing signalling . The enhanced signalling at the cell front results in further Irsp53 activation ( 5 ) and an elevation of pH ( 6 ) . The increase in pH in turn , leads to an elevation in Rac1 activity at this aspect of the cell ( 7 ) . Once established , the front inhibits other parts of the cell from assuming front characteristics including dynamic filopodia formation , elevated pH and Rac1 activation ( 8 ) . Lower panels—Irsp53 inhibition or Afap1L1a overexpression result in loss of dynamic filopodia at the cell front , abrogating the local increase in pH . As a consequence , PGCs migrate less directionally . Lack of a Cxcl12a gradient leads to the formation of numerous , evenly distributed , long and persistent filopodia in cells that migrate in random directions . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 037 An important question that remains to be elucidated is the mechanism by which Cxcl12/Cxcr4b signalling activates Irsp53 function , leading to formation of the Irsp53-dependent filopodia at the cell front . CDC42 was shown to be active at the leading edge of migrating cells and in neuronal growth cones ( Etienne-Manneville and Hall , 2002; Nalbant et al . , 2004 ) , and IRSp53 was shown to be one of the effectors of CDC42 ( Govind et al . , 2001; Krugmann et al . , 2001 ) . A direct or indirect effect of Cxcl12 binding to Cxcr4 on the activation of Cdc42 and thus of Irsp53 would be an interesting option to explore . The finding that at least two populations of filopodia formed on the surface of migrating PGCs and that these populations exhibit differences concerning their distribution and dynamic properties are reminiscent of the findings concerning cytonemes in Drosophila ( Roy et al . , 2011 ) . In this case , tracheal cells were shown to generate morphologically different populations of cytonemes , each responding specifically to different secreted signals . The effect the reversed formation of filopodia had on cell migration speed in PGCs manipulated for Afap1L1 function ( Figure 9D ) raises the option that these protrusions contribute to cell motility in normal cells . This possibility is supported by the observations that some filopodia adhere to somatic cells , appear to be under tension and can pull particles towards the cell ( Kress et al . , 2007 ) . Upon laser-mediated cutting of a filopodium of this kind , its distal tip remained attached to a cell in the environment , while the proximal part attached to the cell rapidly retracted ( see Video 10 ) . It would be interesting to evaluate the significance of this finding relative to the role we demonstrated concerning chemokine signalling and cell polarization . 10 . 7554/eLife . 05279 . 038Video 10 . Some filopodia adhere to somatic cells and appear to be under tension . Laser-induced cut of a filopodium was performed using a Zeiss LSM710 confocal microscope equipped with a Zeiss W Plan-Apochromat 63× objective controlled by the ZEN software ( Zeiss , Germany ) . Following cut , the tip of the filopodium remained in contact with the distant somatic cell , while the part of the protrusion attached to the cell rapidly retracted , signifying tension . The video spans 7 s with snapshots captured at a 200 ms time-interval ( pinhole 208 µm , 128 × 128 ) . Time indicated in seconds and scale bar represents 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05279 . 038 The function of filopodia in decoding the Cxcl12a gradient in the context of PGC migration must bear on cancer cell invasion , since Cxcl12 is known to play a key role in metastasis of several cancer types ( Vila-Coro et al . , 1999; Singh et al . , 2004; Sun et al . , 2010 ) . Furthermore , proteins involved in filopodia formation such as Fascin , Eps8 and Irsp53 have been shown to be associated with increased invasiveness of cancer cells ( Matoskova et al . , 1995; Funato et al . , 2004; Machesky and Li , 2010 ) . The translation of Cxcr4b activation into filopodia formation constitutes the first morphological step in PGC polarization in response to the chemokine . It would thus be interesting to examine this important step in normal and pathological conditions where migrating cells initiate their reaction to guidance signals .
Zebrafish ( Danio rerio ) of the AB background and transgenic fish carrying the kop-egfp-f-nos-3′UTR transgene ( Blaser et al . , 2005 ) , kop-mcherry-f-nos3′UTR transgene , or the kop-lifeact-egfp-nos3′UTR transgene were used as wild type fish . medusaNYO45 ( Valentin et al . , 2007 ) mutant embryos were used for investigating filopodia formation in the absence of the guidance cue Cxcl12a and odysseus ( ody ) homozygous mutant fish were used for the rescue experiment with Cxcr4b . A list of constructs and in situ probes of the Zebrafish genes generated in this work , containing primers and amounts injected , is provided in Supplementary file 1B . Commonly PGCs were labelled with farnesylated mCherry ( mCherry-F' ) or EGFP ( EGFP-F' ) . The dominant-negative form of Irsp53 was generated according to Millard et al . ( Millard et al . , 2005 ) Additional constructs with amounts injected are found in Supplementary file 1C . To direct protein expression to the germ cells , the corresponding open reading frames ( ORFs ) were cloned upstream to the 3′UTR of the nanos1 gene , facilitating translation and stabilization of the RNA in these cells ( Koprunner et al . , 2001 ) . For global protein expression in the embryo , the ORFs were cloned upstream of the 3′UTR of the Xenopus globin gene . Capped sense mRNA was synthesized using the mMessage Machine kit ( Ambion , Foster City , CA ) . 2 nanoliters of RNA and/or morpholino antisense oligonucleotides were microinjected into the yolk of 1-cell stage embryos , unless stated otherwise . To inhibit protein translation in the embryo , morpholino antisense oligonucleotides ( Gene Tools , Philomath , OR ) were injected into the one-cell stage embryos . A list of morpholinos used in this work is provided in Supplementary file 1D . Whole-mount in situ hybridization was performed as previously described ( Weidinger et al . , 1999 ) , using afap1l1a and irsp53 digoxigenin-labeled probes . The sequence of the primers used for PCR amplification of these probes from zebrafish cDNA is provided in the Supplementary file 1B . For the design of the experiment and amount of the injected material see Supplementary file 1A–C . Cells from a 4 hpf donor embryo were transplanted into a 5–6 hpf host on a Zeiss AxioImager . M2 microscope with a 5× objective , immediately followed by the acquisition of a time-lapse video using a 40× objective or followed by acquisition of snapshots using a 63× objective . Time-lapse imaging was performed using a Zeiss AxioImager . M2 microscope equipped with a dual view filter ( MAG Biosystems , Exton , PA ) , Photometrics cameras ( Cascade II and CoolSNAP ES2 ) and VS-Laser Control . Time-lapse videos were generated using 63×- or 40× water-immersion objectives for imaging cell morphology and behaviour , and a 10× objective for the purpose of speed and migration track analysis . Detailed experimental setups are listed in Supplementary file 1A . Experiments were performed between 6–9 hpf . Frames were captured simultaneously with 488 nm and 561 nm lasers at 10-or 12 s intervals for high magnification videos ( 63× ) and at 15-s intervals for 40× videos , with 300 ms exposure time . 24 to 37 images at focal planes 1 µm ( for 63× videos ) or 4 µm ( for 40× videos ) apart were captured to generate the Z-stacks . For migration track and speed analysis time-lapse videos were captured with a 10× objective at 2-min intervals over 68 min with 300 ms exposure time at three focal planes ( 15 µm apart ) to generate the Z-stacks . For filopodia quantification the first time-point of a 2-min video was analysed and filopodia dynamics was followed over a time-period of 10 min . For filopodia measurements the manual mode of the ‘filament tracer’ module of the Imaris software ( Bitplane , Switzerland ) was used . PGC response to transplants was captured for 30 min starting ca . 2 min after transplantation . To calculate the percentage of ectopic germ cells , PGCs were counted at 24 hpf in the GFP-channel . The average number of ectopic PGCs was counted in percentage to the average total PGC number for the given experiment . For gastrulation correction , speed measurements and tracking of migrating germ cells the ‘spots’ module of the Imaris software ( Bitplane , Switzerland ) was used . Analysis of Cxcr4b distribution and protein age on the PGC membrane was done using ImageJ software ( for the analysis protocol see Supplementary file 1F ) . Imaging and analysis for the Rac1-FRET ( as snapshots ) was performed as for the pH-FRET ( as time-lapse ) . To measure the pH distribution in PGCs , pH-lameleon5 ( Esposito et al . , 2008 ) mRNA fused to the 3′UTR of nanos1 was injected into 1-cell stage of kop-mCherry-f-nanos3′UTR transgenic embryos . Imaging was performed at 7–9 hpf using an LSM710 confocal microscope ( 40× , NA 0 . 75 , pinhole 276 µm , 512 × 512 , 7 . 5 s per frame ) controlled by the ZEN software ( Zeiss , Germany ) . Analysis was done using ImageJ software ( for the analysis protocol see Supporting material ) . Embryos that express mCherry on the membrane of PGCs were knocked down for Cxcr7b using a morpholino antisense oligonucleotide by injection at 1-cell stage and were then injected with cxcl12a-venus RNA into a corner cell of the 16-cell stage for a mosaic expression pattern of the chemokine . Time-lapse imaging was performed at 10–15 hpf using either a Zeiss AxioImager . M2 microscope ( 63× water imersion objective ) , equipped with a dual view filter ( MAG Biosystems , Exton , PA ) , Photometrics cameras ( Cascade II and CoolSNAP ES2 ) and VS-Laser Control ( Video 8 ) or a Lightsheet Z1 ( Zeiss , Germany ) equipped with a Zeiss W Plan-Apochromat 20× objective ( 15 s intervals for 20 cycles at 960 × 960 ) and controlled by the ZEN software ( Zeiss , Germany ) ( Video 9 ) . Laser ablation was performed with a Zeiss LSM710 confocal microscope equipped with a Zeiss W Plan-Apochromat 63× objective controlled by the ZEN software ( Zeiss , Germany ) . The ‘bleaching’ module of the Zen software was used to perform ablation within a sub-region ( ROI ) . The laser source for ablation was a femtosecond Titanium-Sapphire laser ( Coherent Chameleon , Santa Clara , CA ) tuned to lambda = 740 nm with laser power reduced to 10% . The power at the objective back aperture during ablation was 90–100 mW . Ablation was monitored with confocal imaging of PGC membrane ( laser 488 nm ) and somatic membrane ( laser 561 nm ) in medNY054 homozygous embryos over 45 s with 199 ms time-interval ( pinhole 208 µm , 128 × 128 ) . All the data in the work were first tested for normal distribution by the Kolmogorov–Smirnov test . Experiments on the percentage of ectopic germ cells and FRET measurements were analysed by the Student's t-test . Experiments on the signal intensity between front and back of a cell were analysed using the paired t-test . All other significance tests are based on the Mann Whitney U test for the statistical difference . Error bars represent S . E . M . n . s . = non significant , *p ≤ 0 . 05 , **p ≤ 0 . 01 , ***p ≤ 0 . 001 . | Some of the cells in an animal embryo have to migrate long distances to reach their final positions; that is to say , to reach the locations where they will participate in the formation of tissues and organs . The migration of cells is also important throughout the entire lifespan of an animal . White blood cells , for example , must be able to move within tissues to search for and fight infections as well as to detect and remove abnormal cells . The front end of a migrating cell typically protrudes . The back of the cell is then pulled and detaches , which allows the whole cell to move forward . Migrating cells generate thin finger-like projections known as filopodia that have been suggested to help the cell sense their external environments and follow chemical cues . It is not clear what happens to a migrating cell in a living organism if the formation of its filopodia is impaired , or even how filipodia help the normal migration of cells in animals . To define how filopodia help to guide migrating cells in an animal , Meyen et al . analyzed the migration of cells called ‘primordial germ cells’ ( or PGCs ) in zebrafish . These cells form very early on in development of a zebrafish embryo at a position that is far away from their final location ( in the testes or ovaries where they will go on to form sperm or egg cells respectively ) . Meyen et al . revealed that cells that are exposed to the guidance cue ( a protein called a chemokine ) form more filopodia at their front compared to their rear . The filopodia formed at the cell front also extend and retract more frequently . Meyen et al . further observed that the specific chemokine that guides the cells can bind to the filopodia and enter the cell . This leads to a signal inside the cell that tells the cell to move in the direction where more of the chemokine is found . Indeed , altering the distribution and number of filopodia around the cell's edge decreases the ability of the primordial germ cells to reach their targets . Together , this work shows that the filopodia at the front end of cells are required for sensing the chemokines that guide cell movement . Further work is required to understand the mechanism that determines the distribution of filopodia on the surface of migrating cells , and the role of chemokines in the process . Moreover , this work may also be relevant for understanding the migration of cancer cells , because several types of cancer can invade new tissues by following directional cues including chemokines . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology"
] | 2015 | Dynamic filopodia are required for chemokine-dependent intracellular polarization during guided cell migration in vivo |
Cells of multi-cellular organisms evolve toward uni-cellularity in the form of cancer and , if humans intervene , continue to evolve in cell culture . During this process , gene dosage relationships may evolve in novel ways to cope with the new environment and may regress back to the ancestral uni-cellular state . In this context , the evolution of sex chromosomes vis-a-vis autosomes is of particular interest . Here , we report the chromosomal evolution in ~ 600 cancer cell lines . Many of them jettisoned either Y or the inactive X; thus , free-living male and female cells converge by becoming ‘de-sexualized’ . Surprisingly , the active X often doubled , accompanied by the addition of one haploid complement of autosomes , leading to an X:A ratio of 2:3 from the extant ratio of 1:2 . Theoretical modeling of the frequency distribution of X:A karyotypes suggests that the 2:3 ratio confers a higher fitness and may reflect aspects of sex chromosome evolution .
Genomes of multi-cellular organisms evolve to ensure the survival and reproduction of the whole organisms . With human interventions akin to domestication , hundreds of cell lines survive as free-living cells that are not organized into tissues , organs or individuals ( Alberts et al . , 2002 ) . Evolution in such a quasi-unicellular state may be very different from the evolution as multi-cellular entities . Most cell lines are cancerous in origin but a few are derived from normal tissues ( Hayflick , 1998 ) . Regardless of their origin , they have all evolved characteristics for survival in the unicellular state that is distinct from their natural environments . Cell lines derived from cancer tissues are usually karyotypically less stable than normal cell lines ( Lengauer et al . , 1997 ) . While this instability may impose a cost , it also permits cancer cell lines to evolve new karyotypes , including polyploidy , more readily than normal cell lines could . Tumorigenesis has been increasingly viewed as a process of evolution , rather than merely pathological conditions ( Nowell , 1976; Merlo et al . , 2006 ) . This ‘ultra-microevolutionary process’ is subjected to similar rules including mutation , genetic drift , migration and selection that govern organismal evolution ( Wu et al . , 2016 ) . While this process usually ends when the organism dies , cell lines in the cultured state will continue to evolve . Much like the diversity unleashed by domestication , cultured cell lines , which can be considered ‘domesticated’ , may be informative about the evolutionary potentials at the cellular level . In this quasi-unicellular state , gene dosage has been observed to change extensively as polyploidy , aneuploidy ( full or partial ) and various copy number variations ( CNVs ) are common in cancer cell lines ( Roschke et al . , 2003 ) . Since these cell lines are derived from somatic tissues of men or women ( referred to as male and female cells , for simplicity ) , they should be different in their sex chromosomes in relation to the autosomes ( A’s ) . Nevertheless , the possibility of separate evolutionary paths has not been raised before . Somatic cells have an inactive X chromosome in females and a Y chromosome in males ( Charlesworth , 1991 ) . Since cell lines presumably do not need sexual characters , we ask how the X:A relationship might have evolved in both male and female cells . More generally , we ask whether the evolution in this relationship may shed light on the emergence of mammalian sex chromosomes and their subsequent evolution . In this study , we analyze 620 cancer cell lines that have been genotyped using SNP arrays ( Greenman et al . , 2010 ) . Among them , 279 are derived from female tissues and 341 from male tissues . We observed the elimination of the Y and the inactive X chromosome , followed by the evolution toward a new equilibrium with two active X chromosomes and 3 sets of autosomes ( 2X:3A ) . We discuss the implication of these findings for the evolution of sex chromosome , the transition between uni- and multi-cellularity and cancers biology .
The most common form of genomic changes in cell lines is the loss of heterozygosity ( LOH ) when one of the two homologous chromosomes is eliminated ( Roschke et al . , 2003 ) . We therefore examine single nucleotide polymorphisms ( SNPs ) across the 620 cell lines for occurrences of LOH on each autosome and the X chromosome . Male and female cell lines are separately analyzed . Figure 1A shows the LOH frequency for each autosome ( black dots ) and the red dot represents the sex chromosomes ( X in female and Y in male ) . For autosomes , the percentages of LOH are remarkably similar between sexes , with a correlation coefficient of 0 . 94 among 620 cell lines . There is a slight tendency for the smaller autosomes to have higher LOH rate ( R =~−0 . 4 , p=~0 . 046 , Figure 1—figure supplement 1 ) . The median percentage of LOH is about 13% for autosomes . However , the losses of X ( 37% in females ) and Y ( 40% in males ) stand out . Given its rank as the 7th largest chromosome , the X is not expected to be lost in more than 15% of cell lines , based on the regression analysis of Figure 1A . Since the expression from the X is not lost , we infer that it’s the inactive X ( or Xi ) that is eliminated . Female lines lose the inactive X ( Xi ) and male lines lose the Y chromosome at a higher rate than other chromosomes . The two sexes may thus be expected to converge toward having a single sex chromosome . Furthermore , given that spontaneous LOH is not infrequent and the loss cannot be regained , long-term cultures might evolve to complete LOH for sex chromosomes as well as autosomes . The genome-wide low rate of LOH suggests selection holding back such changes . The strong correlation between sexes further reflects a balance between the production and elimination of LOH’s , likely involved natural selection . A most unexpected finding is that , accompanying the loss of the Y or Xi , an extra X chromosome is often gained . Figure 1B shows approximately equal numbers of male cell lines with one or two X chromosomes ( partial X aneuploidy not counted ) . This extra X is active because the inactivating XIST lncRNA is silenced in male cell lines ( Figure 1C ) , consistent with previous findings ( Guttenbach et al . , 1995 ) . XIST does not become activated in free-living cells that do not already express this . The expression of X-linked genes is higher in those male lines with two X’s than in those with one X and the up-regulation occurs along the length of the X chromosome ( Figure 1D ) . The pattern is more complex in female lines which , in their original state , contain an Xa and an Xi , the latter expressing XIST ( Chow et al . , 2005; Plath et al . , 2002; Ng et al . , 2007 ) . We use only female lines that show LOH of the whole X chromosome ( ~37% of female lines ) in counting Xa’s for the following reason . In order to count active Xa’s , we require the absence of XIST expression in the line such that all X’s can be assumed active . Figure 1E shows that female lines with LOH indeed rarely express XIST , presumably because LOH lines that survive lost the inactive X and kept the active Xa . In contrast , non-LOH lines tend to express XIST , thus obscuring the counting of Xa’s . Of the 103 LOH female lines , 30 lines have single Xa and 31 lines have whole extra X’s as shown in Figure 1F . Much like male lines of Figure 1B , Figure 1F also shows roughly half of female lines to have gained an extra Xa . Cancer cell lines usually have high rates of aneuploidy and could be heterogeneous within a given line , thus making its status difficult to assess . To assess the level of within-line heterogeneity , we chose two representative cell lines to count the X chromosomes in individual cells using fluorescent in situ hybridization ( FISH ) . The two lines are A549 ( a male cell line from adenocarcinomic alveolar basal epithelium ) and HeLa ( a female cervical cancer cell line ) . Neither line expresses XIST ( Supplementary file 1 ) , suggesting that all X chromosomes are active . Figure 2A–B shows results from individual A549 and HeLa cells with two and three X’s . Figure 2C–D shows the X karyotype distributions . While there is a modest degree of heterogeneity within each line , almost all cells have two or more active X chromosomes . While labor intensity of assays and cell availability limited our sample size , we nevertheless can conclude that within-cell line heterogeneity does not seem to undermine our conclusions . With an extra copy of the active X , the ‘expression phenotype’ is expected to change . The ratio of the median gene expression on the X to that on the autosomes ( EX/A ) is of particular interest . EX/A has been reported to be around 0 . 5 ~ 0 . 8 for normal mammalian tissues ( Xiong et al . , 2010; Deng et al . , 2011; Kharchenko et al . , 2011 ) . We assayed EX/A by separating lines derived from cancerous and normal tissues . Figure 3A shows that EX/A distributions center on ~ 0 . 84 in normal cell lines and on one in cancerous cell lines . Given the controversy in the assay of EX/A , we also varied the threshold for counting expressed transcripts ( see Materials and methods ) . By varying the threshold ( Figure 3B ) , EX/A ranges from 0 . 78 to 1 . 05 in normal cell lines but is consistently higher by approximately 15% in cancer cell lines . The same pattern is seen in the RNA-seq data ( Figure 3—figure supplement 1 ) . While sex chromosomes evolve , autosomes should also evolve . Since the generation of aneuploidy may happen independently for each autosome , a key question is whether selection operates on the autosomes as a set . Does natural selection favor cells that have full sets of autosomes ? Figure 4A shows the distribution of chromosome number across the 620 cell lines we studied . Apparently , cancerous cell lines acquire autosomes during evolution . The distribution of ploidy ( n = 22 ) number shows peaks at 2 and 3 , indicates that many cell lines appear to be in transition between full diploidy and triploidy of 44 and 66 autosomes . Similarly , the majority of sublines of HeLa cells we examined have 55–75 chromosomes centering about the triploid count of 69 ( Figure 4—figure supplement 1 ) . Indeed , autosomes appear to exist as a full complement with n = 22 . Although autosomes may evolve as a set , cells most likely add one autosome at a time . It is hence desirable to track each chromosome individually . Single cells were individually isolated from a HeLa cell line and subsequently grown to a sub-line of 106 cells . We subjected six such sub-lines to whole genome sequencing such that each chromosome could be tracked individually . Smaller chromosomes are indeed more erratic in their numbers in cell lines . Only the largest 14 chromosomes ( 13 autosomes and X ) , which together account for ~75% of the genome , are used to test the convergence of autosomes . The cutoff is based on the observation that chromosome 13 is the largest autosome yielding viable trisomic new-borns ( Taylor , 1968; Patterson , 2009; Kleijer et al . , 2006 ) . We reason that , if whole organisms can survive trisomy , the fitness consequence of the particular aneuploidy would probably be very small at the cellular level . In all 6 lines , each of the 13 autosomes has 2–4 copies , ranging from an average of 2 . 62 to 3 . 23 ( Supplementary file 2 ) . If each autosome behaves independently , the number of autosomes that increase by x copies ( x = 0 , 1 , 2 etc . ) should follow a Poisson distribution with a mean of λ . Two different lines , with λ = 10/13 and λ = 16/13 , are shown in Figure 4B and C . In the former , all cells have x = 0 or x = 1 and , in the latter , all cells have x = 1 or x = 2 ( Supplementary file 2 ) . The data suggest that each autosome increases by one copy and only after all of the 13 autosomes have gained an extra copy do further increases continue . Figure 4—figure supplement 1 shows the composite distribution of the five lines with λ < 1 . The pattern , like that of Figure 4B , is statistically significant ( p=0 . 0021 by the χ2 test ) with an excess at x = 1 . These results suggest that the larger autosomes evolve cohesively as a set . With autosomes evolving as a cohesive unit , X:A can be represented by whole numbers of 1:2 , 2:3 etc . We now summarize the evolution of cell lines by their C ( Xa:A ) genotypes . C ( Xa:A ) is the number of active X chromosomes and the ploidy number of autosomes ( in multiples of 22 ) and is equal to C ( 1:2 ) in normal cells . For the purpose of counting active Xa’s , data from most male lines are usable . For female lines , only data from the LOH lines of the X can be used . Between the two sexes , C ( Xa:A ) distributions are very similar and the combined distribution is used in the analysis ( Figure 4—figure supplement 2 ) . Shown in Figure 4D , most lines have the C ( 1:2 ) or C ( 2:3 ) genotype which together account for 2/3 of the lines . Given that C ( 1:2 ) is the starting genotype , its common occurrence at 37 . 4% is not surprising . The high frequency of C ( 2:3 ) , however , is unexpected . To reach C ( 2:3 ) from the starting point of C ( 1:2 ) , cells should evolve to either C ( 2:2 ) or C ( 1:3 ) first , but neither genotype is commonly seen in these cells lines . In contrast , C ( 2:3 ) at 29 . 2% is the second most common genotype . If we include the two genotypes , C ( 2:4 ) and C ( 3:3 ) , that are derivatives of C ( 2:3 ) , this inclusive C ( 2:3 ) cluster is the most common genotype . The model p the next section helps to interpret the observation . The pathways of chromosomal evolution can be diagrammed as a series steps in Figure 5A . Each node represents a C ( Xa:A ) genotype , the abundance of which is reflected in the size of the node . Thicker arrows indicate faster transitions which add/delete one X while the thinner arrow denotes the slower transition of adding/deleting the whole set of autosomes . The fitness of each genotype , W , is assumed to be determined by the Xa/A ratio . In general , one would expect the wild type ( W1 ) to be the fittest genotype and we particularly wish to know whether that is indeed the case here . We first model the evolution under strict neutrality where all nodes have the same fitness . For simplicity , genotypes are grouped into 3 clusters centering around the 3 dominant genotypes , C ( 1:2 ) , C ( 2:2 ) and C ( 2:3 ) , the frequencies of which are x1 , x2 and x3 , respectively . Each cluster consists of the dominant genotype as well as the less common ones adjacent to it ( see Figure 5A ) . For instance , x2 is the sum of the frequencies of C ( 2:2 ) and C ( 3:2 ) and x1 is those of C ( 1:2 ) , C ( 1:1 ) and half of C ( 1:3 ) . The frequency of the last one , being adjacent to both C ( 1:2 ) and C ( 2:3 ) , is split between the two clusters . Tallying up the numbers in Figure 4D , we obtain x1 = 0 . 41 , x2 = 0 . 092 and x3 = 0 . 482 with a total of 0 . 984 , excluding the marginal genotypes . The analysis below can be expanded to account for each genotype separately . The transitions between clusters are defined as follows:x1Tu⇌au x2Tv⇌bvx3T where u and v are the transition rates and xi ( T ) is the frequency of cluster i at time T . Let X ( T ) be the vector of [x1 ( T ) , x2 ( T ) , x3 ( T ) ] , expressed as ( 1 ) X ( T ) =X ( 0 ) [1−uu0au1−au−vv0bv1−bv]T When T >> 0 , ( 2 ) [x1 ( T ) , x2 ( T ) , x3 ( T ) ]∼[ab , b , 1]/z where z = ab + b+1 . The genotype frequencies evolve toward the equilibrium , [ab , b , 1]/z , which depends on a and b , but not u and v . We posit that a > 1 and b > 1 because , as the chromosome number increases , the probability of chromosome gain/loss increases as well . By - Equation 2 , x1 ( T ) > x2 ( T ) > x3 ( T ) when T >> 0 . In short , the relative frequency should be in the descending order of C ( 1:2 ) , C ( 2:2 ) and C ( 2:3 ) if there is no fitness difference among genotypes . This predicted inequality at T >> 0 is very different from the observed trend . Equation 2 assumes that cell lines have been evolving long enough to approach this equilibrium . A more appropriate representation should be X ( T ) where T reflects the time a cell line has been in culture . It is algebraically simpler if T is measured by the rate of chromosomal changes , u or v , rather than by the actual cell generation ( Equation 1 , Figure 5B and legends ) . We also assume u > v as u involves only the X but v involves the whole set of autosomes . With the initial condition of X ( 0 ) = [1 , 0 , 0] , Figure 5B shows that the C ( 2:3 ) cluster approaches the equilibrium more slowly than the other two clusters . Therefore , the observed high frequency of the C ( 2:3 ) cluster ( x3 = 0 . 482 vs . x1 = 0 . 41 and x2 = 0 . 092 ) is incompatible with a neutrally evolving model of chromosome numbers . The discrepancy is true at all time points and is more pronounced at smaller T’s . Rejecting the neutral evolution model , we now incorporate fitness differences into Figure 5A with W1 = 1 [for C ( 1:2 ) and C ( 2:4 ) ] , W2 = 1 + s [for C ( 2:2 ) ] and W3 = 1 + t [for C ( 2:3 ) ] where s and t can either be positive or negative . Here , we add a fourth genotype , C ( 2:4 ) . In the supplement , we model 4 genotypes with x1 – x4 for the frequencies of C ( 1:2 ) , C ( 2:2 ) , C ( 2:3 ) and C ( 2:4 ) respectively . An expanded transition matrix is used to model selection , followed by a normalization step ( Supplementary file 3 , Equation S1 ) . The solution in the form of X ( T ) =X ( 0 ) MT is given in Supplementary file 3 ( Equation S2 ) and the equilibrium X ( T ) is given in Supplementary file 3 ( Equation S3 ) . We are particularly interested in whether t > 0 in the 4-genotype model , that is , whether C ( 2:3 ) has a higher fitness than the wild type , C ( 1:2 ) . We observe that [x1 , x2 , x3 , x4] = [0 . 374 , 0 . 087 , 0 , 292 , 0 . 128] where x3 = 0 . 292 is more than 3 times higher than x2 = 0 . 087 and is close to x1 = 0 . 374 . Equation S3 shows that s < 0 is necessary for x2 to be smaller than x3 , and t > 0 is necessary for x3 to be close to x1 ( see Supplement ) . Figure 5C is an example in which s = −0 . 5 and t = 0 . 5 . The equilibrium at T >> 0 is indeed close to the observed values . In conclusion , it appears that the extant state in multicellular organisms of C ( 1:2 ) is not the fittest genotype for free-living mammalian cells . The observed genotypic distributions suggest that C ( 2:3 ) may have a higher fitness than the wild type , C ( 1:2 ) .
Free-living mammalian cells like all living things speed up the evolution when the environment changes . The practice of cell culturing , however , is to slow down the evolution to preserve cell lines’ usefulness as proxies for the source tissues . Nevertheless , changes are inevitable and the evolution of sex chromosomes is but one example . It should be noted that cell lines derived from cancerous tissues and normal tissues are different in one important aspect . Cell lines derived from normal tissues generally do not undergo karyotypic changes at an appreciable rate ( Shirley et al . , 2012; Frazer et al . , 2007; Pickrell et al . , 2010 ) . They are therefore much less responsive to selection in cultured conditions that favor new karyotypes . Cancer cell lines , having been through more rounds of passages , have generally experienced stronger selection more frequently than normal cell lines . Our observations suggest that the extant X:A relationship ( C ( 1:2 ) ) may not be optimal for free-living mammalian cells . The highest fitness peak , instead , appears to be closer to the karyotype of C ( 2:3 ) as free-living cells reproducibly evolve toward this new karyotype . The fitness peaks in free-living cells being different from that of the multi-cellular organisms is not unexpected . With many possible conflicts between individual cells and the community of cells ( i . e . , the organism ) , the interest of the community may lie in its ability to regulate the growth potential of its constituents . Free-living cells , on the other hand , are driven by selection to realize their individual proliferative capacity relative to other cells . The convergence among these many cell lines to C ( 2:3 ) is unexpected in the context of cancer evolution . The TCGA project ( reference ) has shown that cancer evolution is a process of divergence , not convergence . Indeed , only two genes have been mutated in more than 10% of all cancer cases and tumors of the same tissue origin from two different patients may often share no mutated genes at all ( Wu et al . , 2016; Kandoth et al . , 2013 ) . Therefore , the karyotypic convergence reported here is rather unusual . We note that C ( 2:3 ) toward which cultured cells evolved happens to be the smallest possible increase in the X/A ratio from C ( 1:2 ) . The higher fitness of C ( 2:3 ) than C ( 1:2 ) in free-living cells may lend new clues to the debate about the evolution of mammalian sex chromosomes ( Kharchenko et al . , 2011; Lin et al . , 2012 ) . With X-inactivation , it has been suggested that EX/A could have been reduced , or even halved ( Xiong et al . , 2010; Lin et al . , 2012 ) . The debate is about whether , and by how much , EX/A might have increased in evolution . Our observation that free-living cells continue to evolve toward C ( 2:3 ) raised the possibility that the evolutionary increase in EX/A has not been complete , in comparison with the ancestral EX/A . Finally , this study of cancerous cell lines may also have medical implications . The common view that tumorigenesis is an evolutionary phenomenon posits that individual cells in tumors evolve to enhance self-interest ( Nowell , 1976; Merlo et al . , 2006; Chen et al . , 2015; Chen and He , 2016 ) . A corollary would be that tumorigenesis may have taken the first few steps toward uni-cellularity . This extended view is supported by many expression studies as well as the higher likelihood of obtaining cell lines from tumors than from normal tissues ( Hayflick , 1998 ) . An alternative view , posits that tumors remain multi-cellular in organization ( Almendro et al . , 2013 ) . These different views have been critically examined recently ( Wu et al . , 2016 ) . It is possible that cancer cells in vivo may have been gradually evolving toward a new optimum . In that case , cancer cells in men and women are converging in their sex chromosome evolution and become more efficient in proliferation in this new de-sexualized state .
The processing of clonal expansion and whole genome sequencing of HeLa lines are described at Zhang et . al . ( https://www . biorxiv . org/content/early/2017/10/05/193482 ) . For each line , the copies of each chromosome are estimated according to the average sequencing depth by Control-FREEC , a tool for assessing copy number using next generation sequencing data ( Boeva et al . , 2012 ) . Three large-scale datasets were used in this study ( Greenman et al . , 2010; Barretina et al . , 2012; Cheung et al . , 2010 ) . Genome-wide SNP array data on cancer cell lines and a normal training set were downloaded from The Wellcome Trust Sanger Institute under the data transfer agreement . Among the 755 cancer cell lines , 620 ( from 279 females and 341 males ) with available gender information were used for genotype information analysis in the present study . The details of these cell lines are shown in Supplementary file 4 . The processed data are in PICNIC output file format , which includes information on genotype , loss of heterozygosity and absolute allelic copy number segmentation ( Greenman et al . , 2010 ) . Greenman et . al . developed the algorithm , PICNIC ( Predicting Integral Copy Number In Cancer ) , to predict absolute allelic copy number variation in cancer ( Greenman et al . , 2010 ) . This algorithm improved the normalization of the data and the determination of the underlying copy number of each segment . It has been used for Affymetrix genome-wide SNP6 . 0 data from 755 cancer cell lines , which were derived from 32 tissues . The Affymetrix Genome-Wide SNP Array 6 . 0 has 1 . 8 million genetic markers , including more than 900 , 000 single nucleotide polymorphism probes ( SNP probes ) and more than 900 , 000 probes for the detection of copy number variation ( CN probes ) . The genome-wide gene expression data for 947 human cancer cell lines from 36 tumor types were generated by Barretina et al ( Barretina et al . , 2012 ) , as part of the cancer cell line Encyclopedia ( CCLE ) project using Affymetrix U133 plus 2 . 0 arrays and are available from the CCLE project website ( CCLE_Expression_Entrez_2012-09-29 . gct , http://www . broadinstitute . org/ccle/home ) . The expression profiles of 768 cell lines with gender information , representing 337 females and 431 males , were used in this study . These cell lines were partially overlapped with the lines used in Greenman et . al . Additionally , RNA-seq data from 41 lymphoblastoid cell lines from 17 females and 24 males were downloaded from GEO database ( GSE16921 ) ( Cheung et al . , 2010 ) . The details of these cell lines are shown in Supplementary file 5 . Human genomes harbor single nucleoid polymorphisms ( SNPs ) at a density of about 0 . 5–1 SNP per kb . When a large segment of chromosome is lost in somatic cells , the corresponding region would be devoid of SNPs , referred to as loss of heterozygosity ( LOH ) . LOH regions may regain the copy number but the lost heterozygosity cannot be regained . We used the genotype information and the allelic copy number estimation generated from PICNIC to infer LOH as well as copy number of a specific chromosome . As for a chromosome , if ≥ 95% of SNP sites were homologous we considered that there was a LOH ( loss of heterogeneity ) event for this chromosome . Similarly , if ≥ 95% of detected alleles on the chromosome had a constant copy number of 0 , 1 , 2 , 3 or 4 , the copy number would be considered as the copy number of the chromosome . The copy number of the Y chromosome was estimated separately . In females , although all sites on Y chromosome should have yielded 0 copies , only ~ 60% of sites detected by the Y chromosome probes showed a copy number of 0 . This result indicated that several X homologous regions on the Y were covered by ~ 30% of Y probes . Therefore , Y chromosome loss was defined as when more than 60% of SNP probes from the Y chromosome showed a copy number of 0 . The expression level of XIST can be used as a proxy to distinguish the active X chromosome from the silent one as this gene was expressed on the inactive X chromosome and functioned in cis ( Richardson et al . , 2006 ) . According to Greenman’s and Barretina’s studies , 496 cancer cell lines have both copy number and expression data . As expected , XIST was silenced in male cell lines , as well as in females with whole X chromosome LOH ( Figure 1C ) . Based on X chromosome LOH and copy number information , we identified five genotypes , including XaO ( female lines with one X = 20 lines ) , XaXa ( female lines with isodisomy of X = 17 lines ) , XaXb ( female lines with heterozygous for the X = 28 lines ) , Xa[Y] ( male lines with one X = 53 lines ) and XaXa[Y] ( male lines with two X's = 69 lines ) . All male ( 341 lines ) and female cell lines with whole X chromosome LOH ( 103 lines ) were employed for C ( Xa:A ) calculation . C ( Xa:A ) was defined as the ratio of absolute X copy number to that of all autosomes . EX/A was defined as the ratio of the expression of X-linked genes to that of autosomal ones . The median values of expressed X-linked and autosomal genes were used to calculate EX/A in both cancerous and normal cell lines . For the datasets from the Affymetrix U133 + 2 . 0 array , genes with signal intensities ≥ 32 ( log2 ≥ 5 ) were considered to be expressed . While as for RNA-seq data , genes with RPKM values ≥ 1 were considered to be expressed . Previous studies have shown that EX/A value may be affected by gene set used ( Deng et al . , 2011 ) . In addition , several silent genes in normal tissues have been shown to be expressed in tumor tissues ( Hofmann et al . , 2008 ) . Those genes were dominant on X chromosome , which could result in an increase of EX/A . To exclude the possibility that EX/A ratios may be biased in cancerous cell lines , gene sets for EX/A calculation were first selected in normal cell lines by three criteria , with the same sets then selected in cancerous cell lines . The three filtering criteria for gene set selection were RPKM > 0 , 1 , and 5 in normal cell lines ( Figure 2C ) . To explore the impact of extra X chromosome on gene expression levels of X-linked genes , 53 cell lines with Xa[Y] and expression data , 69 cell lines with XaXa[Y] and expression data were used . T-test with Benjamini and Hochberg adjusting method was employed to determine genes , the expression of which are significantly changed due to an extra X copy . 648 detected X-linked genes are plotted in Figure 2A . The free statistical programming language R was used for the statistical analysis ( version 3 . 0 . 1 ) . HeLa cells ( from the Culture Collection of the Chinese Academy of Sciences , Shanghai , China ) were cultured in DMEM ( Life Technologies , CA , United State ) supplemented with 10% fetal bovine serum ( FBS ) , 100 U/ml of penicillin , and 100 μg/ml of streptomycin . A549 cells ( from ATCC ) were cultured in RPMI-1640 ( Life Technologies ) with 10% fetal bovine serum ( FBS ) , 100 U/ml of penicillin , and 100 μg/ml of streptomycin at 37°C with 5% CO2 . Approximately 2 × 106 cells were seeded and cultured in 10 cm dishes with 10 ml growth medium as described above . To synchronize the cells , 200 μl of thymidine ( 100 mM ) was added to the cells . After incubating for 14 hr , the cells were washed twice with 10 ml PBS and then supplemented with 10 ml growth medium containing deoxycytidine ( 24 μM ) . After incubating for 2 hr , 10 µl nocodazole ( 100 μg/ml ) was added to the cells . The cells were incubated for an additional 10 hr . After synchronization , cells were harvested and treated with 4 ml hypotonic solution ( 75 mM , KCl ) pre-warmed to 37°C for 30 min . The cells were then fixed via three immersions in fresh fixative solution ( 3:1 methanol:acetic acid ) ( 15 min each time ) . The fixed cell suspension was spotted onto a clean microscope slide and allowed to air dry . We used the ‘‘XCyting Chromosome Paints’’ and ‘Xcyting Centromere Enumeration Probe’ ( MetaSystems , Germany ) for whole X chromosomes and centromere of X chromosome fluorescence in situ hybridization ( FISH ) analysis , respectively . Following the manufacturer's instructions , 10 µl of probe mixture was added to the prepared slide . The slide was then covered with 22 × 22 mm2 cover slip and sealed with rubber cement . Next , the slide was heated at 75°C for 2 min on a hotplate to denature the sample and probes simultaneously , followed by incubation in a humidified chamber at 37°C overnight for hybridization . After hybridization , the slide was washed in 0 . 4 x SSC ( pH 7 . 0 ) at 72°C for 2 min , then in 2 x SSC and 0 . 05% Tween-20 ( pH 7 . 0 ) at room temperature for 30 s , before being rinsed briefly in distilled water to avoid crystal formation . The slide was drained and allowed to air dry . Finally , 5 µl DAPI ( MetaSystems ) was applied to the hybridization region and covered with a coverslip . The slide was processed and captured using fluorescence microscopy as recommended ( Olympus FV1000 , 100X objective ) . The number of Xs were counted for each individual cell . A total of 343 HeLa cells and 170 A549 cells were screened . The identification of HeLa cells was confirmed by genome sequence method and the identification of A549 cells was confirmed by karyotype profile . The mycoplasma contamination status was tested by DNA staining for both HeLa and A549 . In the model , autosomes are treated as an integrated set , labeled ‘A’ and counted as a set . There may be two reasons to do so . One is mechanistic if the entire haploid set of chromosomes increases a unit . While this may happen in organismal evolution , we consider the mechanism dubious for cell lines . In the absence of meiosis , whole-sale changes should involve the entire diploid set ( diploids , tetraploids and octoploids , as in human hepatocytes ) . We therefore suggest that chromosomes are gained and lost individually . They evolve more or less as a cohesive set in the long run thanks to natural selection that imposes a cost on uneven sets . For autosomes , the dynamics is portrayed in Figure 4B–C . When autosomes are gained , say from nA = 2 to nA = 3 ( nA being the number of autosomal haploid set ) , the imbalance within the autosomal set appears to be tolerated only to a point . Let ∆ij ( i , j = 1 , 2 , . . 13 for largest 13 autosomes ) designate the difference in the number between autosome i and autosome j . In a balanced set , all ∆ij = 0 which represents a fitness peak when all autosomes have the same number . During evolution , ∆ij = 1 , having a reduced fitness , can be tolerated but not ∆ij >= 2 . ( The constraint appears to be loosened for the smaller autosomes . ) Thus , the autosomal set evolves between integers of nA = 1–3 , with the occasional nA = 4 ( see Figure 4D ) . Obviously , moving nA from one whole number to the next is a slow process . In Equation ( 1 ) of the main text , u represents the change in the number of X and v represents the change in nA . In testing the model , we let u = 10 v but the results are not sensitive to the ratio . Assuming all genotypes have whole numbers of X and A , we assign a fitness to all genotypes of Figure 4A . Under neutrality , these genotypes have the same fitness . The main goal of the modeling work ( see Figure 5B and C ) is to test the fitness neutrality of these ‘whole number’ genotypes . When we attribute the observed genetic changes associated with cells’ unicellular existence , we do include all environmental factors that make the unicellular existence possible . Without these factors , cell lines cannot live . An analogy is the study of the evolution of social structure , which is also conditional on many environmental factors ( e . g . , food supply ) but one often uses ‘social structure’ as an all-encompassing term . Since the unicellular existence requires a number of environmental factors ( which the cell culture community has been keen to identify ) , it is not possible to separate ‘unicellularty’ and the environments needed to sustain the unicellular existence . It is also important to point out that these environmental factors are often antagonistic to the multicellular living . In this study , we use the model to compare the observations with the neutral expectation . Although we could reject the neutral model and conclude the direction of selection , we refrain from estimating the strength of selection for two reasons – both biological and technical . First , the most important demonstration is that the wildtype C ( Xa:A ) =1:2 is not the fittest genotype . We believe this conclusion in itself is very novel because all evolutionary theories posit the wildtype to be at a local fitness optimum . We could conclude that the wild type C ( Xa:A ) =1:2 is less fit than C ( Xa:A ) =2:3 based on Figure 5B . Second , while we could conclude C ( Xa:A ) =1:2 to be less fit than C ( Xa:A ) =2:3 , estimating the strength of selection is an entirely different proposition . In this case , the main unknown is vT in Figure 5B and C . In other words , we do not know how close each cell line is to the equilibrium . Given the various histories of these cell lines , we suspect that the value may range between 0 . 1 and 10 . The differences are qualitatively consistent but the actual values will require knowing the precise culture history of each cell line . | Multicellular life relies on a group of cells working together for a common interest . To study these cells , researchers take them out of the organism and grow them in the laboratory . Instead of growing as part of organs and tissues , the cells normally have a free-living lifestyle . Because multicellular life evolved from single-celled organisms , laboratory-grown cells can be considered as life forms that are evolving backward from a multicellular to a single-celled existence . Normally , the cells that make up most of the tissues in the human body have 22 pairs of chromosomes known as autosomes and a pair of sex chromosomes . The cells of women have two X sex chromosomes , one of which is inactive , while those of men have one X and one Y chromosome . However , free-living single cells do not need to distinguish between male and female cells . Xu , Peng , Chen et al . have now studied the chromosomes of cancer cells taken from over 600 people and grown in the laboratory . As the cells evolved in response to their free-living lifestyle , they became ‘de-sexualized’; male cells lost their Y chromosome , while female cells abandoned their inactive X chromosome . The cells then evolved toward a new state in which they possessed two active X chromosomes and three sets of autosomes . This new configuration suggests that the current X chromosome to autosome ratio may not be optimal for fitness and hence sheds some light on how mammalian sex chromosomes evolved . It is currently thought that as cancerous tumors grow , their cells evolve to favor their own interests over the common interests of the rest of the organism . In this way , they develop characteristics more like those of single cells . Further research is therefore needed to investigate whether changes occur to the chromosomes of cancer cells growing within the body , and whether this gives them an advantage over normal cells . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology"
] | 2017 | Free-living human cells reconfigure their chromosomes in the evolution back to uni-cellularity |
Senescent cells accumulate in fat with aging . We previously found genetic clearance of senescent cells from progeroid INK-ATTAC mice prevents lipodystrophy . Here we show that primary human senescent fat progenitors secrete activin A and directly inhibit adipogenesis in non-senescent progenitors . Blocking activin A partially restored lipid accumulation and expression of key adipogenic markers in differentiating progenitors exposed to senescent cells . Mouse fat tissue activin A increased with aging . Clearing senescent cells from 18-month-old naturally-aged INK-ATTAC mice reduced circulating activin A , blunted fat loss , and enhanced adipogenic transcription factor expression within 3 weeks . JAK inhibitor suppressed senescent cell activin A production and blunted senescent cell-mediated inhibition of adipogenesis . Eight weeks-treatment with ruxolitinib , an FDA-approved JAK1/2 inhibitor , reduced circulating activin A , preserved fat mass , reduced lipotoxicity , and increased insulin sensitivity in 22-month-old mice . Our study indicates targeting senescent cells or their products may alleviate age-related dysfunction of progenitors , adipose tissue , and metabolism .
A major function of adipose tissue is to store potentially cytotoxic lipids , including fatty acids ( FAs ) , as less reactive neutral triglycerides ( TG ) within fat droplets ( Listenberger et al . , 2003 ) . Lipid storage by adipose tissue appears to constitute a defense against lipotoxicity and metabolic disease ( Wang et al . , 2008; Unger and Scherer , 2010; Gustafson et al . , 2015; Tchkonia et al . , 2010 ) . Fat cells turn over throughout life , with generation of new fat cells through differentiation of fat progenitors ( also known as preadipocytes or adipose-derived stem cells ) ( Tchkonia et al . , 2013; Spalding et al . , 2008; Tchoukalova et al . , 2010 ) . Adipogenesis is orchestrated by a transcription factor cascade involving the two key regulators , peroxisome proliferator-activated receptor-γ ( PPARγ ) and CCAAT/enhancer binding protein-α ( C/EBPα ) ( Wu et al . , 1999; Lin and Lane , 1994 ) and their downstream targets , including fatty acid binding protein 4 ( FABP4 ) and perilipin ( PLIN1 ) ( Bernlohr et al . , 1997; Sun , 2013 ) . Compromised adipogenic capacity can contribute to impaired ability of adipose tissue to store lipids , leading to FA spillover and ectopic lipid accumulation in liver and other sites , insulin resistance , and lipotoxicity ( Garbarino and Sturley , 2009; Slawik and Vidal-Puig , 2006; Tchkonia et al . , 2006; Guo et al . , 2007 ) . By late middle age , capacity for adipogenesis , PPARγ and C/EBPα expression , adipose tissue mass , and metabolic function begin to decline in experimental animals and humans ( Tchkonia et al . , 2010; Slawik and Vidal-Puig , 2006; Fink et al . , 1983; Tchkonia et al . , 2013; Cowie et al . , 2006; North and Sinclair , 2012; Palmer et al . , 2015; Cartwright et al . , 2007; Raguso et al . , 2006; Kuk et al . , 2009; Cartwright et al . , 2010; Tchkonia et al . , 2007; Karagiannides et al . , 2001; Kirkland et al . , 1990 ) . This age-related lipodystrophy likely contributes to the pathogenesis of metabolic dysfunction at older ages ( Gustafson et al . , 2015; Tchkonia et al . , 2010; Tchkonia et al . , 2006; Guo et al . , 2007; Kuk et al . , 2009 ) . We hypothesize that cellular senescence could contribute to impaired adipogenesis and age-related lipodystrophy ( Tchkonia et al . , 2010 ) . Cellular senescence refers to an essentially irreversible arrest of cell proliferation ( Hayflickl and Moorhead , 1961 ) . It can be induced by a variety of stresses , including DNA damage , telomere shortening , radiation , chemotherapeutics , and reactive metabolites ( Tchkonia et al . , 2013; Campisi and d'Adda di Fagagna , 2007 ) . Senescent cells accumulate in adipose tissue with aging across a number of mammalian species ( Tchkonia et al . , 2010; Xu et al . , 2015; Stout et al . , 2014 ) and secrete an array of cytokines , chemokines , proteases , and growth factors—the senescence-associated secretory phenotype ( SASP ) ( Coppé et al . , 2008; Coppé et al . , 2010 ) . Cultures of progenitors isolated from adipose depots of older animals or humans contain senescent cells and exhibit impaired adipogenic capacity , with reduced lipid accumulation and C/EBPα and PPARγ expression after exposure to differentiation-inducing stimuli ( Tchkonia et al . , 2010; Tchkonia et al . , 2007; Park et al . , 2005; Mitterberger et al . , 2014 ) . Senescent cells appear to be able to spread inflammatory activation and perhaps even senescence to nearby non-senescent cells ( Xu et al . , 2015; Acosta et al . , 2013; Nelson et al . , 2012 ) . In previous work , we used a genetically modified INK-ATTAC ( Cdkn2a /p16Ink4a promoter driven apoptosis through targeted activation of caspase ) mouse model to selectively eliminate Cdkn2a ( p16Ink4a ) positive senescent cells through apoptosis by the administration of AP20187 , a drug that induces dimerization of a membrane-bound myristoylated FK506 binding protein fused with caspase 8 ( FKBP–Casp8 ) ( Baker et al . , 2011 ) . We showed that clearance of senescent cells can delay age-related phenotypes including lordokyphosis and cataract formation , and can actually reverse age-related fat loss in progeroid BubR1H/H animals ( Baker et al . , 2011 ) , implicating senescent cells as a driver of age-related phenotypes . Furthermore , interleukin-6 ( IL6 ) ( Gustafson and Smith , 2006; Okada et al . , 2012 ) , tumor necrosis factor α ( TNFα ) ( Tchkonia et al . , 2007; Gustafson and Smith , 2006; Okada et al . , 2012 ) , and interferon γ ( IFNγ ) ( McGillicuddy et al . , 2009 ) can inhibit adipogenesis in vitro . These factors are among the SASP components in senescent fat progenitors and other senescent cell types ( Tchkonia et al . , 2013; Xu et al . , 2015; Coppé et al . , 2008; Coppé et al . , 2010 ) . However , causal links between these paracrine factors and impaired adipogenesis related to cellular senescence have not been demonstrated . We recently reported that the JAK/STAT ( Janus kinase/signal transducer and activator of transcription ) pathway plays a role in regulating the SASP ( Xu et al . , 2015 ) . Therefore , we hypothesized that JAK inhibition might rescue impaired adipogenesis due to senescent cells and thus preserve fat mass and metabolic function in older individuals . We report here that senescent fat progenitors impede differentiation of non-senescent progenitors , in part by secreting activin A , a member of the transforming growth factor superfamily , which can inhibit adipogenesis and interfere with stem cell and progenitor function ( Zaragosi et al . , 2010 ) . Eliminating senescent cells from naturally-aged INK-ATTAC mice reduced activin A and increased adipose tissue C/EBPα and PPARγ . JAK pathway inhibition suppressed production of activin A by senescent fat progenitors and partially rescued adipogenic capacity both in vitro and in vivo . JAK inhibition in aged mice reduced lipotoxicity and increased insulin sensitivity . Our findings provide new insights into the mechanisms of age-related progenitor dysfunction , fat loss , and metabolic dysfunction , as well as potential therapeutic avenues for preventing or alleviating these common conditions .
To determine if senescent cells influence adipogenesis in adjacent non-senescent cells , we devised a co-culture system with non-senescent human primary fat progenitors as 'target' cells and either senescent or non-senescent human progenitors as 'source' cells . Primary cells were isolated from the stromal-vascular fraction of collagenase-digested subcutaneous fat from healthy human subjects undergoing surgery to donate a kidney . Cells were passaged 4–6 times under conditions to enrich for fat progenitors as opposed to endothelial cells or macrophages ( Tchkonia et al . , 2013 ) . These cells were exposed to 10 Gy irradiation , which induced at least 70% of cells to become senescence-associated β-galactosidase ( SABG ) -positive within 20 days , as previously described ( Xu et al . , 2015 ) . Target cells were distinguished from source cells by fluorescent labeling ( CM-DiI ) , which does not independently affect adipogenesis . We differentiated the mixture of cells using an adipogenic differentiation medium ( DM ) for 15 days . Differentiation was assessed by examining lipid accumulation inside the cells . We considered a cell to be differentiated if it contained doubly refractile lipid droplets visible by low power phase contrast microscopy , a change that occurs in fat cell progenitors following DM exposure , but not in other cell types ( Karagiannides et al . , 2006 ) . We found that senescent source cells were less differentiated than control non-senescent source cells ( Figure 1a ) . When co-cultured with senescent source cells , only 20% of target progenitors accumulated lipid compared to more than 50% when co-cultured with non-senescent source cells ( Figure 1b ) , indicating that senescent cells can directly impair lipid accumulation by nearby fat progenitors . 10 . 7554/eLife . 12997 . 003Figure 1 . Adipogenesis in human fat progenitors is impeded by co-culture with senescent cells . Primary subcutaneous human fat progenitors were labelled with DiI and seeded into wells containing either control or radiation-induced senescent preadipocytes . ( a ) Photographs were taken 15 days after initiating differentiation . Representative images are shown . DiI-positive cells are red and DAPI staining is blue . ( b ) Number of differentiated DiI positive cells as a percentage of total DiI positive cells is expressed as mean ± s . e . m . *p<0 . 00001 . Results were obtained using separate strains of fat progenitors harvested from 6 healthy human subjects during surgery to donate a kidney ( N=6 ) . Two-tailed Student's t tests were used to determine statistical significance . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 00310 . 7554/eLife . 12997 . 004Figure 1—source data 1 . Adipogenesis in human fat progenitors is impeded by co-culture with senescent cells . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 004 Next , we examined the nature of the factors responsible for impairing adipogenesis . Non-senescent progenitors were treated with DM in the presence of conditioned medium ( CM ) from cultures of senescent or non-senescent cells . Senescent progenitor CM reduced differentiated cell numbers in target non-senescent cells at all three time points tested ( Figure 2a ) . PPARγ , C/EBPα , FABP4 , and PLIN2 are normally up-regulated during adipogenesis ( Wu et al . , 1999; Lin and Lane , 1994; Bernlohr et al . , 1997; Sun , 2013 ) . Differentiation-dependent expression of these genes was blunted by CM from senescent cells compared to CM from blank culture flasks or control non-senescent cells ( Figure 2b ) . Cellular senescence did not appear to be induced in the target cells by the 15 days of CM exposure , since p16Ink4a and Cdkn1a ( p21Cip ) transcript levels were not increased ( Figure 2—figure supplement 1a ) . Adipogenesis was not impaired when exposure to CM was limited to 24 hours of pretreatment before exposure to DM ( Figure 2—figure supplement 1b ) . This suggests that impaired adipogenesis due to senescent CM depends on continued presence of products secreted by senescent cells . CM from doxorubicin-induced senescent cells suppressed adipogenesis similarly to CM from irradiation-induced senescent cells ( Figure 2—figure supplement 1c ) . 10 . 7554/eLife . 12997 . 005Figure 2 . Conditioned medium from senescent cells impedes adipogenesis in human progenitors . Conditioned medium ( CM ) was collected from a flask with no cells present ( Blank ) , control non-senescent ( CON ) , and senescent ( SEN ) fat progenitor cultures . Pooled human progenitors from subcutaneous fat of 5 healthy subjects were treated with 50:50% CM:differentiation medium ( DM ) for 15 days . ( a ) Representative images are shown at day 5 , 10 , and 15 of exposure to CM + DM . ( b ) Gene expression was analyzed by real-time PCR at day 5 , 10 , and 15 of exposure to CM + DM . Results are shown as fold change relative to the CON group at day 5 . Results were obtained using CM from 5 strains of human primary fat progenitors from different subjects and expressed as mean ± s . e . m . *p<0 . 05 . Two-tailed Student's t tests were used to determine statistical significance . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 00510 . 7554/eLife . 12997 . 006Figure 2—source data 1 . Conditioned medium from senescent cells impedes adipogenesis in human progenitors . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 00610 . 7554/eLife . 12997 . 007Figure 2—figure supplement 1 . Senescent cells impede adipogenesis in fat progenitors . ( a ) CM was collected from control non-senescent ( CON ) and senescent ( SEN ) fat progenitor cultures . Pooled human progenitors from subcutaneous fat of 5 healthy subjects were treated with 50:50% CM:DM for 15 days . Gene expression was analyzed by real-time PCR . Results were obtained using CM from 4 strains of human primary fat progenitors and expressed as mean ± s . e . m . ( b ) Pooled fat progenitors were pre-treated with CM collected from control ( CON 24h ) and senescent ( SEN 24h ) cells for 24 hours . Then they were treated with DM for 15 days . Gene expression was analyzed by real-time PCR . Results were obtained from 4 strains of human primary fat progenitors and expressed as mean ± s . e . m . ( c ) CM was collected from non-senescent ( CON ) and doxorubicin-induced senescent ( DOX ) fat progenitor cultures . Pooled human progenitors were treated with 50:50% CM:DM for 15 days . Gene expression was analyzed by real-time PCR . Results were obtained using CM from 3 strains of human primary fat progenitors and expressed as mean ± s . e . m . *p<0 . 05 ( d ) CM was collected from non-senescent ( CON ) and senescent ( SEN ) fat progenitor cultures . CM from SEN was separated into two fractions using molecular size filters with a cutoff at ~10 kd . The volumes of the fraction larger than ~10kd ( >10k ) and the fraction smaller than ~10kd ( <10k ) were matched to CM from SEN using blank CM . Pooled human fat progenitors were treated with 50:50% CM:DM for 10 days . Gene expression was analyzed by real-time PCR . Results were obtained using CM from 3 strains of human primary fat progenitors and expressed as mean ± s . e . m . *p<0 . 05 . ( e ) CM was collected from control non-senescent ( CON ) and senescent ( SEN ) fat progenitor cultures . Pooled human progenitors from subcutaneous fat of 5 healthy subjects were treated with 50:50% CM:DM for 5 days in presence of 20μg/ml of IGG ( SEN+IGG ) , IL6 antibody ( SEN+IL6 ab ) , IFNγ antibody ( SEN+ IFNγ ab ) or TNFα antibody ( SEN+TNFα ab ) . Gene expression was analyzed by real-time PCR . Results were obtained using CM from 2 strains of human primary fat progenitors and expressed as mean ± s . e . m . Two-tailed Student's t tests were used to determine statistical significance . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 007 We next investigated which factors secreted by senescent cells impair adipogenesis . We found that CM from senescent cells inhibited adipogenesis even after freeze-thaw cycles ( Figure 2a , b ) . Therefore , cell-cell contact or molecules with short half-lives , including many metabolites such as reactive oxygen species ( ROS ) , do not appear to be the sole responsible factors . CM from senescent cells was separated into two fractions using molecular size filters with a cutoff at ~10 kd . The fraction larger than ~10kd impaired adipogenesis while the fraction smaller than ~10kd had no effect ( Figure 2—figure supplement 1d ) . This led us to hypothesize SASP peptides or proteins might play a role in the inhibition of adipogenesis . Using either neutralizing antibodies or specific inhibitors , we inhibited candidate SASP factors in the CM , including IL6 , TNFα , IFN γ , and activin A , which can be secreted by senescent cells and inhibit adipogenesis ( Tchkonia et al . , 2007; Xu et al . , 2015; Gustafson and Smith , 2006; Okada et al . , 2012; McGillicuddy et al . , 2009; Zaragosi et al . , 2010 ) ( Figure 2— figure supplement 1e ) . Among the compounds screened , SB-431542 , an activin A receptor inhibitor ( Inman et al . , 2002 ) , substantially improved adipogenesis in progenitors exposed to CM from senescent cells , while only slightly increasing adipogenesis in control cells ( Figure 3a , b ) . Due to the fact that SB-431542 also inhibits TGFβ signaling ( Inman et al . , 2002 ) , to confirm further the role of activin A , we used activin A-specific neutralizing antibody and observed a similar enhancement of adipogenesis ( Figure 3c , d ) . Together , these findings indicate that activin A plays a role in the impairment of adipogenesis by senescent cells . 10 . 7554/eLife . 12997 . 008Figure 3 . Inhibition of activin A alleviates the impairment of adipogenesis induced by senescent progenitors . CM was collected from control ( CON ) and senescent ( SEN ) fat progenitors . Pooled human progenitors were treated with a 50:50 mixture of CM:DM in the presence of DMSO or 5μM SB431542 ( SB431542 ) . ( a ) Representative images are shown of differentiated cells at day 15 . ( b ) RNA was collected 7 days after differentiation and real-time PCR was performed . Pooled human progenitors were treated with a 50:50 mixture of CM:DM in the presence or absence of 1μg/ml activin A neutralizing antibody ( Activin A AB ) . ( c ) Representative images are shown of differentiated cells at day 15 . ( d ) RNA was collected 7 days after differentiation and real-time PCR was performed . Results are shown as fold change relative to the SEN group . Results were obtained using CM from 5 strains of human primary cells from different subjects and expressed as mean ± s . e . m . *p<0 . 05 . Two-tailed Student's t tests were used to determine statistical significance . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 00810 . 7554/eLife . 12997 . 009Figure 3—source data 1 . Inhibition of activin A alleviates the impairment of adipogenesis induced by senescent progenitors . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 009 After 17–18 months-of-age , mice begin to lose fat mass . We previously found that senescent cells start to accumulate noticeably before 18 months-of-age in mouse fat tissue ( Stout et al . , 2014 ) and senescent cells play a role in age-related loss of subcutaneous fat in animals with progeria ( Baker et al . , 2011 ) . However , it is still unknown whether senescent cell clearance has effect on age-related adipose phenotypes in naturally aged mice . To test this , we treated late middle-aged ( 18-month-old ) INK-ATTAC+/- and wild-type ( WT ) littermates with two 3-day courses of AP20187 , with 14 days between treatments , for 3 weeks ( total 6 days of treatment ) to activate the caspase-8 moiety in the ATTAC suicide gene product that is expressed only in p16Ink4apositive senescent cells . This allowed us to investigate the short-term response to senescent cell clearance , for example effects on adipogenic transcription factors , and to reduce effects of possible long-term compensatory responses . During the three-week treatment period , WT mice lost more fat than INK-ATTAC+/- mice ( Figure 4a ) , while lean mass ( Figure 4b ) and total body weight ( Figure 4c ) were unaffected . Circulating activin A was reduced more than 30% compared to baseline in the INK-ATTAC+/- mice , while activin A increased by 10% in the WT group ( Figure 4e ) . Activin A was also reduced in adipose tissue of the INK-ATTAC+/- mice ( Figure 4f ) . Adipose tissue expression of C/EBPα and PPARγ was higher in he INK-ATTAC+/- than WT mice ( Figure 4f ) , indicative of improved adipogenesis . Lipin-1 , whose expression in fat tissue is positively associated with adipose tissue function ( Nadra et al . , 2012 ) and insulin sensitivity ( Donkor et al . , 2008 ) , was also increased in the INK-ATTAC+/- mice ( Figure 4f ) . The senescence markers , IL6 , p16Ink4a , and p21Cip1 ( Figure 4f ) as well as SABG+ cells ( Figure 4d and Figure 4—figure supplement 1 ) , were reduced in fat tissue of AP20187-treated INK-ATTAC+/- mice . These results suggest that senescent cells are a cause of age-related adipose tissue loss and dysfunction in older mice . 10 . 7554/eLife . 12997 . 010Figure 4 . Genetic clearance of senescent cells blunts fat loss and increases adipogenic markers in fat of 18-month-old mice . Eighteen-month-old wild-type and INK-ATTAC+/- mice were treated with AP20187 for 3 weeks ( 10mg/kg , three consecutive days with 14 days rest between treatments; total 6 treatments ) . Fat mass ( a ) and lean mass ( b ) were measured by MRI along with body weight ( c ) before and after treatment . The percent changes relative to baseline are shown . Results ( N=8 ) are expressed as mean ± s . e . m . *p<0 . 05 for comparison between WT and INK-ATTAC+/- at 3 weeks . ( d ) SABG+ cells were counted in WAT and their percentages as a function of total cells ( N=7 ) are expressed as mean ± s . e . m . *p<0 . 05 . ( e ) Activin A protein in plasma was measured before and after treatment . The percent changes relative to baseline are shown . Results ( N=8 ) are expressed as mean ± s . e . m . *p<0 . 05 for comparison between WT and INK-ATTAC+/- at 3 weeks . ( f ) RNA from white adipose tissue ( WAT ) was collected and real-time PCR was performed . Results ( N=8 ) are expressed as mean ± s . e . m . *p<0 . 05 . Two-tailed Student's t tests were used to determine statistical significance . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 01010 . 7554/eLife . 12997 . 011Figure 4—source data 1 . Genetic clearance of senescent cells blunts fat loss and increases adipogenic markers in fat of 18-month-old mice . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 01110 . 7554/eLife . 12997 . 012Figure 4—figure supplement 1 . Genetic clearance of senescent cells reduced SABG+ cells in adipose tissue . Eighteen-month-old wild-type and INK-ATTAC+/- mice were treated with AP20187 for 3 weeks ( 10 mg/kg , three consecutive days with 14 days rest between treatments; total 6 treatments ) . WAT was collected and assayed for cellular SABG activity and counterstained with DAPI . The SABG+ cells are indicated by red arrows . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 01210 . 7554/eLife . 12997 . 013Figure 4—figure supplement 2 . Senescent cell clearance blunts fat loss in 18-month INK-ATTAC+/- mice . Eighteen-month-old wild-type and INK-ATTAC+/- mice were treated with AP20187 for 3 weeks ( 10mg/kg , three consecutive days with 14 days between treatments; total 6 treatments ) . Changes from baseline for fat mass , lean mass , and body weight are shown . Results ( N=8 ) are expressed as mean ± s . e . m . *p<0 . 05 . Two-tailed Student's t tests were used to determine statistical significance . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 013 We recently reported that JAK inhibition suppresses SASP factors , including IL6 and TNFα , in senescent fat progenitors ( Xu et al . , 2015 ) . We also previously observed that direct addition of recombinant activin A to cultured human fat progenitors impedes adipogenesis ( Zaragosi et al . , 2010 ) . Here , we found that JAK inhibition reduces activin A at both the transcript ( Figure 5a ) and secreted protein levels ( Figure 5b ) in senescent fat progenitors . We therefore tested whether JAK inhibition alleviates impaired adipogenesis related to senescence . CM prepared from senescent progenitors exposed to JAK inhibitor caused less inhibition of adipogenesis in non-senescent target progenitors than CM prepared from senescent cells exposed to vehicle ( Figure 5c , d ) . Since JAK inhibitor was present in the CM , we examined whether the improvement of adipogenesis in the target non-senescent cells was due to the effect of JAK inhibitor on the senescent source cells or if JAK inhibitor had direct effects on the target cells . Addition of JAK inhibitor directly to CM previously collected from either control or senescent cells did not affect adipogenesis in the target non-senescent cells ( Figure 5—figure supplement 1a ) . This indicates that JAK inhibitor alleviated impaired adipogenesis mainly by acting on the senescent source fat progenitors , in turn altering the composition of the CM , rather than having direct effects on the target cells . Moreover , JAK inhibition improved adipogenesis in cultures of fat progenitors isolated from aged rats , which contain senescent cells , but not in cultured progenitors isolated from young rats ( Figure 5—figure supplement 1b and c ) . 10 . 7554/eLife . 12997 . 014Figure 5 . JAK inhibition suppresses activin A production by senescent fat progenitors and partially rescues adipogenesis . Senescent human progenitors were treated with DMSO ( SEN ) or 0 . 6 μM JAK inhibitor 1 ( SEN+JAKi ) for 72 hours . ( a ) RNA was collected from control ( CON ) , SEN , and SEN+JAKi progenitors and real-time PCR was performed . Results ( N=7 ) are expressed as mean ± s . e . m . *p<0 . 05 . ( b ) CM was collected and activin A protein was assayed by ELISA . Results ( N=6 ) are expressed as mean ± s . e . m . *p<0 . 05 . ( c ) Representative images are shown of differentiating cells at day 10 . ( d ) RNA was collected 10 days after initiation of differentiation and real-time PCR was performed . Results are shown as fold change relative to the SEN group . Results were obtained using CM from 7 strains of human primary progenitors from different subjects and expressed as mean ± s . e . m . *p<0 . 05 . Two-tailed Student's t tests were used to determine statistical significance . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 01410 . 7554/eLife . 12997 . 015Figure 5—source data 1 . JAK inhibition suppresses activin A production by senescent fat progenitors and partially rescues adipogenesis . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 01510 . 7554/eLife . 12997 . 016Figure 5—figure supplement 1 . Impaired adipogenesis due to effects of senescent cells is partially rescued by JAK inhibition . ( a ) CM was collected from non-senescent ( CON ) and senescent ( SEN ) fat progenitor cultures . JAK inhibitor 1 ( 0 . 6 µM ) was directly added into CON ( CON+JAKi ) and SEN ( SEN+JAKi ) CM . Pooled human fat progenitors were treated with 50:50% CM:DM for 10 days . Gene expression was analyzed by real-time PCR . Results were obtained using CM from 3 strains of human primary fat progenitors and expressed as mean ± s . e . m . ( b ) Rat fat progenitors were isolated from 3 and 30-month old rats . These cells were differentiated in presence of DMSO or 0 . 6µM JAK inhibitor 1 . Representative pictures were shown 48 hours after initiation of differentiation . ( c ) Gene expression was analyzed by real-time PCR in fat progenitors from 30-month old rats . Results ( N=4 ) are expressed as mean ± s . e . m . *p<0 . 05 . Two-tailed Student's t tests were used to determine statistical significance . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 016 To test effects of JAK inhibition in vivo , we treated 22-24 month-old C57BL/6 male mice with ruxolitinib ( INCB ) , a selective JAK1/2 inhibitor approved by the FDA , or vehicle ( DMSO ) for 2 months . Vehicle-treated mice progressively lost fat over two months , while JAK inhibitor administration prevented this age-related fat loss ( Figure 6a , d ) . The lean mass of both groups remained unchanged ( Figure 6b , e ) . The body weights of the vehicle-treated compared to the INCB-treated mice was not significantly different ( Figure 6c , f ) . This was consistent in two independent cohorts of mice using the same treatment regimen . Inguinal , subscapular , and brown fat mass were reduced in the vehicle-treated group , but were preserved in INCB-treated mice ( Figure 7a ) . The same INCB treatment only exhibited a non-significant trend to alter fat mass in young ( 8-month-old ) mice ( Figure 6—figure supplement 2a ) . 10 . 7554/eLife . 12997 . 017Figure 6 . JAK inhibition reduces age-related fat loss in mice . Twenty-two-month old male mice were treated with vehicle ( CON ) or ruxolitinib ( INCB ) for 8 weeks . Fat mass ( a ) and lean mass ( b ) were measured by MRI along with body weight ( c ) before treatment , as well as 1 month and 2 months after treatment . The percent changes relative to baseline are shown for fat mass ( d ) , lean mass ( e ) , and body weight ( f ) . Results ( N=9 ) are expressed as mean ± s . e . m . *p<0 . 05 . Two-tailed Student's t tests were used to determine statistical significance . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 01710 . 7554/eLife . 12997 . 018Figure 6—source data 1 . JAK inhibition reduces age-related fat loss in mice . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 01810 . 7554/eLife . 12997 . 019Figure 6—figure supplement 1 . JAK inhibition did not affect metabolic rate or food intake in aged mice . Twenty-two-month old male mice were monitored using CLAMS before and after 8 weeks of vehicle ( CON ) or ruxolitinib ( INCB ) treatment . ( a ) Metabolic rate and ( b ) food intake ( N=7 ) are expressed as mean ± s . e . m . Two-tailed Student's t tests were used to determine statistical significance . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 01910 . 7554/eLife . 12997 . 020Figure 6—figure supplement 2 . JAK inhibition had less impact on body composition and adipogenesis in 8-month old mice compared to 22-month old mice . Eight-month old male mice were treated with vehicle ( Y CON ) or ruxolitinib ( Y INCB ) for 8 weeks . ( a ) Fat mass , lean mass , and body weight were measured before and one month after treatment . The percent changes relative to baseline ( N=6 ) are expressed as mean ± s . e . m . ( b ) RNA from WAT was isolated and real-time PCR was performed . Results ( N=6 ) are expressed as mean ± s . e . m . ( c ) WAT was collected from 8-month old ( Young ) and 22-month old mice ( Old ) . RNA was isolated and real-time PCR was performed . Results ( N=6 for Young , N=8 for Old ) are expressed as mean ± s . e . m . *p<0 . 05 . Two-tailed Student's t tests were used to determine statistical significance . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 02010 . 7554/eLife . 12997 . 021Figure 7 . JAK inhibition increases adipogenic markers in adipose tissue and decreases circulating free fatty acids in aged mice . Twenty-two-month old male mice were treated with vehicle ( CON ) or ruxolitinib ( INCB ) for 8 weeks . ( a ) Weights of different fat depots and liver are shown as percent of whole body weight . Results ( N=9 ) are expressed as mean ± s . e . m . *p<0 . 05 . ( b ) RNA from WAT was isolated and real-time PCR was performed . Results ( N=8 ) are expressed as mean ± s . e . m . *p<0 . 05 . ( c ) Plasma activin A protein levels were assayed by ELISA in parallel from 8 six-month-old male mice ( Young ) . Results ( N=15 for CON and INCB , N=8 for Young ) are expressed as mean ± s . e . m . *p<0 . 05 . Plasma TG ( d ) and FA ( e ) levels are expressed as mean ± s . e . m . ( N=8 ) . *p<0 . 05 . ( f ) Hepatic TG/protein levels are expressed as mean ± s . e . m . ( N=11 ) . ( g ) Total hepatic TG levels are expressed as mean ± s . e . m . ( N=11 ) . Two-tailed Student's t tests were used to determine statistical significance . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 02110 . 7554/eLife . 12997 . 022Figure 7—source data 1 . JAK inhibition increases adipogenic markers in adipose tissue and decreases circulating free fatty acids in aged mice . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 02210 . 7554/eLife . 12997 . 023Figure 7—figure supplement 1 . JAK inhibition in aged mice suppressed activin A expression in primary fat progenitors . Twenty-two-month old male mice were treated with vehicle ( CON ) or ruxolitinib ( INCB ) for 8 weeks . Fat progenitors were isolated from WAT and gene expression was analyzed by real-time PCR . Some progenitors were pooled from several mice within the same treatment group due to limited yield of cells . Results ( N=5 pools , each from different sets of mice ) are expressed as mean ± s . e . m . *p<0 . 05 . Two-tailed Student's t tests were used to determine statistical significance . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 023 We next examined the mechanism of fat mass preservation due to JAK inhibition . JAK inhibition increased adipose tissue transcript levels of the adipogenesis markers , PPARγ , C/EBPα , FABP4 , and adipo-Q , as well as GPAT4 ( glycerol-3-phosphate acyltransferase isoform-4 , a TG synthesis marker ) ( Figure 7b ) , suggesting that JAK inhibition may act by enhancing adipogenesis and increasing TG storage in fat in aged mice . Lipin-1 was also increased in fat tissue from JAK inhibitor-treated mice ( Figure 7b ) . Activin A increased with aging in both fat tissues ( Figure 6—figure supplement 2c ) and the circulation ( Figure 7c ) . JAK inhibition suppressed activin A in both whole fat ( Figure 7b ) and progenitors isolated from fat tissue ( Figure 7—figure supplement 1 ) , as well as circulating activin A ( Figure 7c ) . Notably , JAK inhibition did not reduce activin A expression or improve adipogenesis in fat tissue of younger ( 8-month old ) mice ( Figure 6—figure supplement 2b ) . We also examined other potential causes of preservation of fat mass by JAK inhibition . Administering JAK inhibitor did not change metabolic rate or food intake in aged mice ( Figure 6—figure supplement 1 ) , and was previously found by us to actually increase activity of old mice ( Xu et al . , 2015 ) . Expression of two lipolytic enzymes , adipose triglyceride lipase ( ATGL ) and hormone-sensitive lipase ( HSL ) , was induced in fat tissue by JAK inhibition ( Figure 7b ) . This suggests that the fat maintenance we observed was not due to increased food intake , decreased energy expenditure , or decreased lipolysis . Next , we examined whether increased adipogenic capacity was associated with suppressed FA spillover and ectopic lipid accumulation . Plasma free fatty acid ( FFA ) levels were reduced by JAK inhibitor ( Figure 7e ) , while TG was not different ( Figure 7d ) . In addition , JAK inhibitor decreased both liver weight ( Figure 7a ) and hepatic TG ( Figure 7f , g ) in old mice . Lipotoxicity and decreased adipogenic capacity are associated with insulin resistance ( Wang et al . , 2008; Unger and Scherer , 2010; Gustafson et al . , 2015; Tchkonia et al . , 2010 ) . We investigated whether JAK inhibitor administration enhanced insulin sensitivity in aged mice . By conducting glucose and insulin tolerance tests , we found that insulin sensitivity was impaired in 22-month- compared to 8-month-old mice ( Figure 8e ) . JAK inhibitor improved glucose homeostasis ( Figure 8a , b ) and insulin sensitivity ( Figure 8d , e ) in 22-month-old mice , while it had little effect in young mice ( Figure 8—figure supplement 1 ) . Glucose-stimulated insulin secretion capacity was not altered by JAK inhibition in 22-month-old mice ( Figure 8c ) , suggesting that pancreatic islet function might not be affected . Fasting glucose was also unchanged with JAK inhibitor treatment ( Figure 8e ) . To test whether insulin sensitivity in fat tissue of aged mice was improved by JAK inhibitor , we performed an ex vivo insulin challenge test and found that fat tissue isolated from the JAK inhibitor-treated group exhibited more robust induction of p-AKT in response to insulin compared to the control group ( Figure 8g , h ) . Therefore , it appears that improved fat tissue function through JAK inhibition possibly contributed to enhanced insulin sensitivity in aged mice . 10 . 7554/eLife . 12997 . 024Figure 8 . JAK inhibition increases insulin sensitivity in aged mice . Seven-month old and twenty-two-month old male mice were treated with vehicle ( CON ) or ruxolitinib ( INCB ) daily . An oral glucose tolerance test was performed after 5 weeks of treatment . ( a ) Glucose level was monitored over 120 minutes for 22-month old mice ( the results for 7-month old mice are shown in Figure 8—figure supplement 1 ) and ( b ) the area under the curve ( AUC ) was calculated . Results ( N=6 for CON and INCB groups of 8-month-old mice , N=9 for CON and INCB groups of 22-month-old mice ) are expressed as mean ± s . e . m . *p<0 . 05 . ( c ) Plasma insulin was measured at baseline , 20 minutes , and 60 minutes after oral glucose gavage . Results ( N=9 ) are expressed as mean ± s . e . m . *p<0 . 05 . An insulin tolerance test was performed after 6 weeks of the treatment . ( d ) Glucose was monitored over 120 minutes for 22-month old mice ( the results for 7-month old mice are shown in Figure 8—figure supplement 1 ) and ( e ) area over curve ( AOC ) was calculated . Results ( N=9 ) are expressed as mean ± s . e . m . *p<0 . 05 . ( f ) Fasting glucose levels ( N=9 ) are expressed as mean ± s . e . m . *p<0 . 05 . ( g ) WAT tissue was collected and cultured in CM with or without 5 nM insulin for 5 minutes at 37oC and tissue lysates were then prepared . p-AKT ( Ser473 ) and total AKT protein abundance were assayed . Representative images are shown . ( h ) These signals were quantified by densitometry using ImageJ . The ratios of p-AKT/total AKT are expressed as mean ± s . e . m . N=6 . *p<0 . 05 . Two-tailed Student's t tests were used to determine statistical significance . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 02410 . 7554/eLife . 12997 . 025Figure 8—source data 1 . JAK inhibition increases insulin sensitivity in aged mice . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 02510 . 7554/eLife . 12997 . 026Figure 8—figure supplement 1 . JAK inhibition had less impact on glucose tolerance and insulin sensitivity in 8-month old mice compared to 22-month old mice . Eight-month old male mice were treated with vehicle ( CON ) or ruxolitinib ( INCB ) for 6 weeks . ( a ) An oral glucose tolerance test was performed after 5 weeks of treatment . Blood glucose was monitored over 120 minutes . Results ( N=6 ) are expressed as mean ± s . e . m . ( b ) An insulin tolerance test was performed after 6 weeks of the treatment . Blood glucose was monitored over 120 minutes . Results ( N=6 ) are expressed as mean ± s . e . m . Two-tailed Student's t tests were used to determine statistical significance . DOI: http://dx . doi . org/10 . 7554/eLife . 12997 . 026
Adipose tissue is a key metabolic organ , dysfunction of which can be linked to metabolic disease , particularly type 2 diabetes ( Gustafson et al . , 2015 ) . Adipose tissue function declines with age ( Tchkonia et al . , 2010 ) , likely contributing to increased prevalence of metabolic disorders with aging . Thus , potential pharmacotherapies that alleviate age-related adipose tissue dysfunction may lead to important clinical benefit . Previously , we found that senescent cells contribute to age-related adipose tissue dysfunction in a progeroid mouse model ( Baker et al . , 2011 ) . Here , we used progenitor cells isolated from human adipose tissue to demonstrate that senescent cells directly inhibit adipogenesis of non-senescent human fat progenitors . One mechanism by which senescent cells exert this inhibitory effect is through secretion of activin A , a protein that we previously showed inhibits adipogenesis ( Zaragosi et al . , 2010 ) . We tested the role of senescent cells in naturally-aged INK-ATTAC mice and confirmed that senescent cells play a causal role in age-related fat dysfunction in vivo . Moreover , we found that JAK inhibitor reduced activin A secretion both in vitro and in vivo . Two months of JAK inhibitor administration preserved adipose tissue function and restored insulin sensitivity in 22-month-old mice . Our study provides proof-of-concept evidence that senescent cells play an important role in age-related adipose tissue loss and dysfunction . It also suggests that inhibiting the JAK signaling pathway or selectively eliminating senescent cells hold promise as avenues to prevent or treat age-related metabolic dysfunction . The JAK/STAT pathway plays an important role in adipose tissue development and function ( Richard and Stephens , 2014 ) . Previously , we found that JAK inhibitor treatment inhibited production of SASP factors and improved physical function in aged mice ( Xu et al . , 2015 ) . Here , we show that JAK1/2 inhibition has metabolic benefits in aged mice . JAK inhibitor treatment suppressed activin A production by senescent cells in vitro and in fat progenitors , fat tissue , and the circulation in aged mice . These observations are concordant with improved adipogenesis and reduced activin A in fat tissue after genetic clearance of senescent cells from 18-month-old INK-ATTAC+/- mice . One possible mechanism for improved adipogenesis and fat tissue function by JAK inhibitor treatment is reduction of activin A production by senescent cells . Other SASP components ( i . e . IL-6 , TNFα , and IFNγ ) may also contribute to impairment of adipose tissue function in aged animals . It is possible that mechanisms other than those directly affecting senescent cells contributed to the improved adipogenesis we found . Also , senescent cells of many types and in multiple tissues , not only in fat , are likely affected by systemic administration of JAK1/2 inhibitors to mice or AP20187 to INK-ATTAC animals . Very likely , these systemic effects of our interventions contributed to alleviating metabolic dysfunction . To partially address this , rather than conducting an epistasis experiment ( treating senescent cell-depleted INK-ATTAC mice with JAK inhibitor to check for off-target effects ) , we compared effects of JAK inhibitor treatment in old to young mice , since the latter have fewer senescent cells , like AP20187-treated INK-ATTAC mice . We feel this experiment achieves essentially the same goals as would an epistasis experiment , and arguably may even have certain advantages: potential off-target effects of AP20187 are avoided and senescent cells are very few in young mice , unlike older AP20187-treated INK-ATTAC mice , in which more than 50% of senescent cells can remain after treatment with AP20187 ( Figure 4—figure supplement 1 ) . The lack of substantial effects of JAK inhibitor treatment on adipogenesis , fat depot weights , and insulin sensitivity in young animals , but strong effects in old animals with higher senescent cell burden and activin A , coupled with parallel effects between JAK inhibitor treatment in wild type mice to those of genetic clearance with AP20187 in INK-ATTAC mice , suggest that effects of JAK inhibitors on senescent cells may contribute to improved metabolic function in older mice . Our findings are consistent with the speculation that impaired adipogenesis leads to ectopic lipid accumulation and insulin resistance ( Gustafson et al . , 2015 ) . We found that JAK pathway inhibition led to maintained fat mass and enhanced metabolic function in tandem with improved adipogenic capacity in aged mice . Expression of PPARγ and C/EBPα , both of which are essential for insulin sensitivity ( Wu et al . , 1999; El-Jack et al . , 1999 ) , increased in adipose tissue of JAK inhibitor-treated mice . These changes were accompanied by reduced circulating FFAs and hepatic lipid accumulation , two important manifestations of lipotoxicity associated with insulin resistance ( Slawik and Vidal-Puig , 2006; Boden , 2011 ) . Indeed , JAK inhibition improved insulin sensitivity in these mice . In addition to improved adipogenesis , JAK inhibition reduces systemic inflammation ( including reducing circulating IL6 ) in aged mice ( Xu et al . , 2015 ) and promotes 'browning' of adipose tissue ( Moisan et al . , 2015 ) , both of which are known to affect adipogenesis and insulin sensitivity . These mechanisms might also contribute to improved insulin sensitivity in aged mice in addition to reduced activin A level . Importantly , JAK inhibition improved adipogenesis and insulin sensitivity in aged mice but did so much less in younger mice , suggesting that the JAK pathway participates in age- or senescence-related pathogenesis of adipose tissue dysfunction . Furthermore , the effects of JAK inhibitors seem to be similar in mouse , rat , and human models . This is consistent with the speculation that the most fundamental aging mechanisms are conserved across mammalian species . Activin A is a member of the transforming growth factor superfamily and is involved in a variety of biological events ( Xia and Schneyer , 2009 ) . Activin A has widespread effects on multiple types of progenitors ( Zaragosi et al . , 2010; Bowser et al . , 2013 ) both directly and through interaction with the closely related growth and differentiation factors ( GDFs ) , which share receptor and signaling mechanisms with activin A ( Sako et al . , 2010 ) . Our results suggest that circulating activin A levels could be a bio-marker of senescent cell burden since: 1 ) circulating levels of activin A increase with aging , consistent with the increase in senescent cell abundance with aging , 2 ) senescent cells secrete activin A , 3 ) genetic clearance of senescent cells from 18-month-old INK-ATTAC+/- mice reduced circulating activin A , and 4 ) JAK inhibition suppresses activin A production in senescent cells in vitro and in aged mice in vivo . It is important to note that a variety of cell types can regulate activin A production , including macrophages ( Zaragosi et al . , 2010 ) . It is possible that these cell types contribute to increased activin A levels with aging . It will be valuable to study the effect of specific inhibition of activin A during aging . However , most activin A-blocking agents such as follistatin also inhibit myostatin due to structural similarity to activin A ( Cash et al . , 2009; Nakatani et al . , 2008 ) . These agents would therefore be anticipated to alter both muscle and fat mass though additional mechanisms that may be independent of activin A ( Lee , 2004; McPherron and Lee , 2002 ) . JAK inhibitor treatment did not alter lean mass in aged mice ( Figure 6b , e ) . Thus , JAK inhibition might be superior to current activin A-blocking agents for alleviation of age-related adipose tissue dysfunction . Ruxolitinib , the JAK1/2 inhibitor we used in vivo , is approved by FDA for treating myelofibrosis ( Verstovsek et al . , 2010; Harrison et al . , 2012; Verstovsek et al . , 2012 ) . Although it has side-effects in human subjects with myelofibrosis including anemia and thrombocytopenia ( Verstovsek et al . , 2010; Verstovsek et al . , 2012 ) , we and others found that ruxolitinib has minimal effects on peripheral blood cell populations in both young ( Quintás-Cardama et al . , 2010 ) and old mice ( Xu et al . , 2015 ) . Considerable work remains to be done to assess potential side-effects from JAK1/2 inhibitors , especially in older subjects . We stress this needs to be done before contemplating their use for age-related dysfunction in clinical practice . We observed an unusually rapid loss of fat from 18-month old INK-ATTAC+/- mice within 3 weeks . This fat loss could be related to the need to administer AP20187 by intraperitoneal ( ip ) injection , despite our making every effort to reduce this effect . Both WT and INK-ATTAC+/- mice were injected ip with AP20187 for three consecutive days , with 14 days between treatments . Thus , both groups received 6 ip injections within 3 weeks , the stress from which might have accelerated fat loss . Due to limited numbers of naturally aged INK-ATTAC+/- mice , we selected the most closely matched control group to detect an effect of clearing senescent cells on activin A and adipogenesis . The strategy of treating both the WT and INK-ATTAC+/- littermates with AP20187 had the advantages that both the treated and control groups received ip injection of the same drug in parallel . We used 18-month-old INK-ATTAC+/- mice because we have previously observed that 18 month old mice already have a detectable increase in senescent cell burden in their adipose tissue ( Stout et al . , 2014 ) . In addition , we decided to focus on the acute effect of clearance of senescent cells from INK-ATTAC+/- mice on adipogenic transcription factor expression , which can precede other changes . Therefore , we decided to treat these INK-ATTAC+/- mice for 3 weeks . Intermittent clearance of senescent cells with AP20187 was used based on our recent finding that senolytics are effective when administered intermittently , likely because senescent cells do not divide and may be slow to re-accumulate once cleared in the absence of a strong continuing insult ( Zhu et al . , 2015 ) . Furthermore , AP20187 has to be administered i . p . , precluding daily administration . On the other hand , we showed that JAK inhibitors , which blunt the SASP and can be administered orally , need to be continuously present to inhibit the SASP ( Xu et al . , 2015 ) . In summary , we demonstrated a likely causal role for senescent cells in age-related fat dysfunction and discovered a novel mechanism through which senescent cells can directly impair healthy fat progenitor function . Pharmacologic inhibition of the JAK pathway reduced activin A production in vitro and in vivo , alleviated age-related adipose tissue dysfunction , and improved insulin sensitivity in aged mice . Albeit speculative , our findings are consistent with the general hypothesis that senescent cells might exert profound effects on tissue and organismal function by affecting normal progenitors or stem cells through production of TGFβ family members , such as activin A , and potentially other types of factors secreted by senescent cells . Our work suggests that targeting senescent cells or their products could be a promising avenue for delaying , preventing , alleviating , or treating age-related stem cell , progenitor , and adipose tissue dysfunction and metabolic disease .
Primary human fat progenitors were isolated from subcutaneous fat collected from healthy , lean ( BMI 26 . 6 ± 0 . 9 kg/m2 ) kidney donors aged 39 ± 3 . 3 years as previously described ( Tchkonia et al . , 2005 ) . The protocol ( 10-005236 ) was approved by the Mayo Clinic Foundation Institutional Review Board for Human Research . Informed consent and consent to publish was obtained from all human subjects . Rat fat progenitors were isolated from 3- and 30-month-old Brown Norway rats ( purchased from Harlan Sprague Dawley ) as previously described ( Tchkonia et al . , 2007 ) . All rat and mouse experimental procedures ( A21013 , A37715 , and A16315 ) were approved by the Institutional Animal Care and Use Committee ( IACUC ) at Mayo Clinic . Human fat cell progenitors were subjected to 10 Gy of cesium radiation to induce senescence as described previously ( Xu et al . , 2015 ) . Human fat cell progenitors were also treated with 0 . 2 μM doxorubicin for 24 hours to induce senescence . Senescence was induced by irradiation unless otherwise indicated . For co-culture experiments , primary progenitors were stained with CellTracker CM-DiI dye ( Thermo Fisher Scientific , Waltham , MA , USA ) according to the manufacturer’s instructions . These cells were then seeded into wells containing either non-senescent control or senescent progenitors . The mixtures of cells were differentiated for 15 days . Differentiation of progenitors was assessed by observers who were not aware of which treatments the cultures had been exposed to . Cells with multiple doubly-refractile lipid inclusions visible by low power phase contrast microscopy were considered to be differentiated ( Karagiannides et al . , 2006 ) . JAK inhibitor 1 ( CAS 457081-03-7 ) was purchased from EMD Millipore ( Billerica , MA , USA ) . Ruxolitinib ( INCB18424 , CAS 941678-49-5 ) was purchased from ChemieTek ( Indianapolis , IN , USA ) . Amicon Ultra centrifugal filters were purchased from EMD Millipore . Activin A ELISA kits ( catalog number: DAC00B ) and activin A neutralizating antibody ( catalog number: MAB3381 ) were purchased from R&D Systems ( Minneapolis , MN , USA ) . SB 431542 was purchased from Cayman Chemical ( Ann Arbor , MI , USA ) . Cells were washed with PBS 3 times and cultured in medium to be conditioned ( CM ) containing 1 mM sodium pyruvate , 2 mM glutamine , MEM vitamins , MEM non-essential amino acids , and antibiotic ( Thermo Fisher Scientific ) for 24 hours . For JAK inhibitor treatment , cells were treated with 0 . 6µM JAK inhibitor or DMSO for 48 hours in regular medium , washed with PBS 3 times , and then exposed to CM containing JAK inhibitor or DMSO for another 24 hours . For differentiation , confluent human primary progenitors were treated with differentiation medium ( DM ) containing DMEM/F12 , 15 nM HEPES , 15 mM NaHCO3 , 2 mM glutamine , 10 mg/L transferrin , 33 μM biotin , 0 . 5 μM insulin , 17 μM pantothenate , 0 . 1 μM dexamethasone , 2 nM triiodo-L-thyronine ( T3 ) , 540 μM 3-isobutyl-1-methylxanthine ( IBMX ) , 1μM ciglitazone , 1 mg/ml fetuin , and penicillin/streptomycin for 15 days unless indicated otherwise . For conditioned medium experiments , 2x-DM was prepared by doubling the concentration of key differentiation ingredients ( 20 mg/L transferrin , 66μM biotin , 1 μM insulin , 34 μM pantothenate , 0 . 2 μM dexamethasone , 4 nM T3 , 1080 µM IBMX , 2 μM ciglitazone , and 2mg/ml fetuin ) . Pooled cells isolated from several human subjects were then differentiated with CM mixed with 2x-DM at a 1:1 ratio for 15 days unless indicated otherwise . The media were changed every 2 days . To induce differentiation of rat cells , confluent fat progenitors were exposed to DM containing 5 μg/ml insulin , 10 μg/ml transferrin , and 0 . 2 nM triiodothyronine in DMEM/F-12 for 48 hours . DMEM/F12 and glutamine were purchased from Thermo Fisher Scientific . All other reagents were purchased from Sigma-Aldrich ( St . Louis , MO , USA ) . Trizol ( Thermo Fisher Scientific ) was used to extract RNA from tissues or cells . M-MLV Reverse Transcriptase kit ( Thermo Fisher Scientific ) was used for reverse transcription . Real-time PCR was performed using TaqMan fast advanced master mix . All reagents including probes and primers were purchased from Thermo Fisher Scientific . TATA-binding protein ( TBP ) was used as an internal control . Cells or tissues were homogenized in cell lysis buffer ( Cell Signaling , Danvers , MA , USA ) with protease inhibitors ( Sigma-Aldrich ) . Coumassie Plus reagents ( Pierce , Rockford , IL , USA ) were used to determine total protein content . Proteins were loaded on SDS-PAGE gels and transferred to immuno-blot PVDF membranes ( Biorad , Hercules , CA , USA ) . SuperSignal West Pico Chemiluminescent Substrate ( Pierce ) was used to develop signals . p-AKT ( #4060 ) and total-AKT ( #4691 ) antibodies were purchased from Cell Signaling . Metabolic rate and food intake were measured using a Comprehensive Laboratory Animal Monitoring System ( CLAMS ) as previously described ( Xu et al . , 2015 ) . Adipose tissue cellular SABG was assayed as previously described ( Xu et al . , 2015 ) . SABG+ cells were quantified by observers who were not aware of which treatments cultures had been exposed to . Experimental procedures ( A21013 , A37715 , and A16315 ) were approved by the IACUC at Mayo Clinic . Twenty two-month-old C57BL/6 male mice were obtained from the National Institute on Aging ( NIA ) . INK-ATTAC+/- transgenic mice were generated and genotyped as previously described ( Baker et al . , 2011 ) . Briefly , JLK and TT conceived the idea of clearing senescent cells to test if this improves healthspan and devised the experimental strategy of making transgenic mice with a senescence-activated promoter driving the ATTAC drug-inducible suicide gene and GFP to selectively eliminate and identify senescent cells at any time postnatally . The INK-ATTAC mice were produced and phenotyped at Mayo Clinic through a collaboration among the Kirkland , J . van Deursen , and N . LeBrasseur labs . They were bred onto a C57BL/6 background in the JVD lab . KOJ in the Kirkland lab then bred them onto a 50:50 BALB/c:C57BL/6 background , genotyped mice to select INK-ATTAC heterozygotes , and aged them to 18 months . Controls for the INK-ATTAC experiments were INK-ATTAC-null; 50:50 BALB/c:C57BL/6 background mice raised in parallel . Mice were maintained under a 12 hour light and 12 hour dark cycle at 24°C with free access to food ( standard mouse diet , Lab Diet 5053 , St . Louis , MO , USA ) and water in a pathogen-free facility . For drug treatment , ruxolitinib was dissolved in DMSO and then mixed with food . In addition to regular food , each mouse was fed a small amount of food ( 0 . 5g ) containing ruxolitinib 60 mg/kg ( drug/body weight ) or DMSO daily . During the treatment , all mice consumed the drug-containing food completely every day . For AP20187 ( 10mg/kg ) treatment , drug was administered by i . p . injection for three consecutive days , with 14 days between treatments . Intermittent clearance of senescent cells with AP20187 was used based on our recent finding that senolytics are effective when administered intermittently ( Zhu et al . , 2015 ) . For oral glucose tolerance testing , mice were fasted for 6 hours and glucose ( 2 g/kg body weight ) was administrated by oral gavage . For insulin tolerance testing , mice were fasted for 4 hours and insulin ( 0 . 6 unit/kg body weight ) was injected intraperitoneally . Glucose was measured using a handheld glucometer ( Bayer ) in blood from the tail vein . For the glucose-stimulated insulin secretion assay , mice were fasted for 6 hours and glucose ( 2 g/kg body weight ) was administrated by oral gavage . Blood samples were collected at baseline , 20 minutes , and 60 minutes after glucose administration . Plasma insulin levels were measured by ELISA ( ALPCO , Salem , NH , USA ) . Fat and lean mass were measured by MRI ( Echo Medical Systems , Houston , TX , USA ) . Hepatic TG was measured as previously described ( Xu et al . , 2011 ) . FFA and TG were measured using kits from Wako Chemicals ( Richmond , VA , USA ) . In all studies , investigators conducting analyses of animals were not aware of which treatments animals had received . Two-tailed Student's t tests were used to determine statistical significance . p<0 . 05 was considered significant . All values are expressed as mean ± s . e . m . No randomization was used to assign experimental groups . We determined the sample size based on our previous experiments , so no statistical power analysis was used . All replicates in this study were independent biological replicates , which came from different biological samples . | The likelihood of developing metabolic diseases such as diabetes increases with age . This is , in part , because the cells within fat and other tissues become less sensitive to the hormone insulin as people and other animals get older . Also , the stem cells that give rise to new , insulin-responsive fat cells become dysfunctional with increasing age . This is related to the accumulation of “senescent” cells , which , unlike normal fat cell progenitors , release molecules that are toxic to nearby and distant cells . Xu , Palmer et al . have now investigated if senescent cells interfere with the activity of stem cells from human fat tissue , and if getting rid of these senescent cells might restore the normal activity and insulin responsiveness of aged fat tissue . The experiments revealed that human senescent fat cell progenitors release a protein called activin A , which impedes the normal function of stem cells and fat tissue . Additionally , older mice had higher levels of activin A in both their blood and fat tissue than young mice . Xu , Palmer et al . then analyzed older mice that had been engineered to have senescent fat cells that could be triggered to essentially kill themselves when the mice were treated with a drug . Eliminating the senescent cells from these mice led to lower levels of activin A and more fat tissue ( due to improved stem cell capacity to become fully functional fat cells ) that expressed genes required for insulin responsiveness . This showed that senescent cells are a cause of age-related fat tissue loss and metabolic disease in older mice . Next , Xu , Palmer et al . treated older mice with drugs called JAK inhibitors , which they found reduce the production of activin A by senescent cells isolated from fat tissue . After two months of treatment , the levels of activin A in the blood and in fat tissue were indeed reduced . The fat tissue in treated mice also showed fewer features associated with the development of diabetes than the fat tissue of untreated mice . As such , these results paralleled those after selectively eliminating the senescent cells . Together these findings suggest that JAK inhibitors or drugs ( called senolytics ) that selectively eliminate senescent cells may have clinical benefits in treating age-related conditions such as diabetes and stem cell dysfunction . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology"
] | 2015 | Targeting senescent cells enhances adipogenesis and metabolic function in old age |
The ability to form cooperative societies may explain why humans and social insects have come to dominate the earth . Here we examine the ecological consequences of cooperation by quantifying the fitness of cooperative ( large groups ) and non-cooperative ( small groups ) phenotypes in burying beetles ( Nicrophorus nepalensis ) along an elevational and temperature gradient . We experimentally created large and small groups along the gradient and manipulated interspecific competition with flies by heating carcasses . We show that cooperative groups performed as thermal generalists with similarly high breeding success at all temperatures and elevations , whereas non-cooperative groups performed as thermal specialists with higher breeding success only at intermediate temperatures and elevations . Studying the ecological consequences of cooperation may not only help us to understand why so many species of social insects have conquered the earth , but also to determine how climate change will affect the success of these and other social species , including our own .
Social animals , including humans and many insects , have come to dominate the earth , possibly because of their ability to form complex societies ( Laland et al . , 2001; Fuentes et al . , 2010; Boyd et al . , 2011; Wilson , 2012; Lucky et al . , 2013 ) . While studies of animal social evolution often emphasize the environment drivers of group-living ( Emlen , 1982; Rubenstein and Lovette , 2007; Jetz and Rubenstein , 2011; Gonzalez et al . , 2013 ) , the ecological consequences of sociality have received less attention . A rare exception comes from our own species , where cooperation is thought to have played a critical role in allowing modern humans to expand rapidly across the earth to exploit a more diverse range of environments than the African savannas in which our ancestors evolved ( Laland et al . , 2001 ) . This shift from being a habitat specialist to generalist , and the subsequent ecological dominance by social species , has been termed the social conquest hypothesis ( Wilson , 2012 ) . Although this idea has drawn attention from a variety of disciplines , it has proven difficult to test empirically ( Richerson and Boyd , 2008; Fuentes et al . , 2010 ) . Animals derive a variety of cooperative benefits from living in groups ( Alexander , 1974; Shen et al . , 2014 ) . Identifying the specific type of benefit individuals receive may help determine the ecological consequences of sociality . If the primary benefit of grouping is to cope with environmental challenges ( e . g . , predation risk , fluctuating climates , or interspecific competition ) ( Alexander , 1974; Korb and Foster , 2010; Jetz and Rubenstein , 2011; Celiker and Gore , 2012; Shen et al . , 2012; Gonzalez et al . , 2013 ) , cooperation should translate into individuals adopting a generalist strategy that allows them to live in a broad range of conditions and cope with a variety of environmental challenges . In contrast , when species form groups as an adaptation to intraspecific challenges ( e . g . , competition with conspecific groups or with members of their own group over a lack of breeding vacancies or critical resources; Emlen , 1982; West et al . , 2006; Reeve and Hölldobler , 2007; Gonzalez et al . , 2013; Hsiang et al . , 2013 ) , cooperation should enable individuals to specialize in a single environment ( Figure 1 ) . 10 . 7554/eLife . 02440 . 003Figure 1 . Illustration of two different causes of sociality , and their ecological consequences ( i . e . , niche breadth ) . ( A ) If cooperation is for coping with harsh environments or interspecific competition , cooperative phenotypes ( i . e . , forming groups; orange lines ) will have higher fitness than non-cooperative phenotypes in poor environments or when the pressure of interspecific competition is high . However , non-cooperative phenotypes ( i . e . , being solitary; blue lines ) could have higher fitness in favorable environments because there are few benefits of cooperating . ( B ) Under such a scenario , a species' total niche breadth ( black lines ) is expanded due to the cooperative phenotype because a social species' total niche breadth equals to the sum of the cooperative and non-cooperative phenotypes . ( C ) In contrast , if cooperation is the best-of-a-bad-job strategy as a response to intraspecific challenges , the per capita reproductive success will be lower in groups than solitary pairs . This scenario often occurs when grouping occurs because of a lack of critical resources , such as when breeding territories are limited in many cooperatively breeding birds ( Emlen , 1982 ) . Therefore , cooperative phenotypes do not necessarily have higher fitness than non-cooperative phenotypes in either poor or favorable environments . ( D ) As a consequence , cooperative phenotypes will have little influence on the total niche breadth of a species when cooperation is a response to intraspecific challenges . Note that the trade-offs between specialist and generalist strategies occur only in the case of coping with environmental challenges or interspecific competition , and not in the case of adaptation to intraspecific competition . DOI: http://dx . doi . org/10 . 7554/eLife . 02440 . 003 The contrast between habitat specialist and generalist strategies derives from ecological niche theory ( Levins , 1968; Futuyma and Moreno , 1988 ) . Although niche theory has been used to investigate a range of ecological phenomena including species interactions ( Kassen , 2002 ) , geographic distributions ( Peterson et al . , 2011 ) and the ecological consequences of climate change ( Clavel et al . , 2010 ) , to our knowledge it has not yet been applied to social evolution . To understand how sociality influences niche breadth evolution , social and non-social populations from the same species need to be examined in a variety of different environments . That is , rather than focusing upon the ecological interactions of a species as a whole , one could separate a species’ total niche into different phenotypic components and then determine how these phenotypes influence fitness in varying environments ( Roughgarden , 1972; Bolnick et al . , 2010 ) . For social species , total niche breadth can be partitioned into the ‘cooperative’ and ‘non-cooperative’ phenotypes , which correspond to generalist and specialist strategies , respectively if the grouping benefit is to cope with harsh environments or severe interspecific competition . Here we examine how group-living impacts the generalist-specialist behavioral tradeoff and its subsequent effect on niche breadth ( defined as a thermal performance that influences elevational distribution ) in the facultative cooperatively breeding burying beetle ( Nicrophorus nepalensis ) . The primary benefit of cooperative breeding behavior in burying beetles is to jointly prepare and bury carcasses more rapidly than their primary competitor , carrion-feeding flies ( Table 1 ) ( Eggert and Müller , 1992; Scott , 1994; Trumbo , 1995 ) . We consider how intraspecific cooperation drives the evolution of thermal specialist vs generalist strategies along an elevational gradient where the degree of temperature-mediated interspecific competition with flies for resources ( carcasses ) varies with elevation . To determine how temperature influences the degree of interspecific competition , which in turn mediates the cooperative and competitive strategies of N . nepalensis , we first documented the natural patterns of group size , cooperation , breeding success , and the degree of interspecific competition with flies along the elevational gradient . We then experimentally manipulated the group size of N . nepalensis and the degree of interspecific competition with flies to determine the mechanisms underlying the fitness patterns along the elevational gradient . 10 . 7554/eLife . 02440 . 004Table 1 . Identification and abundance of carrion-feeding insects collected on rat carcasses from June to August 2011 . DOI: http://dx . doi . org/10 . 7554/eLife . 02440 . 004OrderPercentage ( % ) FamilyFrequencyColeoptera6 . 18Hydraenidae6Leiodidae11Ptiliidae6Silphidae9Diptera91 . 89Anthomyiidae7Calliphoridae117Carnidae2Drosophilidae33Fanniidae67Muscidae63Mycetophilidae1Phoridae103Psychodidae11Sarcophagidae7Sciaridae5Sphaeroceridae60Hymenoptera1 . 74Formicidae8Vespidae1Lepidoptera0 . 19Tortricidae1Total10019 families518
We began by quantifying the natural patterns of group size , cooperative behavior , and breeding success along an elevational gradient in central Taiwan ( Figure 2 ) where daily minimum air temperature decreased with increasing elevation ( χ²1 = 222 . 50 , p<0 . 001 , n = 116 ) . We found that group size decreased with increasing elevation ( Figure 3A ) and decreasing air temperature ( Figure 3B ) . Furthermore , the probability of breeding successfully varied unimodally along the elevational ( Figure 3C ) and air temperature gradients ( Figure 3D ) , peaking at intermediate elevations and air temperatures . Additionally , cooperative behavior—quantified as levels of cooperative carcass processing ( ‘Materials and methods’ ) —increased with increasing group size ( Figure 4 ) , suggesting that the greater breeding success at higher elevations was due to the cooperative behavior of groups . 10 . 7554/eLife . 02440 . 005Figure 2 . Spatial distribution of study sites ( black triangles ) along an elevational gradient in Nantou , Taiwan ( 24°5' N , 121°10' E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02440 . 00510 . 7554/eLife . 02440 . 006Figure 3 . Natural patterns of group size and the probability of breeding successfully in relation to elevation and daily minimum air temperature . Mean group size in natural populations decreased with ( A ) increasing elevation ( χ²1 = 16 . 26 , p<0 . 001 , n = 54 ) and ( B ) daily minimum air temperature ( χ²1 = 15 . 26 , p<0 . 001 , n = 53 ) . The probability of breeding successfully in natural populations varied unimodally along ( C ) the elevational ( χ²2 = 8 . 68 , p=0 . 013 , n = 70 ) and ( D ) daily minimum air temperature gradients ( χ²2 = 6 . 37 , p=0 . 041 , n = 66 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02440 . 00610 . 7554/eLife . 02440 . 007Figure 4 . The relationship between group size and total investment in cooperative carcass processing in natural groups . Total social investment ( minutes , min ) in cooperative carcass processing increased with the increasing group size ( χ²1 = 1681 . 10 , p<0 . 001 , n = 21 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02440 . 007 To further determine how cooperation influences breeding success in different environments , we created small , non-cooperative groups ( one male and one female , n = 53 ) and large , cooperative groups ( three males and three females , n = 39 ) at 23 sites along the elevational gradient by placing locally trapped beetles on rat carcasses in specially designed breeding chambers that allowed flies and other small insects to move in-and-out of the chambers freely , but that limited the natural access of beetles ( Figure 5 ) . Initial group size simulated the number of beetles attracted to odorants produced by decomposing vertebrate carcasses , and the timing of beetle placement mimicked the natural pattern of arrival times , which are longer at higher elevations . We found that the probability of breeding successfully for small and large groups varied along the elevational gradient such that large groups performed as thermal generalists with similar breeding success at all elevations ( Figure 6A ) and air temperatures ( Figure 6B ) , whereas small groups performed as thermal specialists with high breeding success only at intermediate elevations ( Figure 6A ) and air temperatures ( Figure 6B ) . Moreover , large groups had higher breeding success than small groups at low elevations ( Figure 6A ) and at warmer temperatures ( Figure 6B ) , but small groups had marginally higher breeding success than large groups at intermediate elevations ( Figure 6A ) and temperatures ( Figure 6B ) . There were no significant differences in breeding success between large and small groups at high elevations ( Figure 6A ) and low air temperatures ( Figure 6B ) . 10 . 7554/eLife . 02440 . 008Figure 5 . Diagram of the experimental container . The apparatus consisted of a larger plastic container to isolate the carcass from scavengers , but beetles and flies were allowed to move freely between the chamber and the outside environment . A smaller container with a rat carcass was provided for burial in the center of the larger container . The entire burial process and behavioral assays were recorded with a video-recorder . Dashed lines represent places connected to long pipes , which allowed beetles to leave the box . Cross hatching indicates the soil layer inside the chamber ( unit: cm ) . Our manipulation successfully created different mean group sizes even after some free-living beetles entered the chambers and some experimental beetles left ( t58 = 15 . 08 , p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02440 . 00810 . 7554/eLife . 02440 . 009Figure 6 . Reproductive success varied with group size along elevational and temperature gradients . ( A ) The probability of breeding successfully for small ( blue circles , solid line ) and large groups ( orange circles , dashed line ) varied differently along the elevational gradient ( group size × elevation interaction , χ²2 = 10 . 56 , p=0 . 005 , n = 92; for large groups , χ²2 = 3 . 19 , p=0 . 20 , n = 39; for small groups , χ²2 = 7 . 66 , p=0 . 022 , n = 53 ) , with large groups having higher breeding success than small groups at lower elevations ( χ²1 = 5 . 60 , p=0 . 018 , n = 26 ) , but small groups having marginally higher breeding success than larger groups at intermediate elevations ( χ²1 = 3 . 51 , p=0 . 061 , n = 53 ) . There was no significant difference in breeding success between small and larger groups at high elevations ( χ²1 = 0 . 04 , p=0 . 84 , n = 13 ) . ( B ) The probability of breeding successfully for small and large groups also varied differently along the daily minimum air temperature gradient ( group size × temperature interaction , χ²2 = 7 . 28 , p=0 . 026 , n = 92; for large groups , χ²2 = 1 . 55 , p=0 . 46 , n = 39; for small groups , χ²2 = 6 . 15 , p=0 . 046 , n = 53 ) , with large groups showing higher breeding success than small groups at higher temperatures ( χ²1 = 5 . 60 , p=0 . 018 , n = 26 ) , but small groups having marginally higher breeding success than small groups at intermediate temperatures ( χ²1 = 3 . 46 , p=0 . 063 , n = 53 ) . Again , there was no significance in breeding success between small and larger groups at high temperatures ( χ²1 = 0 . 0001 , p=0 . 99 , n = 13 ) . Open circles indicate failed breeding attempts and closed circles indicate successful breeding events . Solid lines denote predicted relationships from GLMMs , whereas dashed lines denote statistically non-significant relationships . DOI: http://dx . doi . org/10 . 7554/eLife . 02440 . 009 To establish why breeding success varied with elevation and temperature differently in cooperative and non-cooperative groups , we quantified levels of cooperative carcass processing in our group size treatments across the elevational gradient . We found no relationship between cooperative carcass processing and elevation ( Figure 7A ) or air temperature ( Figure 7B ) in small groups . However , investment in cooperative carcass processing in large groups increased with decreasing elevation ( Figure 7A ) and increasing air temperature ( Figure 7B ) , presumably because carcasses decompose more quickly at lower elevations ( Figure 8A ) where fly abundance ( Figure 8B ) and activity ( Figure 8C ) is highest . Experimental exclusion of flies from carcasses confirmed that flies indeed enhance carcass decomposition rates; the mean dry weight of carcasses from which flies were excluded was more than two times heavier than carcasses for which flies had access ( Figure 9 ) . Our data further showed that in large groups , per capita social conflict ( ‘Materials and methods’ ) varied unimodally with a peak at intermediate elevations and air temperatures ( Figure 10 ) . Importantly , only investment in cooperative carcass processing , and not social conflict , increased with increasing temperature in large groups . Together these results indicate that an individual’s cooperative and competitive strategies are not influenced directly by temperature-dependent physiological constraints per se because higher ambient temperatures typically reduce the cost of activity for ectotherms ( Angilletta , 2009 ) . Instead , our experiments suggest that an individual's cooperative and competitive strategies are influenced by temperature-mediated interspecific competition for resources , which increases with increasing temperature . 10 . 7554/eLife . 02440 . 010Figure 7 . Investment in cooperative carcass processing along the elevational and temperature gradients . Investment ( minutes , min ) in large ( closed orange circles , solid line ) and small groups ( closed blue circles , successful trials; open blue circles , failed trials; dashed line ) varied along the ( A ) elevational ( group size × elevation interaction , χ²1 = 7 . 65 , p=0 . 006 , n = 45 ) and ( B ) daily minimum air temperature gradients ( group size × temperature interaction , χ²1 = 9 . 90 , p=0 . 002 , n = 45 ) such that investment in large groups decreased with ( A ) increasing elevation ( χ²1 = 10 . 30 , p=0 . 001 , n = 14 ) and ( B ) decreasing daily minimum temperature ( χ²1 = 9 . 93 , p=0 . 002 , n = 14 ) . There was no relationship between cooperative carcass processing and ( A ) elevation ( χ²1 = 0 . 80 , p=0 . 37 , n = 31 ) or ( B ) daily minimum air temperature ( χ²1 = 0 . 04 , p=0 . 84 , n = 31 ) in small groups . Solid lines denote predicted relationships from GLMs , whereas dashed lines denote statistically non-significant relationships . DOI: http://dx . doi . org/10 . 7554/eLife . 02440 . 01010 . 7554/eLife . 02440 . 011Figure 8 . Carcass decomposition rates and the effect of experimentally heating carcasses on fly abundance and activity along the elevational gradient . ( A ) The time until the carcass was completely consumed by carrion-feeding insects increased with increasing elevation ( χ²1 = 50 . 87 , p<0 . 001 , n = 40 ) . The control treatments ( closed circles ) represent the natural patterns of fly abundance and activity on carcasses . ( B ) Fly abundance decreased with increasing elevation ( χ²1 = 21 . 49 , p<0 . 001 , n = 33 ) , but heated carcass treatments ( open circles ) showed higher fly abundance than controls ( closed circles ) ( χ²1 = 42 . 65 , p<0 . 001 , n = 55 ) . ( C ) Diurnal fly activity decreased with increasing elevation ( χ²1 = 39 . 90 , p<0 . 001 , n = 33 ) , but flies were more active on heated carcass treatments than on controls ( χ²1 = 29 . 85 , p<0 . 001 , n = 55 ) . Solid lines denote predicted relationships from GLMs . DOI: http://dx . doi . org/10 . 7554/eLife . 02440 . 01110 . 7554/eLife . 02440 . 012Figure 9 . Remaining carcasses dry weight after exposure in different treatments . The mean ± SE remaining carcass dry weights in the fresh carcass controls ( white column , n = 9 ) were significantly heavier than those in the net covering ( black column , n = 9 ) and fly access treatments ( grey column , n = 9 ) ( χ²2 = 145 . 66 , p<0 . 001 , n = 27 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02440 . 01210 . 7554/eLife . 02440 . 013Figure 10 . Per capita social conflict in small and large groups along the elevational and temperature gradients . Patterns of per capita social conflict differed between small ( closed blue circles , successful trials; open blue circles , failed trials; dashed line ) and large groups ( closed orange circles , solid line ) along gradients of ( A ) elevation ( group size × elevation interaction , χ²2 = 14 . 73 , p<0 . 001 , n = 45 ) and ( B ) daily minimum air temperature ( group size × temperature interaction , χ²2 = 13 . 98 , p<0 . 001 , n = 45 ) . In large groups , per capita social conflict varied unimodally with elevation ( χ²2 = 9 . 11 , p=0 . 011 , n = 14 ) and daily minimum air temperature ( χ²2 = 6 . 17 , p=0 . 046 , n = 14 ) , peaking at intermediate elevations and temperatures . However , in small groups , per capital social conflict did not vary with elevation ( χ²2 = 4 . 37 , p=0 . 11 , n = 31 ) or daily minimum air temperature ( χ²2 = 0 . 73 , p=0 . 70 , n = 31 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02440 . 013 Experimental exclusion of flies from carcasses confirmed that interspecific competition between beetles and flies reduces beetle breeding success; the probability of beetles breeding successfully in small groups along the elevational gradient ( from 1664 m to 2809 m ) was lower when flies had access to carcasses than when they were excluded ( Figure 11; for additional details see fly competition treatment in ‘Materials and methods’ ) . To determine if temperature mediates this competition , we simultaneously manipulated group size and the degree of competition with flies along the portion of the elevational range where small groups had higher breeding success . We found that experimentally heating carcasses ( ‘Materials and methods’ ) increased fly abundance ( Figure 8B ) and activity ( Figure 8C ) relative to controls . If temperature-mediated competition with flies at low elevations explains why large groups had higher breeding success than small groups , then our heated carcass treatment at higher elevations should have decreased the probability of breeding successfully in small but not large groups . In support of this prediction , we found that heating carcasses differentially affected the breeding success of small and large groups when controlling for elevation such that the probability of breeding successfully in small groups decreased in the heated carcass treatments ( Figure 12A ) , but the probability of breeding successfully for large groups remained the same ( Figure 12A ) . Moreover , individuals were more cooperative in carcass processing in the heated carcass treatments than in the controls ( Figure 12B ) . 10 . 7554/eLife . 02440 . 014Figure 11 . The probability of breeding successfully in relation to fly accessibility . Mean ± SE probability of breeding successfully ( GLMM fitted values ) in small groups was higher when flies were excluded from carcasses ( n = 18 ) than when they had access to carcasses ( n = 23 ) along the elevational gradient ( χ²1 = 12 . 06 , p<0 . 001 , n = 41 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02440 . 01410 . 7554/eLife . 02440 . 015Figure 12 . Investment in cooperative carcass processing in control and heated carcass treatments along the elevational gradient . Heating carcasses differentially affected the breeding success of small and large groups when controlling for elevation ( χ²1 = 6 . 55 , p=0 . 010 , n = 116 ) . ( A ) Mean ± SE probability of breeding successfully ( GLMM fitted values ) for large ( orange columns ) and small ( blue columns ) groups of burying beetles in control and heated carcass treatments . Heating carcasses reduced the probability of breeding successfully in small groups ( χ²1 = 5 . 99 , p=0 . 014 , n = 68 ) , but not in large groups ( χ²1 = 0 . 98 , p=0 . 32 , n = 48 ) . ( B ) Mean ± SE total investment ( minutes , min ) in cooperative carcass processing was higher in heated carcass than control treatments ( χ²1 = 12 . 67 , p<0 . 001 , n = 16 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02440 . 015 In summary , our experiments demonstrated that cooperative beetle groups performed as thermal generalists , but non-cooperative groups performed as thermal specialists . This generalist-specialist behavioral tradeoff along the elevational gradient in N . nepalensis is generated by the tension between an individual's share of the grouping benefit and the group's productivity . At low elevations where the pressure of interspecific competition with flies is highest , individuals in large groups were not only more cooperative at handling carcasses , but they also engaged in lower levels of social conflict , both of which enabled them to outcompete flies . As a consequence , cooperation enables burying beetles to expand their thermal niche to a warmer region where competitors are more abundant . In contrast , we found that the ‘tragedy of the commons’ ( Hardin , 1968; Rankin et al . , 2007 ) —that is the degree of social conflict was higher in large groups , which led to a reduction in breeding success relative to small groups—occurred at intermediate elevations where the pressure of interspecific competition with flies was lower . At these intermediate elevations , non-cooperative groups have marginally higher breeding success than cooperative groups because intraspecific social conflict increased in the absence of interspecific conflict . Nonetheless , this within-group conflict has relatively little influence on the ecological dominance of burying beetles because breeding success is still relatively high in large groups in favorable environments , compared with those at elevations where environments are less favorable . We found a similar pattern in the natural populations ( i . e . , those without group size manipulations ) where breeding success was highest at intermediate elevations even though there are many naturally occurring large groups in this region . This study provides the first experimental evidence consistent with the social conquest hypothesis , which argues that cooperation promotes the evolution of generalist strategies when the primary benefit of living in groups is to cope with environmental challenges , including climate-mediated interspecific competition ( Wilson , 2012 ) . Preliminary support for this hypothesis comes from a recent comparative study of sponge-dwelling snapping shrimp ( Synalpheus spp . ) , showing that eusocial species were more abundant and occupied a broader range of host sponge species than non-social sister species ( Duffy and Macdonald , 2010 ) . We have shown experimentally in burying beetles that cooperative groups performed as thermal generalists , but non-cooperative groups performed as thermal specialists . Being cooperative enables burying beetles to extend their range to lower elevations where temperatures are warmer and where competitors are more abundant because individuals in large groups were more cooperative at handling carcasses , which enabled them to outcompete flies . Thus , cooperation allows burying beetles to expand their thermal niche into an environment from which they would otherwise be competitively excluded . Ultimately , studying the ecological consequences of cooperation may not only help us to understand why so many species of social insects have conquered the earth , but also to determine how climate change will affect the success of these and other social species , including our own .
The elevational gradient in central Taiwan ( Figure 2 ) covers broadleaf forest at lower elevations and mixed conifer-broadleaf forest at higher elevations . We chose study sites primarily in mature forests and avoided cultivated or open areas . We conducted a preliminary investigation of the natural pattern of arrival times of free-ranging beetles on carcasses along the elevational gradient from August to September 2012 and from June to September 2013 . In each trial , a 75 g rat carcass was presented on the soil and covered with a 21 × 21 × 21 cm iron cage with mesh size of 2 × 2 cm to prevent vertebrate scavengers . We video recorded the entire burial process . Because video recordings showed that the number of beetles on the carcass varied with time , we determined the mean group size ( an average group size of the maximum number of beetles sampled every hour ) before the burial was complete . Beetle arrival time was determined when the first burying beetle was observed on the carcass . The arrival time of free-ranging burying beetles on carcasses increased with increasing elevation ( χ²1 = 24 . 41 , p<0 . 001 , n = 73 ) . To confirm that flies ( Diptera ) are the major competitors of burying beetles , we first examined the succession pattern of carrion-feeding insects on 150 g ( n = 5 ) and 200 g ( n = 7 ) rat carcasses . This experiment was conducted at an intermediate elevation ( 2000 m ) from June to August 2011 . Initially , rat carcasses were placed at 50 m intervals along the ground and covered by 21 × 21 × 21 cm iron cages following the previous procedure . Samples were collected daily in the morning ( between 10:00 and 12:00 ) for three days to resemble the insect community at an early successional stage . Mean abundances of carrion-feeding insects on 12 carcasses were examined daily after exposure , continuing for 1 day ( n = 5 ) , 2 days ( n = 3 ) , and 3 days ( n = 4 ) . For each sampling period , we first used an aerial sweep net to collect flying insects before the carcass was moved . We then collected all insects present on the carcass . Finally , the soil beneath each carcass was sampled within a sieve tray ( 2500 cc ) , and insects were extracted by a modified Berlese funnel ( Newell , 1955 ) . All specimens were preserved in 70% ethanol for further identification in the laboratory . Taxonomic determination was made to the family level ( Borror et al . , 1989 ) . In total , 518 adult carrion-feeding insects were collected , representing 29 families in four orders ( Table 1 ) , including necrophagous , saprophagous , and omnivorous species ( Smith , 1986 ) . Of these , Diptera and Coleoptera were the two most represented groups , constituting 98 . 1% of the individuals captured . A GLM was performed to assess if the abundance ( number of individuals per carcass ) differed between insect families ( Diptera and Coleoptera ) using carcass weight and the day after carcass exposure as covariates . We found that the mean abundance of Diptera was significantly higher than that of Coleoptera ( χ²1 = 49 . 85 , p<0 . 001 , n = 12 ) . Burying beetles were collected by hanging pitfall traps baited with 100 ± 10 g of rotting chicken . Pitfall traps were checked each morning . Beetles were housed individually in 320 ml transparent plastic cups and fed with mealworms ( Zophobas morio ) if they were kept more than three days before the experiment . Each beetle was weighed to the nearest 0 . 1 mg and marked with Testors enamel paint on the elytra ( Butler et al . , 2012 ) for individual identification the night before use . Sex was determined by the markings on the clypeus; males have a rectangular , orange marking , whereas females do not . Our experimental chambers consisted of a smaller plastic container ( 21 × 13 × 13 cm with 10 cm of soil ) located inside a larger container ( 41 × 31 × 21 . 5 cm with 11 cm of soil ) ( Figure 5 ) . Multiple holes on the side walls of the smaller container permitted beetle movement between the two containers . The cap of the larger container was fitted with a digital camera and was raised up 2 cm by iron mesh to allow entry by free-ranging flies and beetles , but not by vertebrate scavengers ( Figure 5 ) . Digital cameras were powered by Yuasa lead-acid batteries ( 6V 12Ah ) , which were replaced every morning . We measured air temperature every 30 min for the duration of the experiment using Maxim's iButton devices that were placed within the larger container . Based upon the natural pattern of arrival times from our pilot study ( see ’Group size in natural populations’ in ‘Materials and methods’ ) , we released the marked beetles into the experimental apparatus 1 day , 2 days , and 3 days after the trials began at elevations of 1700–2000 m ( low ) , 2000–2400 m ( intermediate ) and 2400–2800 m ( high ) , respectively . To quantify breeding success , we exhumed the carcasses approximately 14 days after they were buried and collected third instar larvae , if there were any . Across the 92 trials that were completed successfully , 52 trials resulted in successful breeding attempts and 40 trials contained carcasses that were completely consumed by maggots . The 40 failed trials were used to examine the carcass consumption rate by maggots as an indicator of interspecific competition along the elevational gradient ( Figure 8A ) . To assess the effect of fly competition on carcass decomposition rates , we evaluated the difference in carcass weight loss among net-covered treatments ( i . e . , fly access was restricted from the entire cage ) , natural fly access treatments , and fresh carcass controls at intermediate elevations ( 2100 m ) . The carcasses of natural fly access treatments were exposed to flies until maggots finished consuming and left the carcasses . The dried weights of all carcasses were obtained by dehydrating the carcasses to a constant weight in a drying oven at 65°C . We also compared the probability of breeding successfully in treatments where flies had access to the carcasses and those where flies were excluded along the elevational gradient ( from 1664 m to 2809 m ) . To explore temperature-mediated cooperation in response to fly competition in situ , a heating device was continuously applied underneath each carcass to provide a warming effect . To determine if heating carcasses made them more attractive to flies , we compared fly activity and abundance on heated carcasses to those of control treatments on the day we released the beetles in each trial . Fly activity was quantified as the total duration between the first fly arriving at the carcass and the last fly leaving the carcass , whereas fly abundance was quantified as the total number of flies video recorded between 6:00 to 18:00 at 30-min intervals . The heating device was constructed with a series circuit of cement resistors ( 40 Ω ) , which was powered by Yuasa lead-acid batteries ( 6V 12Ah ) . The soil temperature differences between the heated carcass treatment and its ambient environment were measured using thermal probes at a depth of 5 cm daily in the morning in 32 trials . On average , the heated carcass treatment created higher soil temperatures ( 28 . 7 ± 0 . 71°C ) than those of ambient environment ( 17 . 4 ± 0 . 31°C ) ( χ²1 = 212 . 06 , p<0 . 001 , n = 64 ) . Further , a total of 24 heated carcass treatments were conducted along the elevational gradient ( from 2039 m to 2814 m ) where small , non-cooperative groups had higher breeding success . In total , 4488 hr of video were recorded from the 92 successful non-heated ( control ) trials ( n = 39 large groups , 53 small groups ) and 1170 hr from the heated carcass treatments ( n = 9 large groups , 15 small groups ) . A variety of social behaviors , including per capita social conflict and investment in cooperative carcass processing , were scored on the first night ( from 19:00 to 05:00 ) using the Observer Video-Pro software ( Noldus ) for the 34 successful breeding trials ( n = 14 large groups , 20 small groups ) and 11 trials of small groups failed at the lower elevations ( from 1664 m to 1844 m ) . Aggressive interactions were defined as social conflict if a beetle grasped , bit , chased , or escaped from the other same-sexed individual . A sample video of aggressive interaction can be seen in Video 1 . We measured per capita social conflict as the total number of aggressive interactions divided by mean group size for each observation period . To quantify total social investment in cooperative carcass processing , we estimated the cumulative time that each beetle spent depilating rat hair , removing maggots , or digging soil during carcass burial and preparation . A sample video of cooperative carcass processing can be seen in Video 2 . Investment was quantified as the duration of cumulative time sampled for a 10 min observation period in each hour ( 100 min in total ) . 10 . 7554/eLife . 02440 . 016Video 1 . Social investment , Large group , August 15 , 2011 . DOI: http://dx . doi . org/10 . 7554/eLife . 02440 . 01610 . 7554/eLife . 02440 . 017Video 2 . Social conflict , Large group , July 3 2013 . DOI: http://dx . doi . org/10 . 7554/eLife . 02440 . 017 Multivariate analyses were performed using generalized linear models ( GLMs ) . If the random effects of repeated sampling of study sites were required , generalized linear mixed models ( GLMMs ) were used . To test for the differences in the probability of breeding successfully between the two group sizes and carcass heating treatments along the elevational and temperature gradients , the outcome of breeding success ( 1 = Success , 0 = Failure ) was fitted as a binomial response term . The environmental factors ( elevation and daily minimum air temperature ) , group size treatments , and carcass heating treatments were fitted as covariates of interest . For the carcass heating treatments , the fitted value of the probability of breeding successfully was compared between heated carcass and control treatments . All statistic analyses were performed in the R statistical software package ( R Core Team , 2012 ) . | The ability to live and work together in groups likely helped the earliest humans to leave their savannah homes in Africa and successfully settle around the globe . In doing so , humans shifted from being savannah specialists to generalists able to cope with a range of different environments . Cooperation is also believed to be a key to the global success of social insects like bees and ants . However , testing the idea that cooperation allows animals to become generalists that thrive in diverse environments—an idea referred to as the ‘social conquest hypothesis’—is difficult . Climate change has added a new sense of urgency to understanding how species adapt to changing environments , and some studies of humans and other animals have suggested that cooperation may increase or decrease in changing environments . Living in social groups has both benefits and drawbacks: it helps some animals to avoid being eaten by predators , but it also creates more competition for mates , food or other resources . As such , predicting how climate change will impact human and animal societies has also been difficult to test . Sun et al . have now tested the social conquest hypothesis by looking at how changes in environmental conditions affect the social behavior of the burying beetle . These insects find dead animals and then bury them to be eaten by their larvae . Burying beetles often fight each other to ensure that their own young get exclusive access to a food source . However , working together allows the beetles to bury a carcass before flies and other competitors discover it . Sun et al . compared how much the beetles cooperated at different elevations in the mountains of Taiwan . At each elevation the beetles faced different challenges: higher elevations were colder but had fewer flies , while lower elevations were warmer but had more flies . Although burying beetles tended to work together more at warmer elevations , where the competition from flies was the most intense , beetles that cooperated with each other were able to successfully breed at all elevations . On the other hand , beetles that were less cooperative were best adapted to raising their young at more moderate elevations , where the climate and competition were less harsh . Similar results were seen when Sun et al . created non-cooperative and cooperative groups of beetles at different elevations and provided each group with a rat carcass . Further experiments that used heaters to artificially warm the carcasses directly proved that cooperation among beetles was indeed encouraged by higher temperatures . Many studies have suggested that global warming might cause higher levels of conflict in human societies . But by studying how changes in an environment impact cooperation in burying beetles , Sun et al . provide new insights into how climate change may affect the future success of other social animals , including humans . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"ecology"
] | 2014 | Climate-mediated cooperation promotes niche expansion in burying beetles |
Recycling of synaptic vesicles ( SVs ) is a fundamental step in the process of neurotransmission . Endocytosed SV can travel directly into the recycling pool or recycle through endosomes but little is known about the molecular actors regulating the switch between these SV recycling routes . ADP ribosylation factor 6 ( Arf6 ) is a small GTPase known to participate in constitutive trafficking between plasma membrane and early endosomes . Here , we have morphologically and functionally investigated Arf6-silenced hippocampal synapses and found an activity dependent accumulation of synaptic endosome-like organelles and increased release-competent docked SVs . These features were phenocopied by pharmacological blockage of Arf6 activation . The data reveal an unexpected role for this small GTPase in reducing the size of the readily releasable pool of SVs and in channeling retrieved SVs toward direct recycling rather than endosomal sorting . We propose that Arf6 acts at the presynapse to define the fate of an endocytosed SV .
Synaptic vesicles ( SVs ) are continuously recycled at the nerve terminal and this process is tightly controlled to ensure the fidelity of neurotransmission . After endocytosis , SVs are shuttled back to the recycling pool either directly or via endosome-like compartments . The existence of endosome-like compartments has been demonstrated at the synapse since the pioneering experiments on synapse ultrastructure ( Heuser and Reese , 1973 ) , but their functional role in SV cycling is highly debated and the molecular and signaling pathways that define the fate of an endocytosed SV are still largely unknown ( see for recent reviews Morgan et al . , 2013; Kokotos and Cousin , 2015; Jähne et al . , 2015 ) . The ADP ribosylation factor 6 ( Arf6 ) is a small GTPase that localizes primarily to the plasma membrane/endosomal system and is best known as a regulator of endocytic trafficking and actin cytoskeleton dynamics ( D'Souza-Schorey and Chavrier , 2006; Myers and Casanova , 2008 ) . As for others small GTPases , the activity of Arf6 is tightly controlled by two families of regulatory proteins: guanine nucleotide exchange factors ( GEFs ) that function as Arf6 activators by facilitating the conversion from GDP-bound inactive form to GTP-bound active form , and GTPase activating protein ( GAPs ) that function as negative regulators by enhancing GTP hydrolisis . Neurons express several GEFs and GAPs for Arf6 and Arf6 has been reported to play a role in neurite outgrowth and dendritic spine maturation although its function in mature synapses is largely unknown ( Jaworski , 2007 ) . At postsynaptic sites , accumulating evidence has recently shown that Arf6 plays a role in AMPA receptor trafficking and long-term synaptic plasticity ( Scholz et al . , 2010; Myers et al . , 2012; Oku and Huganir , 2013; Zheng et al . , 2015 ) . At presynaptic sites , Arf6 has been predicted to play a role in the assembly of the clathrin coated complex during SV endocytosis ( Krauss et al . , 2003 ) . While overexpression of the Arf6 GEF msec7-1 has been reported to increase basal synaptic transmission at the Xenopus neuromuscular junction ( Ashery et al . , 1999 ) , the function of Arf6 at the presynaptic terminal has never been directly addressed . Interestingly , mutations in Arf6 regulatory genes have been recently associated with intellectual disability and epilepsy in humans ( Shoubridge et al . , 2010; Falace et al . , 2010; Rauch et al . , 2012; Fine et al . , 2015 ) . Here , we investigate the ultrastructural and functional effects of Arf6 silencing in hippocampal synapses and reveal an unexpected presynaptic role for this small GTPase in determining the size of the readily releasable pool of SVs and in promoting direct versus endosomal recycling of SVs .
We first investigated on the expression of the small GTPase Arf6 at synaptic level by biochemical experiments and revealed expression of Arf6 in isolated nerve terminal-extract; differential extraction of synaptosomal proteins ( Phillips et al . , 2001 ) revealed that Arf6 is not tightly associated with presynaptic or postsynaptic membranes , as it is mainly extracted at pH6 similarly to the SV protein synaptophysin . We also evaluated Arf6 expression at synaptic level by immunocytochemistry . Endogenous Arf6 colocalyzed with both presynaptic ( Synaptophysin ) and postsynaptic ( Homer1 ) markers in primary rat hippocampal neurons ( 17 days in vitro , DIV ) and triple labelling showed expression of the small GTPase at the presynaptic and postsynaptic site of single synaptic puncta ( Figure 1—figure supplement 1 ) . To directly examine how Arf6 activity impacts on synapse structure , we performed electron microscopy ( EM ) analysis at Arf6-knockdown ( KD ) synapses . Rat hippocampal neurons were transduced at 12 DIV , after the initial wave of synaptogenesis had occurred , with a lentiviral vector , driving the expression of short hairpin targeting the coding sequence of the rat Arf6 mRNA ( shRNA#1 ) or the respective mismatch control and GFP as a reporter . The silencing efficiency was tested 5 days post transduction by western blotting ( WB ) and immunocytochemistry ( ICC ) ( Figure 1—figure supplement 2 ) . Ultrastructural analysis revealed that Arf6-KD synapses were undistinguishable from control synapses in terms of synaptic area and active zone ( AZ ) length ( Supplementary file 1 ) , but were characterized by a decreased total number of SVs and a significantly increased number of SVs docked at the AZ ( Figure 1A ) . Moreover , intraterminal cisternae , resembling endosome-like structures and occasionally found in control synapses , were dramatically increased in Arf6-silenced synapses ( Figure 1A ) . The observed phenotype was completely rescued by the expression of a rat Arf6 variant resistant to shRNA#1 silencing ( Arf6-res , Figure 1—figure supplement 3 ) . 10 . 7554/eLife . 10116 . 003Figure 1 . Reduced SV density and accumulation of intraterminal cisternae at Arf6 deficient synapses . ( A ) Upper panels , representative electron micrographs of synaptic terminals from cultured hippocampal neurons ( 17 DIV ) transduced with lentiviruses expressing an Arf6 shRNA ( Arf6-KD ) or an inactive mismatched version ( Control ) . Scale bar , 200 nm . Lower panels , morphometric analysis of the density of SVs , docked SVs and cisternae in control ( black ) and Arf6-silenced ( red ) synapses . Data are means ± SEM from 3 independent preparations ( n=134 and 162 synapses for control and Arf6-KD , respectively ) . ( B ) Upper panels , representative 3D synapse reconstructions from 60 nm-thick serial sections obtained from hippocampal neurons transduced as in A . Total SVs , docked SVs , presynaptic plasma membrane , postsynaptic density and cisternae are shown in light blue , yellow , green , blue and red respectively . Lower panels , morphometric analysis of number of SVs , docked SVs and cisternae in control ( black ) and Arf6-KD ( red ) synapses . Right panel shows the percentage of intraterminal cisternae that are connected ( green ) or not-connected ( orange ) with the plasma membrane for Arf6-KD synapses . Data are means ± SEM from 10 ( Control ) and 10 ( Arf6-KD ) reconstructed synapses . Statistical analysis was performed with the unpaired Student’s t-test . **p<0 . 005 , ***p<0 . 001 , versus control . DOI: http://dx . doi . org/10 . 7554/eLife . 10116 . 00310 . 7554/eLife . 10116 . 004Figure 1—figure supplement 1 . Synaptic localization of Arf6 . DOI: http://dx . doi . org/10 . 7554/eLife . 10116 . 00410 . 7554/eLife . 10116 . 005Figure 1—figure supplement 2 . Tools for silencing and rescue Arf6 expression . DOI: http://dx . doi . org/10 . 7554/eLife . 10116 . 00510 . 7554/eLife . 10116 . 006Figure 1—figure supplement 3 . Rescue of the Arf6-KD ultrastructural phenotype . DOI: http://dx . doi . org/10 . 7554/eLife . 10116 . 006 Three-dimensional ( 3D ) reconstruction of Arf6-silenced and control synapses confirmed the decreased SV density , the increased number of docked SVs and intraterminal cisternae and revealed that most if not all of the latter structures were separated from the plasma membrane ( Figure 1B ) . This observation rules out the possibility that the intraterminal cisternae represent membrane infoldings in continuity with the plasma membrane . To investigate on the nature of intraterminal cisternae , Arf6-silenced and control neurons were stained with the synaptic marker VAMP2 and the synaptic endosome markers Rab5 and Vti1A . At Arf6-KD VAMP2-positive synapses , a significant increase of both Rab5 and Vti1A signals was observed ( Figure 2 ) . The increased immmunoreactivity for endosomal markers at synaptic sites together with the direct observation that cisternae that accumulate upon Arf6 silencing are detached from the plasma membrane lead us to consider them bona fide endosome-like organelles ( Elos ) . 10 . 7554/eLife . 10116 . 007Figure 2 . Increased expression of endosomal markers at Arf6-deficient synapses . ( A ) Representative images of synapses from rat hippocampal neurons ( 17 DIV ) transduced with either Arf6 shRNA ( Arf6-KD ) or an inactive mismatched version ( Control ) and immunostained with anti-Vamp2 ( blue ) and either anti-Rab5 or anti-Vti1A ( red ) antibodies . Scale bar , 5 µm . ( B ) Intensity values for Rab5 and Vti1A signal at VAMP2-positive puncta in control ( black ) and Arf6-silenced ( red ) synapses . Data are means ± SEM from 3 independent preparations . 500 synapses have been counted for each preparation . Statistical analysis was performed with the unpaired Student's t-test . **p<0 . 005 versus control . DOI: http://dx . doi . org/10 . 7554/eLife . 10116 . 007 To assess whether the morphological phenotype was due to the loss of Arf6 protein per se or rather to its impaired activation we tested the effect of SecinH3 , an inhibitor of the GEF of the cytohesin family ( Hafner et al . , 2006 ) that prevents the activation of Arf proteins . We first tested the efficiency of SecinH3 to impair Arf6 activation in hippocampal primary cultures ( 17 DIV ) by pull-down experiments and revealed a significant decrease of GTP-bound Arf6 in SecinH3-treated neurons with respect to control ( Figure 3—figure supplement 1 ) . Wild-type neurons were therefore treated with either SecinH3 or vehicle ( DMSO ) and analyzed for synapse morphology . SecinH3 treatment did not result in any variations in the synaptic area or AZ length ( Supplementary file 1 ) , but the complex ultrastructural modifications observed at Arf6-silenced synapses ( i . e . decreased number of SVs , increased number of docked SVs and increased number of intraterminal cisternae ) were completely phenocopied ( Figure 3A ) . The possibility that the SecinH3 treatment results in delocalization of Arf6 from synaptic sites was assessed by staining control and treated cultures with the presynaptic marker Synaptophysin and Arf6 . No significant difference was observed in the intensity of the Arf6 signal at Synaptophysin-positive puncta in control and SecinH3-treated neurons ( Figure 3B ) . These data demonstrate that the lack of Arf6 activation , rather than of its expression , is sufficient to cause the accumulation of both SVs at the AZ and Elos at the expense of total SVs ( Figure 3 ) . Moreover , the acute SecinH3 treatment suggests that accumulating Elos represent transient structures rather than stable organelles recruited at the synapse upon loss of Arf6 activation . 10 . 7554/eLife . 10116 . 008Figure 3 . SecinH3 treatment phenocopies Arf6 silencing . ( A ) Upper panels , representative electron micrographs of synaptic terminals from cultured hippocampal neurons ( 17 DIV ) treated with either SecinH3 ( 30 µM ) or DMSO ( Vehicle ) . Lower panels , morphometric analysis of the density of SVs , docked SVs and cisternae in control ( black ) and SecinH3 treated ( red ) synapses . Data are means ± SEM from 3 independent preparations ( n=75 for experimental group ) . Statistical analysis was performed with the unpaired Student’s t-test . ***p<0 . 001 , versus respective control . Scale bar , 200 nm . ( B ) Left panels , representative images of synapses from rat hippocampal neurons ( 17 DIV ) treated with SecinH3 ( 30 µM ) or DMSO ( Vehicle ) and immunostained with anti-Arf6 and anti-synaptophysin ( p38 ) antibodies . Scale bar , 5 µm . Right panel , intensity values for Arf6 signal at p38-positive puncta in control ( black ) and SecinH3-treated ( red ) synapses . Data are means ± SEM from 3 independent preparations . 300 synapses have been counted for each preparation . DOI: http://dx . doi . org/10 . 7554/eLife . 10116 . 00810 . 7554/eLife . 10116 . 009Figure 3—figure supplement 1 . SecinH3 treatment inhibits Arf6 activation in hippocampal primary neurons . DOI: http://dx . doi . org/10 . 7554/eLife . 10116 . 009 To test this hypothesis we challenged control and Arf6-KD neurons with tetrodotoxin ( TTX ) to block spontaneous activity in the network and the related SV trafficking . As expected TTX had no effect on control synapses , because the performed treatments were too short to induce homeostatic changes ( Turrigiano , 2012; Verstegen et al . , 2014 ) . In contrast , the blockade of spontaneous activity with 1 µM TTX for either 12 or 24 hrs , completely reverted the increase in intraterminal Elos and the decrease in SV density observed upon Arf6 silencing ( Figure 4A , B ) . TTX treatments did not result in any variations in the synaptic area or AZ length in both experimental groups ( Supplementary file 1 ) . These data unequivocally demonstrate that , at Arf6-KD synapses , intraterminal Elos represent an activity-dependent intermediate station in the SV recycling pathway . 10 . 7554/eLife . 10116 . 010Figure 4 . The formation of intraterminal cisternae in Arf6-deficient synapses is activity dependent . ( A ) Representative electron micrographs of synaptic terminals from cultured hippocampal neurons ( 17 DIV ) transduced with Arf6-KD or an inactive mismatched version ( Control ) . Neurons were either left untreated or treated with 1 µM TTX for 12 or 24 hr . Scale bar , 200 nm . ( B ) Morphometric analysis of the density of SVs , docked SVs and cisternae in control ( black ) and Arf6-KD ( red ) synapses . Data ( 80 synapses/group ) are means ± SEM from 3 independent preparations . Statistical analysis was performed with two-way ANOVA followed by the Bonferroni’s multiple comparison test . ***p<0 . 001 versus control; °°°p<0 . 001 versus the respective untreated sample . DOI: http://dx . doi . org/10 . 7554/eLife . 10116 . 010 Interestingly , the increased number of docked SVs at Arf6-KD synapses , was unaffected by TTX treatment ( Figure 4A , B ) , suggesting a distinct mechanism by which Arf6 activation regulates the abundance of docked SVs , or a differential sensitivity to network activity . To couple the synaptic ultrastructure with functional analysis of SV trafficking , we employed SynaptophysinpHluorin ( Syphy ) , a fluorescent probe exquisitely designed to monitor the SV exo-endocytosis cycle at single synapses ( Miesenböck et al . , 1998; Fassio et al . , 2011 ) . Arf6-shRNA ( or the mismatched control ) together with the RFP reporter , and SypHy were transiently co-transfected in rat hippocampal neurons in order to record from a silenced neuron embedded in a network of non-silenced ones ( Figure 5—figure supplement 1 , Figure 5—figure supplement 2 ) . Considering the emerging role of Arf6 in AMPA receptor trafficking ( Scholz et al . , 2010; Myers et al . , 2012; Oku and Huganir , 2013; Zheng et al . , 2015 ) , this approach allows ruling out the contribution of postsynaptic effects on the observed phenotype . Experiments were performed at 17–18 DIV ( 3–4 days after transfection ) and RFP/SypHy-positive axonal processes were selected for stimulation and analysis ( Figure 5A ) . When the amount of surface-expressed SypHy was evaluated , we observed a significant increase of extracellularly exposed SypHy at Arf6-silenced synapses compared with control synapses , suggestive of a global impairment of SV cycling ( Figure 5—figure supplement 2 ) . Neurons were first stimulated with 40 action potentials ( APs ) at 20 Hz , a protocol widely employed to measure the readily releasable pool ( RRP ) of SVs and , therefore , suitable to reveal if the increased docked SVs observed by EM were functional or represented mistargeted/unprimed SVs , not competent for fusion . Both the shRNA #1 ( Figure 5B , C ) and the shRNA #2 ( Figure 5—figure supplement 2 ) significantly increased the peak fluorescence at the end of the 2 s . Stimulation without affecting the kinetics of fluorescence return to basal level at the end of the stimulation , which describe ‘post-stimulus’ endocytosis . To reveal whether the increased peak fluorescence was due to higher exocytosis or impairment of ‘during-stimulus’ endocytosis we employed the H+ATPase inhibitor Bafilomycin . This drug prevents reacidification of internalized SVs and allows monitoring net exocytosis . The peak fluorescence was equally increased when neurons were stimulated in the presence of Bafilomycin , suggesting that the observed peak fluorescence increase was indeed attributable to an increased number of exocytosed SVs . ( Figure 5D ) . Next , to exclude the involvement of shRNA-mediated off-target effects , we performed rescue experiments . When hippocampal neurons were triple transfected with SypHy , Arf6 shRNA#1 and a V5-tagged Arf6 rat variant resistant to shRNA#1 silencing ( Arf6-res , Figure 5—figure supplement 1 ) the peak fluorescence at the end of the 40APs@20Hz stimulation returned to the control level ( Figure 5E ) . These data demonstrate that increased docked SVs in Arf6-depleted synapses are indeed functional releasable SVs and that Arf6 silencing causes an alteration in the process of neurotransmission in the absence of any effects on the kinetics of RRP endocytosis . 10 . 7554/eLife . 10116 . 011Figure 5 . Increased readily releasable pool at Arf6-deficient synapses . ( A ) Representative images from neuronal processes cotransfected with Arf6-shRNA #1 ( RFP ) and SypHy . The merge panel shows colocalization at synaptic puncta . Scale bar , 10 µm . ( B ) Ensemble average traces of SypHy fluorescence plotted for control ( n=18 , black trace ) and Arf6-KD ( n=20 , red trace ) neurons stimulated with 40APs@20Hz ( dotted line ) . Data are means with SEM values shown every 5 time points . ( C ) Means ± SEM of the peak fluorescence at the end of the stimulus ( left ) and of the time constant of endocytosis ( τ endo , right ) , evaluated by fitting the fluorescence decay after stimulation by a single exponential function , for control ( black ) and ARF6-KD ( red ) synapses . ( D ) Left panel , ensemble average traces of SypHy fluorescence plotted for control ( n=18 , black trace ) and Arf6-KD ( n=20 , red trace ) neurons stimulated with 40APs@20Hz ( dotted line ) in the presence of 1 µM Bafilomycin ( +Baf ) . Data are means with SEM values shown every 5 time points . Right panel , means ± SEM of the peak fluorescence at the end of the stimulus in the presence of 1 µM Baf . ( E ) Left panel , ensemble average traces of SypHy fluorescence plotted for Arf6-KD ( n=10 , red trace ) and Arf6-KD neurons cotransfected with the Arf6 shRNA#1 resistant form ( rescue , n=13 , gray trace ) and stimulated with 40APs@20Hz ( dotted line ) . Data are means with SEM values shown every 5 time points . Right panel , mean ± SEM of peak fluorescence at the end of the stimulus in Arf6-KD ( red ) or rescued ( gray ) synapses . Statistical analysis was performed with the unpaired Student’s t-test . *p<0 . 05; **p <0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 10116 . 01110 . 7554/eLife . 10116 . 012Figure 5—figure supplement 1 . Tools for silencing and rescue Arf6 expression by transfection . DOI: http://dx . doi . org/10 . 7554/eLife . 10116 . 01210 . 7554/eLife . 10116 . 013Figure 5—figure supplement 2 . SypHy expression at Arf6-silenced synapses and effect of Arf6 shRNA#2 . DOI: http://dx . doi . org/10 . 7554/eLife . 10116 . 013 Then , to reveal if accumulation of intraterminal Elos also impacts on neurotransmission , a longer stimulation protocol ( 400APs@20Hz ) , which allows to recruit and recycle a larger SV pool , have been employed . Bafilomycin was concomitantly used to dissect out ‘during-stimulus’ recycling . At control synapses , Bafilomycin increased the response to the stimulation , indicative of the existence of an active ‘during-stimulus’ recycling ( Figure 6A ) . At the end of the stimulation , and in the absence of Bafilomycin , fluorescence returned to the basal level due to ‘post-stimulus’ endocytosis , with an average time constant of endocytosis ( τ ) of 36 s . Strikingly , at Arf6-KD synapses , the peak fluorescence was not modified by Bafilomycin , suggesting a strong impairment of ‘during-stimulus’ endocytosis ( Figure 6A ) . To measure the net exocytosis ( EXO ) during stimulation , the areas below the curves in the presence of Bafilomycin were calculated and no differences were observed between control and Arf6 KD synapses ( Figure 6B ) . To quantify net ‘during-stimulus’ endocytosis ( ENDO ) , additional curves were obtained by plotting the point-by-point differences between the curves run in the presence and absence of Bafilomycin . The time course of ‘during- stimulus’ endocytosis revealed that recycling was virtually absent in the first 12 s of stimulation , and started to take place only in the last 8 s of stimulation , resulting in a significant decrease in ENDO upon Arf6 silencing ( Figure 6B ) . Moreover , Arf6 silencing slowed down also ‘post-stimulus’ endocytosis as revealed by the significantly higher τ ( 59 s , Figure 6A ) , suggesting an involvement of the small GTPase in the endocytosis of recycling SVs , differently from what has been observed for SVs belonging to the RRP ( Figure 5C ) . Remarkably , all the described defects on ‘during-stimulus’ and ‘post-stimulus’ recycling in Arf6-KD synapses were completely rescued by the concomitant expression of the shRNA#1-resistant Arf6 ( Figure 6A , B ) . The functional data demonstrate that SV recycling is impaired at Arf6 depleted synapses and this impairment is paralleled by an accumulation of synaptic Elos . 10 . 7554/eLife . 10116 . 014Figure 6 . Decreased SV recycling at Arf6-deficient synapses . ( A ) Upper panels , ensemble average traces of SypHy fluorescence plotted for control ( n=9 , left ) , Arf6-KD ( n=8 , center ) and rescued ( n=8 , right ) neurons stimulated with 400APs@20Hz ( dotted line ) in the absence or presence of 1 µM Bafilomycin ( + Baf . ) . Data are means with SEM values shown every 5 time points . Lower panels ( from left to right ) , peak fluorescence at the end of the stimulus in the absence of Baf . ; peak fluorescence at the end of the stimulus in the presence of Baf . ; ‘during-stimulus’ recycling fraction calculated as the difference between the peak fluorescence in the presence and absence of Baf . ; time-constants of ‘post-stimulus’ endocytosis evaluated by fitting the fluorescence decay after stimulation in the absence of Baf . by a single exponential function . Dara are means ± SEM . ( B ) Left panel , Time courses of endocytosis during stimulation with 400 APs at 20 Hz in control ( black ) , Arf6-KD ( red ) and rescued ( gray ) synapses for the same experiments shown in A . Time courses were derived by subtracting the fluorescence trace in the presence of Bafilomycin ( ΔFexo ) from the trace in its absence ( ΔFexo-ΔFendo ) . Middle and right panels , Rate of exocytosis ( EXO ) and endocytosis ( ENDO ) during stimulation were calculated as the area under the curves . Statistical analysis was performed with one-way ANOVA followed by the Bonferroni’s multiple comparison test . *p<0 . 05 , **p<0 . 01 versus Arf6-KD . DOI: http://dx . doi . org/10 . 7554/eLife . 10116 . 014 We next asked whether the accumulation of synaptic Elos at Arf6 depleted synapses is attributable to a block of SV reformation from intermediate endosomal structures or to the fact that SVs are constrained to pass through intermediate endosomal structures instead of recycling directly . To challenge these hypotheses , we visualized SVs and intraterminal cisternae formation upon AP firing by stimulating control or Arf6-KD neurons ( 17–18 DIV ) in the presence of soluble HRP . Samples were fixed at the end of the stimulation ( 400APs@20Hz ) and after 2 and 20 min . wash in the absence of HRP . The decreased number of SVs and the increased number of docked SVs and intraterminal cisternae were fully confirmed in Arf6 deficient synapses at all time point ( Figure 7—figure supplement 1 ) . However , when the HRP-positive ( HRP+ ) structures at the end of the stimulus , representative of active cycling during stimulation , were analyzed , a significant decrease HRP+ SVs accompanied by an increase of HRP+ cisternae in Arf6-KD terminals compared to control was observed ( Figure 7 ) . This result correlates with the observed defect in ‘during-stimulus’ endocytosis revealed by the SypHy experiments ( Figure 6 ) and suggests that Elos preferentially form when Arf6 signalling is occluded . Interestingly , after 2 min . wash in the absence of HRP , HRP+ SVs remained substantially unchanged in control synapses , while they significantly increased in Arf6-depleted terminals ( Figure 7 ) . This increase demonstrates that , in Arf6-depleted terminals , SVs efficiently form from intermediate endosomal structures , since the additional HRP+SVs can only derive from intraterminal HRP+-Elos after HRP had been removed from the extracellular medium . The data suggest that Arf6 regulates the direct formation and recycling of SVs so that , in its absence , Elos form upon stimulation and SVs lately bud from the newly formed structures . Finally , after 20 min-wash in the absence of HRP , SVs and intraterminal cisternae significantly lose their HRP content in both experimental groups , suggesting that both SVs and cisternae are actively recycled , thus ruling out the possibility that SVs are stuck at Elos in Arf6-depleted synapses ( Figure 7 ) . 10 . 7554/eLife . 10116 . 015Figure 7 . Stimulus-induced formation of intraterminal cisternae is increased at the expense of SV formation at Arf6-silenced synapses . ( A ) Representative EM images for Control and Arf6-KD synapses in neurons stimulated with 400APS@20hz in the presence of HRP ( 10 mg/ml ) and then incubated for additional 2 or 20 min , in the absence of HRP . Scale bar , 200 nM . ( B ) Quantitative analysis of SV and cisternae dynamics measured by HRP labelling in control and Arf6-KD synapses treated as in A . Data ( 120 synapses/group ) are means ± SEM from 4 independent preparations . Statistical analysis was performed with two-way ANOVA followed by the Bonferroni’s multiple comparison test . **p<0 . 01 versus control; °p<0 . 05 , °°p<0 . 01 , °°°p<0 . 001 versus the respective HRP-loaded sample . DOI: http://dx . doi . org/10 . 7554/eLife . 10116 . 01510 . 7554/eLife . 10116 . 016Figure 7—figure supplement 1 . Morphometric analysis at the different time point in the HRP functional assay . DOI: http://dx . doi . org/10 . 7554/eLife . 10116 . 016
In the present study , we uncovered a crucial function for the small GTPase Arf6 in the maintenance of RRP and determining recycling route of endocytosed SVs at the presynaptic terminal . Our morpho-functional analysis demonstrated that upon Arf6 depletion synaptic Elos preferentially form during activity at the expense of SVs . Considering that the molecular nature , as well as the function of synaptic endosomal structures is a matter of debate , our 3D analysis coupled with staining for endosomal markers demonstrate that the structures that form upon Arf6-silencing are bona fide synaptic Elos . Moreover , these organelles unequivocally participate in SV recycling as their formation is abrogated by TTX treatment . The same phenotype is observed when blocking Arf6 activation by pharmacological treatment , demonstrating that synaptic Elos form due to the loss of Arf6 activation , that results therefore essential for the direct recycling of endocytosed SV . The increased travelling of SVs via synaptic Elos at Arf6-depleted synapses , results in recycling defects during long-lasting stimulation ( 20 s ) , when multiple rounds of exo-endocytosis are required , suggesting that direct , rather than endosomal , recycling is the favorite and most efficient recycling route during repetitive stimulation . Moreover , synaptic Elos formation is accompanied by an increased RRP demonstrating that , while defining the recycling route of endocytosed SV , Arf6 also regulates the abundance of release competent SVs at the AZ . The function of endosomal structures at the synaptic terminal is still elusive , but a role in both regeneration of SVs and SV protein sorting and renewal has been described ( Hoopmann et al . , 2010; Watanabe et al . , 2014; Wucherpfennig et al . , 2003; Uytterhoeven et al . , 2011; Fernandes et al . , 2014 ) . Our data reveal that the small GTPase Arf6 is a component of the machinery that regulates the fate of an endocytosed SV at the synapse . We speculate that active Arf6 shifts SVs toward direct recycling , away from synaptic Elos , through which only a subset of SVs , either devoid of Arf6 or in which Arf6 is inactive , travel to undergo protein turnover . This scenario is reminiscent of the role of active Arf6 in endosomal recycling of transferrin receptors , where the small GTPase mediates fast recycling of the receptor , and confirms the central role of Arf6 in determining the intracellular trafficking pathway of endocytic cargoes ( Montagnac et al . , 2011; Grant and Donaldson , 2009 ) . Intriguingly , upstream regulators of Arf6 activity have been recently demonstrated to act in SV recycling at the Drosophila neuromuscular junction with morphological and functional phenotypes reminiscent of the Arf6-depleted hippocampal synapses described in this paper ( Uytterhoeven et al . , 2011; Podufall et al . , 2014 ) . Several downstream effectors for active GTP-bound Arf6 have been described that could play a role in SV recycling at the plasma membrane ( Donaldson and Jackson , 2011 ) , and among them the adaptor protein AP-2 is of particular relevance ( Paleotti et al . , 2005 ) . Neuron-specific deletion of AP-2 has been recently described to result in the accumulation of presynaptic endosome-like vacuoles and loss of SVs at hippocampal synapses ( Kononenko et al . , 2014 ) , resembling the Arf6-KD phenotype described here . Accumulation of synaptic endosomal structures can be attributed to defective SVs reformation or to an increase in the trafficking of SVs via intermediate endosomal compartments because of a defect in the direct recycling route bypassing intermediate endosomes . For Arf6-KD neurons , we favor the second hypothesis as morpho-functional experiments revealed that Arf6-depleted synapses show deficient SV formation upon stimulation but functional SV reformation from intraterminal Elos during the post-stimulus period . Endosome-like structures seem therefore to preferentially form at the expense of SVs in the absence of Arf6 , which instead does not appear to play a major role on SV reformation from Elos . Additional Arf6 downstream targets , other than AP-2 , could also be involved , such as the phosphatidylinositolphosphate-kinase type Iγ , already described as a SV cycle regulator ( Di Paolo et al . , 2004 ) , or the scaffold protein JIP3 . Interestingly , JIP3 has been identified as an Arf6 target mediating the fast component of constitutive endocytosis ( Montagnac et al . , 2011 ) and the C . elegans JIP3 homologue UNC-16 mutant shows endosomal structure accumulation and SV depletion at motor neuron synapses ( Brown et al . , 2009 ) . The hypothesis that SVs preferentially travel via endosomes when Arf6 expression/activation is impaired is also in accordance with the increased RRP , which was reported to go through endosomal sorting ( Hoopmann et al . , 2010 ) . Indeed , RRP is increased at Arf6-depleted synapses but the kinetics of endocytosis for the SVs belonging to RRP was unaffected , supporting the idea that endosomal recycling is the retrieval mechanisms selected by SVs from the RRP . In line with these observations , clathrin depletion was recently reported to be ineffective on endocytotic decay after RRP depletion ( Kononenko et al . , 2014 ) , suggesting a clathrin/Arf6-independent SV retrieval for RRP vesicles . However , the increased RRP due to Arf6 silencing was insensitive to TTX , opening the possibility that it is a distinct activity-independent mechanism . Differently from what we observed after the short ( 2 s ) stimulation , Arf6 depletion impacts on SV endocytosis after the longer ( 20 s ) stimulation , suggesting that not only the frequency , but also the duration of the stimulus can challenge the membrane retrieval mechanisms . Whereas the Rab family of small GTPases has been extensively implicated in various steps of SV cycling , ranging from the docking/priming phases during exocytosis to recycling phases ( Binotti et al . , 2015; see for a review Pavlos and Jahn , 2011 ) , our work reveals additional roles for the small GTPase of the Arf family , Arf6 , in the SV cycling pathway and in the definition of the size of the SV pools . Interestingly , Rab proteins can act both as regulators or effectors for activated Arf6 , as demonstrated in various cellular systems ( Chesneau et al . , 2012; Shi et al . , 2012; Pelletán et al . , 2015 ) . In particular Arf6 and Rab35 have been shown to turn each other off by recruitment of reciprocal GAPs ( Chaineau et al . , 2013; Dutta and Donaldson , 2015 ) and active Rab35 is reported to regulate SV endosomal sorting ( Uytterhoeven et al . , 2011 ) . Future work is needed to unravel the Arf6 interactome at the presynapse in order to define the specific GEFs and GAPs that regulate Arf6 activity in discrete subsynaptic compartments and the Arf6 effectors that mediate the described functions . The use of pharmacological tools to mimic or revert Arf6 function at synapse opens new therapeutic approaches to correct dysregulations of the Arf6 pathway occurring in human neurological diseases associated with mutations in the Arf6 regulatory genes ( Shoubridge et al . , 2010; Falace et al . , 2010; Rauch et al . , 2012; Fine et al . , 2015 ) .
Sprague-Dawley rats were from Charles River . All experiments were performed in accordance with the guidelines established by the European Community Council ( Directive 2010/63/EU of March 4th 2014 ) and approved by the Italian Ministry of Health . Amino-5-phosphonopentanoic acid ( D-APV ) , 6-cyano-7-nitroquinoxaline-2 , 3-dione ( CNQX ) , tetrodotoxin ( TTX ) , BafilomycinA1 , and SecinH3 were from Tocris Bioscience ( Cookson , USA ) . Restriction enzymes were from New England Biolabs ( Ipswich , USA ) . Cell culture media were from Invitrogen ( Carlsbad , CA , USA ) . Soluble horseradish peroxidase ( HRP ) type VI , 3 , 3′-Diaminobenzidine ( DAB ) ( #P6782 and #D8001 ) and all other chemicals were from Sigma-Aldrich , St . Louis , MO . The following primary antibodies have been used: anti-GFP ( #MAB3580; Millipore , Darmstadt , Germany ) ; anti-synaptobrevin-2 ( Syb2; #104 202; Synaptic Systems , Goettingen , Germany ) ; anti-actin ( #A4700 ) and anti-Arf6 ( #A5230 , Sigma ) ; anti-Arf6 ( #ARP54598_P050; Aviva , San Diego , CA ) ; anti-Rab5 ( #108 011 Synaptic Systems ) ; anti-Arf1 ( #05–1427; EMD Millipore ) ; anti-V5 ( #60708; Invitrogen ) ; anti-β-tubulin ( #T8328; Sigma ) anti-p38 ( Mab5258 , EMD Millipore ) ; anti-Homer ( #106 011; Synaptic Systems ) ; anti-Vglut1 ( #135 304; Synaptic Systems ) . Two shRNAs targeting the coding sequence ( shRNA#1 , CGCCAGGAGCTGCACCGCATTATCAATGA , shRNA#2 , CGGATTCGGGACAGGAACTGGTATGTGCA ) of Rattus norvegicus ADP-ribosylation factor 6 ( Arf6 ) mRNA ( GenBank accession no . NM_024152 ) cloned into pRFP-C-RS plasmid have been acquired from OriGene Technologies ( Rockville , MD ) . Non-effective mismatch shRNA was obtained by inserting four point mutations in the nucleotide sequence of shRNA#1 ( CGCGAGCACCTGCAGCGCATTATGAATGA ) . RNAi resistant Arf6 isoform , harboring nine silent point mutations into the rat Arf6 mRNA sequence targeted by shRNA#1 ( CGGCAAGAACTCCATCGGATAATTAACGA ) , has been provided by Biomatik ( Cambridge , ON , Canada ) and then subcloned into a pcDNA3 . 1/V5-His-Topo plasmid ( Life Technologies , Carlsbad , CA ) or into lentiviral vector pLVX-Ires-mCHerry ( Clontech , Mountain View , CA ) . Oligonucleotides corresponding to shRNA#1 ( Forward: 5'-AATT—CGCCAGGAGCTGCACCGCATTATCAATGA—CTCGAG—TCATTGATAATGCGGTGGCAGCTCCTGGCG—TTTTTTTAT-3; Reverse 5'-AAAAAAA— CGCCAGGAGCTGCACCGCATTATCAATGA —CTCGAG— TCATTGATAATGCGGTGGCAGCTCCTGGCG -3’ ) and the mismatch control ( Forward: 5'-AATT—CGCGAGCACCTGCAGCGCATTATGAATGA—CTCGAG—TCATTCATAATGCGCTGCAGGTGCTCGCG—TTTTTTTAT-3'; Reverse: 5'-AAAAAAA— CGCGAGCACCTGCAGCGCATTATGAATGA —CTCGAG— TCATTCATAATGCGCTGCAGGTGCTCGCG -3’ ) were inserted into the lentiviral vector pLKO . 3G ( Addgene , Cambridge , MA ) using PacI and EcoRI restriction sites . Insertions were confirmed by single digestion with PshAI followed by direct Sanger sequencing . The production of vesicular stomatitis virus-pseudotyped third-generation lentiviruses was performed as described previously ( Verstegen et al . , 2014 ) . Viral titers ranging from 1 . 0 to 8 . 0 x 108 transforming units/ml were obtained . Pregnant rats were killed by inhalation of CO2 , and 18-day-old embryos were immediately removed by Cesarean section . Hippocampi were dissected and dissociated by enzymatic digestion in 0 . 125% trypsin for 20 min at 37°C and then triturated with a fire-polished Pasteur pipette . No antimitotic drugs were added to prevent glia proliferation . Neurons were plated on poly-L-lysine ( 0 . 1 mg/ml; Sigma-Aldrich ) -treated 25-mm glass coverslips at a density of 40 , 000–60 , 000 cells per coverslip and on poly-L lysine ( 0 . 01 mg/ml ) -treated 35-mm plastic wells at a density of 300 , 000 cells per well for immunoblots . HeLa cells were transfected with Lipofectamine 2000 reagent ( Life Technologies ) according to manufacturer’s instructions and processed after 48 hr for immunoblotting . Primary hippocampal neurons were nucleofected at 0 DIV with 3 μg of sh-Arf6-RFP or mismatched-Arf6-RFP constructs with AMAXA P3 primary cell 4D-Nucleofector kit ( Lonza ) and processed for western blotting at 7 DIV . For live imaging experiments neurons were transfected at 12–14 DIV with 0 . 5 μg of sh-Arf6-RFP or mismatched-Arf6-RFP , resistant Arf6-V5 and SypHy constructs using Lipofectamine 2000 and analyzed 3–5 days post-transfection . Primary hippocampal neurons were infected with lentiviruses for sh-Arf6 and respective mismatched version at 12 DIV at 10 multiplicity of infection ( MOI ) . After 20 h , half of the medium was replaced with fresh medium and EM experiments were performed between 17 and 18 DIV . Primary hippocampal neurons were fixed with 4% paraformaldehyde in PBS for 20 min , permeabilized with 0 . 1% Triton X-100 in PBS for 2 min , blocked 30 min with 5% BSA in PBS before incubation with primary antibodies , followed by Alexa Fluor-conjugated 405 , 488 , 568 and 647 secondary antibodies . For immunoblotting , neurons or HeLa cells were scraped in lysis buffer ( 150 mM NaCl , 50 mM Tris , NP-40 1% , SDS 0 . 1% ) plus protease and phosphatase inhibitors ( 0 . 2 mM PMSF , 2 μg/ml pepstatin , 1 mM NaF , and 1 mM NaVO4 ) . After centrifugation at 16 , 000 x g for 10 min at 4°C , samples were loaded on SDS-PAGE gel , transferred to nitrocellulose membranes , and immunoblotted with primary antibodies followed by HRP-conjugated secondary antibodies . Synaptosomes were purified from the P2 fractions by centrifugation on discontinuous Percoll gradient from the pooled cortex of two adult mice ( 3–6 months old ) . The tissue was homogenized in 10 ml of ice-cold HB buffer ( 0 . 32 M sucrose , 1 mM EDTA , 10 mM Tris , pH 7 . 4 ) containing protease inhibitors , using a glass-Teflon homogenizer . The resultant homogenate was centrifuged at 1000 × g for 5 min at 4°C in order to remove nuclei and debris and the supernatant was further centrifuged at 18 , 900 × g for 10 min at 4°C to obtain the P2 fraction . The pellet was resuspended in 2 ml of HB buffer and gently stratified on a discontinuous Percoll gradient ( 3%–10%–23% ) . After a 10 min centrifugation at 18 , 900 × g at 4°C , the synaptosomal fraction from the 10% and 23% Percoll interface was collected , washed in Krebs buffer ( 140 mM NaCl , 5 mM KCl , 5 mM NaHCO3 , 1 . 3 mM MgSO4 , 1 mM phosphate buffer pH 7 . 4 , 10 mM Tris/Hepes pH 7 . 4 ) to eliminate Percoll and used as the starting material for subsequent ultrafractionation . Ultrasynaptic fractionation was performed as previously described ( Phillips et al . , 2001 ) . Briefly , synaptosomes were pelleted by centrifugation ( 16 , 000 × g , 5 min , 4°C ) and resuspended in 300 µl of 0 . 32 M sucrose , 0 . 1 mM CaCl2 . An aliquot was removed and kept as total . Protease inhibitors were used in all purification steps . Synaptosomes were then diluted 1:10 in ice-cold 0 . 1 mM CaCl2 and mixed with an equal volume of 2X solubilization buffer ( 2% Triton X-100 , 40 mM Tris , pH 6 , 4°C ) . After a 30 min incubation at 4°C , the insoluble material ( synaptic junction ) was pelleted by centrifugation ( 40 , 000 × g , 30 min , 4°C ) . The supernatant ( 1% TX-100 pH 6 soluble ) was decanted and the proteins precipitated with 6 volumes of acetone at -20°C overnight and then centrifuged ( 18 , 000 × g , 30 min , -15°C ) . The synaptic junction pellet ( containing the insoluble postsynaptic density and the presynaptic active zone ) was resuspended in 10 volumes of 1X solubilization buffer ( 1% Triton X-100 , 20 mM Tris , pH 8 , 4°C ) , incubated for 30 min at 4°C and then centrifuged ( 40 , 000 × g , 30 min , 4°C ) . The proteins in the supernatant ( 1% TX-100 pH 8 soluble ) were precipitated in 6 volumes of acetone at -20°C overnight and then centrifuged ( 18 , 000 × g , 30 min , -15°C ) , while the pellet was kept as postsynaptic density fraction ( 1% TX-100 pH 8 insoluble ) . All pellets were resuspended in 5% SDS , quantified by BCA assay , and loaded on SDS-PAGE gels for electrophoresis and consecutive western blotting . Arf6 activation was determined using an Arf6-GTP-specific pull-down assay ( Cytoskeleton , Denver , CO , USA ) , according to manufacturer instructions . Samples were fractionated by SDS-12% PAGE and Arf6-GTP and total Arf6 level were evaluated by western blotting using an anti-Arf6 antibody ( Sigma Aldrich ) . Primary hippocampal neurons were fixed with 1 . 2% glutaraldehyde in 66mM sodium cacodylate buffer , pH 7 . 4 , postfixed in 1% OsO4 , 1 . 5% K4Fe ( CN ) 6 , and 0 . 1M sodium cacodylate , en bloc stained with 1% uranyl acetate , dehydrated , and flat embedded in epoxy resin ( Epon 812 , TAAB ) . Ultrathin sections ( 70 nm ) were collected on copper mesh grids ( EMS ) and observed with a Jeol JEM-1011 microscope at 100 kV equipped with an ORIUS SC1000 CCD camera ( Gatan , Pleasanton , CA ) . Morphometric analysis was done using NIH ImageJ . Structures with sagittal diameter comprised between 20 and 80 nm were classified as SVs , while those with a sagittal diameter bigger than 80nm were classified as intra-terminal cisternae . SVs touching AZ were classified as docked SVs . For 3D reconstructions , 60-nm serial sections were collected on carbon-coated copper slot formwar grids and serial synaptic profiles were acquired as described for mono-dimensional TEM . Serial electron micrographs were aligned with Midas of IMOD software ( Colorado University , CO , USA ) . Primary hippocampal neurons were maintained in Tyrode’s solution ( 140 mM NaCl , 3 mM KCl , 2 mM CaCl2 , 1 mM MgCl2 , 10 mM HEPES-buffered to pH 7 . 4 , 10 mM glucose , CNQX 10 μM; APV 50 μM ) through a laminar flow perfusion system . For electrical field stimulation , coverslips were mounted into the imaging chamber ( ~100 μl volume; Quick Exchange Platform; Warner Instruments ) and APs were evoked by passing 1 msec current pulses through platinum iridium electrodes using an AM2100 stimulator ( AM-Systems ) . After 10 s of baseline acquisition ( F0 ) , neurons were stimulated with a train of 40/400 APs at 20Hz in the presence or absence of 1 μM Bafilomycin . At the end of the stimulation protocol , cells were perfused with 50 mM NH4Cl ( Fmax ) . Images were analyzed using the eXcellence software ( Olympus ) . The total increase in the fluorescence signal ( ΔF ) was calculated by subtracting F0 and the ΔF was normalized to the fluorescence value obtained by alkalization of the entire vesicle pool using NH4Cl ( ΔFmax ) . The time constant of endocytosis ( τ endo ) was calculated by the fitting the post-stimulus decay with a single-exponential function . To quantify net exocytosis ( EXO ) ongoing during stimulation , the areas below the curves obtained in the presence of Bafilomycin were measured . To quantify the net during stimulus’ endocytosis ( ENDO ) , additional curves were built from the differences between the curves run in the presence and those in the absence of Bafilomycin and the areas below the curves were measured . To evaluate the SypHy surface fraction , neurons were perfused for 20 sec with Tyrode Solution ( pH 7 . 4 ) to measure the basal fluorescence ( Fbasal ) , then with MES buffer ( pH 5 . 5 ) ( Facid ) to measure the fluorescence under the acidic condition , and lastly with a Tyrode solution with NH4Cl ( Falkalin ) to evaluate the total amount of probes at the synaptic boutons . SypHy surface fraction was evaluated by subtracting the fluorescence change between basal and acidic conditions ( Fbasal-Facid ) from the fluorescence change between alkaline and acidic conditions ( Falkalin-Facid ) and reported as percent values , as follows: Fextra =[ ( Fbasal-Facid ) / ( Falkalin-Facid ) ]*100 . Data are expressed as means ± SEM for the number of coverslips analyzed . The number of synapses analyzed per coverslip ranged between 20 and 40 . Three to five cells preparation were performed for each experimental condition . Primary hippocampal neurons were stimulated as above in presence of 10mg/ml soluble HRP . At the end of the stimulus , soluble HRP was washed out and neurons were fixed either immediately or after 2 or 20 min . After chemical fixation , neurons were washed in 0 . 1 M cacodylate buffer and incubated for 10 min in a solution containing 0 . 3 mg/ml of 3 , 3-Diaminobenzine ( DAB ) in 0 . 1 M cacodylate buffer . Neurons were incubated in a solution containing 0 . 3 mg/ml of DAB + 0 . 003% H2O2 in 0 . 1 M cacodylate buffer until a brown substrate developed , to allow for HRP peroxidation . Neurons were then rinsed in cold distilled water , to block DAB peroxidation and post-fixed in 1% OsO4 , 1 . 5% K4Fe ( CN ) 6 , 0 . 1 M sodium cacodylate . The numbers of HRP-positive SVs and cisternae were evaluated using NIH ImageJ . Data were analyzed using the Student’s t-test or , in case of more than two experimental groups , by one- or two-way ANOVA , followed by post hoc multiple-comparison tests ( Bonferroni’s or Dunnet’s ) using GraphPad Prism or SigmaPlot software . The significance level was set at p<0 . 05 . | Communication between neurons takes place at cell-to-cell contacts called synapses . Each synapse is formed between one neuron that sends the message , and another neuron that receives it . The neuron before the synapse – called the presynaptic neuron – contains packets called synaptic vesicles , which are full of chemical messengers ready to be released upon activity . Accurate communication between neurons relies on the exact composition , and organized trafficking , of the synaptic vesicles when the neuron is active . Synapses also contain bigger structures , called endosomal structures , which may represent an intermediate station in which synaptic vesicle composition is controlled . However , the trafficking of synaptic vesicles through the endosomal structures is poorly understood . Now , Tagliatti , Fadda et al . have revealed that a protein called Arf6 plays an important role in presynaptic neurons . The experiments involved rat neurons grown in the laboratory , and showed that Arf6 controls both the number of synaptic vesicles ready to be released and the trafficking of synaptic vesicles via endosomal structures in active neurons . The next step following on from these findings is to understand how Arf6 exerts its effects and how this protein is regulated in the presynaptic neuron . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | Arf6 regulates the cycling and the readily releasable pool of synaptic vesicles at hippocampal synapse |
A guiding principle in self-assembly is that , for high production yield , nucleation of structures must be significantly slower than their growth . However , details of the mechanism that impedes nucleation are broadly considered irrelevant . Here , we analyze self-assembly into finite-sized target structures employing mathematical modeling . We investigate two key scenarios to delay nucleation: ( i ) by introducing a slow activation step for the assembling constituents and , ( ii ) by decreasing the dimerization rate . These scenarios have widely different characteristics . While the dimerization scenario exhibits robust behavior , the activation scenario is highly sensitive to demographic fluctuations . These demographic fluctuations ultimately disfavor growth compared to nucleation and can suppress yield completely . The occurrence of this stochastic yield catastrophe does not depend on model details but is generic as soon as number fluctuations between constituents are taken into account . On a broader perspective , our results reveal that stochasticity is an important limiting factor for self-assembly and that the specific implementation of the nucleation process plays a significant role in determining the yield .
We model the assembly of a fixed number of well-defined target structures from limited resources . Specifically , we consider a set of S different species of constituents denoted by 1 , … , S which assemble into rings of size L . The cases S=1 and 1<S≤L ( S=L ) are denoted as homogeneous and partially ( fully ) heterogeneous , respectively . The homogeneous model builds on previous work on virus capsid ( Chen et al . , 2008; Hagan et al . , 2011 ) , linear protein filament assembly ( Michaels et al . , 2016; Michaels et al . , 2017; D'Orsogna et al . , 2012 ) and aggregation and polymerization models ( Krapivsky et al . , 2010 ) . The heterogeneous model in turn links to previous model systems used to study , for example , DNA-brick-based assembly of heterogeneous structures ( Murugan et al . , 2015; Hedges et al . , 2014; D'Orsogna et al . , 2013 ) . We emphasize that , even though strikingly similar experimental realizations of our model exist ( Gerling et al . , 2015; Wagenbauer et al . , 2017; Praetorius and Dietz , 2017 ) , it is not intended to describe any particular system . The ring structure represents a general linear assembly process involving building blocks with equivalent binding properties and resulting in a target of finite size . The main assumption in the ring model is that the different constituents assemble linearly in a sequential order . In many biological self-assembling systems like bacterial flagellum assembly or biogenesis of the ribosome subunits the assumption of a linear binding sequence appears to be justified ( Peña et al . , 2017; Chevance and Hughes , 2008 ) . In order to test the validity of our results beyond these constraints we also perform stochastic simulations of generalized self-assembling systems that do not obey a sequential binding order: i ) by explicitly allowing for polymer-polymer bindings and ii ) by considering the assembly of finite sized squares that grow independently in two dimensions ( see Figures 6 and 7 ) . The assembly process starts with N inactive monomers of each species . We use C=N/V to denote the initial concentration of each monomer species , where V is the reaction volume . Monomers are activated independently at the same per capita rate α , and , once active , are available for binding . Binding takes place only between constituents of species with periodically consecutive indices , for example 1 and 2 or S and 1 ( leading to structures such as …1231… for S=3 ) ; see Figure 1 . To avoid ambiguity , we restrict ring sizes to integer multiples of the number of species S . Furthermore , we neglect the possibility of incorrect binding , for example species 1 binding to 3 or S-1 . Polymers , that is incomplete ring structures , grow via consecutive attachment of monomers . For simplicity , polymer-polymer binding is disregarded at first , as it is typically assumed to be of minor importance ( Zlotnick et al . , 1999; Chen et al . , 2008; Murugan et al . , 2015; Haxton and Whitelam , 2013 ) . To probe the robustness of the model , later we consider an extended model including polymer-polymer binding for which the results are qualitatively the same ( see Figure 6 and the discussion ) . Furthermore , it has been observed that nucleation phenomena play a critical role for self-assembly processes ( Ke et al . , 2012; Wei et al . , 2012; Reinhardt and Frenkel , 2014; Chen et al . , 2008 ) . So it is in general necessary to take into account a critical nucleation size , which marks the transition between slow particle nucleation and the faster subsequent structure growth ( Michaels et al . , 2016; Lazaro and Hagan , 2016; Morozov et al . , 2009; Murugan et al . , 2015 ) . We denote this critical nucleation size by Lnuc , which in terms of classical nucleation theory corresponds to the structure size at which the free energy barrier has its maximum . For l<Lnuc attachment of monomers to existing structures and decay of structures ( reversible binding ) into monomers take place at size-dependent reaction rates μl and δl , respectively ( Figure 1 ) . Here , we focus on identical rates μl=μ and δl=δ . A discussion of the general case is given in Appendix 4 . Above the nucleation size , polymers grow by attachment of monomers with reaction rate ν≥μ per binding site . As we consider successfully nucleated structures to be stable on the observational time scales , monomer detachment from structures above the critical nucelation size is neglected ( irreversible binding ) ( Murugan et al . , 2015; Chen et al . , 2008 ) . Complete rings neither grow nor decay ( absorbing state ) . We investigate two scenarios for the control of nucleation speed , first separately and then in combination . For the ‘activation scenario’ we set μ=ν ( all binding rates are equal ) and control the assembly process by varying the activation rate α . For the ‘dimerization scenario’ all particles are inherently active ( α→∞ ) and we control the assembly process by varying the dimerization rate μ ( we focus on Lnuc=2 ) . It has been demonstrated previously in Chen et al . ( 2008 ) and ( Endres and Zlotnick , 2002; Hagan and Elrad , 2010; Morozov et al . , 2009 ) that either a slow activation or a slow dimerization step are suitable in principle to retard nucleation and favour growth of the structures over the initiation of new ones . We quantify the quality of the assembly process in terms of the assembly yield , defined as the number of successfully assembled ring structures relative to the maximal possible number NS/L . Yield is measured when all resources have been used up and the system has reached its final state . We do not discuss the assembly time in this manuscript , however , in Appendix 5 we show typical trajectories for the time evolution of the yield in the activation and dimerization scenario . If the assembly product is stable ( absorbing state ) , the yield can only increase with time . Consequently , the final yield constitutes the upper limit for the yield irrespective of additional time constraints . Therefore , the final yield is an informative and unambiguous observable to describe the efficiency of the assembly reaction . We simulated our system both stochastically via Gillespie’s algorithm ( Gillespie , 2007 ) and deterministically as a set of ordinary differential equations corresponding to chemical rate equations ( see Appendix 1 ) .
First , we consider the macroscopic limit , N≫1 , and investigate how assembly yield depends on the activation rate α ( activation scenario ) and the dimerization rate μ ( dimerization scenario ) for Lnuc=2 . Here , the deterministic description coincides with the stochastic simulations ( Figure 2a and b ) . For both high activation and high dimerization rates , yield is very poor . Upon decreasing either the activation rate ( Figure 2a ) or the dimerization rate ( Figure 2b ) , however , we find a threshold value , αth or μth , below which a rapid transition to the perfect yield of 1 is observed both in the deterministic and stochastic simulation . By exploiting the symmetries of the system with respect to relabeling of species , one can show that , in the deterministic limit , the behavior is independent of the number of species S ( for fixed L and N , see Appendix 1 ) . Consequently , all systems behave equivalently to the homogeneous system and yield becomes independent of S in this limit . Note , however , that equivalent systems with differing S have different total numbers of particles SN and hence assemble different total numbers of rings . Decreasing the activation rate reduces the concentration of active monomers in the system . Hence growth of the polymers is favored over nucleation , because growth depends linearly on the concentration of active monomers while nucleation shows a quadratic dependence . Likewise , lower dimerization rates slow down nucleation relative to growth . Both mechanisms therefore restrict the number of nucleation events , and ensure that initiated structures can be completed before resources become depleted ( see Figure 2c and d ) . Mathematically , the deterministic time evolution of the polymer size distribution c ( l , t ) is described by an advection-diffusion equation ( Endres and Zlotnick , 2002; Yvinec et al . , 2012 ) with advection and diffusion coefficients depending on the instantaneous concentration of active monomers ( see Appendix 2 ) . Solving this equation results in the wavefront of the size distribution advancing from small to large polymer sizes ( Figure 2e ) . Yield production sets in as soon as the distance travelled by this wavefront reaches the maximal ring size L . Exploiting this condition , we find that in the deterministic system for Lnuc=2 , a non-zero yield is obtained if either the activation rate or the dimerization rate remains below a corresponding threshold value , that is if α<αth or μ<μth , where ( 1 ) αth=PανμνC ( L-L ) 3andμth=Pμν ( L-L ) 2 ( see Appendix 3 ) with proportionality constants Pα=[πΓ ( 2/3 ) /Γ ( 7/6 ) ]3/3≈5 . 77 and Pμ=π2/2≈4 . 93 . These relations generalize previous results ( Morozov et al . , 2009 ) to finite activation rates and for heterogeneous systems . A comparison between the threshold values given by Equation 1 and the simulated yield curves is shown in Figure 2a , b . The relations highlight important differences between the two scenarios ( where α→∞ and μ=ν , respectively ) : While αth decreases cubically with the ring size L , μth does so only quadratically . Furthermore , the threshold activation rate αth increases with the initial monomer concentration C . Consequently , for fixed activation rate , the yield can be optimized by increasing C . In contrast , the threshold dimerization rate is independent of C and the yield curves coincide for N≫1 . Finally , if α is finite and μ<ν , the interplay between the two slow-nucleation scenarios may lead to enhanced yield . This is reflected by the factor ν/μ in αth , and we will come back to this point later when we discuss the stochastic effects . In summary , for large particle numbers ( N≫1 ) , perfect yield can be achieved in two different ways , independently of the heterogeneity of the system - by decreasing either the activation rate ( activation scenario ) or the dimerization rate ( dimerization scenario ) below its respective threshold value . Next , we consider the limit where the particle number becomes relevant for the physics of the system . In the activation scenario , we find a markedly different phenomenology if resources are sparse . Figure 3a shows the dependence of the average yield on the activation rate for different , low particle numbers in the completely heterogeneous case ( S=L ) . Here , we restrict our discussion to the average yield . The error of the mean is negligible due to the large number of simulations used to calculate the average yield . Still , due to the randomness in binding and activation , the yield can differ between simulations . A figure with the average yield and its standard deviation is shown in Appendix 6 . For very low and very high average yield , the standard deviation has to be small due to the boundedness of the yield . For intermediate values of the average , the standard deviation is highest but still small compared to the average yield . Thus , the average yield is meaningful for the essential understanding of the assembly process . Whereas the deterministic theory predicts perfect yield for small activation rates , in the stochastic simulation yield saturates at an imperfect value ymax<1 . Reducing the particle number N decreases this saturation value ymax until no finished structures are produced ( ymax→0 ) . The magnitude of this effect strongly depends on the size of the target structure L if the system is heterogeneous . Figure 3c shows a diagram characterizing different regimes for the saturation value of the yield , ymax ( N , L ) , in dependence of the particle number N and the size of the target structure L for fully heterogeneous systems ( S=L ) . We find that the threshold particle number Nyth necessary to obtain a fixed yield y increases nonlinearly with the target size L . For the depicted range of L , the dependence of the threshold for nonzero yield , N>0th , on L can approximately be described by a power-law: N>0th∼Lξ , with exponent ξ≈2 . 8 for L≤600 . Consequently , for L=600 already more than 105 rings must be assembled in order to obtain a yield larger than zero . In Appendix 8 we included two additional plots that show the dependence of ymax on N for fixed L and the dependence on L for fixed N , respectively . The suppression of the yield is caused by fluctuations ( see explanation below ) and is not captured by a deterministic description . Because these stochastic effects can decrease the yield from a perfect value in a deterministic description to zero ( see Figure 3a ) , we term this effect ‘stochastic yield catastrophe’ . For fixed target size L and fixed maximum number of target structures NSL , ymax increases with decreasing number of species , see Figure 3d . In the fully homogeneous case , S=1 , a perfect yield of 1 is always achieved for α→0 . The decrease of the maximal yield with the number of species S thus suggests that , in order to obtain high yield , it is beneficial to design structures with as few different species as possible . In large part this effect is due to the constraint SN=const , whereby the more homogeneous systems ( small S ) require larger numbers of particles per species N and , correspondingly , exhibit less stochasticity . If N is fixed instead of SN , the yield still initially decreases with increasing number of species S but then quickly reaches a stationary plateau and gets independent of S for S≫1 , see Appendix 7 . Moreover , increasing the nucleation size Lnuc , and with it the reversibility of binding , also increases ymax , see Figure 3 ( d ) . This indicates that , beside heterogeneity of the target structure , irreversibility of binding on the relevant time scale makes the system susceptible to stochastic effects . The stochastic yield catastrophe is mainly attributable to fluctuations in the number of active monomers . In the deterministic ( mean-field ) equation the different particle species evolve in balanced stoichiometric concentrations . However , if activation is much slower than binding , the number of active monomers present at any given time is small , and the mean-field assumption of equal concentrations is violated due to fluctuations ( for S>1 ) . Activated monomers then might not fit any of the existing larger structures and would instead initiate new structures . Figure 4a illustrates this effect and shows how fluctuations in the availability of active particles lead to an enhanced nucleation and , correspondingly , to a decrease in yield . Due to the effective enhancement of the nucleation rate , the resulting polymer size distribution has a higher amplitude than that predicted deterministically ( Figure 4b ) and the system is prone to depletion traps . A similar broadening of the size distribution has been reported in the context of stochastic coagulation-fragmentation of identical particles ( D'Orsogna et al . , 2015 ) . In the dimerization scenario , in contrast , there is no stochastic activation step . All particles are available for binding from the outset . Consequently , stochastic effects do not play an essential role in the dimerization scenario and perfect yield can be reached robustly for all system sizes , regardless of the number of species S ( Figure 3 ( b ) ) . So far , the two implementations of the ‘slow nucleation principle’ have been investigated separately . Surprisingly , we observe counter-intuitive behavior in a mixed scenario in which both dimerization and activation occur slowly ( i . e . , μ<ν , α<∞ ) . Figure 5 shows that , depending on the ratio μ/ν , the yield can become a non-monotonic function of α . In the regime where α is large , nucleation is dimerization-limited; therefore activation is irrelevant and the system behaves as in the dimerization scenario for α→∞ . Upon decreasing α we then encounter a second regime , where activation and dimerization jointly limit nucleation . The yield increases due to synergism between slow dimerization and activation ( see μ/ν dependence of αth , Equation 1 ) , whilst the average number of active monomers is still high and fluctuations are negligible . Finally , a stochastic yield catastrophe occurs if α is further reduced and activation becomes the limiting step . This decline is caused by an increase in nucleation events due to relative fluctuations in the availability of the different species ( ‘fluctuations between species’ ) . This contrasts the deterministic description where nucleation is always slower for smaller activation rate . Depending on the ratio μ/ν , the ring size L and the particle number N , maximal yield is obtained either in the dimerization-limited ( red curves , Figure 5 ) , activation-limited ( blue curve , Figure 5b ) or intermediate regime ( green and orange curves , Figure 5 ) . In our model , the reason for the stochastic yield catastrophe is that - due to fluctuations between species - the effective nucleation rate is strongly enhanced . Hence , if binding to a larger structure is temporarily impossible , activated monomers tend to initiate new structures , causing an excess of structures that ultimately cannot be completed . Natural questions that arise are whether ( i ) relaxing the constraint that polymers cannot bind other polymers or ( ii ) abandoning the assumption of a linear assembly path , will resolve the stochastic yield catastrophe . To answer these questions , we performed stochastic simulations for extensions of our model system showing that the stochastic yield catastrophe indeed persists . We start by considering the ring model from the previous section but take polymer-polymer binding into account in addition to growth via monomer attachment ( Figure 6 ) . In detail , we assume that two structures of arbitrary size ( and with combined length ≤L ) bind at rate ν if they fit together , that is if the left ( right ) end of the first structure is periodically continued by the right ( left ) end of the second one . Realistically , the rate of binding between two structures is expected to decrease with the motility and thus the sizes of the structures . In order to assess the effect of polymer-polymer binding , we focus on the worst case where the rate for binding is independent of the size of both structures . If a stochastic yield catastrophe occurs for this choice of parameters , we expect it to be even more pronounced in all the ‘intermediate cases’ . Figure 6 shows the dependence of the yield on the activation rate in the polymer-polymer model . As before , yield increases below a critical activation rate and then saturates at an imperfect value for small activation rates . Decreasing the number of particles per species , decreases this saturation value . Compared to the original model , the stochastic yield catastrophe is mitigated but still significant: For structures of size S=L=100 , yield saturates at around 0 . 87 for N=100 particles per species and at around 0 . 33 for N=10 particles per species . We thus conclude that polymer-polymer binding indeed alleviates the stochastic yield catastrophe but does not resolve it . Since binding only happens between consecutive species , structures with overlapping parts intrinsically can not bind together and depletion traps continue to occur . Taken together , also in the extended model , fluctuations in the availability of the different species lead to an excess of intermediate-sized structures that get kinetically trapped due to structural mismatches . Note that in the extreme case of N=1 , incomplete polymers can always combine into one final ring structure so that in this case the yield is always 1 . Analogously , for high activation rates yield is improved for N=10 compared to N≥50 ( Figure 6b ) . Kinetic trapping due to structural mismatches can occur in every ( partially ) irreversible heterogeneous assembly process with finite-sized target structure and limited resources . From our results , we thus expect a stochastic yield catastrophe to be common to such systems . In order to further test this hypothesis , we simulated another variant of our model where finite sized squares assemble via monomer attachment from a pool of initially inactive particles , see Figure 7 . In contrast to the original model , the assembled structures are non-periodic and exhibit a non-linear assembly path where structures can grow independently in two dimensions . While the ring model assumes a sequential order of binding of the monomers , the square allows for a variety of distinct assembly paths that all lead to the same final structure . Note that , because of the absence of periodicity , the square model is only well defined for the completely heterogeneous case . Figure 7 depicts the dependence of the yield on the activation rate for a square of size S=100 . Also in this case , we find that the yield saturates at an imperfect value for small activation rates . Hence , we showed that the stochastic yield catastrophe is not resolved neither by accounting for polymer-polymer combination nor by considering more general assembly processes with multiple parallel assembly paths . This observation supports the general validity of our findings and indicates that stochastic yield catastrophes are a general phenomenon of ( partially ) irreversible and heterogeneous self-assembling systems that occur if particle number fluctuations are non-negligible .
Our results show that different ways to slow down nucleation are indeed not equivalent , and that the explicit implementation is crucial for assembly efficiency . Susceptibility to stochastic effects is highly dependent on the specific scenario . Whereas systems for which dimerization limits nucleation are robust against stochastic effects , stochastic yield catastrophes can occur in heterogeneous systems when resource supply limits nucleation . The occurrence of stochastic yield catastrophes is not captured by the deterministic rate equations , for which the qualitative behavior of both scenarios is the same . Therefore , a stochastic description of the self-assembly process , which includes fluctuations in the availability of the different species , is required . The interplay between stochastic and deterministic dynamics can lead to a plethora of interesting behaviors . For example , the combination of slow activation and slow nucleation may result in a non-monotonic dependence of the yield on the activation rate . While deterministically , yield is always improved by decreasing the activation rate , stochastic fluctuations between species strongly suppress the yield for small activation rate by effectively enhancing the nucleation speed . This observation clearly demonstrates that a deterministically slow nucleation speed is not sufficient in order to obtain good yield in heterogeneous self-assembly . For example , a slow activation step does not necessarily result in few nucleation events although deterministically this behavior is expected . Thus , our results indicate that the slow nucleation principle has to be interpreted in terms of the stochastic framework and have important implications for yield optimization . We showed that demographic noise can cause stochastic yield catastrophes in heterogeneous self-assembly . However , other types of noise , such as spatiotemporal fluctuations induced by diffusion , are also expected to trigger stochastic yield catastrophes . Hence , our results have broad implications for complex biological and artificial systems , which typically exhibit various sources of noise . We characterize conditions under which stochastic yield catastrophes occur , and demonstrate how they can be mitigated . These insights could usefully inform the design of experiments to circumvent yield catastrophes: In particular , while slow provision of constituents is a feasible strategy for experiments , it is highly susceptible to stochastic effects . On the other hand , irrespective of its robustness to stochastic effects , the experimental realization of the dimerization scenario relies on cooperative or allosteric effects in binding , and may therefore require more sophisticated design of the constituents ( Sacanna et al . , 2010; Zeravcic et al . , 2017 ) . Our theoretical analysis shows that stochasticity can be alleviated either by decreasing heterogeneity ( presumably lowering realizable complexity ) or by increasing reversibility ( potentially requiring fine-tuning of bond strengths and reducing the stability of the assembly product ) . Alternative approaches to control stochasticity include the promotion of specific assembly paths ( Murugan et al . , 2015; Gartner , Graf and Frey , in preparation ) and the control of fluctuations ( Graf , Gartner and Frey , in preparation ) . One possibility to test these ideas and the ensuing control strategies could be via experiments based on DNA origami . Instead of building homogeneous ring structures as in Wagenbauer et al . ( 2017 ) , one would have to design heterogeneous ring structures made from several different types of constituents with specified binding properties . By varying the opening angle of the ‘wedges’ ( and thus the preferred number of building blocks in the ring ) and/or the number of constituents , both the target structure size L as well as the heterogeneity of the target structure S could be controlled . Moreover , the ideas presented in this manuscript are relevant for the understanding of intracellular self-assembly . In cells , provision of building blocks is typically a gradual process , as synthesis is either inherently slow or an explicit activation step , such as phosphorylation , is required . In addition , the constituents of the complex structures assembled in cells are usually present in small numbers and subject to diffusion . Hence , stochastic yield catastrophes would be expected to have devastating consequences for self-assembly , unless the relevant cellular processes use elaborate control mechanisms to circumvent stochastic effects . Further exploration of these control mechanisms should enhance the understanding of self-assembly processes in cells and help improve synthesis of complex nanostructures .
We start with the general Master equation and derive the chemical rate equations ( deterministic/mean-field equations ) for the heterogeneous self-assembly process . We renounce to show the full Master equation here but instead state the system that describes the evolution of the first moments . To this end , we denote the random variable that describes the number of polymers of size ℓ and species s in the system at time t by nℓs ( t ) with 2≤ℓ<L and 1≤s≤S . The species of a polymer is defined by the species of the respective monomer at its left end . Furthermore , n0s and n1s denote the number of inactive and active monomers of species s , respectively , and nL the number of complete rings . We signify the reaction rate for binding of a monomer to a polymer of size ℓ by νℓ . α denotes the activation rate and δℓ the decay rate of a polymer of size ℓ . By ⟨…⟩ we indicate ( ensemble ) averages . The system governing the evolution of the first moments ( the averages ) of the {nℓs} is then given by: ( 2a ) ddt⟨n0s⟩=−α⟨n0s⟩ , ( 2b ) ddt⟨n1s⟩=α⟨n0s⟩−∑ℓ=1L−1νℓ ( ⟨n1snℓs+1⟩+⟨n1snℓs−ℓ⟩ ) +∑ℓ=2Lnuc−1∑k=s+1−ℓk=sδℓ⟨nℓk⟩ , ( 2c ) ddt⟨n2s⟩=ν1⟨n1sn1s+1⟩−ν2⟨n2sn1s+2⟩−ν2⟨n2sn1s−1⟩−δ2⟨n2s⟩1{2<Lnuc} , ( 2d ) ddt⟨nℓs⟩=νℓ−1⟨nℓ−1sn1ℓ+s−1⟩+νℓ−1⟨nℓ−1s+1n1s⟩−νℓ⟨nℓsn1s+ℓ⟩−νℓ⟨nℓsn1s−1⟩−δ⟨nℓs⟩1{ℓ<Lnuc } , ( 2e ) ddt⟨nLs⟩=νL−1⟨nL−1sn1L+s−1⟩+νL−1⟨nL−1s+1n1s⟩ . The different terms of this equation are illustrated graphically in Figure 8 . The first equation describes loss of inactive particles due to activation at rate α . Equation 2b gives the temporal change of the number of active monomers that is governed by the following processes: activation of inactive monomers at rate α , binding of active monomers to the left or to the right end of an existing structure of size ℓ at rate νℓ , and decay of below-critical polymers of size ℓ into monomers at rate δℓ ( disassembly ) . Equations 2c and 2d describe the dynamics of dimers and larger polymers of size 3≤ℓ<L , respectively . The terms account for reactions of polymers with active monomers ( polymerization ) as well as decay in the case of below-critical polymers ( disassembly ) . The indicator function 𝟏{x<Lnuc} equals 1 if the condition x<Lnuc is satisfied and 0 otherwise . Note that a polymer of size ℓ≥3 can grow by attaching a monomer to its left or to its right end whereas the formation of a dimer of a specific species is only possible via one reaction pathway ( dimerization reaction ) . Finally , polymers of length L – the complete ring structures – form an absorbing state and , therefore , include only the respective gain terms ( cf Equation 2e ) . We simulated the Master equation underlying Equation 2 stochastically using Gillespie’s algorithm . For the following deterministic analysis , we neglect correlations between particle numbers {nℓs} , which is valid assumption for large particle numbers . Then the two-point correlator can be approximated as the product of the corresponding mean values ( mean-field approximation ) ( 3 ) ⟨nisnjk⟩=⟨nis⟩⟨njk⟩∀s , k Furthermore , for the expectation values it must hold ( 4 ) ⟨nℓs⟩=⟨nℓ1⟩∀sbecause all species have equivalent properties ( there is no distinct species ) and hence the system is invariant under relabelling of the upper index . By ( 5 ) cℓ:=⟨nℓs⟩V , we denote the concentration of any monomer or polymer species of size ℓ , where V is the reaction volume . Due to the symmetry formulated in Equation 4 , the heterogeneous assembly process decouples into a set of S identical and independent homogeneous assembly processes in the deterministic limit . The corresponding homogeneous system then is described by the following set of equations that is obtained by applying ( Equation 3 , Equation 4 ) and ( Equation 5 ) to ( Equation 2 ) ( 6a ) ddtc0=-αc0 , ( 6b ) ddtc1=αc0-2c1∑ℓ=1L-1νℓcℓ+∑ℓ=2Lnuc-1lδℓcℓ , ( 6c ) ddtc2=ν1c12−2ν2c1c2−δ2c21{2<Lnuc} , ( 6d ) ddtcℓ=2νℓ−1c1cℓ−1−2νℓc1cℓ−δℓcℓ1{ℓ<L nuc} , for 3≤ℓ<L , ( 6e ) ddtcL=2νL-1c1cL-1 . The rate constants νℓ in Equations 6 and 2 differ by a factor of V . For convenience , we use however the same symbol in both cases . The rate constants νℓ in Equation 6 can be interpreted in the usual units [litermol sec] . Due to the symmetry , the yield , which is given by the quotient of the number of completely assembled rings and the maximum number of complete rings , becomes independent of the number of species S ( 7 ) yield ( t ) =ScL ( t ) VSNL-1=cL ( t ) VLN . Hence , it is enough to study the dynamics of the homogeneous system , Equation 6 , to identify the condition under which non zero yield is obtained . The dynamical properties of the evolution of the polymer-size distribution become evident if the set of ODEs , Equation 6 , is rewritten as a partial differential equation . This approach was previously described in the context of virus capsid assembly ( Zlotnick et al . , 1999; Morozov et al . , 2009 ) . For simplicity , we restrict ourselves to the case Lnuc= 2 and let ν1=μ and νℓ≥2=ν . Then , for the polymers with ℓ>2 we have ( 8 ) ∂tcℓ=2νc1[cℓ-1-cℓ] . As a next step , we approximate the index ℓ∈{2 , 3 , … , L} indicating the length of the polymer as a continuous variable x∈[2 , L] and define c ( x=ℓ ) :=cℓ . By A:=c1 we denote the concentration of active monomers in the following to emphasize their special role . Formally expanding the right-hand side of Equation 8 in a Taylor series up to second order ( 9 ) c ( ℓ-1 ) =c ( ℓ ) -∂xc ( ℓ ) +12∂x2c ( ℓ ) , one arrives at the advection-diffusion equation with both advection and diffusion coefficients depending on the concentration of active monomers A ( t ) ( 10 ) ∂tc ( x ) =-2νA∂xc ( x ) +νA∂x2c ( x ) . Equation 10 can be written in the form of a continuity equation ∂tc ( x ) =-∂xJ ( x ) with flux J= 2νAc-νA∂xc . The flux at the left boundary x= 2 equals the influx of polymers due to dimerization of free monomers J ( 2 , t ) =μA2 . This enforces a Robin boundary condition at x= 2 ( 11 ) 2νAc ( 2 , t ) -νA∂xc ( 2 , t ) =μA2 . At x=L we set an absorbing boundary c ( L , t ) = 0 so that completed structures are removed from the system . The time evolution of the concentration of active monomers is given by ( 12 ) ∂tA=αCe-αt-2μA2-2νA∫2Lc ( x , t ) 𝑑x . The terms on the right-hand side account for activation of inactive particles , dimerization , and binding of active particles to polymers ( polymerization ) . Qualitatively , Equation 10 describes a profile that emerges at x= 2 from the boundary condition Equation 11 , moves to the right with time-dependent velocity 2νA ( t ) due to the advection term , and broadens with a time-dependent diffusion coefficient νA ( t ) . In Appendices 2–3 we show how the full solution of Equations 10 and 11 can be found assuming knowledge of A ( t ) . Here , we focus only on the derivation of the threshold activation rate and threshold dimerization rate that mark the onset of non-zero yield . Yield production starts as soon as the density wave reaches the absorbing boundary at x=L . Therefore , finite yield is obtained if the sum of the advectively travelled distance dadv and the diffusively travelled distance ddiff exceeds the system size L-2 ( 13 ) dadv+ddiff≥L-2 . According to Equation 10 , dadv=2ν∫0∞A ( t ) 𝑑t and ddiff=2ν∫0∞A ( t ) 𝑑t , giving as condition for the onset of finite yield ( 14 ) 2ν∫0∞A ( t ) 𝑑t= ! 14 ( 1+4 ( L-2 ) -1 ) 2≈L-L , where the last approximation is valid for large L . In order to obtain ∫0∞A ( t ) 𝑑t we derive an effective two-component system that governs the evolution of A ( t ) . To this end , we denote the total number of polymers in Equation 12 by B ( t ) :=∫2∞c ( x , t ) 𝑑x ( as long as yield is zero the upper boundary is irrelevant and we can consider L=∞ ) . Equation 12 then reads ( 15 ) ddtA=αCe-αt-2μA2-2νAB , and the dynamics of B is determined from the boundary condition , Equation 11 ( 16 ) ddtB=∫2∞∂tc ( x , t ) dx=∫2∞−∂xJ ( x , t ) dx=−J ( ∞ , t ) ⏟=0+J ( 2 , t ) =μA ( t ) 2 . Measuring A and B in units of the initial monomer concentration C and time in units of ( νC ) -1 the equations are rewritten in dimensionless units as ( 17a ) ddtA=ωe−ωt−2ηA2−2AB , ( 17b ) ddtB=ηA2 , where ω=ανC and η=μν . Equation 17 describes a closed two-component system for the concentration of active monomers A and the total concentration of polymers B . It describes the dynamics exactly as long as yield is zero . In order to evaluate the condition ( 14 ) we need to determine the integral over A ( t ) as a function of ω and η ( 18 ) ∫0∞Aω , η ( t ) 𝑑t:=g ( ω , η ) . To that end , we proceed by looking at both scenarios separately . The numerical analysis , confirming our analytic results , is given in Appendix 3 . The activation rate in the dimerization scenario is α→∞ , and instead of the term ωe-ωt in dA/dt , we set the initial condition A ( 0 ) =1 ( and B ( 0 ) =0 ) . Furthermore , η=μ/ν≪1 and we can neglect the term proportional to η in dA/dt . As a result , dAdB=-2BηA . Solving this equation for A as a function of B using the initial condition A ( B=0 ) =1 , the totally travelled distance of the wave is determined to be ( 19 ) 2g ( ω , η ) =2π221η , where for the evaluation of the integral we used the substitution ηA2dt=dB . In the activation scenario , yield sets in only if the activation rate and thus the effective nucleation rate is slow . As a result , in addition to ω≪1 , we can again neglect the term proportional to η in dA/dt . This time , however , we have to keep the term ωe-ωt . As a next step , we assume that dA/dt is much smaller than the remaining terms on the right-hand side , ωe-ωt and -2AB . This assumption might seem crude at first sight but is justified a posteriori by the solution of the equation ( see Appendix 3 ) . Hence , we get the algebraic equation A ( t ) =ωe-ωt/ ( 2B ( t ) ) . Using it to solve dB/dt=ηA2 for B , and then to determine A , the totally travelled distance of the wave is deduced as ( 20 ) 2g ( ω , η ) =232/3πΓ ( 2/3 ) 6Γ ( 7/6 ) ( ωη ) -1/3 . Taken together , we therefore obtain two conditions out of which one must be fulfilled in order to obtain finite yield ( 21 ) 2a ( ηω ) −13≥L−L⇒α<αth:=PανμνC ( L−L ) 3 ( 22 ) or2bη−12≥L−L⇒μ<μth:=Pμν ( L−L ) 2 , where a and b are numerical factors , and Pα= 8a3≈5 . 77 and Pμ= 4b2≈4 . 93 . This verifies Equation 1 in the main text . | The self-assembly of a large biological molecule from small building blocks is like finishing a puzzle of magnetic pieces by shaking the box . Even though each piece of the puzzle is attracted to its correct neighbours , the limited control makes it very hard to finish the puzzle in a short amount of time . The problem becomes even more difficult if several copies of the same puzzle are assembled in one box . If several puzzles start at the same time , the different parts might steal pieces from each other , making it impossible to successfully complete any of the puzzles . This is called a depletion trap . If the box is only shaken and there is no real control over individual pieces , these traps occur at random . Overcoming these random depletion traps is an important challenge when assembling nanostructures and other artificial molecules designed by humans without wasting many , potentially expensive , components . Previous studies have shown that when multiple copies of the same structure are assembled simultaneously , slowing the rate of initiation increases the yield of correctly-made structures . This prevents new structures from stealing pieces from existing structures before they are fully completed . Now , Gartner , Graf , Wilke et al . have used a mathematical model to show that changing the way initiation is delayed leads to different yields . This was especially true for small systems where fluctuations in the availability of the different pieces strongly enhanced the initiation of new structures . In these cases , the self-assembly process terminated undesirably with many incomplete structures . Nanostructures have various applications ranging from drug delivery to robotics . These findings suggest that in order to efficiently assemble biological molecules , the concentrations of the different building blocks need to be tightly controlled . A question for further research is to investigate strategies that reduce fluctuations in the availability of the building blocks to develop more efficient assembly protocols . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"physics",
"of",
"living",
"systems"
] | 2020 | Stochastic yield catastrophes and robustness in self-assembly |
Biofilms are communities of microorganisms attached to a surface or each other . Biofilm-associated cells are the etiologic agents of recurrent Staphylococcus aureus infections . Infected human tissues are hypoxic or anoxic . S . aureus increases biofilm formation in response to hypoxia , but how this occurs is unknown . In the current study we report that oxygen influences biofilm formation in its capacity as a terminal electron acceptor for cellular respiration . Genetic , physiological , or chemical inhibition of respiratory processes elicited increased biofilm formation . Impaired respiration led to increased cell lysis via divergent regulation of two processes: increased expression of the AtlA murein hydrolase and decreased expression of wall-teichoic acids . The AltA-dependent release of cytosolic DNA contributed to increased biofilm formation . Further , cell lysis and biofilm formation were governed by the SrrAB two-component regulatory system . Data presented support a model wherein SrrAB-dependent biofilm formation occurs in response to the accumulation of reduced menaquinone .
Staphylococcus aureus is a commensal bacterium that is estimated to colonize between 20–50% of the healthy human population ( Naimi et al . , 2003; Graham et al . , 2006; Enright et al . , 2002; Ohara-Nemoto et al . , 2008; Zafar et al . , 2007 ) . Colonization typically occurs in the nares , throat , or on the skin ( Ohara-Nemoto et al . , 2008; Zafar et al . , 2007; Hamdan-Partida et al . , 2010 ) . Under select conditions , S . aureus is capable of causing both invasive as well as non-invasive infections ( Klevens et al . , 2007; Tong et al . , 2015; Williamson et al . , 2013 ) . The dominant fraction of invasive infections caused by this bacterium occur in the context of bacteremia ( Klevens et al . , 2007 ) . In addition , S . aureus can infect and cause diseases of the lungs ( penumonia ) , skin ( cellulitis ) , skeletal tissues ( ostoemyelitis ) , and heart tissue ( endocarditis ) , as well as septic shock ( Klevens et al . , 2007; Tong et al . , 2015 ) . In the United States , pneumonia and septic shock are rapidly progressing infections and are often fatal with mortality rates in the United States ( US ) of 30–55% ( Klevens et al . , 2007 ) . While bacteremia and endocarditis infections have a lower degree of mortality , they are associated with a higher degree of recurrence , suggestive of therapeutic recalcitrance ( Klevens et al . , 2007 ) . A recent epidemiological analysis of ~8 , 700 cases of invasive S . aureus infections in the US found that nearly 92% cases required hospitalization ( Klevens et al . , 2007 ) . Historically , S . aureus infections in the US were largely nosocomial in origin; however , their onset or occurrence increasingly transpires in community settings ( Klevens et al . , 2007; Tenover et al . , 2006 ) . In the United States , pulsed-field type USA300 methicillin-resistant S . aureus ( MRSA ) has emerged as the dominant etiologic agent of community-associated invasive infections ( Klevens et al . , 2007 ) . Treatment of S . aureus infections is often problematic due to the increasing prevalence of antibiotic resistance . S . aureus strains have been isolated that are resistant to nearly all clinically available antibiotics , including the last-line antibiotics linezolid and daptomycin ( Sass et al . , 2012; Sánchez García et al . , 2010 ) . Biofilms are architecturally complex , multicellular communities of microorganisms of either mono- or poly-microbial compositions ( Costerton , 1995; Costerton et al . , 1995 ) . It has been theorized , based upon studies using direct techniques , such as microscopy , that ~99% of bacteria establish biofilms in their natural environments ( Costerton et al . , 1995 ) . A number of persistent and chronic infections in humans , such as periodontis and cystic fibrosis , are associated with the ability of the microorganisms to establish biofilms ( Sedghizadeh et al . , 2009; Costerton et al . , 1999 ) . In addition , biofilms of infectious agents are well characterized to form upon biomedical devices such as prosthetics , heart valves , catheters , and contact lenses ( Costerton et al . , 1999 , 2005; Bispo and Haas , 2015 ) . A number of staphylococcal infections , such as osteomyelitis , are also intimately connected to the ability of the bacterium to form biofilms ( Joo and Otto , 2012; Otto , 2008 ) . Reflective of their clinical significance , biofilms are considered to be the etiologic agents of recurrent staphylococcal infections ( Joo and Otto , 2012; Otto , 2008 ) . S . aureus biofilms are typically composed of one or more extracellular polymeric molecules ( DNA , proteins , or polysaccharides ) that provide structural integrity and may also facilitate intercellular adhesion ( Rice et al . , 2007; Schwartz et al . , 2012; Boles and Horswill , 2008; Cramton et al . , 1999 ) . The polymers interact to facilitate the formation an extracellular matrix . This matrix provides protection from environmental stress , innate immunity , as well as therapeutic agents ( Davies , 2003 ) . The polymer ( s ) utilized to facilitate biofilm formation can vary between staphylococcal isolates with some favoring DNA and/or proteins and others polysaccharides ( Rice et al . , 2007; Schwartz et al . , 2012; Boles and Horswill , 2008; Cramton et al . , 1999 ) . The complexity of biofilm formation results in this process being highly regulated and deterministic . Biofilm formation in S . aureus is responsive to diverse signals including nutrient limitation and quorum sensing ( Joo and Otto , 2012; Otto , 2008; Boles and Horswill , 2008; Majerczyk et al . , 2008 ) . Oxygen concentrations vary greatly between healthy human tissues ( between 19 . 7 to ~1 . 5%; normoxia ) ( Carreau et al . , 2011 ) . Oxygen concentrations also vary between healthy and infected or necrotic tissues , as well as in wounds , where concentrations are estimated to be below 1% ( hypoxic ) or anoxic ( Carreau et al . , 2011; Vogelberg and König , 1993; Arnold et al . , 1987 ) . A recent study found that S . aureus infections in skeletal tissues ( osteomyelitis ) cause an ~3 fold decrease in oxygen concentrations resulting in increasing hypoxia as infection proceeds ( Wilde et al . , 2015 ) . Multiple studies have focused upon the human systems that are active under hypoxia or anoxia and aid in combating bacterial infections . However , relatively little is known about how S . aureus mount a response to hypoxia or anoxia . A study by Cramton et al . found that decreased oxygen concentrations result in increased biofilm formation in S . aureus ( Cramton et al . , 2001 ) . An alternate study found that S . aureus growing in biofilms are starved for oxygen and that the rate of oxygen depletion is proportional to the rate of biofilm maturation ( Zhu et al . , 2007 ) . Cramton et al . also found that decreased oxygen concentrations lead to increased production of the polysaccharide intercellular adhesin ( PIA ) , which is a polymer used by some S . aureus isolates to facilitate intercellular adhesion ( Cramton et al . , 2001 ) . However , the role or requirement of PIA in low oxygen biofilms is unclear since biofilm formation in a PIA deficient strain was not examined ( Cramton et al . , 2001 ) . It is also unclear how the lack of oxygen , a cell permeable molecule , translates into increased biofilm formation . Two-component regulatory systems ( TCRS ) are modular signal transduction pathways that facilitate the integration of multiple stimuli into cellular signaling circuits , allowing for a rapid and robust response to environmental alterations ( Stock et al . , 2000; Stephenson and Hoch , 2002 ) . In S . aureus , which encodes for classical TCRS , the systems are predicted to be composed of a histidine kinase ( HK ) and a DNA-binding response regulator ( RR ) . The HK interacts with the environmental stimulus and can be either membrane associated or cytosolic . Upon stimulation , the HK alters the levels of the phosphoryl group upon the RR . In the case of most ( but not all ) DNA-binding RRs , altered phosphoryl levels modify the affinity of the RR for DNA resulting in altered gene transcription and a tailored physiological response ( Stock et al . , 2000; Stephenson and Hoch , 2002 ) . The goal of this study was to examine the mechanisms by which oxygen affects S . aureus biofilm formation . Data presented show that oxygen impacts biofilm formation in its capacity as a terminal electron acceptor in cellular respiration . Consequently , growth conditions that diminish respiration elicit increased biofilm formation . Impaired respiration leads to increased cell lysis via increased expression of the AltA murein hydrolase and a concomitant decrease in the expression of wall-teichoic acids . The regulatory tuning of these two processes in a divergent manner affects cell lysis . Increased biofilm formation and cell lysis is a programmed mechanism that is governed by the SrrAB TCRS . Genetic evidence suggests that SrrAB-dependent biofilm formation occurs in response to the accumulation of reduced menaquinone .
The influence of anaerobiosis upon biofilm formation of S . aureus was examined . Regulatory networks integral to staphylococcal physiology differ between S . aureus isolates ( Herbert et al . , 2010; Memmi et al . , 2012 ) . Biofilm formation was examined in diverse S . aureus isolates that vary in their ability to form biofilms ( LAC , SH1000 , MW2 , N315 ) . Strains were cultured aerobically , with a seal that allows free diffusion of gases , or anaerobically ( in a COY anaerobic chamber equipped with an oxygen scavenging catalyst , O2 <1 ppm ) prior to quantifying biofilms . Biofilm formation increased substantially for each strain during anaerobic growth ( between ~4–30 fold ) ( Figure 1A and B ) . Unless specifically mentioned , the experiments described henceforth were conducted using the community-associated MRSA strain LAC ( hereafter wild-type; WT ) . 10 . 7554/eLife . 23845 . 003Figure 1 . Oxygen impacts biofilm formation in its capacity as a terminal electron acceptor . Panels A and B; Anaerobic growth elicits increased biofilm formation in multiple S . aureus isolates . Biofilm formation of the LAC ( JMB1100; hereafter wild-type ( WT ) ) , SH1000 ( JMB 1323 ) , MW2 ( JMB1324 ) and N315 ( JMB 7570 ) isolates following aerobic or anaerobic growth is displayed . MRSA denotes methicillin resistance , MSSA denotes methicillin sensitivity , CC denotes clonal complex type and the USA number denotes the pulsed-field gel electrophoeresis type . Panel C; Supplementing growth media with the alternate terminal electron acceptor nitrate results in decreased biofilm formation during anaerobic growth . Biofilm formation for WT following aerobic or anaerobic growth and in media containing between 0–2 mM sodium nitrate is displayed . Panel D; A strain incapable of respiration upon oxygen forms increased biofilms when cultured aerobically , but not fermentatively . Biofilm formation for the WT and hemB::Tn ( JMB6037 ) strains following aerobic or anaerobic growth is displayed . Panel E; Nitrate supplementation does not decrease anaerobic biofilm formation in a nitrate reductase mutant . Biofilm formation for the WT and narG::Tn ( JMB7277 ) strains following anaerobic growth and in media containing between 0–2 mM sodium nitrate . Panel F; Chemical inhibition of respiration elicits increased biofilm formation during aerobic growth . Biofilm formation for the WT following aerobic growth in media supplemented with 0–250 µM sodium azide . The data represent the average values of eight wells ( Panels A , C-E ) or quadruplicates ( Panel F ) and error bars represent standard deviations . Representative photographs of biofilms formed upon the surface of a 96-well microtiter plate and stained with crystal violet are displayed in Panel B or insets in Panel C and D . Error bars are displayed for all data , but on occasion may be too small to see . Statistical significance was calculated using a two-tail Student's t-test and p-values>0 . 05 were considered to be not significant while * indicates p-value of <0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 23845 . 003 The principal influence of oxygen upon staphylococcal physiology is achieved in its capacity as a terminal electron acceptor ( TEA ) for respiration . Increased biofilm formation during anaerobic growth occurred upon culture in a medium lacking a terminal electron acceptor ( fermentative growth ) . We tested the hypothesis that impaired respiration is a signal that elicits biofilm formation . In addition to oxygen , S . aureus can utilize nitrate as a TEA . Anaerobic biofilm formation decreased , as the concentration of nitrate provided in the medium was increased ( Figure 1C ) . The addition of nitrate to aerobic cultures did not significantly alter biofilm formation ( Figure 1C ) . We reasoned that strains incapable of respiration would display increased biofilm formation . Heme auxotrophs have non-functional terminal oxidases and are unable to respire . They , form small colonies when cultured in the presence of oxygen , and therefore are termed small-colony variants ( Hammer et al . , 2013 ) . A hemB::Tn strain formed considerably more biofilm than the WT when cultured aerobically , but displayed biofilm formation similar to the WT when cultured fermentatively ( Figure 1D ) . Likewise , nitrate supplementation did not decrease anaerobic biofilm formation in a nitrate reductase ( narG::Tn ) mutant , which is unable to utilize nitrate as a TEA ( Figure 1E ) ( Schlag et al . , 2008; Burke and Lascelles , 1975 ) . To further test our premise , biofilm formation was examined in the WT cultured aerobically with varying amounts of the respiratory poison sodium azide . Biofilm formation increased in synchrony with the concentration of sodium azide in the growth medium ( Figure 1F ) . From Figure 1 we concluded that decreased cellular respiration results in increased biofilm formation . Further , biofilm formation was responsive to the concentration of a terminal electron acceptor or the ability of cells to respire . We sought to understand the mechanisms underlying the formation of fermentative biofilms . We examined the dependence of fermentative biofilms upon one or more of the described structural polymers: intercellular polysaccharide adhesin ( PIA ) , high-molecular weight extracellular DNA ( eDNA ) , or proteins ( Rice et al . , 2007; Schwartz et al . , 2012; Boles and Horswill , 2008; Cramton et al . , 1999; Foulston et al . , 2014 ) . The icaABCD operon encodes for proteins required to biosynthesize PIA ( Cramton et al . , 1999 ) . Strains lacking functional IcaA , IcaB , or IcaC were not attenuated in fermentative biofilm formation , suggesting that PIA is dispensable for this phenotype ( Figure 2—figure supplement 1 ) . However , supplementation of the growth medium with DNase , which degrades DNA , substantially attenuated biofilm formation suggesting that DNA is an integral component of fermentative biofilms ( Figure 2A ) . Consistent with this theory , the accumulation of high-molecular weight extracellular DNA ( eDNA ) increased appreciably in the matrix of fermenting biofilms ( Figure 2B ) . 10 . 7554/eLife . 23845 . 004Figure 2 . Impaired respiration results in AtlA-dependent release of high-molecular weight DNA , cytoplasmic proteins and an increase in biofilm formation . Panel A; Fermentative biofilm formation is attenuated upon supplementation of growth medium with DNase . Biofilm formation of the WT ( JMB 1100 ) following fermentative growth in media with or without 20 µg/mL DNase is displayed . Panel B; High-molecular weight DNA ( eDNA ) accumulation is increased in the biofilm matrix of fermenting cells . Biofilms of the WT were cultured aerobically or fermentatively , eDNA was extracted , and analyzed using agarose gel electrophoeresis ( inset photograph ) . The data were normalized to the viable cell count , and thereafter , to eDNA accumulation in fermenting WT . Panel C; Fermentative biofilm formation is dependent upon the AtlA murein hydrolase . Biofilm formation for the WT and the atlA::Tn ( JMB 6625 ) strains cultured aerobically or fermentatively is displayed . Panel D; eDNA accumulation in fermenting biofilms is dependent upon AtlA . Biofilms of the WT and atlA::Tn strains were cultured fermentatively and eDNA accumulation assessed . The data were normalized to the viable cell count , and thereafter , to eDNA accumulation in WT . Panel E; Fermentative biofilm formation is attenuated upon supplementation of growth medium with Proteinase K . Biofilm formation for the WT following fermentative growth in media with or without 10 µg/mL Proteinase K is displayed . Panel F; Fermentative growth results in AtlA-dependent release of a cytosolic protein into the extracellular milleu . Biofilms of the WT and atlA::Tn strains were cultured fermentatively and the activity of the cytosolic protein catalase ( Kat ) was measured in the spent media supernatant . The data were normalized to intracellular Kat activity , and thereafter to WT levels . The data represent the average values of eight wells ( Panels A , C and E ) , sextuplets ( Panel B ) or triplicates ( Panels D and F ) and error bars represent standard deviations . Representative photographs of high-molecular weight eDNA are displayed in Panel B or inset in Panel D . Error bars are displayed for all data , but might be too small to see on occasion . Statistical significance was calculated using a two-tail Student's t-test and p-values>0 . 05 were considered to be not significant while * indicates p-value of <0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 23845 . 00410 . 7554/eLife . 23845 . 005Figure 2—figure supplement 1 . Polysaccharide intercellular adhesin ( PIA ) is dispensable for fermentative biofilm formation . Biofilm formation of the WT ( JMB 1100 ) , icaA::Tn ( JMB 5577 ) , icaB::Tn ( JMB 5579 ) , and icaC::Tn ( JMB 5578 ) strains following fermentative growth is displayed . Data represent the average value of eight wells and error bars represent standard deviation . Statistical significance was calculated using a two-tail Student's t-test and p-values>0 . 05 were considered to be not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 23845 . 00510 . 7554/eLife . 23845 . 006Figure 2—figure supplement 2 . Supplementing growth media with the autolysis inhibitor polyanethole sulfonate ( PAS ) attenuates fermentative biofilm formation . Biofilm formation of the WT ( JMB 1100 ) cultured fermentatively in the presence of vehicle or 300 µg/mL PAS is displayed . Data represent the average value of eight wells and error bars represent standard deviation . Statistical significance was calculated using a two-tail Student's t-test and * indicates p-value of <0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 23845 . 00610 . 7554/eLife . 23845 . 007Figure 2—figure supplement 3 . Fermentative biofilm formation is dependent on the AtlA murein hydrolase . Biofilm formation of the WT ( JMB 2977 ) , atlA::Tn ( JMB 6625 ) , lytN::Tn ( JMB 7265 ) , sle1::Tn ( JMB 7266 ) , lytZ::Tn ( JMB 7269 ) , lytM::Tn ( JMB 7271 ) , lytY::Tn ( JMB 7268 ) , and hmrA::Tn ( JMB 7270 ) strains cultured fermentatively is displayed . Data represent the average value of eight wells and error bars represent standard deviation . Statistical significance was calculated using a two-tail Student's t-test and * indicates p-value of <0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 23845 . 00710 . 7554/eLife . 23845 . 008Figure 2—figure supplement 4 . Cytosolic protein release is increased upon fermentative growth . Biofilms of the WT ( JMB 1100 ) were cultured aerobically or fermentatively and the activity of the cytosolic protein catalase ( Kat ) was measured in the spent media supernatant . Extracellular Kat activity was normalized to intracellular Kat activity and thereafter to levels under aerobic growth . Statistical significance was calculated using a two-tail Student's t-test and * indicates p-value of <0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 23845 . 008 Prevailing models suggest that eDNA in staphylococcal biofilms arises as a consequence of a self digestive cell-lysis process ( autolysis ) , which results in the release of high-molecular weight genomic DNA ( Rice et al . , 2007; Foulston et al . , 2014 ) . Polyanethole sulfonate ( PAS ) inhibits S . aureus autolysis ( Wecke et al . , 1986; Yabu and Kaneda , 1995 ) . Supplementing growth media with PAS diminished fermentative biofilm formation ( Figure 2—figure supplement 2 ) . Peptidoglycan ( murein ) cleavage would be necessary for autolysis . The S . aureus genome encodes for multiple murein hydrolases ( Navarre et al . , 1999; Frankel et al . , 2011 ) . Fermentative biofilm formation was examined in a set of strains that each lacked one predicted murein hydrolase . One strain , with a disruption in the gene encoding for the AtlA murein hydrolase ( atlA::Tn ) , was attenuated in biofilm formation ( Figure 2—figure supplement 3 ) . AtlA has been previously implicated to be required for biofilm formation during aerobic growth ( Bose et al . , 2012; Houston et al . , 2011; Biswas et al . , 2006 ) . The atlA::Tn strain displayed decreased biofilm formation in the presence of oxygen ( ~1 fold decrease ) and this phenotype was exacerbated ( ~10 fold decrease ) in fermenting cultures ( Figure 2C ) suggesting that the role of AtlA in biofilm formation is increased during fermentative growth . Moreover , eDNA accumulation was greatly decreased in the biofilm matrix of the fermentatively cultured atlA::Tn strain ( Figure 2D ) . A recent study found that cytosolic proteins form a significant portion of staphylococcal biofilm matrixes ( Foulston et al . , 2014 ) . AtlA has been implicated in the release of cytosolic proteins into the extracellular milleu ( Pasztor et al . , 2010 ) . The supplementation of media with proteinase K , which degrades proteins , attenuated fermentative biofilm formation , suggesting that in addition to eDNA , proteins also form an integral part of the biofilm matrix in fermenting cells ( Figure 2E ) . To further examine this , the activity of catalase ( Kat ) ( Cosgrove et al . , 2007; Mashruwala et al . , 2016a ) , an abundant intracellular protein ( Cosgrove et al . , 2007 ) , was measured in the spent media supernatants . The spent media supernatant from fermenting WT had ~5 fold increased Kat activity relative to aerobically cultured WT ( Figure 2—figure supplement 4 ) . Kat activity was decreased by ~5 fold in the spent media supernatant from the fermentatively cultured altA::Tn strain ( Figure 2F ) . These data were normalized to intracellular Kat activity to negate for potential changes in Kat expression . From Figure 2 and Figure 2—figure supplements 1–4 we concluded that fermenting cells release an increased quantity of DNA and cytoplasmic proteins , into their extracellular mileu , in an AtlA-dependent manner . The eDNA and proteins are incorporated into the biofilm matrix and contribute to biofilm formation . Three scenarios could underlie the increased role of AtlA in fermentative biofilm formation . First , the expression of AtlA is increased leading to increased autolysis . Second , cell walls are altered in order to make them more amenable to AtlA-dependent lysis . Third , a combination of scenarios one and two . To discern which of these scenarios is operative in fermenting cells , the abundance of the atlA transcript was assessed in WT cultured aerobically or fermentatively . The atlA transcript was increased ~5 fold upon fermentative culture ( Figure 3A ) . Subsequently , AtlA activity was examined within the context of intact whole cells using autolysis assays ( Bose et al . , 2012 ) . Fermentatively cultured WT cells underwent autolysis faster than those cultured aerobically . The atlA::Tn strain , cultured aerobically or fermentatively , was severely deficient in undergoing autolysis suggesting that AtlA was the dominant murein hydrolase contributing to autolysis under the growth conditions examined ( Figure 3B ) . 10 . 7554/eLife . 23845 . 009Figure 3 . Impaired respiration elicits increased expression of AtlA and alterations that make cells more amenable to cleavage by AtlA . Panel A; The atlA transcript is increased upon fermentative growth . Biofilms of the WT ( JMB 1100 ) were cultured aerobically or fermentatively , mRNA was extracted , and the abundance of the atlA transcript was quantified . The data were normalized to 16S rRNA levels , and thereafter , to levels observed aerobically . Panel B; Fermenting cells undergo increased autolysis in an AtlA-dependent manner . The WT and atlA::Tn ( JMB 6625 ) strains were cultured aerobically or fermentatively and autolysis was examined in intact whole cells . Panel C; AtlA-dependent bacteriolytic activity is increased in fermenting cells . Murein-hydrolase activity in cell-wall associated proteins ( CW-extracts ) detached from the WT or atlA::Tn strains cultured aerobically or fermentatively is displayed ( pH of 7 . 5 ) . Heat-killed Micrococcus luteus was used as a substrate . Panel D-G; Fermenting cells are more amenable to AtlA and N-acetylmuramyl-L-alanine amidase ( AM ) -dependent cleavage . Murein-hydrolase activity using CW-extracts detached from a ΔatlA strain ( KB 5000 ) carrying plasmids encoding for empty vector control ( Panel D ) , GL only ( patlAGL ) ( Panel E ) , full-length AtlA ( patlA ) ( Panel F ) , or AM only ( patlAAM ) ( Panel G ) upon heat-killed cells of the WT cultured aerobically or fermentatively or M . luteus as substrates is displayed ( pH of 7 . 5 ) . The data in Panel A represent the average values of triplicates . Statistical significance was calculated using a two-tail Student's t-test and * indicates p-value of <0 . 05 . The data in Panels B-G represent the average value of technical duplicates from one set of substrate preparation , autolysis experiments , or CW extract preparations . Autolysis experiments or the preparation of heat-killed substrates or CW-extracts were conducted on least three separate occasions and similar results were obtained . Error bars in all panels represent standard deviations . Error bars are displayed for all data , but might be too small to see on occasion . DOI: http://dx . doi . org/10 . 7554/eLife . 23845 . 00910 . 7554/eLife . 23845 . 010Figure 3—figure supplement 1 . Representation of the full-length AtlA precursor protein and of the plasmid encoded variants used in this study . S . aureus AtlA is a bifunctional protein with an amidase ( AM ) and a glucosaminidase ( GM ) domain . The schematic is a modification of previous illustrations ( Bose et al . , 2012; Götz et al . , 2014 ) . AtlA is post-translationally processed ( indicated by arrows ) between the propeptide and AM domain and between the repeat domains to free AM-R1-2 and R3-GM . The four allelic variants used in this study were constructed previously ( Bose et al . , 2012 ) . The alleles are carried upon multi-copy plasmids that encode for full length AtlA ( patlA ) , the amidase and repeat domains ( AM-R1-R2 ) ( patlAAM ) , catalytically inactivated amidase and repeat domains via a point mutation H263A ( AMH263A-R1-R2 ) ( patlAAMH263A ) ( not displayed ) , or glucosaminidase ( R3-GM ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23845 . 010 Murein hydrolase assays were used to quantify AtlA-dependent bacteriolytic activity . The WT and atlA::Tn strains were cultured aerobically or fermentatively , cell-wall associated proteins were detached ( hereafter CW-extracts ) , and bacteriolytic activity was examined using heat-killed Micrococcus luteus as a substrate . CW-extracts from fermenting WT lysed M . luteus more rapidly than CW-extracts from WT cultured aerobically ( Figure 3C ) . Bacteriolytic activity was nearly undetectable when using CW-extracts from the atlA::Tn strain cultured aerobically or fermentatively . These data confirmed that AtlA was the dominant murein hydrolase in the extracts and increased AtlA activity was associated with the WT cultured fermentatively ( Figure 3C ) . We next examined whether cell walls were altered in order to make them more amenable to AtlA . The WT strain was cultured aerobically or fermentatively , heat-killed to inactivate native autolysins , and the cells were subsequently provided as substrates in murein hydrolase assays . AtlA is a bifunctional enzyme that is proteolytically cleaved into a N-acetylmuramyl-L-alanine amidase ( AM ) and endo-β-N-acetylglucosaminidase ( GL ) ( ( Oshida et al . , 1995 ) and illustrated in Figure 3—figure supplement 1 ) . The use of M . luteus and S . aureus cells as substrates allows for differentiation between AM and GL activities ( Oshida et al . , 1995; Wadström and Hisatsune , 1970 ) . GL displays poor activity against S . aureus , but is capable of cleaving M . luteus . Murein hydrolase assays were conducted using CW-extracts obtained from a ΔatlA strain carrying empty vector or plasmids encoding for full length AtlA ( patlA ) , AM only ( patlAAM ) and GL only ( patlAGL ) ( Bose et al . , 2012 ) . Lysis of heat-killed S . aureus , as well as M . luteus , was undetectable with CW-extracts from the ΔatlA strain carrying empty vector verifying that bacteriolytic activity under the conditions examined was dependent upon AtlA , AM , or GL ( Figure 3D ) . CW-extracts from the ΔatlA strain carrying patlAGL did not lyse S . aureus , but proficiently lysed M . luteus , confirming that S . aureus are poor substrates for GL ( Figure 3E ) . CW-extracts from the ΔatlA strain carrying patlA or patlAAM lysed fermentatively cultured heat-killed WT at a faster rate than aerobically cultured WT , suggesting fermenting S . aureus cells are more amenable to cleavage by AtlA and AM ( Figure 3F–G ) . Wall-teichoic acids ( WTA ) are cell-surface glycopolymers that are covalently attached to peptidoglycan . The biosynthetic pathway for WTA in S . aureus is illustrated in Figure 4A . WTA negatively modulate AtlA activity ( Biswas et al . , 2012; Schlag et al . , 2010 ) . Decreased expression of WTA during fermentative growth could result in cells that are more amenable to AtlA-dependent lysis . Consistent with this logic , the transcription of genes encoding for proteins in the WTA biosynthetic pathway ( tarA , tarO , tarB , tarH ) was decreased during fermentative growth ( between 6–50 fold ) ( Figure 4B ) . 10 . 7554/eLife . 23845 . 011Figure 4 . Decreased expression of wall-teichoic acids during fermentative growth makes S . aureus more amenable to cleavage by AtlA . Panel A; Schematic of wall-teichoic acid ( WTA ) biosynthesis in S . aureus . The diagram displays select proteins involved in WTA biosynthesis and is redrawn as initially presented by Campbell et al . ( 2012 ) . The initial transformations in the pathway catalyzed by TarO and TarA are non-essential , while the latter steps are essential . Tunicamycin inhibits TarO , as well as the 2-epimerase MnaA , which modulates the substrate levels for TarO ( Campbell et al . , 2011; Mann et al . , 2016 ) . MnaA is not displayed . Panel B; Transcript levels corresponding to genes encoding for WTA biosynthesis proteins are decreased upon fermentative growth . Biofilms of the WT ( JMB 1100 ) were cultured aerobically or fermentatively , mRNA was extracted , and the abundances of the tarO , tarA , tarB , and tarH transcripts were quantified . The data were normalized to 16S rRNA levels , and thereafter to levels observed aerobically . Panel C; AtlA-dependent cleavage of heat-killed cells at a decreased pH is modulated via wall-teichoic acids . Murein-hydrolase activity at pH of 5 for cell-wall associated proteins ( CW-extracts ) detached from a ΔatlA strain ( KB 5000 ) carrying patlA and incubated with heat-killed cells of the WT cultured aerobically or fermentatively in the presence or absence of 100 ng/mL tunicamycin as substrates is displayed . Panel D; AtlA-dependent autolysis of intact whole cells at decreased pH is modulated via wall-teichoic acids . The WT and atlA::Tn ( JMB 6625 ) strains were cultured aerobically or fermentatively in the presence or absence of 100 ng/mL tunicamycin . Autolysis was examined in intact cells resuspended in a buffer with pH of 5 . Panels E; Heat-killed aerobic or fermenting WT bind similar amounts of AtlA . CW-extract detached from a ΔatlA strain ( KB 5000 ) carrying patlA was incubated at pH of 5 with heat-killed WT , cultured aerobically or fermentatively in the presence or absence of 100 ng/mL tunicamycin , or in the absence of cells ( control ) for 8 min . The cells were separated by centrifugation and bacteriolytic activity in the resultant supernatant was assessed upon heat-killed M . luteus as a substrate is displayed . Data in Panel B represents the average value of triplicates . Statistical significance was calculated using a two-tail Student's t-test and * indicates p-value of <0 . 05 . Data in Panels C-E represent the average value of technical duplicates from one set of substrate preparation or autolysis assays . The heat-killed substrates were prepared or autolysis assays were conducted on least three separate occasions and similar results were obtained . Error bars in all panels represent standard deviations . Error bars are displayed for all data , but might be too small to see on occasion . DOI: http://dx . doi . org/10 . 7554/eLife . 23845 . 01110 . 7554/eLife . 23845 . 012Figure 4—figure supplement 1 . AtlA- and AM-dependent cleavage of heat-killed cells is modulated via altered expression of wall-teichoic acids . Murein hydrolase assays were conducted using cell-wall associated proteins ( CW-extracts ) detached from a ΔatlA ( KB 5000 ) strain carrying patlAAM ( Panel A ) or patlA ( Panel B ) and upon heat-killed cells of the WT ( JMB 1100 ) cultured aerobically or fermentatively and in the presence or absence of 100 ng/mL tunicamycin as substrates ( pH 7 . 5 ) . Data represent the average value of technical duplicates from one set of substrate preparation . The heat-killed substrates were prepared on least three separate occasions and similar results were obtained . Error bars represent standard deviations . DOI: http://dx . doi . org/10 . 7554/eLife . 23845 . 01210 . 7554/eLife . 23845 . 013Figure 4—figure supplement 2 . AtlA-dependent lysis rates of heat-killed tunicamycin treated cells are not altered upon alterations in the assay buffer pH . Murein hydrolase assays were conducted at pH 7 , pH 6 , or pH 5 using cell-wall associated proteins ( CW-extracts ) detached from a ΔatlA ( KB 5000 ) strain carrying patlA upon heat-killed cells of the WT cultured aerobically in the presence of 100 ng/mL tunicamycin as substrates . Data represent the average value of technical duplicates from one set of substrate preparation . The heat-killed substrates were prepared on least three separate occasions and similar results were obtained . Error bars represent standard deviations . DOI: http://dx . doi . org/10 . 7554/eLife . 23845 . 013 Tunicamycin is an inhibitor of TarO and MnaA , which are necessary for WTA biosynthesis ( Figure 4A ) ( Campbell et al . , 2011; Mann et al . , 2016; Hancock et al . , 1976 ) . S . aureus cultured in the presence of tunicamycin do not synthesize WTA ( Campbell et al . , 2011 ) . WT was cultured aerobically or fermentatively in the presence or absence of tunicamycin , the cells were heat-killed , and used as substrates in murein hydrolase assays . WT cells cultured aerobically with tunicamycin were lysed at a rate similar to that of fermentatively cultured cells by CW-extracts from a ΔatlA strain carrying either patlAAM or patlA . This confirmed that changes in WTA expression alter the amenability of fermenting cells to cleavage by AtlA and AM ( Figure 4—figure supplement 1A and B ) . Two models have been proposed to explain the influence of WTA upon AtlA activity . Schlag et al . found that the presence of WTA interferes with the binding of AtlA to the cell surface ( Schlag et al . , 2010 ) . Biswas et al . found that WTA contributes to proton binding on the cell surface . AtlA activity decreases substantially below pH 6 . 5 ( Biswas et al . , 2012 ) , and therefore , it was proposed that binding of protons by WTA leads to a decrease in the local pH of the cell surface thereby inhibiting AtlA activity ( Biswas et al . , 2012 ) . We examined the contribution of these two mechanisms in the lysis of fermenting S . aureus . First , the effect of proton binding by WTA upon AtlA activity was examined by decreasing the pH of the murein hydrolase and autolysis assays . We reasoned that an increased concentration of protons would exacerbate the effect of proton binding by WTA . Under this scenario , cells containing an increased abundance of WTA would be expected to be resistant towards AtlA-dependent cleavage at decreased pH . Consistent with this premise , AtlA-dependent lysis of heat-killed WT was dramatically decreased in murein hydrolase assays conducted at a pH of 5 ( Figure 4C and Figure 4—figure supplement 1A and B ) . Importantly , lysis of fermenting WT cells was still observed while it was nearly absent for those cultured aerobically ( Figure 4C ) . In contrast , lysis rates for tunicamycin treated WT were unaltered upon decreasing the pH ( Figure 4C , Figure 4—figure supplements 1 and 2 ) , confirming that the influence of pH upon AtlA activity was observed entirely as a result of alterations in WTA expression . The results from autolysis assays conducted at pH 5 lent further support to the findings of the murein hydrolase assays ( Figure 4D ) . Strikingly , autolysis was abrogated in aerobically cultured WT , while fermentatively cultured cells or those cultured in the presence of tunicamycin underwent proficient AtlA-dependent autolysis ( Figure 4D ) . Second , we examined whether fermenting WT bind an increased amount of AtlA and whether this is dependent upon WTA expression ( Fournier and Hooper , 2000 ) . Various heat-killed cells were incubated at pH 5 with CW-extract from a ΔatlA strain carrying patlA . The cells were subsequently removed , and the bacteriolytic activity remaining in the supernatants was quantified using heat-killed M . luteus cells as substrate . Aerobically or fermentatively cultured heat-killed WT cells did not bind bacteriolytic enzymes while tunicamycin treated cells bound a majority of the bacteriolytic enzymes ( Figure 4E ) . We concluded that the complete loss of WTA expression does indeed increase binding of AtlA to the cell surface confirming and extending the findings of Schlag et al . ( 2010 ) . However , altered AtlA binding to WTA was unlikely to underlie the increased lysis of fermenting cells . From Figures 3 , 4 , and Figure 4—figure supplements 1 and 2 , we concluded that fermenting S . aureus had increased expression of AtlA and concomitantly decreased expression of wall-teichoic acids . The combination of these two divergent responses facilitates increased autolysis . Since the changes in expression were accompanied by similar changes in transcription we concluded that impaired respiration elicits programmed cell lysis ( PCL ) . Respiration is predominantly mediated by membrane-associated factors . Regulatory system ( s ) that perceive respiratory status were likely to be membrane-associated . S . aureus encodes for 16 two-component regulatory systems ( TCRS ) . Of these , 14 are predicted to employ a membrane-associated histidine kinase . Fermentative biofilm formation was examined in strains that each lacked one individual TCR system ( except WalKR , which is essential ) ( Dubrac and Msadek , 2004; Pang et al . , 2014 ) . A strain lacking the staphylococcal respiratory regulatory system ( SrrAB ) was attenuated in fermentative biofilm formation ( Figure 5A ) . Reintroduction of srrAB into the ΔsrrAB strain upon an episome restored fermentative biofilm formation ( Figure 5A ) . Consistent with SrrAB mediated changes in biofilm formation occurring as a result of altered respiratory status , the introduction of a ΔsrrAB mutation into a hemB::Tn strain attenuated the increased biofilm formation of the hemB::Tn strain during aerobic growth ( Figure 5B ) . Unlike the WT , anaerobic biofilms formed by the ΔsrrAB strain were largely unaltered when the growth medium was supplemented with nitrate ( Figure 5—figure supplement 1 ) . 10 . 7554/eLife . 23845 . 014Figure 5 . Programmed cell lysis and biofilm formation in fermenting cells are governed by the SrrAB two-component regulatory system . Panel A; Fermentative biofilm formation is dependent upon SrrAB . Biofilm formation is displayed following aerobic or fermentative growth in the WT ( JMB 1100 ) carrying pLL39 ( pEV ) or the ΔsrrAB ( JMB 1467 ) strains carrying either pLL39 ( pEV ) or pLL39_srrAB ( psrrAB ) . Panel B; A hemB mutant forms SrrAB-dependent biofilms aerobically . Biofilm formation following aerobic growth is displayed for the WT , ΔsrrAB , hemB::Tn ( JMB 6037 ) , and ΔsrrAB hemB::Tn ( JMB 6039 ) strains . Panel C; Transcript levels corresponding to genes involved in programmed cell lysis and biofilm formation are altered in a ΔsrrAB strain . Biofilms of the WT and ΔsrrAB strains were cultured fermentatively , mRNA was extracted , and the abundances of the atlA , tarO , tarA , tarB , and tarH transcripts were quantified . Data were normalized to 16S rRNA levels , and thereafter , to levels observed in the WT . Panel D; The fermentative biofilm formation phenotypes associated with the ΔsrrAB and atlA::Tn mutations are not additive . Biofilm formation is displayed following fermentative growth for the WT , ΔsrrAB , atlA::Tn ( JMB 6625 ) , and ΔsrrAB atlA::Tn ( JMB 6624 ) strains . Panel E; Autolysis of fermenting S . aureus is decreased in a strain lacking SrrAB . The WT , ΔsrrAB , and atlA::Tn strains were cultured fermentatively and autolysis was examined ( pH of 5 ) . Panel F; eDNA accumulation is decreased in a strain lacking SrrAB . Biofilms of the WT , ΔsrrAB , and atlA::Tn strains were cultured fermentatively and eDNA was quantified . The data were normalized to the viable cell count and thereafter to the levels in the WT . Panel G; atlA in multicopy partially suppresses the biofilm formation defect of the ΔsrrAB strain . Fermentative biofilm formation is displayed for the WT and ΔsrrAB strains carrying either patlAAM H263A or patlA . Panel H; Heat-killed cells of a ΔsrrAB strain are less amenable towards AtlA-dependent lysis . Murein-hydrolase activity for cell-wall associated proteins ( CW-extracts ) detached from a ΔatlA strain ( KB 5000 ) carrying patlA and combined with fermentatively cultured and heat-killed WT or ΔsrrAB strains as substrates are displayed . Data presented represent the average value of eight wells ( Panels A , B , D-G ) or biological triplicates ( Panel C and F ) . Data in Panels E and H represent the average value of technical duplicates from one set of autolysis assays or substrate preparations . The heat-killed substrates were prepared or autolysis assays were conducted on least three separate occasions and similar results were obtained . Error bars in all panels represent standard deviations . Error bars are displayed for all data , but might be too small to see on occasion . Statistical significance was calculated using a two-tail Student's t-test and p-values>0 . 05 were considered to be not significant while * indicates p-value of <0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 23845 . 01410 . 7554/eLife . 23845 . 015Figure 5—figure supplement 1 . Biofilm formation of a ΔsrrAB strain is largely unaltered upon supplementing anaerobic biofilms with the alternate terminal electron acceptor nitrate . Biofilm formation following anaerobic growth in the presence or absence of varying concentrations of sodium nitrate is displayed for the WT ( JMB1100 ) and ΔsrrAB ( JMB1467 ) strains . Data represent the average value of eight wells and error bars represent standard deviation . Statistical significance was calculated using a two-tail Student's t-test and p-values>0 . 05 were considered to be not significant while * indicates p-value of <0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 23845 . 015 The influence of SrrAB upon the transcription of genes encoding for factors involved in PCL and biofilm formation was examined . The abundance of the atlA transcript was decreased ( ~5 fold ) in the ΔsrrAB strain ( Figure 5C ) . In contrast , the abundances of transcripts corresponding to genes required for WTA biosynthesis were increased in the ΔsrrAB strain ( ~2 . 5–5 fold ) . A strain lacking SrrAB displayed phenotypes consistent with decreased expression of AtlA . The fermentative biofilm formation phenotype of the ΔsrrAB atlA::Tn strain was similar to that of the atlA::Tn strain , suggesting that SrrAB influences biofilm formation , in part , via AtlA ( Figure 5D ) . Moreover , the ΔsrrAB strain was deficient in autolysis ( Figure 5E ) and had decreased accumulation of eDNA in its biofilm matrix when cultured fermentatively ( Figure 5F ) . To further examine the influence of AtlA upon SrrAB-dependent biofilm formation we introduced multicopy plasmids with alleles encoding for either full length AtlA ( patlA ) or an enzymatically inactivated AM ( patlAAM H263A ) into the ΔsrrAB strain and examined biofilm formation . The presence of patlA partially suppressed the fermentative biofilm formation defect of the ΔsrrAB strain when compared to the strain carrying patlAAM H263A ( Figure 5G ) . Additionally , fermentatively cultured , heat-killed , ΔsrrAB cells were lysed at a slower rate by CW-extracts from the ΔatlA strain carrying patlA , consistent with increased expression of WTA in the ΔsrrAB strain ( Figure 5H ) . The cellular molecule ( s ) that influence SrrAB activity are unidentified . S . aureus synthesizes menaquinone and strains lacking menaquinone are unable to respire ( Wakeman et al . , 2012 ) . Upon analyzing previous studies we observed that the transcription of genes positively regulated by SrrAB were reduced in a menaquinone auxotroph ( Kohler et al . , 2008; Kinkel et al . , 2013; Yarwood et al . , 2001; Pragman et al . , 2004 ) . A hemB mutant is also unable to respire ( Hammer et al . , 2013 ) and data presented in Figure 5B suggest that SrrAB activity , with respect to biofilm formation , is stimulated in a hemB::Tn strain . These seemingly conflicting pieces of information could be readily explained if menaquinone is necessary for SrrAB stimulation . We reasoned that if SrrAB activity is diminished in the absence of menaquinone then a hemB::Tn menF::Tn strain should phenocopy a ΔsrrAB hemB::Tn strain for biofilm formation . Biofilm formation was examined during aerobic growth in a hemB::Tn menF::Tn double mutant , a ΔsrrAB hemB::Tn menF::Tn triple mutant , as well as their parental strains . The hemB::Tn strain displayed increased biofilm formation relative to the menF::Tn strain ( Figure 6A ) . Importantly , the ΔsrrAB hemB::Tn , hemB::Tn menF:Tn , and ΔsrrAB hemB::Tn menF::Tn strains phenocopied the biofilm formation of the menF::Tn strain ( Figure 6A ) . These data confirmed that the presence of menaquinone is necessary for SrrAB-dependent biofilm formation in a hemB::Tn strain . 10 . 7554/eLife . 23845 . 016Figure 6 . SrrAB-dependent biofilm formation is responsive to the oxidation state of the cellular menaquinone pool . Panel A; SrrAB-dependent biofilm formation is inactivated in strains lacking the ability to synthesize menaquinone . Biofilm formation following aerobic growth is displayed for the menF::Tn ( JMB6219 ) , hemB::Tn ( JMB6037 ) , ΔsrrAB menF::Tn ( JMB6221 ) , ΔsrrAB hemB::Tn ( JMB6039 ) , hemB::Tn menF::Tn ( JMB6217 ) , and ΔsrrAB hemB::Tn menF::Tn ( JMB6673 ) strains . Panel B; SrrAB-dependent biofilm formation is not stimulated in strains enriched for oxidized menaquinone . Biofilm formation following aerobic growth is displayed for the WT ( JMB 1100 ) , ΔsrrAB ( JMB 1467 ) , ΔndhC ndhF::Tn sdh:Tn ( JMB 6613 ) , and ΔsrrAB ΔndhC ndhF::Tn sdh:Tn ( JMB 6614 ) strains . Data in both panels represent the average value of eight wells and the errors bars represent standard deviation . Statistical significance was calculated using a two-tail Student's t-test and p-values>0 . 05 were considered to be not significant while * indicates p-value of <0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 23845 . 016 Menaquinone functions as both an electron acceptor and an electron donor in the electron transfer chain ( ETC ) ( Kohler et al . , 2008 ) . Inactivation of heme biosynthesis results in defective terminal oxidases ( Proctor et al . , 2006 ) and the accumulation of reduced menaquinone . We examined whether a strain enriched for oxidized menaquinone also displayed an increase in the formation of SrrAB-dependent biofilms . S . aureus encodes for two NADH::menaquinone oxidoreductases ( NdhC and NdhF ) and one succinate dehydrogenase ( Sdh ) ( Schurig-Briccio et al . , 2014; Gaupp et al . , 2010 ) . A ΔndhC ndhF::Tn sdh:Tn strain is deficient in the passage of electrons to menaquinone and consequently enriched in oxidized menaquinone . The ΔndhC ndhF::Tn sdh:Tn strain displayed a negligible increase in aerobic biofilm formation ( ~1 . 4 fold increase ) , which was phenocopied by the ΔsrrAB ΔndhC ndhF::Tn sdh:Tn strain ( Figure 6B ) . Taken together , the data in Figure 6 led us to infer that with respect to biofilm formation ( 1 ) menaquinone influences SrrAB activity , ( 2 ) the absence of menaquinone results in SrrAB being non-responsive , ( 3 ) SrrAB activity is increased upon enrichment of reduced menaquinone , and ( 4 ) SrrAB is non-responsive to the enrichment of oxidized menaquinone .
Biofilms are the etiologic agents of recurrent staphylococcal infections . Previous work found that hypoxic growth results in increased biofilm formation of S . aureus . However , the molecular and regulatory mechanism ( s ) translating the lack of oxygen into biofilm formation were unknown . We report that oxygen impacts biofilm formation in its capacity as a terminal electron acceptor ( TEA ) for cellular respiration . Consistent with this premise , supplementing the growth medium with the alternate TEA nitrate decreased biofilm formation during anaerobic growth . Moreover , genetic or chemical inhibition of respiratory processes resulted in increased biofilm formation even in the presence of a TEA . TEA availability in the natural microenvironments of S . aureus varies , leading to the supposition that biofilm formation would be responsive to the concentration of TEA . Consistent with this logic , biofilm formation was titratable with respect to the concentration of a TEA or a molecule that inhibits respiration . Fermenting biofilms were dependent upon the presence of high-molecular weight DNA . High-molecular weight DNA in S . aureus biofilm matrixes ( eDNA ) has been shown to originate from genomic DNA , and thus , its presence suggested that fermenting cells undergo increased autolysis ( Rice et al . , 2007 ) . Lending support to this concept , fermentative biofilm formation was attenuated upon chemical inhibition of autolysis or genetic inactivation of the AtlA murein hydrolase . Fermenting cells underwent increased autolysis in a AtlA-dependent manner and the matrix from the atlA::Tn strain had nearly undetectable levels of eDNA . S . aureus biofilms incorporate cytosolic proteins into their matrixes and AtlA has been implicated in the release of cytosolic proteins via a process that is not completely understood ( Foulston et al . , 2014; Pasztor et al . , 2010 ) . We found that fermenting cells had increased activity for a cytosolic protein in the extracellular mileu and an atlA::Tn strain was deficient in the release of this protein . Fermenting biofilms were also readily disrupted upon supplementing media with proteinase K suggesting that , in addition to eDNA , proteins are integral components of the fermentative biofilm matrix . The increased role of AtlA in fermenting biofilms was due to a combination of two divergent cellular responses . First , fermenting cells increased the transcription of atlA and autolysis and murein hydrolase assays confirmed that this was translated into increased AtlA activity . Second , fermenting WT cells that had been heat-killed displayed an increased amenability to AtlA-dependent cleavage when used as substrates in murein hydrolase assays . These findings suggested that the cell surface was being altered to facilitate cell lysis . Wall-teichoic acids ( WTA ) are cell surface glycopolymers that are covalently attached to peptidoglycan and negatively impact AtlA activity ( Biswas et al . , 2012; Schlag et al . , 2010 ) . The transcription of WTA biosynthesis genes was decreased during fermentative growth . Autolysis and murein hydrolase assays , as well as the WTA synthesis inhibitor tunicamycin , confirmed that WTA expression was decreased during fermentative growth . Since two cellular processes are divergently modulated at the transcriptional level in response to an environmental stimulus ( TEA availability ) to affect autolysis , we propose that this process be termed as programmed cell lysis ( PCL ) , which is illustrated in our working model shown in Figure 7 . 10 . 7554/eLife . 23845 . 017Figure 7 . A working model for the influence of respiration upon autolysis and biofilm formation in S . aureus . A decreased capacity to respire results in an enrichment of reduced menaquinone effecting altered activity of the SrrAB two-component regulatory system . Altered SrrAB activity leads to increased transcription of atlA and decreased transcription of genes ( tar ) encoding for wall-teichoic acid ( WTA ) biosynthesis . The consequent decrease in WTA expression and increase in AtlA expression results in the release of DNA and proteins , cell lysis and biofilm formation . Since cell lysis is effected via regulatory tuning of two divergent processes we term this mechanism as programmed cell lysis ( PCL ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23845 . 017 The cell walls of gram-positive bacteria have been long recognized to serve as proton reservoirs ( Koch , 1986; Calamita et al . , 2001 ) . The walls of respiring cells have a low pH and calculations estimate that the local pH can decrease by 3–4 units ( Koch , 1986; Calamita et al . , 2001 ) . Further , energy-limiting conditions , such as fermentative growth , or proton trapping , influence bacterial autolysis ( Kemper et al . , 1993; Jolliffe et al . , 1981 ) . Thus , it has been clear that cell wall composition , the localized pH of the cell wall , and cellular autolysis are interconnected . However , the mechanisms underlying these interconnections have remained elusive . A recent study by Biswas et al . shed light on these processes in S . aureus ( Biswas et al . , 2012 ) . Biswas et al . found that WTA traps protons at the cell surface and they proposed that this results in decreased pH of the microenvironment , and thereby , inhibits AtlA activity ( Biswas et al . , 2012 ) . We found that the influence of pH upon AtlA activity , in both murein hydrolase , as well as autolysis assays , was almost entirely as a result of alterations in WTA expression . These findings both confirm and extend the model put forth by Biswas et al . ( 2012 ) . An alternate study by Schlag et al . proposed that WTA negatively affects AtlA activity by interfering with its binding to the cell surface ( Schlag et al . , 2010 ) . We found that at a pH of 5 , tunicamycin treated cells bound a majority of the bacteriolytic activity corresponding to AtlA . In contrast , binding was absent in cells not treated with tunicamycin , regardless of whether they were cultured aerobically or fermentatively . Thus , our findings also confirmed and extended the findings of Schlag et al . However , the complete absence of WTA synthesis is unlikely to be a phenomenon that would be physiologically encountered . Therefore , in fermenting S . aureus , where the final pH of the culture medium is ~5 , we propose that the model of Biswas et al . would dominate with respect to autolysis and biofilm formation . Acidic pH has long been recognized to elicit biofilm formation in S . aureus ( Regassa et al . , 1992 ) ; however , the mechanisms underlying this phenotype have been unclear . Foulston et al . found that cytoplasmic proteins released into the extracellular mileu associate with the exterior of cells , in a pH-dependent and reversible manner , facilitating matrix formation ( Foulston et al . , 2014 ) . The association of the proteins with the cells increases with decreasing pH ( Foulston et al . , 2014 ) . Foulston et al . conducted their study in a medium that leads to a decrease in pH over growth ( Foulston et al . , 2014 ) . Thus , it was unclear whether low pH was necessary for the release of the cytoplasmic proteins . Data presented herein suggest that low pH optimizes AtlA function and thereby effects the release of the cytoplasmic proteins , extending the findings of Foulston et al . Further , the physiological condition ( s ) under which this mechanism would be relevant was not entirely clear . In the present study we demonstrate that this mechanism is pertinent in the context of an environmental signal ( oxygen ) that is crucial in infection progression . Finally , we note that the pH of the skin and nares , which are sites colonized by S . aureus , is lower than the homeostatic 7 . 4 ( Weinrick et al . , 2004 ) . However , to our knowledge , it is unknown if low pH contributes to S . aureus biofilm formation in vivo . Respiration is a process mediated predominantly by membrane associated cellular factors . A strain lacking the SrrAB TCRS , consisting of a transmembrane histidine kinase ( SrrB ) ( Pragman et al . , 2004 ) , was attenuated in biofilm formation . A strain lacking SrrAB had decreased transcription of atlA , increased transcription of WTA biosynthesis genes , and displayed multiple phenotypes consistent with the transcriptional data . Further , the biofilm deficient phenotype of the ΔsrrAB strain was partially suppressed by the introduction of atlA in multicopy . These data suggest that SrrAB influences PCL and biofilm formation by divergently influencing AtlA and WTA expression . SrrAB output was previously shown to be altered under conditions of hypoxia and nitric oxide stress ( Kinkel et al . , 2013 ) . However , the cellular molecule ( s ) that influence SrrAB activity are unidentified . We found that SrrAB-dependent biofilms increased as a function of decreased respiratory activity . SrrAB-dependent biofilms were formed upon accumulation of reduced , but not oxidized menaquinone , and SrrAB output was abrogated in the absence of menaquinone . These findings suggest that ( 1 ) menaquinone is necessary for stimulus transmission to SrrAB , and ( 2 ) the oxidation state of the cellular menaquinone pool influences SrrAB output . We also considered the possibility of two alternate signals that could affect SrrAB output: culture pH and decreased proton motive force . Fermentative growth of S . aureus upon TSB results in the release of acidic by-products , which decrease the pH of the extracellular mileu ( Somerville et al . , 2003 ) . Diminished respiration also decreases the proton-motive force . However , heme and menaquinone auxotrophs are both deficient in respiration and the concentration of fermentative by-products and the pH in the spent media is similar in these strains ( ( Hammer et al . , 2013 ) and data not shown ) . These strains also display a similar decrease in membrane potential ( Hammer et al . , 2013 ) . Yet , only a heme auxotroph forms SrrAB-dependent biofilms . Thus , we deem it unlikely that pH or alterations in proton motive force alter SrrAB activity with respect to biofilm formation . It is worth noting the similarities that exist between the Escherichia coli ArcAB TCRS and SrrAB . Although these TCRS do not display significant homology , the stimuli influencing their activity are similar . ArcB is proposed to donate electrons from conserved cysteine residues to oxidized quinones resulting in silencing of kinase activity ( Malpica et al . , 2004 ) . Similar to ArcB , SrrB contains three conserved cysteine residues , which may facilitate redox interactions with the menaquinone pool . While this leads to the supposition that the molecular mechanism of SrrB signaling may be similar to ArcB , further biochemical analyses are required to make this conclusion . The Bacillus subtilis TCRS ResDE displays similarities to SrrAB and it also responds to changes in the respiratory status . However , unlike SrrB , ResE does not contain cysteine residues and studies have deemed it unlikely that the menaquinone pool influences ResDE activity ( Geng et al . , 2007 ) . Similar to S . aureus , B . subtilis increases biofilm formation under hypoxic growth and this phenotype is reversed upon supplementation with the alternate TEA nitrate ( Kolodkin-Gal et al . , 2013 ) . Biofilm formation in B . subtilis coincided with increased transcription of genes required for matrix production , which was mediated via the membrane-associated kinases KinA and KinB ( Kolodkin-Gal et al . , 2013 ) . B . subtilis ResD binds to the promoter regions or within the coding regions of lytF and cwlO , which encode for two major bacillus autolysins , suggesting it modulates the transcription of these genes ( Henares et al . , 2014; Ohnishi et al . , 1999; Ishikawa et al . , 1998; Yamaguchi et al . , 2004 ) . Further , the binding of ResD to these DNA regions was limited to fermentative growth ( Henares et al . , 2014 ) . However , to our knowledge , it is currently unknown whether ResDE has a role in respiration dependent biofilm formation . The gram-negative bacterium Pseudomonas aeruginosa also increases biofilm formation under hypoxic growth and this phenotype is also reversed upon supplementation with the alternate TEA nitrate ( Dietrich et al . , 2008 , 2013 ) . However , the regulatory mechanisms driving respiration dependent biofilm formation in P . aeruginosa are unknown . Thus , it seems likely that increased biofilm formation in response to TEA limitation is conserved among diverse bacteria . However , the genetic and regulatory bases underlying biofilm formation may differ . Clinical isolates of S . aureus that are incapable of respiration , termed as small colony variants ( SCV ) , display increased resistance towards antibiotics and cause persistent infections ( Proctor et al . , 2006; Melter and Radojevič , 2010 ) . The SCV phenotype often , but not always , arises as a result of mutations in genes necessary for heme biosynthesis resulting in non-functional terminal oxidases ( Hammer et al . , 2013; Proctor et al . , 2006 ) . Our finding that a heme auxotroph forms SrrAB-dependent biofilms lends considerable insight into the mechanisms that may predominate within clinical SCV strains . While we suggest the usage of the term PCL in the context of the mechanisms outlined herein , we note that this should not be confused with the process of programmed cell death ( PCD ) in bacteria or in eukaryotes ( Rice and Bayles , 2008; Kerr et al . , 1972; Kroemer et al . , 2009 ) . Mechanistically , these are distinctly unique processes . Moreover , the morphological and biochemical markers determined in our study do not satisfy the criteria set forth by the committee on cell death ( Kroemer et al . , 2009 ) . However , in the holistic view there are intriguing parallels between S . aureus PCL and eukaryotic PCD . PCD occurs as a homeostatic measure in multicellular organisms , whereby a genetically programmed mechanism of cellular catabolism eliminates select quantities and types of cells ( Kerr et al . , 1972; Kroemer et al . , 2009 ) . PCD is crucial for a variety of processes ranging from proper cell turnover and embryonic development to the functioning of the immune system ( Kerr et al . , 1972; Kroemer et al . , 2009 ) . While PCD occurs at the cellular level , and typically in a localized environment , it provides benefits at the organismal level ( Kerr et al . , 1972 ) . Similar to PCD , the findings presented herein suggest that PCL may provide bacteria with a population-level advantage by facilitating biofilm establishment , thereby imparting protection from the immune system and therapeutic agents . Respiration in eukaryotic cells relies upon using oxygen as a substrate . Similar to PCL , hypoxia or anoxia trigger PCD in eukaryotes ( Shimizu et al . , 1996; Weinmann et al . , 2004 ) . PCD occurs as one of two distinct biochemical modalities: apoptosis or necrosis . Hypoxia triggered PCD manifests as a mixture of apoptosis and necrosis ( Shimizu et al . , 1996 ) . Anoxia triggered PCD is largely an apoptotic process ( Weinmann et al . , 2004 ) . Interestingly , anoxia-triggered PCD is dependent upon mitochondrial membrane permeabilization by the pro-apoptotic Bcl-2 family proteins Bax and Bak ( Weinmann et al . , 2004; Kuwana et al . , 2002 ) . Recent evidence suggests that Bax and Bak function as holin-like proteins and facilitate the formation of oligomeric membrane pores ( Kuwana et al . , 2002; Pang et al . , 2011 ) . S . aureus also encodes for two holin-like proteins termed CidA and LrgA ( Ranjit et al . , 2011 ) . The cid operon genes , cidA and cidB have been implicated in programmed cell death in aerobically cultured cells ( Chaudhari et al . , 2016 ) . CidA was previously proposed to have role in cell lysis ( Rice et al . , 2007 ) . This role was predicated upon the phenotype of a cidA mutant; however , recent studies suggest that this was likely an outcome of a secondary mutation ( Rice et al . , 2007; Chaudhari et al . , 2016 ) . CidB , is predicted to be a membrane-associated protein , however its precise function and biochemical activity ( s ) are yet to be defined ( Rice et al . , 2003; Windham et al . , 2016; Chaudhari et al . , 2016 ) . In our hands , cidA::Tn , cidB::Tn , and lrgA::Tn strains were not attenuated in fermentative biofilm formation suggesting a functional separation of the S . aureus PCD and PCL pathways , with respect to biofilm formation . In summary , we report that oxygen impacts S . aureus biofilm formation in its capacity as a terminal electron acceptor . Decreased respiration results in programmed cell lysis via increased expression of AtlA and decreased expression of wall-teichoic acids . These processes are governed by the SrrAB TCRS and evidence suggests this occurs in response to the accumulation of reduced menaquinone . The AtlA-dependent release of cytosolic components facilitates biofilm formation .
Restriction enzymes , quick DNA ligase kit , deoxynucleoside triphosphates , and Phusion DNA polymerase were purchased from New England Biolabs . The plasmid mini-prep kit , gel extraction kit and RNA protect were purchased from Qiagen . DNase I was purchased from Ambion . Lysostaphin was purchased from Ambi products . Oligonucleotides were purchased from Integrated DNA Technologies and sequences are listed in Supplementary file 1 . Trizol and High-Capacity cDNA Reverse Transcription Kits were purchased from Life Technologies . Tryptic Soy broth ( TSB ) was purchased from MP biomedical . Unless otherwise specified all chemicals were purchased from Sigma-Aldrich and were of the highest purity available . Overnight cultures of S . aureus were grown at 37°C in 10 mL culture tubes containing 1 mL of TSB or 30 mL culture tubes containing 5 mL TSB . Difco BiTek agar was added ( 15 g L−1 ) for solid medium . When selecting for or against plasmids , antibiotics where added to the following concentrations: 150 μg mL−1 ampicillin; 30 μg mL−1 chloramphenicol ( Cm ) ; 10 μg mL−1 erythromycin ( Erm ) ; 3 μg mL−1 tetracycline ( Tet ) ; kanamycin , 125 μg mL−1 ( Kan ) ; anhydrotetracycline 150 ng mL−1 . Unless otherwise stated , the S . aureus strains used in this study ( Table 1 ) were constructed in the community-associated S . aureus USA300 LAC strain that was cured of the native plasmid pUSA03 that confers erythromycin resistance ( Boles et al . , 2010 ) . Transposon insertions were obtained from the NARSA library that is housed at BEI resources . All S . aureus mutant strains and plasmids were verified using PCR , sequencing of PCR products or plasmids ( Genewiz , South Plainfield , NJ ) , or genetic/chemical complementation of phenotypes . Escherichia coli DH5α was used as a cloning host for plasmid construction . All constructs were passaged through RN4220 ( Kreiswirth et al . , 1983 ) and subsequently transduced into the appropriate strains using bacteriophage 80α ( Novick , 1991 ) . 10 . 7554/eLife . 23845 . 018Table 1 . Strains and plasmids used in this study . DOI: http://dx . doi . org/10 . 7554/eLife . 23845 . 018Strains used in this studyS . aureus StrainsGenotype/DescriptionGenetic BackgroundSource/ReferenceJMB1100Wild-type; USA300_LAC ( erm sensitive ) ; MRSA; CC8LACBoles et al . ( 2010 ) RN4220Restriction minus; MSSA; CC8NCTC8325Kreiswirth et al . ( 1983 ) JMB 1467ΔsrrAB ( SAUSA300_1441–42 ) LACPang et al . ( 2014 ) JMB 2047ΔsrrAB::tetLACThis workJMB 2078katA::Tn ( ermB ) ( SAUSA300_1232 ) LACV . TorresSH1000parent; MSSA; CC8SH1000Horsburgh et al . ( 2002 ) JMB 1324parent , MRSA , USA400 , CC1MW2Alex Horswill and Centers for Disease Control and Prevention ( 1999 ) JMB 7570parent , MRSA , USA100; CC5N315Ann Stock and Kuroda et al . ( 2001 ) JMB 1432Δfur::tetMLACHorsburgh et al . ( 2001 ) JMB 6231sdhA::Tn ( ermB ) LACBEI resources and Fey et al . ( 2013 ) JMB 6232ΔsrrAB sdhA::Tn ( ermB ) LACThis workJMB 6384ndhF::Tn ( ermB ) ( SAUSA300_0841 ) LACThis work; BEI resources and Fey et al . ( 2013 ) JMB 2057ΔndhC::tet ( SAUSA300_0844 ) LACThis workJMB 6614ΔsrrAB sdhA::Tn ( ermB ) ΔndhC::tet ndhF::Tn ( ermB ) LACThis workJMB 6613sdhA::Tn ( ermB ) ΔndhC::tet ndhF::Tn ( ermB ) LACThis workJMB 6037hemB::Tn ( ermB ) LACBEI resources and Fey et al . ( 2013 ) JMB 6039ΔsrrAB hemB::Tn ( ermB ) LACThis workJMB 6029menF::Tn ( ermB ) LACBEI resources and Fey et al . ( 2013 ) JMB 6033ΔsrrAB menF::Tn ( ermB ) LACThis workJMB 6219menF::Tn ( tet ) LACThis workJMB 6221ΔsrrAB menF::Tn ( tet ) LACThis workJMB 6217hemB::Tn ( ermB ) menF::Tn ( tet ) LACThis workJMB 6673ΔsrrAB hemB::Tn ( ermB ) menF::Tn ( tet ) LACThis workJMB 6625atlA::Tn ( ermB ) LACBEI resources and Fey et al . ( 2013 ) KB5000ΔatlAUAMS-1Bose et al . ( 2012 ) JMB 6624ΔsrrAB atlA::Tn ( ermB ) LACThis workJMB 5577icaA::Tn ( ermB ) LACThis work; BEI resources and Fey et al . ( 2013 ) JMB 5579icaB::Tn ( ermB ) LACThis work; BEI resources and Fey et al . ( 2013 ) JMB 5578icaC::Tn ( ermB ) LACThis work; BEI resources and Fey et al . ( 2013 ) JMB 7270hmrA::Tn ( ermB ) JE2BEI resources and Fey et al . ( 2013 ) JMB 7265lytN::Tn ( ermB ) JE2BEI resources and Fey et al . ( 2013 ) JMB 7267lytX::Tn ( ermB ) JE2BEI resources and Fey et al . ( 2013 ) JMB 7266sle1::Tn ( ermB ) JE2BEI resources and Fey et al . ( 2013 ) JMB 7268lytY::Tn ( ermB ) JE2BEI resources and Fey et al . ( 2013 ) JMB 7269lytZ::Tn ( ermB ) JE2BEI resources and Fey et al . ( 2013 ) JMB 7271lytM::Tn ( ermB ) JE2BEI resources and Fey et al . ( 2013 ) JMB2977parentJE2BEI resources and Fey et al . ( 2013 ) JMB7277narG::Tn ( ermB ) LACBEI resources and Fey et al . ( 2013 ) JMB 1148ΔhptRSLACPang et al . ( 2014 ) JMB 1357ΔlytSRLACPang et al . ( 2014 ) JMB 1330graS::ermLACBoles et al . ( 2010 ) JMB 1335ΔsaePQRS::specLACNygaard et al . ( 2010 ) JMB 1219ΔSAUSA300_1219–1220LACPang et al . ( 2014 ) JMB 1383ΔarlSRLACPang et al . ( 2014 ) JMB 1358ΔphoSRLACPang et al . ( 2014 ) JMB 1241ΔairSRLACPang et al . ( 2014 ) JMB 1377ΔvraSRLACPang et al . ( 2014 ) JMB 1333Δagr::tetMLACKiedrowski et al . ( 2011 ) JMB 1223ΔkdpSRLACPang et al . ( 2014 ) JMB 1359ΔhssSRLACPang et al . ( 2014 ) JMB 1145ΔnreSRLACPang et al . ( 2014 ) JMB 1232ΔSAUSA300_2558–2559LACPang et al . ( 2014 ) Other StrainsEscherichia coli PX5Sacchromyces cerevisiae FY2Plasmids used in this studyPlasmid nameInsert Locus/functionSource/ReferencepJB38Insertless vector for cloning chromosomal gene deletionsBose et al . ( 2013 ) pJB38_srrAB::tetConstruction of srrAB::tet alleleThis workpCM28Insertless cloning vectorA . HorswillpCM28_srrABsrrAB complementing vectorMashruwala and Boyd ( 2017 ) pLL39Insertless cloning vector for genetic complementationLuong and Lee ( 2007 ) pLL39_srrABsrrAB complementing vectorThis workpJB141atlA complementing vectorBose et al . ( 2012 ) pJB135atlAGL complementing vectorBose et al . ( 2012 ) pJB122atlAAMH263A complementing vectorBose et al . ( 2012 ) pJB128Insertless cloning vectorBose et al . ( 2012 ) pJB111atlAAM complementing vectorBose et al . ( 2012 ) pTnTetConstruction of menF::Tn ( Tet ) Bose et al . ( 2013 ) The erythromycin resistance cassette in a menF::Tn ( ermB ) strain was exchanged to a tetracycline resistance cassette as described earlier , with minor changes ( Bose et al . , 2013 ) . The menF::Tn ( ermB ) strain was transduced with the pTnTet plasmid and Tet resistance was selected at 30°C . A single colony was used to inoculate 5 mL of TSB medium and cultured with shaking overnight at 30°C in the presence of Cm . To initiate recombination , cells from the overnight culture were spread onto a TSB agar plate containing Tet and incubated at 42°C ( replication non-permissive ) . Single recombinants were inoculated into 5 mL of TSB and incubated at 30°C in the absence of antibiotic to promote recombination and plasmid loss . These overnights were re-diluted 1:1 , 000 fold into TSB medium containing 30 ng mL−1 of Atet and cultured overnight at 30°C . The overnight culture was diluted of 1:50 , 000 before plating 20–100 μL onto TSA containing Atet to select against plasmid containing cells . Colonies were screened by replica plating for Cm sensitivity and Tet resistance . The resultant strain , once reconstructed , was verified to be deficient in menaquinone biosynthesis by chemical complementation using menaquinone-4 ( MK4 ) . Where mentioned , strains interrupted in hemB were verified using chemical complementation by supplementing growth medium with hemin . The ΔndhC::tetM strain was constructed as described earlier ( Mashruwala et al . , 2015 ) . The pJB38_ΔsrrAB::tet plasmid was created by using PCR to amplify the tetM allele from strain JMB1432 using primers G+tetMluI and G+tetNheI . The PCR product was digested with MluI and NheI and ligated into similarly digested pJB38_ΔsrrAB ( pJB38_ΔsrrAB::tetM ) ( Joska et al . , 2014 ) . The ΔsrrAB::tetM strain was created as outlined above . The pLL39_srrAB plasmid , containing srrAB under the transcriptional control of their native promoter , was constructed using yeast recombinational cloning as previously described ( Joska et al . , 2014; Mashruwala and Boyd , 2016; Mashruwala et al . , 2016b ) . Amplicons were generated using the following primer pairs: pLL39_yeastF and yeast_srrProR , yeast_srrProF and srrAB_pLL39R . The srrAB alleles and the upstream promoter region were amplified from the LAC chromosome and the pLL39 vector was linearized using SalI . The resultant pLL39_srrAB plasmid was integrated as an episome into the chromosome of the ΔsrrAB strain ( JMB1467 ) . Biofilm formation was examined as described earlier , with minor changes ( Mashruwala et al . , 2016a ) . Overnight cultures were diluted into fresh TSB to a final optical density of 0 . 05 ( A590 ) . 200 µL aliquots of diluted cultures were added to the wells of a 96-well microtitre plate ( Corning 3268 ) and the plate was subsequently incubated statically at 37°C for 22 hr . Prior to harvesting the biofilm , the optical density ( A590 ) of the cultures was determined . The plate was subsequently washed twice with water , biofilms were heat fixed at 60°C , and the plates were allowed to cool to room temperature . The biofilms were stained with 0 . 1% crystal violet , washed thrice with water , destained with 33% acetic acid and the absorbance of the resulting solution was recorded at 570 nm , standardized to an acetic acid blank , and subsequently to the optical density of the culture upon harvest . Finally , the data were normalized with respect to the WT or as described in the figure legends to obtain relative biofilm formation . Biofilms were cultured in the presence or absence of oxygen for eight hours . At point of harvest the spent medium was discarded and the remaining culture was immediately resuspended in RNAProtect reagent ( Qiagen ) and treated according to manufacturer instructions . The treated culture was subjected to centrifugation , the supernatant was discarded , and the cell pellet was resuspended in RNase free 50 mM Tris , pH 8 . Cell-free extracts were generated using bead beating . RNA was extracted using Trizol , as per manufacturer instructions . Downstream treatments of the purified RNA and construction of cDNA libraries was as described earlier ( Mashruwala et al . , 2015 ) . Primers for PCR were designed manually or using the Primer Express 3 . 0 software from Applied Biosystems . Quantitative real time PCR reactions ( Table S1 ) were conducted as described earlier ( Mashruwala et al . , 2015 ) . eDNA was analyzed as described earlier with some changes ( Kaplan et al . , 2012 ) . Overnight cultures were diluted into TSB to a final optical density of 0 . 05 ( A600 ) in a final volume of 6 mL per well of a six-well plate . The cultures were incubated statically at 37°C for 22 hr . At point of harvest , the spent media supernatant was aspirated out of each well . One mL of 1X phosphate buffered saline ( PBS ) was immediately added to the wells and a cell scraper was used to transfer the contents to an eppendorf tube . The biomass was pelleted by centrifugation and the supernatant was removed by aspiration . The pellets were thoroughly resuspended in 1X PBS and vortexed for 5 min using a Vortex Genie 2 ( Scientific Industries ) at the highest speed possible using a vertical micro-tube adapter . Aliquots were removed for determination of the viable cell count ( colony forming units ) and samples were pelleted by centrifugation . Control experiments verified that the viable cell counts were not affected by the vortexing procedure ( data not shown ) . Equal volumes of the supernatants were assessed for the presence of high molecular weight DNA ( >10 kilobases ) using agarose gel electrophoresis . To assess the extracellular DNA in a semi-quantitative manner , the gels were photographed and the bands were subjected to density analysis using Image J software . For each sample , the spot densities were normalized to the viable cell count ( colony forming units ) and subsequently as mentioned in the figure legends . Strains were cultured as described under eDNA analyses . The samples were vortexed briefly , biomass was transferred into a microcentrifuge tube , and cell pellets and spent media supernatants were partitioned by centrifugation . The spent media supernatant was retained for further analyses . The cell pellets were resuspended in lysis buffer ( 50 mM Tris , 150 mM NaCl , 4 μg lysostaphin , 8 μg DNAse , pH 7 . 5 ) and incubated at 37°C until confluent lysis was observed . Cell lysates were clarified using centrifugation to obtain cell-free extracts . Catalase ( Kat ) activity was assayed , in both the cell-free extracts as well as spent medium supernatants as described elsewhere ( Mashruwala et al . , 2016a; Beers and Sizer , 1952 ) . The ratio of extracellular to intracellular Kat activity was utilized to determine protein release . In control experiments , Kat activity was undetectable in a katA::Tn strain ( data not shown ) . Overnight cultures were diluted into TSB to a final optical density of 0 . 05 ( A600 ) and cultured for four hours . Whole cell autolysis assays were conducted as described elsewhere with minor changes ( Bose et al . , 2012 ) . Briefly , the cultures were harvested by centrifugation , cell pellets were washed twice , and resuspended in autolysis buffer ( 50 mM HEPES , 150 mM NaCl , 0 . 05% Triton X-100 , pH 7 . 5 ) . For analyses conducted at pH 5 , HEPES was replaced with 0 . 2 M sodium acetate buffer and all other components remained unaltered . The cell suspensions were then incubated at 37°C with shaking and optical densities were recorded periodically . Biofilms were cultured for four hours and cells were harvested as mentioned under eDNA analyses . Thereafter , cell-wall associated protein extracts ( CW-extracts ) were prepared and murein hydrolase activity determined as described elsewhere with minor changes ( Mani et al . , 1993 ) . Briefly , cell pellets were washed and CW-extracts were prepared by resuspension in 3 M lithium chloride and incubation for 25 min ( Mani et al . , 1993 ) . Protein concentrations of the extracts were determined and between 0 . 1–0 . 5 µg of an individual extract was combined with heat-killed cell substrates ( 0 . 35 optical density ( A600 ) ) in assay buffer ( 50 mM Hepes , 150 mM NaCl , 0 . 01% Triton X-100 , pH 7 . 5 ) . For analyses conducted at pH 5 , HEPES was replaced with 0 . 2 M sodium acetate and all other components remained unaltered . Samples were incubated with shaking at 37°C and optical densities were recorded periodically . Binding assays were conducted as earlier ( Fournier and Hooper , 2000 ) . | Millions of bacteria live on the human body . Generally these bacteria co-exist with us peacefully , but sometimes certain bacteria may enter the body and cause infections , such as gum disease or a bone infection called osteomyelitis . Many of these infections are thought to occur when the bacteria become able to form complex communities called biofilms . Bacteria living in a biofilm cooperate and make lifestyle choices as a community , so in this way , they behave like a single organism containing many cells . A sticky glue-like material called the matrix holds the bacteria in a biofilm together . This matrix protects the bacteria in the biofilm from both the human immune system and antibiotics , allowing infections to develop and making them difficult to treat . Previous research has shown that the supply and level of oxygen in infected tissues decreases as an infection gets worse . One bacterium that typically lives peacefully on our bodies , called Staphylococcus aureus , can sometimes cause serious biofilm-associated infections . S . aureus forms biofilms more readily when oxygen is in short supply , but it was not known how these biofilms form . Understanding how S . aureus forms biofilms could help scientists develop better treatments for bacterial infections . Most bacterial cells have a cell wall to provide them with structural support . Mashruwala et al . found that , when oxygen levels are low , S . aureus decreases the production of a type of sugar that makes up the cell wall . At the same time , the bacteria produce more of an enzyme that breaks down cell walls . Together , these processes cause some of the bacteria cells to break open . The contents of these broken cells , including their DNA , help form the matrix that will hold together and protect the other bacterial cells in the biofilm . The experiments also identified a protein called SrrAB that switches on the process that ruptures the cells when oxygen is low . The findings of Mashruwala et al . show how bacteria grown in the laboratory form biofilms when they are starved of oxygen . The next steps following on from this work are to find out whether the same thing happens when bacteria infect animals and whether drugs that block the rupturing of bacterial cells could be used to treat infections . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2017 | Impaired respiration elicits SrrAB-dependent programmed cell lysis and biofilm formation in Staphylococcus aureus |
Energy conservation in microorganisms is classically categorized into respiration and fermentation; however , recent work shows some species can use mixed or alternative bioenergetic strategies . We explored the use of extracellular electron transfer for energy conservation in diverse lactic acid bacteria ( LAB ) , microorganisms that mainly rely on fermentative metabolism and are important in food fermentations . The LAB Lactiplantibacillus plantarum uses extracellular electron transfer to increase its NAD+/NADH ratio , generate more ATP through substrate-level phosphorylation , and accumulate biomass more rapidly . This novel , hybrid metabolism is dependent on a type-II NADH dehydrogenase ( Ndh2 ) and conditionally requires a flavin-binding extracellular lipoprotein ( PplA ) under laboratory conditions . It confers increased fermentation product yield , metabolic flux , and environmental acidification in laboratory media and during kale juice fermentation . The discovery of a single pathway that simultaneously blends features of fermentation and respiration in a primarily fermentative microorganism expands our knowledge of energy conservation and provides immediate biotechnology applications .
The ways in which microorganisms extract energy to maintain cellular functions are directly linked to their environment , including the availability of nutrients and cooperative or antagonistic interactions with other organisms ( Haruta and Kanno , 2015 ) . Microorganisms must also maintain redox homeostasis by responding to oxidative and reductive changes inside and outside the cell ( Sporer et al . , 2017 ) . Ultimately , microorganisms that can effectively generate cellular energy while also managing redox requirements will maintain higher growth and survival rates , and therefore exhibit greater ecological fitness . All organisms possess mechanisms to conserve energy , that is , to convert light or chemical energy into cellular energy in the form of ATP ( Russell and Cook , 1995 ) . During respiration , microorganisms rely on either oxygen ( aerobic respiration ) or other exogenous substrates ( anaerobic respiration ) as terminal electron acceptors . Some microorganisms , most notably Geobacter spp . , can anaerobically respire using electron acceptors outside the cell , such as iron ( III ) oxides or an electrode ( Renslow et al . , 2013; Richter et al . , 2012 ) . This process is called extracellular electron transfer ( EET ) . Regardless of the identity of the electron acceptor , ATP synthesis during respiration occurs via oxidative phosphorylation ( Kim and Gadd , 2019 ) . In oxidative phosphorylation , electrons from electron carriers are transported by an electron transport chain , which creates a proton motive force ( PMF ) for ATP generation . Under anaerobic conditions , some cells can also conserve energy using fermentation . In fermentation , microorganisms use internally supplied electron acceptors , and ATP is generated mainly through substrate-level phosphorylation ( Kim and Gadd , 2019 ) . In substrate-level phosphorylation , ATP is generated in the cytoplasm by transfer of phosphate from metabolic intermediates to ADP ( Kim and Gadd , 2019 ) . Lactic acid bacteria ( LAB ) are a diverse group of aerotolerant , saccharolytic microorganisms in the Firmicutes phylum which mainly use fermentation for energy conservation . LAB are essential for many food fermentations , including fermented milk and meats , fruits and vegetables , and grains ( Tamang et al . , 2020 ) . Strains of LAB are also used for industrial chemical production ( Sauer et al . , 2017 ) and as probiotics to benefit human and animal health ( Vinderola et al . , 2019 ) . LAB are generally grouped by their differences in hexose metabolism ( Salvetti et al . , 2013 ) . Some species perform homofermentation , reducing pyruvate to lactate as the sole metabolic end-product from glycolysis . Other LAB perform heterofermentation , producing lactate along with ethanol , acetate , and CO2 by the phosphoketolase pathway . However , for redox balancing , homofermentative LAB can also shift to a mixed acid fermentation and heterofermentative LAB use alternative electron acceptors , like fructose or citrate ( Hansen , 2018 ) . Although some LAB can respire in the presence of heme and menaquinone , those bacteria are unable to synthesize heme and many are also auxotrophic for menaquinone ( Pedersen et al . , 2012 ) . Even those species capable of respiration still use fermentation metabolism as the primary mechanism to conserve energy ( Pedersen et al . , 2012 ) . Therefore , LAB growth rates and cell yields are constrained by access to electron acceptors used to maintain intracellular redox balance during substrate-level phosphorylation . The bioenergetics of anaerobic bacteria have been tightly linked to oxidative phosphorylation for anaerobic respiration and substrate-level phosphorylation for fermentation . However , experimental evidence shows a concurrent use of oxidative phosphorylation and substrate-level phosphorylation . For instance , some yeasts perform respiro-fermentation to enhance ATP production ( Pfeiffer and Morley , 2014 ) . Another example is the electron bifurcating mechanism used by some fermentative microorganisms such as Clostridium spp . ( Herrmann et al . , 2008; Li et al . , 2008 ) . Through that energy conservation strategy , cells can generate extra ATP through oxidative phosphorylation ( Buckel and Thauer , 2013; Müller et al . , 2018 ) . Along with other examples that are not fully understood ( Hunt et al . , 2010Kracke et al . , 2018 ) , these observations suggest metabolisms that combine aspects of fermentation and respiration may exist . We recently discovered that Listeria monocytogenes , a facultative anaerobic pathogen known to rely on respiratory metabolism , uses EET to reduce Fe3+ or an anode through a flavin-based extracellular electron transfer pathway ( Light et al . , 2018 ) . Use of this pathway allowed L . monocytogenes to maintain intracellular redox balance via NADH oxidation . This capacity was associated with the presence of a gene locus , called a flavin-based EET ( FLEET ) locus , that was identified in many Gram-positive species in the Firmicutes phylum , including LAB . Studies in individual species of LAB such as Lactococcus lactis ( Freguia et al . , 2009; Masuda et al . , 2010 ) , Enterococcus faecalis ( Hederstedt et al . , 2020; Keogh et al . , 2018 ) , and Lactiplantibacillus pentosus ( Vilas Boas et al . , 2015 ) show that they can perform EET with an anode endogenously , that is without addition of molecules foreign to their native niches . These observations are quite surprising because endogenous EET has been mainly associated with respiratory organisms , even though some of these organisms also possess fermentative-type metabolism ( Glasser et al . , 2014 ) . Those observations also raise the question of whether the FLEET locus is functional in LAB and what , if any , role it plays in energy conservation and metabolism . Here , we explored EET across LAB and studied the implications of this trait at a metabolic and energetic level in Lactiplantibacillus plantarum , a homofermentative LAB capable of mixed acid fermentation and which can respire in the presence of exogenous heme and menaquinone . L . plantarum is of particular interest as it is a generalist LAB species found in insect , animal , and human digestive tracts and is essential for the production of many fermented foods ( Behera et al . , 2018; Duar et al . , 2017 ) . These findings have significance for the understanding of energy conservation strategies in primarily fermentative microorganisms and on lactic acid fermentations in food biotechnology .
To determine whether L . plantarum can reduce extracellular electron acceptors , we first measured its ability to reduce insoluble ferrihydrite ( iron ( III ) oxyhydroxide ) . Incubation of the model strain L . plantarum NCIMB8826 in the presence of ferrihydrite showed that this strain reduces Fe3+ to Fe2+ ( Figure 1A and Figure 1—figure supplement 1A ) . Viable cells are required for iron reduction and this activity is dependent on the presence of exogenous quinone ( DHNA , 1 , 4-dihydroxy-2-naphthoic acid ) ( Figure 1A and Figure 1—figure supplement 1A-B ) . The requirement for DHNA was hypothesized because DHNA is a precursor of demethylmenaquinone ( DMK ) , a membrane electron shuttle utilized by L . monocytogenes for EET ( Light et al . , 2018 ) , and L . plantarum lacks a complete DHNA biosynthetic pathway ( Brooijmans et al . , 2009a ) . For full activity , an electron donor ( such as mannitol or glucose ) was required to be present ( Figure 1A and Figure 1—figure supplement 1A ) . Like L . monocytogenes ( Light et al . , 2018 ) , the addition of riboflavin during the iron reduction assay also increased Fe3+ reduction in a dose-dependent manner ( Figure 1—figure supplement 1C ) . Thus , L . plantarum reduces insoluble iron in a manner similar to L . monocytogenes . Next , we investigated whether the ability of L . plantarum to reduce insoluble iron was altered by growth media . L . plantarum was able to reduce iron after growth in either complete ( MRS ) medium or chemically defined medium ( CDM ) ( Figure 1—figure supplement 1B ) . Iron reduction was greater when mannitol , a sugar alcohol , rather than glucose , was provided as the sole carbon source in MRS ( Figure 1—figure supplement 1B ) . However , reduction was highest when L . plantarum was incubated in mannitol-containing MRS ( mMRS ) with both DHNA and ferric ammonium citrate present ( Figure 1—figure supplement 1D ) . The addition of riboflavin to the growth medium did not further increase iron reduction by L . plantarum ( Figure 1—figure supplement 1E ) , potentially because riboflavin is already present in high quantities in MRS , a medium containing yeast extract ( Tomé , 2021 ) . Thus , L . plantarum was grown in mMRS supplemented with DHNA and iron before ferrihydrite reduction assays in all subsequent experiments . L . plantarum EET activity was confirmed in a bioelectrochemical reactor by quantifying electron output as current ( Figure 1B ) . L . plantarum reduced a carbon electrode ( anode ) polarized to +200 mVAg/AgCl in the presence of both DHNA and an electron donor ( mannitol ) ( Figure 1C ) . No current was observed in the absence of L . plantarum ( Figure 1—figure supplement 2A ) , indicating that current production stems from a biological process . L . plantarum produced a maximum current of 129 ± 19 µA/cm2 in mCDM ( Figure 1C ) and 225 ± 9 µA/cm2 in mMRS ( Figure 1—figure supplement 2B ) . Under EET conditions in mCDM , the L . plantarum biomass was 2 . 7 mg ( dry cell mass ) . Assuming 50% of the dry cell mass was protein , the specific electron transfer rate was 57 µmol electrons/mg- protein/hr and the current production was 1 . 5 mA/mg-protein . This value is lower than that reported for Geobacter sulfurrenducens ( 4–8 mA/mg-protein ) ( Marsili et al . , 2010; Rose and Regan , 2015 ) , the model species for direct EET , and higher than that of Shewanella oneidensis ( 0 . 67 mA/mg-protein ) ( Marsili et al . , 2008 ) , the model species for mediated EET . It should be noted that these species , unlike L . plantarum , can synthesize riboflavin and quinones and do not require the addition of either for EET activity . Similar to our iron reduction experiments , EET to an anode occurred with different electron donors and growth media ( Figure 1—figure supplement 2B-C ) , and current increased after supplementation of riboflavin when it was omitted from the growth medium ( Figure 1—figure supplement 2D ) . Because of these differences , CDM was amended with mannitol and riboflavin in subsequent experiments . DHNA is found in concentrations of 0 . 089–0 . 44 μg/mL in commercial fermented beverages ( Eom et al . , 2012 ) , and under laboratory conditions , microbes can synthesize and secrete DHNA leading to concentrations of 0 . 37–48 μg/mL ( Isawa et al . , 2002; Furuichi et al . , 2006; Kang et al . , 2015 ) . To test whether EET in L . plantarum is relevant under these physiological concentrations , we probed whether L . plantarum can perform EET with a sub-physiological DHNA concentration of 0 . 01 μg/mL . Indeed , L . plantarum can reduce iron and produce significant current density ( Figure 1—figure supplement 3A-B ) , although the magnitude of iron reduction and current was smaller than what was observed with 20 μg/mL . These results show that the concentrations of DHNA found in niches of L . plantarum can support EET and suggest the magnitude of EET will depend on the DHNA concentration . Because iron reduction by L . monocytogenes requires the genes in a 8 . 5 kb gene locus encoding a flavin-based EET ( FLEET ) pathway ( Light et al . , 2018 ) , we looked for the presence of these genes in 1 , 788 LAB genomes deposited in NCBI . Homology searches identified the complete FLEET locus in 11 out of 38 genera including diverse LAB such as Enterococcus and Lacticaseibacillus ( Figure 2A ) . The other LAB genera either lack multiple FLEET pathway genes or , as was observed for all 68 strains of Lactococcus , contain all genes except for pplA , which is predicted to encode an extracellular flavin-binding reductase . Among the lactobacilli , genomes of 19 out of 94 species contain the entire FLEET system ( Figure 2—figure supplement 1 ) . The lactobacilli species with the entire FLEET locus are homofermentative and are distributed between different phylogenetic groups ( Zheng et al . , 2020 ) . These data show that the FLEET locus is conserved across LAB genera besides L . plantarum , including other homofermentative LAB species known to colonize host and food environments . To determine whether LAB FLEET gene presence was associated with EET activity , a diverse collection of LAB strains were examined for their capacity to reduce ferrihydrite . The assay showed that isolates of L . plantarum , Lactiplantibacillus pentosus , Lacticaseibacillus rhamnosus , Lacticaseibacillus casei , Enterococcus faecium , and Enterococcus faecalis are capable of Fe3+ reduction ( Figure 2B ) . The genomes of those species also contain a complete FLEET locus ( Figure 2A and Figure 2—figure supplement 1 ) . Conversely , strains of Lactococcus lactis , Ligilactobacillus murinus , Levilactobacillus brevis , Pediococcus pentosaceus , and Streptococcus agalactiae showed little to no iron reduction activity ( Figure 2B ) . The presence of FLEET-associated genes varied between those species , but only strains of species found to contain both ndh2 , a predicted membrane-bound , type-II NADH dehydrogenase , and pplA were able to reduce iron under the conditions tested . L . plantarum NCIMB8826 exhibited the highest EET activity resulting in at least 2 . 5-fold greater Fe3+ reduction than the other L . plantarum strains ( Figure 2B ) . Remarkably , however , the L . plantarum NCIMB8826 genome and the genomes of 138 other L . plantarum strains queried all harbored a complete FLEET locus including ndh2 and pplA ( Figure 2—figure supplement 1 and Figure 2—figure supplement 2A ) . Among those strains tested for the capacity to reduce Fe3+ , L . plantarum NCIMB700965 and 8 . 1 could not reduce Fe3+ but possessed all genes in the FLEET pathway . Closer examination of both strains by aligning their FLEET loci with NCIMB8826 revealed unique IS30-family transposons in the intergenic promoter regions spanning ndh2 and pplA ( Figure 2—figure supplement 2A ) . These genes were minimally expressed in L . plantarum NCIMB700965 and 8 . 1 in comparison to NCIMB8826 ( Figure 2—figure supplement 2B ) . ndh2 and pplA were also the only two genes in the FLEET gene locus that were induced when L . plantarum NCIMB8826 was incubated in mMRS supplemented with DHNA and iron ( Figure 2C and Figure 1—figure supplement 1D ) . Both ndh2 and pplA were induced ( ~1 . 6 fold , p < 0 . 05 ) in MRS containing mannitol , DHNA , and ferric ammonium citrate ( Figure 2C ) , but were not upregulated when either DHNA or ferric ammonium citrate were omitted from the culture medium ( Figure 2—figure supplement 2C ) . Taken together , these data show that widespread iron reduction in LAB is tightly associated with the presence and upregulation of ndh2 and pplA , suggesting they are required for EET . In order to confirm the necessity of ndh2 and pplA for EET in L . plantarum , we constructed ndh2 and ppA deletion mutants of L . plantarum NCIMB8826 . Both mutants were significantly impaired in their capacities to reduce ferrihydrite compared with the wild-type strain ( Figure 3A ) . The ndh2 and pplA deletion mutants also had different effects on the oxidation-reduction potential ( ORP ) of mMRS . ORP is defined as the ratio of all oxidative to reductive components in an environment ( Killeen et al . , 2018 ) and is an important environmental condition which influences the outcome of LAB fermentations such as flavor development in cheese ( Morandi et al . , 2016 ) and the growth of spoilage microorganisms ( Olsen and Pérez-Díaz , 2009 ) . Expectedly for the L . plantarum EET phenotype , significant reductions in mMRS ORP only occurred during L . plantarum growth when DHNA was included in the culture medium ( Figure 3—figure supplement 1A ) . Although ORP declined for all three strains in a manner consistent with other ORP-reducing enzymatic activities ( for example the reduction of oxygen by NADH oxidase ) ( Tachon et al . , 2010 ) , wild-type L . plantarum resulted in greater reductions in ORP compared to either mutant in mMRS , and these differences were significant at most time points measured over a 12 hr period ( Figure 3B ) . The effects on ORP occurred in the absence of changes in growth rates , cell yields , and medium pH ( Figure 3—figure supplement 1A-D ) . The ΔmVmax was reached during mid-exponential phase ( approximately 5 hr ) ( Figure 3—figure supplement 1B ) , and at that time , wild-type L . plantarum cells but not the Δndh2 or ΔpplA strains were active in the ferrihydrite reduction assay ( Figure 3—figure supplement 1E ) . This difference in ferrihydrite reduction activity similarly persisted in stationary phase cells ( 12 hr ) ( Figure 3—figure supplement 1F ) . These observations show that ndh2 and pplA contribute to the capacity of L . plantarum to reduce iron and have relevance to the ORP-dependent activities occurring during food fermentations ( van Dijk et al . , 2000 ) . Use of an anode as an external electron acceptor instead of ferrihydrite showed a similar , but not identical genetic dependency . L . plantarum Δndh2 produced a significantly lower current density ( Figure 3C ) and a lower peak current ( Figure 3—figure supplement 2A ) . Surprisingly , L . plantarum ΔpplA was able to produce the same amount of current as the wild-type strain , suggesting that the lipoprotein PplA is not essential and might not be involved in anode reduction through EET . This observation led us to investigate the anodic-EET ability of other LAB species lacking pplA like Lactococcus lactis ( Figure 3D ) . DHNA was not provided to these strains because they can synthesize demethylmenaquinone and other quinones ( Rezaïki et al . , 2008 ) . Both L . lactis strain IL1403 and strain KF147 were capable of current generation , confirming that PplA is not essential for LAB to produce current . This is consistent with the finding that other extracellular reductases besides PplA are responsible for EET activity in Gram-positive bacteria ( Light et al . , 2019 ) . Taken together these results show that EET activity is dependent upon the presence of the putative FLEET locus , and specifically ndh2 and conditionally pplA . Building from studies in E . faecalis ( Keogh et al . , 2018 ) , it has been suggested that EET improves growth by either enabling iron to be acquired as a macronutrient or by enhancing respiration ( Jeuken et al . , 2020 ) . It is worth noting that several studies have shown that L . plantarum does not require iron to grow ( Elli et al . , 2000; Weinberg , 1997 ) . To test whether EET allowed increased iron acquisition by L . plantarum , we measured intracellular iron by Inductively Coupled Plasma-Mass Spectrometry ( ICP-MS ) . There was no significant difference in intracellular iron concentrations between L . plantarum growth in mMRS supplemented with DHNA and iron compared to growth in mMRS alone ( Figure 4—figure supplement 1 ) . Moreover , deletion of ndh2 did not significantly change the amount of intracellular iron ( Figure 4—figure supplement 1 ) . ICP-MS showed that other redox-active metals used for EET , such as manganese and copper ( Kouzuma et al . , 2012; Fan et al . , 2018 ) were also not affected ( Figure 4—figure supplements 1 and 2 ) . In contrast to studies in E . faecalis in which iron supplementation leads to intracellular accumulation of this metal ( Keogh et al . , 2018 ) , these data show that L . plantarum does not use EET to increase its acquisition of iron or other redox-active metals , suggesting EET may instead play a role in energy conservation . We next sought to understand if EET impacts energy conservation in L . plantarum by comparing its growth and ATP levels in the presence of a polarized anode . The highest current density ( i . e . greatest EET activity ) produced by L . plantarum in mannitol CDM typically occurred within 24 hr after inoculation into the bioreactor ( Figure 1C ) . At this point , there was an approximately 4-fold higher dry cell weight and 2-fold higher numbers of viable cells compared to L . plantarum incubated in open circuit ( OC ) conditions ( Figure 4A–B ) Current density declined from its maximum value when L . plantarum cells performing EET were in exponential growth ( Figures 1C and 4C ) . By comparison , growth was not observed until two days later under OC conditions ( Figure 4C ) . During peak current production , intracellular ATP levels were significantly higher ( 4 . 5-fold ) under EET compared to OC conditions ( Figure 4D and Table 1 ) . These results strongly suggest faster ATP accumulation under EET conditions allowed L . plantarum to exit lag phase more rapidly . ATP levels were also greater in L . plantarum when in the presence of both mannitol and DHNA , compared to either mannitol or DHNA separately ( Figure 4D ) . Thus , EET allows L . plantarum to initiate growth and accumulate ATP more rapidly , indicating that EET significantly increases energy conservation in L . plantarum . Because fermentation , anaerobic respiration , and aerobic respiration are each associated with a different NAD+/NADH ratio , energy conservation is linked to intracellular redox homeostasis ( Holm et al . , 2010 ) . Therefore , we probed redox homeostasis in L . plantarum under EET conditions by measuring intracellular NAD+/NADH at the point of maximum current density ( Figure 4E ) . L . plantarum showed an 8-fold higher NAD+/NADH ratio under EET conditions compared to OC ( Figure 4E ) . This result was not limited to the presence of a polarized anode as L . plantarum also contained a significantly higher NAD+/NADH ratio when Fe3+ was available as a terminal electron acceptor ( Figure 4—figure supplement 3 ) . These NAD+/NADH ratios are more similar to those found for in E . coli performing aerobic respiration ( de Graef et al . , 1999 ) or G . sulfurreducens performing anaerobic respiration than in LAB performing fermentation ( Guo et al . , 2017 ) . Taken together , our data indicate that EET is involved in energy conservation , and the intracellular redox balance during EET mimics a respiratory rather than a fermentative process . Metal-reducing bacteria use EET in anaerobic respiration ( Richter et al . , 2012; Shi et al . , 2007 ) . Ndh2 is considered an anaerobic respiratory protein , and L . plantarum can perform anaerobic respiration with exogenous menaquinone and heme using an electron transport chain ( Brooijmans et al . , 2009b ) . This led us to hypothesize that those electron transport proteins could also be involved for EET to conserve energy as part of anaerobic respiration . To test this hypothesis , we examined whether any of the known electron transfer proteins needed for PMF generation in aerobic and anaerobic respiration are required for L . plantarum EET . Neither the addition of heme to restore bd-type cytochrome ( cydABCD ) used in aerobic respiration , nor deletion of the respiratory nitrate reductase ( ΔnarGHJI ) significantly altered current production ( Figure 5—figure supplement 1A-B ) . Because Ndh2 is a type-II NADH dehydrogenase which does not contribute to a proton gradient ( Lin et al . , 2008; Nakatani et al . , 2020 ) , these observations show that while EET does involve a respiratory protein , it does not involve any of the known PMF-generating electron transfer proteins in L . plantarum . Respiration is also associated with the tricarboxylic acid ( TCA ) cycle . L . plantarum , like other LAB , does not possess an oxidative branch of the TCA cycle and only contains a reductive branch ( Tsuji et al . , 2013 ) . To probe whether the reductive branch was active during EET , we also examined production of succinate , the terminal end-product of the reductive branch . EET did not increase the succinate concentration ( Figure 5—figure supplement 2 ) . Moreover , we did not detect any intermediates of the reductive branch of the TCA cycle , that is oxalacetate , malate , or fumarate . This indicates that EET did not cause additional metabolic flux through its TCA cycle . Thus , none of the known metabolic pathways or electron transport proteins associated with anaerobic respiration , besides Ndh2 , are required for EET . These results suggest increased energy conservation during EET in L . plantarum is not through canonical anaerobic respiration . An alternative hypothesis is that increased energy conservation under EET conditions is driven by changes in fermentation . L . plantarum uses glycolysis to convert mannitol to two molecules of pyruvate which are then converted mainly to lactate or ethanol via NADH-consuming steps , or acetate via an ATP-generating reaction using substrate-level phosphorylation ( Dirar and Collins , 1972 ) . Thus , shifting toward production of acetate from to lactate or ethanol production can increase ATP yield during fermentation . Additionally , NADH can be re-generated by oxidizing pyruvate to yield 2 , 3-butanediol , using acetoin as an intermediate . Fermentation in L . plantarum also decreases the pH of the surrounding media . To probe changes in fermentation , we measured the concentrations of mannitol , acetate , lactate , ethanol , acetoin , 2 , 3-butanediol , formate , and pyruvate and the pH in L . plantarum cultures during OC and EET conditions . After four days , we accounted for ~80% and ~ 55% of the total carbon under EET and OC conditions ( for all metabolite concentrations see Supplementary file 1 ) , giving us a quantitative view of metabolism under EET conditions . Surprisingly , under EET conditions , the distribution of major end-fermentation products ( acetate , lactate , and ethanol ) did not change , but their yield per cell was 2 . 6-fold higher compared to OC conditions ( Figure 5A ) . While we did not detect acetoin or 2 , 3-butanediol , formate was found at trace levels , and pyruvate was found at similar , low levels under EET and OC conditions ( Figure 5—figure supplement 2 ) . After accounting for mannitol consumption , we observed that EET allowed cells to produce ~1 . 75 x more fermentation products per each mol of mannitol utilized ( Yfermentation , Table 1 ) . The culture medium pH was also significantly lower than under OC ( Figure 5B ) , a result which may indicate that EET conferred higher levels of acid stress on L . plantarum , and therefore , reductions in cell viability , despite EET leading to higher cell numbers overall ( as measured by dry cell weight ) ( Figure 4A–B ) . A similar acidification of the medium was observed for ΔpplA , but not for Δndh2 , when an anode was present as electron acceptor , indicating that ndh2-dependent EET is needed to decrease the pH ( Figure 3—figure supplement 2B ) . When much lower , sub-physiological levels of DHNA were supplied ( 0 . 01 μg/mL ) , a smaller but significant decrease in the pH of the medium was also observed ( Figure 1—figure supplement 3C ) . Overall , these results show that EET allows L . plantarum to ferment to ~1 . 75 x greater extent and to acidify the medium to a greater extent as well . We also observed that EET led to higher cellular metabolic fluxes , that is , higher changes in metabolites per cell per unit time . Although the final OD600nm and dry cell weight were not significantly different ( Supplementary file 1 ) , L . plantarum utilized mannitol and produced acetate and lactate more rapidly under EET than OC conditions ( Figure 5C–F ) . Cells performing EET were ~2 fold faster at consuming mannitol ( Figure 5C ) between days 1 and 3 . Mannitol consumption increased between day 1 and day 2 , approximately when the cells transitioned to higher current density ( Figure 3C ) , suggesting that increased EET drove that increased consumption . The overall rates of acetate and lactate production also increased 3 . 4 and 3 . 6 times ( Figure 5D–E ) , respectively . Measurements of metabolites produced by ΔpplA and Δndh2 strains confirmed that , like for current production to an anode , the EET-associated increased metabolic flux in L . plantarum requires the presence of Ndh2 , but not PplA ( Figure 3—figure supplement 2C ) . Overall , these data indicate that ndh2-dependent EET increases both the flux and final yield of fermentation in L . plantarum . Because the production of acetate yields ATP , these results also suggested that the increase in ATP generation under EET conditions may be due to substrate-level phosphorylation . To probe whether EET-associated increase in fermentative flux could account for the changes in ATP generation , we calculated fermentation balances ( Table 1 ) . Our measurements account for 80% of the carbon under EET conditions ( see Supplementary file 1 ) , leaving a maximum of ~20% systematic uncertainty in these calculations . The concentrations of fermentation products detected ( Supplementary file 1 ) were used to estimate the total ATP in the presence and absence of EET . The estimated ATP was 3-fold higher under EET conditions than OC conditions ( Table 1 ) , a result that is consistent with the ~2 . 5 fold higher accumulation of ATP measured at maximum current density ( Figure 4D ) . Overall , this quantitative analysis shows that the vast majority of the increased energy conservation under EET conditions can be accounted for by an increase in fermentation yield and substrate-level phosphorylation . Thus far , our results provided an unusual picture of the energy metabolism of L . plantarum under EET conditions; while EET significantly shifted the intracellular redox state to a more respiratory-like balance , its increased ATP yield was mainly accounted for by an increased fermentative yield . Another major difference in fermentation and anaerobic respiration is the use of the endogenous versus exogenous electron acceptors . To more deeply understand how L . plantarum uses organic molecules and the anode as electron acceptors when performing EET , the electron balances under EET and OC conditions were calculated ( Table 1 and Supplementary file 1 ) . We estimated the NADH produced using two different methods ( see Supplementary file 1 for methodology ) and the NADH re-oxidized through the reduction of the anode ( measured as current ) and via substrate-level phosphorylation . This allowed us to obtain a global balance of the NAD+/NADH ratio . Under OC conditions between 35% and 66% of the NADH produced from the oxidation of mannitol to pyruvate ( a range is given using the two methods used ) was re-oxidized to NAD+ ( 5 . 5 mM NADH consumed , Table 1 ) , qualitatively agreeing with the low NAD+/NADH ratios measured ( Figure 4E ) . In contrast , electron balance calculations showed that between 77% and 96% of the NADH produced under EET conditions was re-oxidized ( 17 mM NADH consumed , Table 1 ) , a result that is consistent with the significantly higher NAD+/NADH ratios measured ( Figure 4E ) . Interestingly , these calculations estimate that 55–69% of the total NADH generated was oxidized through fermentation , while 21–28% of the NADH was oxidized using the electrode as a terminal electron acceptor ( Table 1 ) . Thus , L . plantarum growing under EET conditions achieves a more oxidized intracellular redox balance by more completely fermenting mannitol to lactate and ethanol and by using the electrode as a terminal electron acceptor ( Figure 5G ) . These observations reinforce that the energy metabolism of L . plantarum under EET conditions utilizes elements of both fermentation and anaerobic respiration . In rapidly dividing cells , energy conservation , a catabolic process , is associated with growth , an anabolic process . However , catabolism need not be coupled with anabolism ( Russell and Cook , 1995 ) . To determine how catabolic and anabolism are connected under EET conditions , the ATP requirements to grow biomass ( YATP ) were estimated using the calculated ATP and the measured dry weight . Under OC conditions , the YATP obtained ( 8 . 06 ± 0 . 8 g dw/mol ATP ) for L . plantarum was similar to that observed previously ( 10 . 9 g-dw/mol ATP ) ( Dirar and Collins , 1972 ) . Hence , without EET , the ATP generated from fermentation was converted into biomass nearly at the expected efficiency . In contrast , a significantly lower YATP was reached for L . plantarum performing EET ( 3 . 07 ± 0 . 35 g dw/mol ATP ) ( Table 1 ) . This observation indicates that under EET conditions , either more ATP is required to produce biomass or more ATP is utilized by other functions such as for PMF-generation and intracellular pH regulation ( Russell and Cook , 1995 ) . EET conditions also resulted in 79% more ATP per mol of fermented mannitol ( Ymannitol ) . Consequently , molar biomass yields ( g-dw/mol-mannitol ) under EET conditions were significantly lower ( Table 1 ) , in agreement with previous observations in respiratory electroactive species ( Esteve-Núñez et al . , 2005 ) . These calculations show that when L . plantarum performs EET , anabolism and catabolism processes are differently coupled than under OC conditions . ATP is produced more efficiently , but this it is less efficiently utilized to make biomass . Overall , these results show an intriguing pattern of coupling between anabolism and catabolism , indicative of a novel energy metabolism in L . plantarum during EET . Our results inspired us to explore whether EET could occur in a physiological niche of LAB such as fermented foods . LAB are necessary for the making of many fermented fruit and vegetable foods and the properties of those foods depend on the metabolic diversity of the LAB strains present ( Gänzle , 2015 ) . Plant tissues also contain a much wider variety of carbon substrates and potential electron acceptors than the CDM used in our prior experiments . To study the physiological and biotechnological relevance of EET in food fermentations , kale juice was fermented using L . plantarum as a starter culture ( Figure 6A ) . The fermentation of kale juice was measured under EET conditions ( a polarized anode with or without DHNA ) , and an OC control ( a nonpolarized anode with DHNA ) was used to separate the role of DHNA and electron flow to the anode on the fermentation process . An additional bioreactor without cells , but with DHNA , was operated to identify any possible electrochemical-driven conversion of substrates . When L . plantarum was added to the prepared kale juice , approximately 10-fold more current was generated during EET conditions with DHNA ( EET+ DHNA ) as compared to abiotic and biological non-EET promoting conditions ( no DHNA ) ( Figure 6B ) . This current was comparable to the current generated in laboratory medium ( Figure 3C ) , indicating that robust EET by LAB is possible in the complex physiological conditions of a food fermentation . We next investigated the impact of EET on L . plantarum growth and metabolism in the kale juice fermentation . Significant changes in the pH and fermentation products were detected under EET conditions ( Figure 6C–D ) . These differences occurred in the absence of significant changes in viable cell numbers ( Figure 6—figure supplement 1A ) at the time points measured . As previously observed using laboratory culture media , an approximately 2-fold greater accumulation of total end-fermentation products per cell was obtained when cells interacted with an anode in the presence of DHNA ( Figure 6C ) . In the kale juice fermentation , EET+ DHNA conditions enhanced both lactate and acetate production per cell without changing the distribution of metabolites ( Figure 6E and Figure 6—figure supplement 1B ) . Thus , when DHNA was provided , EET enhanced the overall yield of fermentation end-products and their production rates per cell , mimicking our observations in laboratory medium ( Figure 5C–D ) . EET also led to a significantly higher acidification of the kale juice compared to OC conditions , and the presence of DHNA dramatically enhanced this pH drop ( Figure 6D ) . In general , when no DHNA was supplied but an anode was present as an electron acceptor , the fermentation process was very similar to OC conditions . This means in kale juice , a source of quinones is essential to support L . plantarum EET activity . Overall , these results show that EET under physiological conditions impacts cellular metabolism in L . plantarum by increasing metabolic flux which ultimately can affect the flavor profile of fermented foods ( Chen et al . , 2017 ) .
When performing EET , the metabolism of L . plantarum , a primarily fermentative bacterial species , is fundamentally different from EET-driven , anaerobic respiration of metal-reducing bacteria . Although aspects of EET in L . plantarum and metal-reducing Geobacter spp . are similar , such as the upregulation of NADH dehydrogenase , the reduction rate of extracellular electron acceptors , and the high NAD+/NADH ratio , other aspects of energy metabolism during EET in these two organisms are starkly different ( see comparison in Supplementary file 2 ) . Geobacter spp . direct their metabolic flux through the TCA cycle , rely almost exclusively on extracellular electron acceptors to regenerate NADH , and produce ATP exclusively through oxidative phosphorylation . In contrast , L . plantarum regenerates a substantial fraction of its NADH by directing metabolic flux through fermentative pathways . Additionally , oxidative phosphorylation is not a major mechanism of energy conservation in L . plantarum during EET , as supported by three lines of evidence: the marginal metabolic flux through the reduced branch of TCA cycle , no involvement of known PMF-generating proteins , and that increased ATP levels can be accounted for by increased substrate-level phosphorylation . While additional data are required to eliminate the possibility that oxidative phosphorylation is occurring in L . plantarum during EET , we can qualitatively state that substrate-level phosphorylation is the major mechanism for ATP generation . Comparing EET and respiratory metabolism in LAB also reveals substantial differences in these metabolisms ( Supplementary file 2 ) . While both metabolisms require quinones , respiration also requires exogenous heme . Our findings and similar findings in E . faecalis ( Pankratova et al . , 2018 ) confirm that heme is not required for EET . Moreover , EET also differs from respiration in LAB because it occurs at the start of or prior to exponential phase growth , does not change the final cell density , and increases fermentation with no effect on the resultant proportions of lactate , acetate , and ethanol ( Duwat et al . , 2001 ) . Thus , EET in LAB diverges from respiration in metal-reducing bacteria or LAB in its metabolic pattern and energetic consequences . In addition , while EET provokes a shift in fermentative metabolism in other bacteria upon the addition of artificial mediators ( Vassilev et al . , 2021; Emde and Schink , 1990 ) , L . plantarum EET is active upon the presence of a mediator present in a complex food system . This EET mechanism is also a novel energy conservation strategy compared to known fermentative metabolisms in LAB ( comparison in Supplementary file 2 ) . L . plantarum and other LAB , reduce alternative intracellular electron acceptors like citrate , fructose , and phenolic acids , resulting in increased intracellular NAD+/NADH ratios ( Hansen , 2018 ) . This metabolic activity is especially important for heterofermentative LAB in order to synthesize additional ATP through acetate kinase ( Gänzle , 2015 ) . Unlike these examples , however , the reduction of extracellular Fe3+ or an anode by EET requires a respiratory protein ( Ndh2 ) and the shuttling of electrons outside of the cell . In addition , the reduction of the oxygen and organic compounds for cofactor regeneration by LAB leads to a metabolic shift toward acetate production and altered metabolic end-product ratios ( Gänzle , 2015 ) , which does not occur during EET . These differences show how the hybrid metabolism under EET conditions is distinct from other pathways that alleviate reduced intracellular conditions in LAB . Previous studies have reported a simultaneous use of fermentation and electron transport elements , such as in respiro-fermentation in Saccharomyces cerevisiae ( Blom et al . , 2000 ) . However , respiro-fermentation produces ATP and maintains intracellular redox balance through substrate-level phosphorylation and/or oxidative phosphorylation using separate pathways . Our data strongly suggests a single pathway is responsible for both ATP generation and intracellular redox balance . This hybrid fermentation mode is also different from the electron bifurcating mechanism , in which the extra ATP generation is driven by the creation of a H+ or Na+ potential from the oxidation of a ferredoxin ( Buckel and Thauer , 2013 ) . Unlike in this example , EET in LAB does not involve PMF creating elements and EET drives ATP generation through substrate-level phosphorylation . Another poorly understood example of the use of substrate-level phosphorylation and electron transport chains to balance intracellular redox state is found in the non-fermentative bacterium S . oneidensis . Although this species is a respiratory bacterium , it relies predominately on substrate-level phosphorylation to grow anaerobically with the exogenous electron acceptor fumarate ( Hunt et al . , 2010 ) . In this scenario , it is unclear if there are changes in intracellular redox state or metabolism in this species . In contrast to and expanding upon these studies , our work elucidates a qualitatively and quantitatively different blending of fermentation and respiration . L . plantarum EET-associated metabolism contains features of both fermentation ( e . g . substrate-level phosphorylation , high fermentation product yields ) and respiratory metabolisms ( e . g . NAD+/NADH ratios , NADH dehydrogenase required ) ( Supplementary file 2 ) . Quantitatively , this hybrid metabolism leads to an overall ~1 . 75x-more efficient and ~1 . 75x-faster energy conservation ( increased Ymannitol , mannitol flux ) , but an overall ~1 . 5 fold weaker coupling between anabolism and catabolism ( lower YATP ) . Additionally , the increased NAD+/NADH ratio arises from using an ~2:1 ratio of endogenous to extracellular electron acceptors . Thus , to our knowledge , the hybrid strategy that L . plantarum uses to generate ATP performing EET constitutes a novel mode of energy conservation in a primarily fermentative microorganism . Based on our observations and others , we propose that EET is widespread in LAB and occurs by different mechanisms . Besides L . plantarum , we showed that L . lactis is able to generate current despite lacking pplA . Current generation by L . lactis was observed previously , found to be riboflavin dependent , and resulted in a small metabolic shift ( yet to be defined ) in which the flux through NADH-oxidizing pathways was reduced and ATP generating pathways were increased ( Freguia et al . , 2009; Masuda et al . , 2010 ) . L . lactis can also perform EET by reduction of tetrazolium violet and this activity depends on the presence of both quinones and an NADH dehydrogenase ( NoxAB ) ( Tachon et al . , 2009 ) . E . faecalis is another LAB that performs EET , and similar to L . plantarum , it requires quinones ( Pankratova et al . , 2018 ) and a type-II NADH dehydrogenase ( Ndh3 ) ( Hederstedt et al . , 2020 ) for Fe3+ reduction . In contrast to this mechanistic similarity , E . faecalis performs EET using matrix-associated iron resulting in both increased final cell biomass and intracellular iron ( Keogh et al . , 2018 ) . Moreover , unlike L . plantarum and L . lactis , the presence of PplA is not necessary for anode reduction or Fe3+ reduction ( Hederstedt et al . , 2020 ) . The conditional need for PplA in EET may be explained by the different prior growth conditions used and/or related to the existence of different mechanisms and proteins depending on the redox potential of the extracellular electron acceptor . Other flavin-binding , extracellular reductases amongst Gram-positive organisms , such as FrdA ( acting on fumarate ) have been identified in L . monocytogenes and UrdA ( acting on urocanate ) in Enterococcus rivorum ( Light et al . , 2019 ) . Thus , there may exist a yet unidentified extracellular reductase in L . plantarum and L . lactis required for anode reduction . Thus , our findings elucidate a new pattern of metabolic changes associated with EET . It seems likely that these many mechanisms reflect the ability of EET to alleviate constraints of intracellular redox balance in fermentative metabolism across LAB . Conservation of the FLEET locus among different LAB species supports the premise that this hybrid fermentation with EET provides an important metabolic strategy for these bacteria in their natural habitats . LAB with a complete FLEET locus are homofermentative , thus underscoring the distinct ways homofermentative and heterofermentative LAB have evolved for energy conservation ( Salvetti et al . , 2013 ) . L . plantarum and other LAB with FLEET systems such as L . casei are genetically and metabolically diverse and grow in a variety of nutrient rich environments including dairy and plant foods and the digestive tract ( Cai et al . , 2009; Martino et al . , 2016; Siezen et al . , 2010a ) . Those environments also are rich sources of sources of quinones , flavins , and extracellular electron acceptors such as iron ( Cataldi et al . , 2003; Fenn et al . , 2017; Kim , 2017; Roughead and McCormick , 1990; Walther et al . , 2013 ) . Increased organic acid production and environmental acidification by LAB with this hybrid metabolism would provide an effective mechanism to inhibit competing microorganisms and confer a competitive advantage for growth . The increased ATP relative to biomass generation observed during growth on mannitol might also give sufficient readiness for using this energy later on to outcompete neighboring organisms ( Russell and Cook , 1995 ) . These effects of EET may be particularly important on plant tissues and intestinal environments , wherein LAB tend to be present in low numbers . Besides our observation that L . plantarum performs EET in kale juice , the FLEET pathway is important for intestinal colonization by both L . monocytogenes ( Light et al . , 2018 ) and E . faecalis ( Lam et al . , 2019 ) , and L . plantarum FLEET genes including ndh2 and pplA were highly induced in the small intestine of rhesus macaques ( Golomb et al . , 2016 ) . The hybrid fermentation metabolism of LAB also has technological relevance . For many LAB food fermentations , acidification of the food matrix is required to prevent the growth of undesired microorganisms and result in a more consistent and reproducible product ( Marco et al . , 2021 ) . Starter cultures are frequently selected based on their capacity for rapid growth and acid production ( Bintsis , 2018 ) . In the presence of an anode , exposure of L . plantarum to EET conditions during kale juice fermentation increased the acidification rate . Thus , this shows that EET metabolism is active in complex nutritive environments such as kale leaf tissues that contain other potential electron acceptors besides the anode and diverse electron donors ( glucose , fructose , sucrose ) ( Thavarajah et al . , 2016 ) . This example also shows how electro-fermentation , the technological process by which fermentation is modulated using electrodes , can be used to control food fermentations ( Moscoviz et al . , 2016; Schievano et al . , 2016; Vassilev et al . , 2021 ) . Because L . plantarum also increased fermentation flux when the electrode was available as an electron sink , higher quantities of organic acid flavor compounds were formed . Therefore , by the manipulation of extracellular redox potential , food electro-fermentations may be used to control microbial growth . This would allow the creation of new or altered sensory profiles in fermented foods , such as through altered organic acid production and metabolism or synthesis of other compounds that alter food flavors , aromas , and textures . We expect that our study will improve the current understanding of energy conservation in primarily fermentative microorganisms and contribute to establishing the ecological relevance of EET in lactic acid bacteria . This work will ultimately allow the use of EET to electronically modulate the flavor and textural profiles of fermented foods and expand the use of lactic acid bacteria in bioelectronics , biomedicine , and bioenergy ( Moscoviz et al . , 2016 ) . The identification of the precise components and full bioenergetics involved in L . plantarum EET will be key to unravel physiological and ecological questions and to develop other biotechnological applications .
All strains and plasmids used in this study are listed in Supplementary file 3 . Standard laboratory culture medium was used for routine growth of bacteria as follows: Lactiplantibacillus spp . , Lacticaseibacillus spp . , Levilactobacillus brevis , Ligilactobacillus murinus , and Pediococcus pentosaceus , MRS ( BD , Franklin Lakes , NJ , USA ) ; Lactococcus lactis and Streptococcus agalactiae , M17 ( BD ) with 2% w/v glucose; Enterococcus faecalis , and Enterococcus faecium , BHI ( BD ) ; and Escherichia coli , LB ( Teknova , Hollister , CA , USA ) . Bacterial strains were incubated without shaking except for E . coli ( 250 RPM ) and at either 30 or 37 °C . Where indicated , strains were grown in filter-sterilized MRS ( De MAN et al . , 1960 ) lacking beef extract with either 110 mM glucose [gMRS] or 110 mM mannitol [mMRS] , or a chemically defined minimal medium ( Supplementary file 4 ) with 125 mM glucose [gCDM] or 125 mM mannitol [mCDM] for 18 hr ( Aumiller et al . , 2021 ) . Riboflavin ( 1 mg/L ) was routinely added to the CDM . When indicated , culture medium was supplemented with 20 μg/mL of the quinone 1 , 4-dihydroxy-2-naphthoic acid ( DHNA ) ( Alfa Aesar , Haverhill , MA , USA ) , 1 . 25 mM ferric ammonium citrate ( C6H8FeNO7 ) ( 1 . 25 mM ) ( VWR , Radnor , PA , USA ) , riboflavin ( Sigma-Aldrich , St . Louis , MO , USA ) , or 5 μg/mL erythromycin ( VWR ) . The FLEET gene locus of L . plantarum NCIMB8826 was identified using NCBI basic local alignment search tool ( BLAST ) ( McGinnis and Madden , 2004 ) using the L . monocytogenes 10403S FLEET genes ( lmo2634 to lmo2641 ) as a reference . L . plantarum genes were annotated based on predicted functions within the FLEET pathwa ( Light et al . , 2018 ) . FLEET locus genes were identified in other LAB by examining 1 , 788 complete Lactobacillales genomes available at NCBI ( downloaded 02/25/2021 ) . A local BLAST ( ver 2 . 10 . 1 ) database containing these genomes was queried using tBLASTx with NCIMB8826 FLEET genes a reference . A gene was considered to be present in the Lactobacillales strain genome if the Bit-score was >50 and the E-value was <10–3 ( Pearson , 2013 ) . Heatmaps showing the percentage of strains in Lactobacillales genera and the Lactobacillus-genus complex ( Zheng et al . , 2020 ) identified to contain individual FLEET genes were visualized using the R-studio package ggplot2 ( Wickham , 2011 ) with clustering done through UPGMA . The FLEET loci of L . plantarum strain 8 . 1 and NCIMB700965 were aligned to the NCIMB8826 genome in MegAlign Pro ( DNAstar Inc , Madison , WI , USA ) . Cells were collected by centrifugation at 10 , 000 g for 3 min , washed twice in phosphate-buffered saline ( PBS ) , pH 7 . 2 ( http://cshprotocols . cshlp . org ) , and adjusted to an optical density ( OD ) at 600 nm ( OD600nm ) of 2 in the presence of 2 . 2 mM ferrihydrite ( Schwertmann and Fischer , 1973; Stookey , 2002 ) and 2 mM ferrozine ( Sigma-Aldrich ) . Where indicated , 55 mM glucose or mannitol , 20 μg/mL DHNA , and riboflavin were added . After 3 hr incubation at 30 °C , the cells were collected by centrifugation at 10 , 000 g for 5 min and the absorbance of the supernatant was measured at 562 nm with a Synergy 2 spectrophotometer ( BioTek , Winooski , VT , USA ) . Quantities of ferrihydrite reduced were determined using a standard curve containing a 2-fold range of FeSO4 ( Sigma-Aldrich ) ( 0 . 25 mM to 0 . 016 mM ) and 2 mM ferrozine . The FeSO4 was dissolved in 10 mM cysteine-HCl ( RPI , Mount Prospect , IL , USA ) to prevent environmental re-oxidation of Fe2+ to Fe3+ in the standard curve . For testing iron reduction activity of cells with a DHNA concentration of 0 . 01 μg/mL in the medium , iron ( III ) oxide nanoparticles < 50 nm ( Sigma-Aldrich ) were used as insoluble iron form ( Figure 1—figure supplement 3 ) . L . plantarum NCIMB8826 ndh2 , pplA , and narGHIJ deletion mutants were constructed by double-crossover homologous recombination with the suicide plasmid pRV300 ( Leloup et al . , 1997 ) . For mutant construction , upstream and downstream flanking regions of these genes were amplified using the A/B and C/D primers , respectively , listed in Supplementary file 5 . Splicing-by-overlap extension ( SOEing ) PCR was used to combine PCR products as previously described ( Heckman and Pease , 2007 ) . PCR products were digested with restriction enzymes EcoRI , SacI , SacII , or SalI ( New England Biolabs , Ipswich , MA , USA ) for plasmid ligation and transformation into E . coli DH5α . The resulting plasmids were then introduced to L . plantarum NCIMB8826 by electroporation . Erythromycin-resistant mutants were selected and confirmed for plasmid integration by PCR ( see Supplementary file 5 for primer sequences ) . Subsequently , deletion mutants were identified by a loss of resistance to erythromycin , PCR ( see Supplementary file 5 for primer sequences ) confirmation , and DNA sequencing ( http://dnaseq . ucdavis . edu ) . L . plantarum NCIMB8826 strains were grown overnight ( approximately 16–18 hr ) from glycerol stocks in MRS . Cells were harvested by centrifugation ( 5200 g , 12 min , 4 °C ) and washed twice in PBS . When L . plantarum wild-type EET activity versus the Δndh2 ( MLES100 ) and ΔpplA ( MLES101 ) deletion mutants was compared , cells were grown as described and the number of cells was normalized across the three strains prior to inoculation in the BES . The bioreactors consisted of double-chamber electrochemical cells ( Adams & Chittenden , Berkeley , CA ) ( Figure 1B ) with a cation exchange membrane ( CMI-7000 , Membranes International , Ringwood , NJ ) that separated them . A three-electrode configuration was used consisting of an Ag/AgCl sat KCl reference electrode ( BASI , IN , USA ) , a titanium wire counter electrode , and a 6 . 35-mm-thick graphite felt working electrode ( anode ) of 4 × 4 cm ( Alfa Aesar , MA , USA ) with a piece of Ti wire threaded from bottom to top as a current collector and connection to the potentiostat . We used a Bio-Logic Science Instruments ( TN , USA ) potentiostat model VSP-300 for performing the electrochemical measurements ( chronoamperometry ) . The bioreactors were sterilized by filling them with ddH2O and autoclaving at 121 °C for 30 min . The water was then removed and replaced with 150 mL of filter sterilized mMRS or mCDM media for the working electrode chamber , and 150 mL of M9 medium ( 6 . 78 g/L Na₂HPO₄ , 3 g/L KH2PO4 , 0 . 5 g/L NaCl , 1 g/L NH4Cl ) ( BD ) for the counter electrode chamber . Both media of the working electrode chamber were supplemented with 20 μg/mL DHNA or 0 . 01 μg/mL diluted 1:1 in DMSO:ddH2O where appropriate . To test the role of bd-cytochrome , heme was added in a final concentration of 10 μg/mL ( diluted 1:1 in DMSO: ddH2O ) . The medium in the working electrode chamber was continuously mixed with a magnetic stir bar and N2 gas was purged to maintain anaerobic conditions for the course of the experiment . The applied potential to the working electrode was of +0 . 2 V versus Ag/AgCl ( sat . KCl ) ( BASI , IN , USA ) . Reactors run under OC conditions were similarly assembled but kept at open circuit and used as control for non-current circulating conditions . Once the current stabilized , the electrochemical cells were inoculated to a final OD600 of 0 . 12–0 . 15 with the cell suspensions prepared in PBS . Current densities are reported as a function of the geometric surface area of the electrode ( 16 cm2 ) . The bioreactors were sampled by taking samples under sterilized conditions at different time points for subsequent analysis . The samples for organic acids analyses were centrifuged ( 15 , 228 g , 7 min ) and the supernatant was separated for High-Performance Liquid Chromatography ( HPLC ) assessments . Samples for ATP and NAD+/NADH analyses were flash frozen in a dry ice/ethanol bath . Organic acids , ethanol , and sugar concentrations were measured by HPLC ( Agilent , 1260 Infinity ) , using a standard analytical system ( Shimadzu , Kyoto , Japan ) equipped with an Aminex Organic Acid Analysis column ( Bio-Rad , HPX-87H 300 × 7 . 8 mm ) heated at 60 °C . The eluent was 5 mM of sulfuric acid , used at a flow rate of 0 . 6 mL min–1 . We used a refractive index detector 1260 Infinity II RID . A five-point calibration curve based on peak area was generated and used to calculate concentrations in the unknown samples . The following standards were included in the HPLC measurements: acetate , formate , pyruvate , malate , lactate , succinate , oxalacetate , fumarate , ethanol , acetoin , butanediol , mannitol , and glucose . No gaseous products were measured . Bioreactors were shaken to remove the cells attached to the working electrode and afterwards sampled to measure viable cells ( colony forming units [CFUs] ) and total biomass ( dry weight ) . Samples for CFU enumeration were collected under sterile conditions at the time of inoculation and at the time of approximately maximum current density . Samples were serially diluted ( 1:1000 to 1:1000000 ) in sterile PBS and plated on MRS for CFUs enumeration after overnight incubation at 30 °C . Dry weight was determined using a 25 mL sample collected at approximately maximum current density . Cells were harvested by centrifugation ( 5250 g , 12 min , 4 °C ) and washed twice in 50 mL ddH2O . Afterwards cells were resuspended in 1 mL of ddH2O and transferred to microfuge tubes ( previously weighted ) . Cells were harvested by centrifugation ( 5250 g , 12 min , 4 °C ) , and the tubes were then transferred to an evaporator to remove humidity . The microfuge tubes were then cooled in a desiccator for 30 min and the weight of each tube was measured to determine cell weight . The difference between the weight of each tube with the pellet and before containing it allowed us to determine the dry weight/mL . L . plantarum NCIMB8826 was grown in triplicate to exponential phase ( OD600 1 . 0 ) at 37 °C in mMRS with or without the supplementation of 20 μg/mL DHNA and 1 . 25 mM ferric ammonium citrate . Cells were collected by centrifugation at 10 , 000 g for 3 min at 4 °C , flash frozen in liquid N2 and stored at –80 °C prior to RNA extraction as previously described ( Golomb et al . , 2016 ) . Briefly , frozen cell pellets were resuspended in cold acidic phenol:chloroform:isoamyl alcohol ( pH 4 . 5 ) [125:24:1] ( Invitrogen , Carlsbad , CA , USA ) before transferring to 2 mL screw cap tubes containing buffer ( 200 mM NaCl , 20 mM EDTA ) , 20% SDS , and 300 mg 0 . 1 mm zirconia/silica beads . RNA was extracted by mechanical lysis with an MP Fastprep bead beater ( MP Biomedicals , Santa Ana , CA , USA ) at 6 . 5 m/s for 1 min . The tubes were centrifuged at 20 , 000 g at 4 °C for 3 min and the upper aqueous phase was transferred to a new tube . The aqueous phase was extracted twice with chloroform:isoamyl alcohol [24:1] ( Fisher Scientific , Waltham , MA , USA ) , The aqueous phase was then transferred to a new tube for RNA ethanol precipitation ( Green and Sambrook , 2020 ) . RNA was then quantified on a Nanodrop 2000c ( ThermoFisher ) , followed by double DNAse digestion with the Turbo DNA-free Kit ( Invitrogen ) according to the manufacturer’s protocols . The quality of the remaining RNA was checked using a Bioanalyzer RNA 6000 Nano Kit ( Agilent Technologies , Santa Clara , CA , USA ) ( all RIN values > 9 ) and then quantified with the Qubit 2 . 0 RNA HS Assay ( Life Technologies , Carlsbad , CA , USA ) . For reverse-transcription PCR ( RT-PCR ) , 800 ng RNA was converted to cDNA with the High Capacity cDNA Reverse Transcription Kit ( Applied Biosystems , Foster City , CA , USA ) according to the manufacturer’s protocols . Quantitative RT-PCR was performed on a 7 , 500 Fast Real-Time PCR System ( Applied Biosystems ) using the PowerUp SYBR Green Master Mix ( ThermoFisher ) and RT-PCR primers listed in Supplementary file 5 . The 2-ΔΔCt method was used for relative transcript quantification using rpoB as a control ( Livak and Schmittgen , 2001 ) . For sequencing , ribosomal-RNA ( rRNA ) was depleted from 4 μg RNA using the RiboMinus Eukaryote Kit v2 with specific probes for prokaryotic rRNA ( ThermoFisher ) following the manufacturer’s instructions . RNA was then fragmented to approximately 200 bp , converted to cDNA , and barcoded using the NEBnext Ultra-directional RNA Library Kit for Illumina ( New England Biolabs , Ipswitch , MA , USA ) with NEBnext Multiplex Oligos for Illumina ( Primer Set 1 ) ( New England Biolabs ) following the manufacturer’s protocols . cDNA libraries containing pooled barcoded samples was run across two lanes of a HiSeq400 ( Illumina , San Diego , CA , USA ) on two separate runs for 150 bp paired-end reads ( http://dnatech . genomecenter . ucdavis . edu/ ) . An average of 36 , 468 , 428 raw paired-end reads per sample was collected ( Supplementary file 6 ) . The DNA sequences were quality filtered for each of the 12 samples by first visualizing with FastQC ( ver . 0 . 11 . 8 ) ( Andrews , 2010 ) to check for appropriate trimming lengths , followed by quality filtering with Trimmomatic ( ver . 0 . 39 ) ( Bolger et al . , 2014 ) . Remaining reads then were aligned to the NCIMB8826 chromosome and plasmids using Bowtie2 ( ver . 2 . 3 . 5 ) in the [-sensitive] mode ( Langmead and Salzberg , 2012 ) . The resulting ‘ . sam’ files containing aligned reads from Bowtie2 were converted to ‘ . bam’ files with Samtools ( ver 1 . 9 ) ( Li et al . , 2009 ) before counting aligned reads with FeatureCounts in the [-- stranded = reverse] mode ( ver . 1 . 6 . 4 ) ( Liao et al . , 2014 ) . Reads aligning to noncoding sequences ( e . g . rRNA , tRNA , trRNA , etc . ) were excluded for subsequent analyses . Differential gene expression based on culture condition was determined with DESeq2 ( Love et al . , 2014 ) using the Wald test in the R-studio shiny app DEBrowser ( ver 1 . 14 . 2 ) ( Kucukural et al . , 2019 ) . Differential expression was considered significant with a False-discovery-rate ( FDR ) -adjusted p-value < 0 . 05 and a Log2 ( fold-change ) >0 . 5 . Clusters of Orthologous Groups ( COGs ) were assigned to genes based on matches from the eggNOG ( ver . 5 . 0 ) database ( Huerta-Cepas et al . , 2019 ) . Hamilton oxidation-reduction potential ( ORP ) probes ( Hamilton Company , Reno , NV , USA ) were inserted into air-tight Pyrex ( Corning Inc , Corning , NY , USA ) bottles containing mMRS supplemented with 20 μg/mL DHNA and/or 1 . 25 mM ferric ammonium citrate and incubated in a water bath at 37 °C . A custom cap for the Pyrex bottles was 3D printed with polylactic acid filament ( 2 . 85 mm diameter ) such that the ORP probe threads into the cap and an o-ring seal can be used to provide an air-tight seal between the probe and the cap . The ORP was allowed to equilibrate over 40 min before L . plantarum NCIMB8826 , Δndh2 ( MLES100 ) , or ΔpplA ( MLES101 ) were inoculated at an OD600 of 0 . 10 . Two uninoculated controls were used to measure baseline ORP over time . The ORP data was collected via Modbus TCP/IP protocol ( Stride Modbus Gateway , AutomationDirect , Cumming , GA , USA ) into a database ( OSIsoft , San Leandro , CA , USA ) and analyzed in MATLAB ( Mathworks , Nantick , MA , USA ) . pH was measured using a Mettler Toledo SevenEasy pH meter ( Mettler Toledo , Columbus , OH , USA ) . Cells were collected at either 24 hr or at the greatest ORP difference between the wild-type and mutant strains ( ΔmVmax ) by centrifugation at 10 , 000 g for 3 min and used for ferrihydrite reduction analyses . Frozen cell pellets were suspended in PBS and lysed by mechanical agitation in a FastPrep 24 ( MP Biomedicals ) at 6 . 5 m/s for 1 min . The cell lysates were then centrifuged at 20 , 000 g for 3 min at 4 °C . ATP and NAD+ and NADH in the supernatants were then quantified with the Molecular Probes ATP Quantification Kit ( ThermoFisher ) and the Promega NAD/NADH-Glo Kit ( Promega , Madison , WI , USA ) , respectively according to the manufacturers’ instructions . L . plantarum was inoculated in mMRS with or without 20 μg/mL DHNA and 1 . 25 mM ferric ammonium citrate at an OD600 of 0 . 10 for 3 . 5 hr . Cells were then collected by centrifugation at 10 , 000 x g for 3 min and washed twice in PBS to remove cell-surface-associated metals . Viable cell numbers were enumerated by plating serial dilutions on MRS laboratory culture medium and the resulting cell materials were digested by incubating at 95 °C for 45 min in a 60% concentrated trace metal grade HNO3 , allowed to cool , then diluted with MilliQ water to a final concentration of 6% HNO3 . The contents were quantified with internal standards with an Agilent 7 , 500Ce ICP-MS ( Agilent Technologies , Palo Alto , CA ) for simultaneous determination of select metals ( Na , Mg , Al , K , Ca , Cu , Zn , Ba , Mn , Fe ) at the UC Davis Interdisciplinary Center for Plasma Mass Spectrometry ( http://icpms . ucdavis . edu/ ) . Green organic kale purchased from a market ( Whole Foods ) was washed with tap water and air dried for 1 hr as previously recommended ( Kim , 2017 ) . A total of 385 g of the leaves and stems were shredded with an electric food processor in 1 L ddH20 . The kale juice was then diluted with 0 . 35 L ddH2O and autoclaved ( 121 °C , 15 min ) . The juice was then centrifuged under sterile conditions at 8000 rpm for 20 min and the supernatant was collected . A rifampicin-resistant variant of L . plantarum NCIMB8826-R ( Tachon et al . , 2014 ) ( grown for 19 hr in MRS medium at 37 °C , 50 μg Rif/mL ) was inoculated to an estimated final OD of approximately 0 . 05 , and DHNA ( 20 μg/mL ) was added where appropriate . Cells were collected and washed as previously described for the bioelectrochemical assays in mCDM . The anodic chambers of bioreactors assembled as previously described ( anode of 4 . 3*6 cm ) were filled with 125 mL of the inoculated kale juice and incubated at 30 °C purged with N2 . After 1 hr , the anodes were polarized to 0 . 2 V versus Ag/AgCl ( sat . KCl ) ( EET conditions ) or kept at open circuit ( OC , no EET ) . Viable cells were measured by plating 10-fold serial dilutions in MRS agar plates with 50 μg/mL of Rif . The total electrons harvested on the anode were estimated by integrating the area ( charge ) under the chronoamperometric curve ( current response ( A ) over time ( s ) ) , which was corrected by subtracting the current baseline obtained before L . plantarum was added to the system . This obtained charge was then converted to mol of electrons using the Faraday constant ( 96 , 485 . 3 A*s/mol electrons ) . L . plantarum RNA-seq data are available in the NCBI Sequence Read Archive ( SRA ) under BioProject accession no . PRJNA717240 . A list of the completed Lactobacillales genomes used in the DNA sequence analysis is available in the Harvard Dataverse repository at https://doi . org/107910/DVN/IHKI0C . | Bacteria produce the energy they need to live through two processes , respiration and fermentation . While respiration is often more energetically efficient , many bacteria rely on fermentation as their sole means of energy production . Respiration normally depends on the presence of small soluble molecules , such as oxygen , that can diffuse inside the cell , but some bacteria can use metals or other insoluble compounds found outside the cell to perform ‘extracellular electron transfer’ . Lactic acid bacteria are a large group of bacteria that have several industrial uses and live in many natural environments . These bacteria survive using fermentation , but they also carry a group of genes needed for extracellular electron transfer . It is unclear whether they use these genes for respiration or if they have a different purpose . Tejedor-Sanz , Stevens et al . used a lactic acid bacterium called Lactiplantibacillus plantarum to study whether and how this group of bacteria use extracellular electron transfer . Analysis of L . plantarum and its effect on its surroundings showed that these bacteria use a hybrid process to produce energy: the cells use aspects of extracellular respiration to increase the yield and efficiency of fermentation . Combining these two approaches may allow L . plantarum to adapt to different environments and grow faster , allowing it to compete against other species . Tejedor-Sanz , Stevens et al . provide new information on a widespread group of bacteria that are often used in food production and industry . The next step will be to understand how the hybrid system is controlled and how it varies among species . Understanding this process could result in new biotechnologies and foods that are healthier , produce less waste , or have different tastes and textures . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"microbiology",
"and",
"infectious",
"disease"
] | 2022 | Extracellular electron transfer increases fermentation in lactic acid bacteria via a hybrid metabolism |
Integrins are transmembrane receptors composed of α and β subunits . Although most integrins contain β1 , canonical activation mechanisms are based on studies of the platelet integrin , αIIbβ3 . Its inactive conformation is characterized by the association of the αIIb transmembrane and cytosolic domain ( TM/CT ) with a tilted β3 TM/CT that leads to activation when disrupted . We show significant structural differences between β1 and β3 TM/CT in bicelles . Moreover , the ‘snorkeling’ lysine at the TM/CT interface of β subunits , previously proposed to regulate αIIbβ3 activation by ion pairing with nearby lipids , plays opposite roles in β1 and β3 integrin function and in neither case is responsible for TM tilt . A range of affinities from almost no interaction to the relatively high avidity that characterizes αIIbβ3 is seen between various α subunits and β1 TM/CTs . The αIIbβ3-based canonical model for the roles of the TM/CT in integrin activation and function clearly does not extend to all mammalian integrins .
Integrins are the principal receptors of cells for the extracellular matrix ( ECM ) and are heterodimeric transmembrane proteins consisting of α and β subunits ( Hynes , 2002; Mathew et al . , 2012a; Pozzi and Zent , 2013 ) . The 18 α and 8 β integrin subunits associate non-covalently to form 24 heterodimeric integrins , which are expressed in different combinations on all cell types throughout the body . All integrin subunits , except β4 , have a large extracellular ligand-binding domain , a single transmembrane domain ( TM ) that spans the cell membrane and a short cytosolic C-terminal domain ( CT ) that regulates integrin function by binding with cytoskeletal and signaling proteins . β1 integrins , in which the β1 subunit can be paired with any one of 12 different α isoforms , are the principal integrins found in solid organs . For integrins to mediate cell adhesion to the ECM and transduce signals from the outside to the inside of the cell and vice-versa they must be able to respond to ligand binding and must also be modulatable in terms of ligand binding affinity . The paradigm for regulation of integrin affinity is set by the platelet-specific integrin αIIbβ3 found in the inactive state under normal conditions and activated to mediate platelet adhesion to fibrinogen following injury ( Xiao et al . , 2004 ) ( Coller and Shattil , 2008 ) . In the low affinity ‘inactive’ state integrin αIIbβ3 has a bent conformation that is stabilized by an outer clasp involving contacts between αIIb and β3 located near the middle and ectoplasmic ends of their TMs and an inner clasp involving αIIb/β3 contacts located in the cytosolic juxtamembrane domains , the latter of which includes a critical salt bridge . Within the inactive state heterodimer , the αIIb transmembrane helix is thought be reasonably well-aligned with the bilayer , while the long TM helix of β3 is thought to have a pronounced ( 25° ) tilt angle that is required to maintain simultaneous formation of the outer and inner clasps ( Lau et al . , 2009; Yang et al . , 2009; Zhu et al . , 2009 ) . It has been proposed that the tilt of the β3 TM in the inactive state is promoted by a ‘snorkeling’ interaction of the side chain ε-NH3+ moiety of the membrane-buried lysine-716 with a negatively charged lipid phosphodiester group located at the adjacent membrane-water interface ( Kim et al . , 2012 ) . Following platelet activation , integrin αIIbβ3 is converted to a high affinity state that is mediated by the binding of intracellular cytoplasmic proteins to the CT of β3 . The best studied of these interactions is the binding of the cytoskeletal protein talin to the β CT . Talin is a key regulator of integrin activation and is comprised of head and rod domains . The head has an atypical FERM ( band 4 . 1 , ezrin , radixin , and moesin ) domain containing four subdomains: F0-F3 . F3 contains a phosphotyrosine-binding ( PTB ) signature that binds to the NPxY motif shared by both integrin β1 and β3 CT , while F0-F2 enhances binding through favorable electrostatic interactions with anionic lipids in the vicinity of the integrin on the membrane surface ( Moore et al . , 2012; Saltel et al . , 2009; Anthis et al . , 2010 ) . This leads to a reduction of the tilt angle of the β3 TM and destabilization of the inner membrane clasp between the β3 and αIIb subunits , promoting dissociation of the transmembrane domains and integrin activation ( Ye et al . , 2014 ) . Whether kindlins , a second FERM domain containing protein family with key roles in integrin activation , also contribute to the TM dissociation is an area of active investigation ( Ye et al . , 2013; Theodosiou et al . , 2016; Moretti et al . , 2013; Lefort et al . , 2012; Montanez et al . , 2008; Moser et al . , 2009a , 2009b , 2008 ) . Despite extensive knowledge about the structural basis of αIIbβ3 integrin activation , data is limited as to whether the widely expressed β1 containing integrins also adopt both inactive and affinity-modulated active states and , if so , how this happens . While there is evidence that the fibronectin α5β1 integrin may resemble αIIbβ3 ( Takagi et al . , 2003; Luo et al . , 2004 ) , it is unclear whether all the 12 possible αβ1 integrin combinations are activated similarly to αIIbβ3 , or whether there is mechanistic heterogeneity underlying their various adhesive functions . That the paradigm of integrin activation established by αIIbβ3 might extend to β1 integrins is suggested by fairly high sequence homology between the β1 and β3 integrin TMs and CTs ( Figure 1 ) . In addition , the putative snorkeling lysine K752 of β1 integrin was shown to regulate activation of integrin α5β1 in the same manner as proposed for K716 of β3 ( Kim et al . , 2012 ) . By contrast , disruption of the inner membrane clasp in β1 integrin did not cause any phenotype in knockin mice , nor did it alter integrin activation as assessed by activation-dependent antibodies or cell adhesion and migration on fibronectin ( Czuchra et al . , 2006 ) . Moreover , contrary to the canonical model , ligand ( i . e . collagen ) binding to the integrins α1β1 and α2β1 has been reported to occur in the absence of integrin activation ( Abair et al . , 2008a; Nissinen et al . , 2012 ) . Thus , our understanding of the role of the TM interactions between α and β1 integrin subunits in regulating integrin activation and function remains incomplete . 10 . 7554/eLife . 18633 . 003Figure 1 . The β1-K752E mutation decreases collecting duct cell adhesion to collagens . ( A ) The sequences of the β1 and β3 TM/CTs are annotated . The highly conserved lysine ( K752 in integrin β1and K716 in integrin β3 ) is colored in red . The β3 TM-only construct used in previous studies ( Kim et al . , 2012 ) ends at F727 , which was colored in blue along with the corresponding residue in β1 , F763 . ( B–C ) The adhesion of CD cells to collagen I ( 0 . 5 μg/ml ) ( B ) or collagen IV ( 0 . 25 μg/ml ) ( C ) for 1 hr was measured . These assays were performed in the presence and absence of blocking antibodies to α1 or α2 respectively . The error bars indicate SD and * indicates a statistically significant difference ( p<0 . 01 ) between WT and K752E CD cells . These assays were performed at least three times . ( D–F ) The spreading of wild type and K752E CD cells on collagen I and collagen IV at 15 , 30 and 45 min was measured on cells stained with rhodamine phalloidin . All the images were taken close to the substrate . Representative cells at 30 min on both collagen I and IV is shown ( D ) . The area of at least 35 cells per time point were quantified using ImageJ software and expressed graphically . Scale bar represents 10 microns ( E and F ) . The error bars indicate SD and * indicates a statistically significant difference ( p<0 . 05 ) between WT and K752E CD cells . These assays were performed at least three times . ( G–H ) . The active conformation of integrin β1 on adherent wild type ( WT ) and K752E expressing CD cells that were allowed to adhere to collagen I for 1 hr was determined using 12 G10 antibody . Total integrin β1 surface expression was determined using AIIB2 antibody . All the images were taken close to the substrate . ( G ) . The relative fluorescence intensity of 12 G10/AIIB2 as described in the Materials and methods is expressed graphically ( H ) . The error bars indicate SD and * indicates a statistically significant difference ( p<0 . 01 ) between WT and K752E CD cells . These experiments were performed at least three times each . Intensity measurements were performed on over 40 different fields for each of the samples . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 003 In this work , we demonstrate that there are significant differences between the β1 and β3 integrin TM/CT structures as well as in the structure-function relationships between the αIIbβ3 and α1β1 and α2β1 integrins . These results challenge the assumption that all integrins are functionally modulated via mechanisms similar to the well-characterized αIIbβ3 integrin .
‘Snorkeling’ of the positively charged side chain of the membrane-buried integrin β3-K716 to form an ion pair with an anionic phosphodiester moiety of a lipid head group at the membrane-water interface is thought to regulate the integrin affinity by stabilizing the inactive state ( Kim et al . , 2012 ) . We wished to assess whether this mechanistic component of integrin regulation is conserved in β1 integrins , specifically for the collagen binding integrins α1β1 and α2β1 . Integrin β1-K752 corresponds to β3-K716 ( Figure 1A ) . We therefore generated populations of integrin β1 null renal collecting duct ( CD ) polarized epithelial cells expressing comparable levels of human integrin β1-WT , β1-K752R and β1-K752E . The CD cells were sorted for equal levels of expression of β1 integrin and endogenous integrin α1 or α2 subunits . The different cell populations were then subjected to adhesion on collagens I or IV , which are the preferred ligands for integrins α2β1 and α1β1 , respectively . The β1-K752R mutation ( which maintained the positive amino acid charge ) did not significantly change CD cell adhesion to collagen I ( 0 . 5 μg/ml ) in the absence or presence of an integrin α1 blocking antibody ( Figure 1B ) . Similar results were found over a concentration range of collagen I from 0 . 125–20 μg/ml ( data not shown ) . However , under the same conditions , CD cells expressing the β1-K752E mutation ( which changed the amino acid charge from positive to negative ) exhibited little adhesion to collagen I in the absence or presence of an integrin α1 blocking antibody ( Figure 1B ) . Comparable results were found for CD cell adhesion to collagen IV ( 0 . 25 μg/ml ) ( Figure 1C ) in the presence or absence of an integrin α2 blocking antibody as well as over a concentration of collagen IV from 0 . 0625–10 μg/ml ( data not shown ) . Due to the severe collagen adhesion defect of CD cells expressing the K752E-β1 mutant , we investigated the effect of this mutation on cell spreading on collagen I and collagen IV over short periods of time . CD cells expressing the β1-K752E mutation spread significantly less on collagen I than CD cells expressing wild type β1 integrin at 15 , 30 and 45 min ( Figure 1D and E ) . Similar results were seen when the cells were plated on collagen IV , although the difference was no longer statistically different at 45 min ( Figure 1D and F ) . These results were completely contrary to expectations based on the activating nature of the K716E-β3 mutation described for αIIbβ3 ( Kim et al . , 2012 ) . We next investigated whether this unexpected major decrease in cell adhesion to the collagens by the β1-K752E mutant was due to alterations in β1 integrin activation as assessed by the amount of the β1 integrin activation epitope-reporting 12G10 antibody binding relative to the total amount of β1 integrin ( as measured by AIIB2 ) in CD cells that adhered to collagen I . The intensity of 12G10/AIIB2 antibody binding was significantly less in the β1-K752E mutant compared to WT CD cells . This difference was no longer present when the cells were adhered in the presence of Mn2+ , which is known to artificially activate integrins ( Figure 1G and H ) . Thus , reversing the basic charge of the snorkeling lysine near the TM domain of the β1 integrin significantly decreased integrin α1β1 and α2β1-dependent adhesion to and spreading on collagens by polarized CD epithelial cells . In addition , it resulted in a lower affinity state of β1 integrin in the context of CD cells adherent on collagen I . We next performed structural studies to probe the basis for the discrepancies between our results and those found when the snorkeling lysine K716 was mutated in β3 . We carried out two different sets of experiments in which the buffer and bicelle model membranes were varied . The first experiment involved 0 . 5 mM ( 0 . 57 mol% ) integrin TM/CT in q = 0 . 3 D6PC/DMPC bicelles and a buffer containing 250 mM imidazole , pH 6 . 5 . DMPC/D6PC bicelles were selected as optimal relative to POPC/D6PC , DMPC/D7PC , and DMPC/cyclofos6 based on comparing the 1H , 15N-TROSY NMR spectra ( Figure 2—figure supplement 1 ) from each composition . D6PC/DMPC bicelles yielded the most optimal β1 TM/CT spectrum in terms of the total number of peaks , evidence for conformational homogeneity , and the relative intensities of peaks from the TM to those of peaks from the cytosolic domain . The second experiment employed bicelle conditions that matched those used in critical prior studies of the isolated TM of the platelet integrin αIIb/β3 heterodimer and constituent monomers ( Kim et al . , 2012; Lau et al . , 2008a; Lau et al . , 2009 ) : 0 . 3 mM ( 0 . 34 mol% ) integrin TM/CT , 20% q = 0 . 3 D6PC/POPC/POPS bicelles ( where POPC:POPS = 2:1 mol:mol ) in a buffer composed of 25 mM HEPES , pH 7 . 4 . For a number of key experiments ( below ) both sets of conditions were employed and usually yielded similar results . TROSY-based 3-D NMR experiments were carried out to assign the backbone amide 1H , amide 15N , 13CA , 13CO , and ( when possible ) 13CB NMR resonances for the integrin β1 TM/CT in bicelles , as illustrated in Figure 2A . While peaks have previously been assigned for the isolated β3 TM ( Lau et al . , 2009 ) , this is not the case for the combined TM/CT in bicelles . We therefore completed assignments for this protein as well ( Figure 2B ) . For both the β1 and β3 TM/CT the assigned chemical shifts have been deposited into the BMRB database ( Ulrich et al . , 2008 ) ( access codes: 26623 and 26624 ) . Using the CA , CO , and CB chemical shifts for the assigned resonances , the secondary structures of both the β1 and β3 integrin TM/CT were determined using chemical shift index and TALOS-N analyses ( Figure 2C and Figure 2—figure supplement 2 ) . For β1 , there is an α-helix extending from the N-terminal I732 at the beginning of the TM to K765 located in the cytosol , roughly eight sites beyond the end of the TM . For the β3 TM/CT , this helix also starts at the beginning of the TM ( at I693 ) , but extends roughly 16 residues into the cytosol , terminating only at A737 . Beyond the TM/CT helix , the remaining intracellular domains of both integrin β1 and integrin β3 are unstructured . The degree to which the TM helix of the β3 integrin extends further into the CT relative to β1 ( nine residues longer for β3 ) is striking . The stability of this helix was probed by examining backbone amide/water hydrogen exchange rates using the CLEANEX-PM NMR experiment ( Hwang et al . , 1998 ) and by measuring backbone amide 15N NMR relaxation rates . Results from these experiments ( Figure 2D and Figure 2—figure supplement 2 ) show that the transmembrane helix is rigid and , not surprisingly , exchange-resistant . The CT segments identified as helical by analysis of the chemical shifts are seen to exhibit a gradient of low exchange/low motion to significantly higher exchange and motion going from the end of the TM to the end of the chemical-shift identified CT helical segments . There is also a modest spike in exchange and motion observed at the TM/CT interface . These data are consistent both with the notion that the CT helices are prone to significant fraying and also that there is a modest degree of hinge motion at the TM/CT helix interface . It is very clear from the topological analysis presented in the following sections that the CT helix is extended away from the membrane into the cytosol and that any hinge motions at the TM/CT interface are not of sufficient magnitude to enable significant interactions of the CT helix with the membrane surface . 10 . 7554/eLife . 18633 . 004Figure 2 . The β1 and β3 transmembrane and cytosolic domains have distinct structures . ( A ) 800 MHz 1H-15N TROSY spectrum of the WT integrin β1 TM/CT with peak assignments shown . This spectrum was collected at 45°C and the sample contains 500 µM β1 TM/CT ( 0 . 57 mol% ) , 20% q = 0 . 3 D6PC-DMPC bicelles , 1 mM EDTA , 250 mM IMD at pH 6 . 5% and 10% D2O . ( B ) 900MHz 1H-15N TROSY spectrum of the WT integrin β3 TM/CT with peak assignments shown . This spectrum was collected at 45°C and the sample contained 300 µM β3 TM/CT , 20% q = 0 . 3 D6PC-DMPC bicelles , 1 mM EDTA , 250 mM IMD buffer at pH 6 . 5 , with 10% D2O . ( C ) Structural comparison of integrin β1 and β3 TM/CT in D6PC/DMPC bicelles . These structural models are based on backbone dihedral angles as determined by TALOS-N and chemical shift index analysis of backbone NMR chemical shift values . The figure was made in PYMOL . As described in the results section , the CT segments of the helices observed for both β integrins appear to be subject to significant dynamic fraying . ( D ) Assessment of exchange of protons between backbone amide sites and water at 100 msec , as determined by the CLEANEX-PM NMR experiment . Sites with low peak intensities after the 100 msec mixing period relative to control conditions ( I/I0 ) are resistant to hydrogen exchange , while sites with high I/I0 values exchange rapidly on the time scale of 100 msec . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 00410 . 7554/eLife . 18633 . 005Figure 2—figure supplement 1 . 900 MHz 1H , 15N-TROSY spectra of integrin β1 TM/CT . 900 MHz 1H , 15N-TROSY spectra of integrin β1 TM/CTat 45o C in four different bicelle mixtures: ( A ) Cyclofos-6/DMPC , ( B ) D7PC/DMPC , ( C ) D6PC/POPC , ( D ) D6PC/DMPC . The NMR samples contain ~500 µM β1 TM/CT , 20% q = 0 . 3 bicelles , 1 mM EDTA , 250 mM IMD , at pH 6 . 5 , with 10% D2O . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 00510 . 7554/eLife . 18633 . 006Figure 2—figure supplement 2 . NMR chemical shifts to determine alpha helical content of the integrin β1 and β3 TM/CT . ( A ) TALOS-N analysis of backbone 13C NMR chemical shifts to determine α-helical content of the integrin β1 and β3 TM/CT . Sample conditions were as described in the captions for Figure 2A and B . ( B ) 600 MHz backbone 15N NMR transverse relation rates ( R2 ) for the integrin β1 and β3 TM/CT . Sample conditions were as described in the captions for Figure 2A and B . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 006 The membrane topologies of the β1 and β3 TM/CT were probed using NMR spectroscopy in the presence of hydrophilic ( Gd ( III ) -DTPA ) and hydrophobic ( 16-DSA ) paramagnetic probes . Probe accessibility to backbone amide 1H sites was assessed as the degree of site-specific backbone amide TROSY peak broadening due to the proximity of the paramagnetic probe , which was quantitated as the ratio between the peak intensity in a probe-containing sample versus the corresponding peak intensity from a matched diamagnetic ( probe-free ) control sample . A low ratio corresponds to high probe access . We first investigated the membrane topology of WT integrin β1 TM/CT in DMPC/D6PC bicelles at pH 6 . 5 by comparing peak intensities ( Figure 3 ) in NMR spectra acquired in the absence ( black peaks ) and presence ( red peaks ) of either water soluble Gd ( III ) -DTPA or lipophilic 16-DSA paramagnetic probes . The plateau in peak ratios observed when Gd ( III ) -DTPA was used , combined with the corresponding trough in the ratio when 16-DSA was used indicates that the beginning and ends of the TM are near sites I732 and I757 , respectively . Both the extracellular segment and the cytosolic domains are largely exposed to the hydrophilic Gd ( III ) -DTPA ( Figure 3A and C ) , although the modest degree of protection to Gd ( III ) -DTPA near the P781IY segment ( see spike in right side of Figure 3C ) suggests that this segment undergoes transient interactions with the bicelle surface . Importantly , these results show that , similar to integrin β3 , the β1 TM includes both the putative ‘snorkeling’ Lys site ( 752 in β1 , 716 in β3 ) and the five hydrophobic residues following it . The β1 TM is terminated by the HDRRE762 motif , where D759 corresponds to D723 in β3 , which forms a critical salt bridge with R995 in the αIIb subunit as part of the ‘inner clasp’ that stabilizes the inactive integrin heterodimeric state ( Lau et al . , 2009 ) . 10 . 7554/eLife . 18633 . 007Figure 3 . Examination of the bilayer topology of integrin WT β1 TM/CT and two mutants ( K752E , K752R ) in bicelles using NMR and paramagnetic probes . Peak intensity changes in the 600MHz 1H-15N TROSY spectra are reported for the integrin WT β1 TM/CT and two mutants ( K752E , K752R ) as induced either by 4 mol% 16-DSA as the hydrophobic paramagnet or by 10 mM Gd-DTPA as the water soluble paramagnet . The spectra were collected at 45°C and the samples contained ~500 µM β1 TM/CT , 20% q = 0 . 3 D6PC-DMPC bicelles , 1 mM EDTA , 250 mM IMD buffer at pH 6 . 5 , with10% D2O . The overlaid spectra for WT β1 TM/CT containing no paramagnetic probe ( black , bottom ) and with Gd-DTPA ( red , on top ) are shown ( A ) , while the corresponding spectra for the two mutants are shown in supporting Figure 3 . The absence of a red peak indicates that the peak is broadened beyond detection by the presence of the paramagnetic probe . The overlay of spectra for WT β1 TM/CT containing 16-DSA ( red ) versus no paramagnetic probe ( black ) are shown ( B ) . Panel C shows the peak intensity changes induced by 10 mM Gd-DTPA on the three forms of the integrin β1 TM/CT , while panel D shows the results for 16-DSA . The intensity ratio ( I/I0 ) reported for each TROSY amide resonance is the observed peak intensity in the presence of the paramagnetic probe divided by the intensity of the corresponding peak under probe-free diamagnetic conditions . Two replicates of this experiment giving consistent results were performed . We show one representative example . The error is estimated from the noise level , or 5% of value , whichever is larger . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 00710 . 7554/eLife . 18633 . 008Figure 3—figure supplement 1 . 1H , 15N-TROSY spectral overlays of integrin β1 TM/CT in D6PC/DMPC bicelle with no paramagnetic probe ( black ) and with either 10 mM Gd-DTPA ( red ) or 4 mol% 16-DSA ( red ) . Panels A–B are for integrin β1 K752E and panels C–D are for the K752R mutant form . The NMR samples contain ~500 µM β1 TM/CT , 20% q = 0 . 3 D6PC/DMPC bicelles , 1 mM EDTA 250 mM IMD , at pH 6 . 5 , with 10% D2O . All the spectra were collected with a 600MHz NMR spectrometer at 45°C . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 008 Probe accessibility measurements were repeated for the TM/CT of both the β1-K752E and β1-K752R mutants . Surprisingly , both mutants yielded almost exactly the same patterns of accessibility to both paramagnetic probes as WT ( Figure 3A and C , Figure 3—figure supplement 1 ) , indicating that both charge-conservative and charge reversal mutations at K752 induce little change in the membrane topology or tilt of the TM helix of β1 TM/CT . The same results were obtained when the lipophilic paramagnetic probe 16-DSA was used ( Figure 3B and D , Figure 3—figure supplement 1 ) . The fact that the β1-K752E mutation did not impact the membrane topology/tilt for the β1 subunit was surprising in light of a previous study that presented Mn ( II ) -EDDA ( polar ) probe accessibility data showing that the β3-K716E mutation induced a major change in membrane topology , at least for the isolated integrin β3 TM ( Kim et al . , 2012 ) . Specifically , a pronounced tilt of the TM for the WT β3 protein was reported to be largely reversed by the K->E mutation to yield a TM that is much more closely aligned with the bilayer normal . It was proposed in that work that the tilt of the β3 TM is stabilized by a ‘snorkeling’ interaction between the side chain amino moiety of the membrane-buried K716 site and oxyanions in the nearby lipid head groups ( Kim et al . , 2012 ) . Because of the surprising discord between the β1 results of this work versus the prior β3 results we repeated the Mn ( II ) -EDDA probe experiment for the isolated β3 TM using the same methods as originally used by Kim , Schmidt , et al ( Kim et al . , 2012 ) . As shown in Figure 4A and Figure 4—figure supplement 1 , our β3-WT vs . β3-K716E results are very similar to the previously reported results . Use of the hydrophilic Gd-DTPA probe led to a similar result as for Mn-EDDA ( Figure 4B and Figure 4—figure supplement 1 ) , yielding results consistent with a major change in tilt for the TM domain due to the β3-K716E mutation . However , a discordant result was obtained when the lipophilic 16-DSA was used as the paramagnetic probe ( note that an apolar probe was not employed in the previous study ( Kim et al . , 2012 ) of the isolated β3 TM ) : for 16-DSA , little difference was observed between the membrane topology and tilt of β3-WT and the β3-K716E mutant ( Figure 4C and Figure 4—figure supplement 1 ) . This led us to recall literature showing that chelate probes such as Mn ( II ) -EDDA and Gd ( III ) -DTPA sometimes have an exposed ligand site that can transiently associate with anions such as Glu and Asp side chain carboxylates ( Hocking et al . , 2013 ) . To test this possibility , we repeated the Gd ( III ) -DTPA probe experiment under conditions in which excess ( 10 mM ) free EDTA was added to the solution to cap any free ligand sites in Gd ( III ) -DTPA , thereby suppressing any direct binding of protein carboxyl sides to the open ligand site in the lanthanide ion chelate . Under these conditions the β3-WT and β3-K716E mutant exhibited the same probe accessibility patterns ( Figure 4D and Figure 4—figure supplement 1 ) , fully consistent with the 16-DSA results . These results indicate that the previously reported differences seen for the isolated β3 TM using Mn ( II ) -EDDA reflected an experimental artifact based on the tendency of free carboxyl groups in proteins ( i . e . , the introduced Glu716 side chain carboxyl ) to transiently serve as ligands to available metal ion coordination sites in chelate complexes . 10 . 7554/eLife . 18633 . 009Figure 4 . Examination of the bilayer topology of isolated WT integrin β3 TM and of β3 K716E TM in bicelles using NMR and paramagnetic probes . Paramagnetic probe-induced intensity changes are reported for the peaks in 900 MHz 1H-15N TROSY spectra of WT integrin β3 TM and β3 K716E TM at 45°C: ( A ) 1 mM Mn-EDDA , ( B ) 10 mM Gd-DTPA , ( C ) 10 mM Gd-DTPA plus 10 mM EDTA , ( D ) 4 mol % 16-DSA . The NMR samples contained 0 . 3 mM protein , 20% D6PC/POPC/POPS bicelles ( q = 0 . 3 , 2:1 POPC:POPS ) , 25 mM HEPES at pH 7 . 4 , with 10% D2O . The intensity ratio ( I/I0 ) reported for each TROSY amide resonance is the observed peak intensity in the presence of the paramagnetic probe divided by the intensity of the corresponding peak under probe-free diamagnetic conditions . Three replicates of these experiments giving consistent results were performed . We show one representative example . The error is estimated from noise level , or 5% of value , whichever is larger . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 00910 . 7554/eLife . 18633 . 010Figure 4—figure supplement 1 . 900 MHz 1H , 15N-TROSY spectral overlays of WT integrin β3 TM in D6PC/POPC/POPS ( 2:1 ) bicelles with no paramagnetic probe ( black ) and with either 10 mM Gd-DTPA ( red ) ( A ) , 1 mM Mn-EDDA ( red ) ( B ) , both 10 mM Gd-DTPA and 10 mM EDTA ( red ) ( E ) , or with 4 mol% 16-DSA ( red ) ( F ) . Panels C–D and G–H are the corresponding overlays for the integrin β3 K716E TM-only . The samples contained ~300 µM β3 TM , 20% q = 0 . 3 D6PC/POPC/POPS ( 2:1 ) bicelles , 1 mM EDTA 250 mM IMD at pH 7 . 4 , with 10% D2O . All the spectra were collected at 900MHz at 35°C . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 010 We used both additional paramagnetic probe experiments and backbone amide 15N NMR R2 relaxation rate measurements to confirm that the similarity of the membrane topology/tilt of β3-WT vs . β3-K716E is maintained when the full cytosolic domain is included ( TM/CT , Figure 5A–5D and Figure 5—figure supplement 1 ) . We also verified that these results were not altered by major changes in bicelle or buffer composition , pH , or temperature ( Figure 5A and B vs . 5C and 5D , also Figure 5—figure supplement 2 ) . Finally , the data of Figure 5E and F and Figure 5—figure supplements 2 and 3 complements that of Figure 3C and D by showing that membrane tilt/topology for both β1-WT and β1-K752E TM/CT does not strongly depend on lipid composition , buffer type , pH , or temperature . 10 . 7554/eLife . 18633 . 011Figure 5 . Examination of the bilayer topology of WT integrins in bicelles using NMR and paramagnetic probes . Peak intensity changes are reported as induced by 10 mM Gd-DTPA ( A ) or 4% 16-DSA ( B ) in the 900MHz 1H-15N TROSY spectra of WT integrin β3 TM/CT and of the K716E mutant at 45°C in 20% q = 0 . 3 D6PC/DMPC bicelles , 1 mM EDTA , 250 mM IMD pH 6 . 5 , with 10% D2O . ( C ) and ( D ) are the corresponding plots for integrin β3 TM/CT and for the K716E β3 TM/CT mutant in 20% q = 0 . 3 D6PC/POPC/POPS bicelles , 25 mM HEPES pH 7 . 4 . In these cases 10 mM EDTA was also included to sequester any free Gd3+ and to cap any open metal ion ligand sites of the Gd-DTPA complex . ( E ) and ( F ) are the corresponding plots for WT integrin β1 TM/CT and for the K752E mutant form in 20% q = 0 . 3 D6PC/POPC/POPS bicelles , 1 mM EDTA , 25 mM HEPES pH 7 . 4 , with 10% D2O . The intensity ratio ( I/I0 ) reported for each TROSY amide resonance is the observed peak intensity in the presence of the paramagnetic probe divided by the intensity of the corresponding peak under probe-free diamagnetic conditions . At least two replicates of these experiments giving consistent results were performed . We show representative examples . The error is estimated from the noise level , or 5% of value , whichever is larger . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 01110 . 7554/eLife . 18633 . 012Figure 5—figure supplement 1 . TROSY spectra of β3 TM/CT with and without paramagentic probes . 900MHz 1H , 15N-TROSY spectral overlay of integrin β3 TM/CT in D6PC/DMPC bicelles with no paramagnetic probe ( black ) versus 10 mM Gd-DTPA plus 10 mM EDTA ( red ) ( A ) . 900MHz 1H , 15N-TROSY spectral overlay of integrin β3 TM/CT with no paramagnetic probe ( black ) versus conditions with 4 mol% 16-DSA ( red ) ( B ) . ( C ) and ( D ) are the corresponding overlays for the integrin β3 K716E TM/CT . All the spectra were collected at 45°C and the NMR samples contained ~300 µM integrin β3 TM/CT , 20% q = 0 . 3 D6PC/DMPC bicelles , 1 mM EDTA , 25 mM HEPES pH 7 . 4 , with 10% D2O . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 01210 . 7554/eLife . 18633 . 013Figure 5—figure supplement 2 . TROSY spectra of β3 TM/CT with and without paramagentic probes . 900MHz 1H , 15N-TROSY spectral overlay of integrin β3 TM/CT in D6PC/POPC/POPS ( 2:1 ) bicelles with no paramagnetic probe ( black ) versus conditions with 10 mM Gd-DTPA plus 10 mM EDTA ( red ) ( A ) . 900MHz 15N-TROSY spectral overlay of integrin β3 TM/CT with no paramagnetic probe ( black ) versus conditions with 4 mol% 16-DSA ( B ) . ( C ) and ( D ) are the corresponding overlays for the integrin β3 K716 TM/CT . The NMR samples for panels A-D contained ~300 µM integrin β3 TM/CT , 20% q = 0 . 3 D6PC/POPC/POPS ( 2:1 ) bicelles , 25 mM HEPES pH 7 . 4 , with 10% D2O . All the spectra were collected at 35°C . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 01310 . 7554/eLife . 18633 . 014Figure 5—figure supplement 3 . TROSY spectra of of β1 TM/CT with and without paramagentic probes . 900MHz 1H , 15N-TROSY spectral overlay of integrin β1 TM/CT in D6PC/POPC/POPS ( 2:1 ) bicelle with no paramagnetic probe ( black ) versus conditions with10mM Gd-DTPA ( red ) ( A ) . 900MHz 1H , 15N-TROSY spectral overlay of integrin β1 TM/CT with no paramagnetic probe ( black ) versus conditions with 4% 16-DSA ( B ) . ( C ) and ( D ) are the corresponding overlays for the integrin β1 K752E TM/CT . The NMR samples contained ~500 µM integrin β1 TM/CT in 20% q = 0 . 3 D6PC/POPC/POPS ( POPC:POPS 2:1 ) bicelles , 1 mM EDTA , 25 mM HEPES pH 7 . 4 , with 10% D2O . All the spectra were collected at 35°C . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 014 These results indicate that for both the β1 and β3 integrins , the lysine buried in the membrane at the membrane/cytosol interface does not play a major role in dictating its TM topology or tilt , at least not for the free β subunits . This finding does not rule out the possibility of snorkeling ion pairing of the β3-K716 and β1-K752 side chains with the lipid head group , but does rule out the notion that such interactions play a decisive role in determining TM helix tilt . The β3-K716E mutation results in constitutive activation of the αIIbβ3 integrin and α5β1 integrins ( Kim et al . , 2012 ) . To quantitate the impact of the β3-K716E and β1-K752E mutations on heterodimer formation NMR was used to monitor titrations in which WT and mutant forms of 15N-labeled integrin β1 and β3 TM/CT were titrated by various unlabeled WT α TM/CT subunits ( Figure 6A–D and Figure 6—figure supplement 1 ) . Binding of integrin α5 to either the WT or K752E β1 TM/CT and binding of integrin αIIb to either WT or K716E β3 resulted in the disappearance of the β subunit TROSY NMR peaks rather than shifts , indicating ‘slow exchange’ between free and complexed subunits on the NMR time scale ( Figure 6—figure supplement 1 ) . For each of these titrations a 1:1 binding model was fitted to the concentration dependence of the disappearance of multiple peaks using a global fitting , in which the traces from multiple peaks were simultaneously fit . It can be seen that this approach generated reasonably good fits of the data ( Figure 6A–D ) . However , traces for a few peaks , such as that of G744 in Figure 6B , exhibited deviation from ideal 1:1 behavior , most likely because the presence of a second concentration-dependent phenomenon such as non-specific interactions between subunits that also contributed to the observed shifts for such peaks . Eliminating the outliers from the global fitting process did not dramatically change the Kd estimated from the fitting . For example , global fitting of all four traces shown in Figure 6B led to a Kd of 0 . 63 ± 0 . 10 mol% , while a repeat fit minus the G744 outlier data led to a Kd of 0 . 73 ± 0 . 1 mol% . 10 . 7554/eLife . 18633 . 015Figure 6 . Fits of at 1:1 binding model to the data from NMR-monitored titrations of wild type and mutant 15N-labeled integrin β TM/CT subunits by unlabeled wild type α TM/CT subunits , with the best fit Kd determined in each case as shown . The residue assignments for the TROSY peaks used for these analyses are indicated . In some cases ( panels E-H ) the binding was so weak that it was only possible to determine a lower limit to the Kd . Measurements were carried out at 45°C in 20% ( w/v ) D6PC/POPC/POPS ( POPC:POPS = 2:1 ) , q = 0 . 3 , 50 mM phosphate buffer with 1 mM EDTA in 10% D2O , pH 6 . 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 01510 . 7554/eLife . 18633 . 016Figure 6—figure supplement 1 . Superimposed 900 MHz 1H-15N-TROSY spectra from titrations of 15N-labeled integrin β TM/CTs with unlabeled α subunit TM/CTs: α5β1 ( WT ) , α5β1 ( K752E ) αIIbβ3 ( WT ) , and αIIbβ3 ( K716E ) . The concentrations of WT β3 , WT β1 and the K752E β1 mutant were fixed at 0 . 17 mol% ( 150 µM ) for all the samples , while the concentration for the K716E β3 mutant was fixed at 0 . 10% ( 90 µM ) . For these titrations , no peak shifts were observed , but some β1 1H-15N-TROSY peaks disappeared during the course of the titrations , indicating subunit association/dissociation exchange that is slow on the NMR time scale . These spectra were used as the basis for the binding analyses of Figure 6A–D . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 01610 . 7554/eLife . 18633 . 017Figure 6—figure supplement 2 . Superimposed 600 MHz 1H-15N-TROSY spectra from titrations of 15N-integrin β TM/CTs with unlabeled α subunit TM/CTs: α1β1 ( WT ) , α2β1 ( WT ) , α1β1 ( K752E ) , and α2K752Eβ1 ( K752E ) . The concentrations of the WT and K752E mutant integrin β1 subunits were fixed at 0 . 23 mol% ( 200 µM ) for all samples . For these titrations , peak shifts were in some cases observed , indicating subunit association/dissociation exchange that is rapid on the NMR time scale . These spectra were used as the basis for the binding analyses of Figure 6E–H . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 017 Because the β1 integrin subunit pairs with many other α subunits besides α5 we tested for the generality of the above results for α5β1 by examining binding of other α subunit TM/CT , namely α1 and α2 . NMR titrations of wild type and mutant β1 TM/CT with the wild type α1 and α2 subunits were carried out and it was observed that the β1 peaks shifted during the titration instead of disappearing ( Figure 6—figure supplement 2 ) , indicative of rapid exchange on the NMR time between free and complexed species . Global fitting a 1:1 binding model to the concentration dependence of peaks seen to shift the most in each titration yielded good fits ( Figure 6E–H ) but revealed that binding did not approach saturation over the concentration range accessible by NMR , such that only lower limits for the Kd could be determined . It is interesting that , for a given concentration , titration with α1 TM/CT generated significantly larger resonance shifts in certain peaks ( Figure 6E and F ) than α2 ( Figure 6G and H ) . It is also notable the largest shifts seen in the titration of K752E β1 TM/CT by α1 ( Figure 6F ) are significantly larger than the corresponding α1-induced shifts seen for the WT β1 TM/CT . Finally , it was surprising that the largest shifts seen in titrations of both WT and mutant forms of β1 are for peaks from sites on the distal cytosolic end of the CT . However , the exact interpretation of these observations and whether they are truly significant in terms informing on integrin activation is difficult to assess in light of the fact that heterodimerization in all four cases ( Figure 6E–H ) was too weak to quantify . That the NMR titration results of Figure 6 report primarily on the formation of stoichiometric complex formation between α and β TM/CT subunits is supported by the results of corresponding fluorescent anisotropy titrations in which fluorescently-labeled α TM/CT subunits were titrated by unlabeled β TM/CT subunits , followed by fitting of the data by a 1:1 binding model ( Figure 7 ) . The various Kd determined by both NMR and fluorescence methods are presented in Table 1 , where it is seen that for each subunit combination the NMR titrations conducted at pH 6 . 5°C and 45°C and fluorescence anisotropy titrations conducted at pH 7 . 4°C and 35°C yielded similar results . 10 . 7554/eLife . 18633 . 018Table 1 . Dissociation constant ( Kd ) of integrin β1 and β3 subunits ( WT and K752E mutant ) with α subunits in D6PC/POPC/POPS ( POPC:POPS=2:1 ) q=0 . 3 bicelles . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 018TitrationKd ( mol% ) Monitored subunitMethodα5β10 . 17±0 . 115N-β1NMRα5β10 . 07±0 . 02α5Anisotropyα5β1KE0 . 63±0 . 115N-β1-KENMRα5β1KE 0 . 80±0 . 27α5AnisotropyαIIbβ30 . 15±0 . 115N-β3NMRαIIbβ3 0 . 09±0 . 03αIIbAnisotropyαIIbβ3KE 0 . 51±0 . 115N-β3-KENMRαIIbβ3KE 0 . 33±0 . 05αIIbAnisotropyα1β1>2 . 4815N-β1NMRα1β1 >3 . 2α1Anisotropyα1β1KE>2 . 4815N-β1-KENMRα1β1KE>3 . 2α1Anisotropyα2β1>2 . 4815N-β1NMRα2β1>3 . 2α2Anisotropyα2β1KE>2 . 4815N-β1-KENMRα2β1KE>3 . 2α2Anisotropy The binding results of Table 1 show that , as expected , the αIIb TM/CT forms a relatively avid complex with the β3 TM/CT ( Kd of ca . 0 . 1 mol% ) ( Table 1 ) . Also as expected , the β3-K716E mutation reduced the affinity of this complex by roughly 4-fold ( to a Kd of ca . 0 . 4 mol% ) . Binding of WT α5 TM/CT to β1 TM/CT was also seen to be avid ( Kd of ca . 0 . 1 mol% ) ( Table 1 ) , while the K752E mutation in the β1 TM/CT reduces its binding affinity for α5 by a factor of 7 ( Kd of ca . 0 . 7 mol% ) . This is consistent with a previous study that showed the β1-K752E mutation induces constitutive activation of the α5β1 integrin ( Kim et al . , 2012 ) . It was very surprising that for all 4 combinations of α/β association of β1-WT and β1-K752E TM/CT with the α1 and α2 TM/CT that binding was so weak it could in no case be quantitated ( Kd > 3 . 2 mol% for both α1β1 and α2β1 TM/CT ) ( Table 1 ) . 10 . 7554/eLife . 18633 . 019Figure 7 . Use of fluorescence anisotropy to determine Kd for complex formation between α and β integrin TM/CTs . Measurements were carried out in D6PC/POPC/POPS bicelles ( 2% total amphiphile , q = 0 . 3 ) in 25 mM HEPES buffer pH 7 . 4 at 35°C . α subunits were labeled with IAEDAN and titrated by unlabeled β subunits . Dissociation constants were reported in mol% of total protein to total lipid present in the bicelles for: ( A ) αIIbβ3 , ( B ) α5β1 , ( C ) α1β1 , ( D ) α2β1 . At least three independent experiments were carried out . Shown is an example of each experiment where at least five measurements were made at each time point . The error bars depict the standard deviation of these five measurements . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 019 Together these data suggest that the TM-K>E mutation decreases the affinity of the αβ heterodimers of both αIIbβ3 and α5β1 resulting in increased ‘activity’ of these integrins . These results also suggest that full length α1β1 and α2β1 integrins may be ‘constitutively active’ under physiological conditions and cannot be further activated . Due to these major discrepancies in the biophysical behavior of α5β1 and the collagen-binding integrins α1β1 and α2β1 , we assessed whether integrin α5β1-dependent adhesion and spreading on fibronectin was increased in response to the introduction of the β1-K752E mutation . CD cells with equal surface expression of the α5 , β1 and β1-K752E mutation were generated . Both the wild type and β1K752E CD cells adhered equally to fibronectin ( 0 . 5 μg/ml ) ( Figure 8A ) . As these cells express integrin αv , the same experiment was carried out in the presence of an integrin αv blocking antibody . Under these conditions CD cell adhesion was decreased , albeit less than the adhesion defect observed with the CD cells adhering to collagen I or IV . No differences in cell adhesion were seen when the cells adhered to fibronectin in the presence of an anti-β1 antibody , suggesting that the β1-K752E mutation decreased integrin α5β1-dependent adhesion of CD cells to fibronectin . When we assessed the spreading of these cells on fibronectin over different time periods , we observed a consistent but non-significant decrease in spreading of the β1K752E CD cells at 15- , 30- and 45 min time points ( Figure 8B ) . We also assessed the relative intensity of 12 G10 compared to AIIB2 on cells plated on fibronectin . Similar to when the cells adhered to collagen I there was decreased intensity of 12 G10 ( Figure 8C and D ) , which correlated with the adhesion and spreading defects . Thus a β1-K752E mutation decreases CD cell adhesion , spreading and integrin α5β1 activation when CD cells bind to fibronectin . 10 . 7554/eLife . 18633 . 020Figure 8 . The β1-K752E mutation decreases collecting duct cell adhesion to fibronectin . ( A ) The adhesion of CD cells to fibronectin ( 0 . 5 μg/ml ) for 1 hr was measured . These assays were carried out in the presence or absence of blocking antibodies directed against the αv or β1 subunits . The error bars indicate SD . * indicates a statistically significant difference ( p<0 . 01 ) between WT and K752E CD cells . These assays were performed at least three times . ( B ) The spreading of wild type and K752E CD cells on fibronectin at 15 , 30 and 45 min was measured on cells stained with rhodamine phalloidin . The area of at least 35 cells per time point were quantified using ImageJ software and expressed graphically . The error bars indicate SD . These assays were performed at least three times . ( C–D ) The active conformation of integrin β1 on adherent wild type ( WT ) and K752E expressing CD cells that were allowed to adhere to fibronectin for 1 hr was determined using 12 G10 antibody . Total integrin β1surface expression was determined using AIIB2 antibody ( C ) . ( D ) The fluorescence intensity was measured as described in the Materials and methods . The error bars indicate SD and * indicates a statistically significant difference ( p<0 . 01 ) between WT and K752E CD cells . These experiments were performed at least three times each . Intensity measurements were performed on over 40 different fields for each of the samples . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 020 The F3 domain of talin contains a phosphotyrosine-binding ( PTB ) signature that binds to the NPxY motif shared by the CT of both integrins β1 and β3 , while the linked F2 domain enhances binding through favorable electrostatic interactions with anionic lipids in the vicinity of the integrin on the membrane surface ( Moore et al . , 2012; Saltel et al . , 2009 ) . We probed whether affinity of the interactions of the isolated talin F3 domain with the β1 and β3 TM/CT are altered by the β1-K752E and β3-K716E mutations . NMR was used to monitor titrations of 15N-labeled integrin TM/CT subunits with unlabeled talin F3 in bicelles containing POPS , an anionic lipid ( Figure 9 ) . During the course of the titrations it was observed that certain peaks disappeared , indicating that binding between the two proteins is slow on the NMR time scale , as is typical when Kd ≤ 100 μM . Binding was therefore quantitated based on monitoring the F3 domain concentration-dependence of peak intensity reductions . As shown in Figure 9 , the talin F3 bound to both β1-WT and β1-K752E TM/CT in POPC/POPS/D6PC bicelles with similar ( moderate ) affinity ( Kd of 27 ± 4 µM and 18 ± 3 µM for WT and K752E β1 TM/CT , respectively ) . In contrast , the β3-K716E mutation results in a 6-fold increase in the Kd for talin-1 binding from 5 ± 2 µM to 30 ± 8 µM , suggesting the β3-K716E mutation alters the structure and/or dynamics of bicelle-associated β3 TM/CT to reduce talin-1 F3 binding affinity ( Figure 10 ) . These results indicate that mutation of the snorkeling lysine of β3 , but not of β1 , significantly reduces the affinity of the talin F3 domain for the isolated integrin subunit . The affinity of F3 for both the β1 and β3 TM/CT is higher by over an order of magnitude than the affinity of F3 for the isolated CT of these integrins in buffer minus either membranes or a membrane-mimetic ( Kd of 490 µM and 270–600 µM for β1 and β3 , respectively ) ( Moore et al . , 2012; Saltel et al . , 2009; Anthis et al . , 2010 ) . On the other hand , it should be acknowledged that the full affinity of the complex between talin and the integrin β3 CT in membranes containing an anionic lipid ( Kd ca . 0 . 9 μM ) requires the presence of a linked talin F2 domain ( Moore et al . , 2012 ) . 10 . 7554/eLife . 18633 . 021Figure 9 . NMR-monitored titrations of integrin β1 TM/CT by the talin-F3 domain . 15N-labeled WT and K752E mutant integrin β1 TM/CT samples were titrated with unlabeled talin1-F3 . The concentration of all integrins was 100 µM and the samples contained 5% q = 0 . 3 D6PC/POPC/POPS ( POPC:POPS = 2:1 ) bicelles in 25 mM HEPES buffer at pH 7 . 4 . The 1H , 15N-TROSY NMR spectra were collected at 35°C at 900MHz ( upper panels ) . In each case six titration points were collected in which the mole ratio between talin1-F3 and the integrin was varied: 0 , 0 . 5 , 1 , 2 , 3 and 4 . The 1H , 15N-TROSY spectra of 15N-labeled integrin from all the titration points were overlaid and plotted . The intensity changes for selected resonances , all thought to be at or in the vicinity of the Talin-F3 binding site , were plotted ( bottom panels ) versus concentration and the Kd was obtained by a single global fit of all the data . left: integrin β1 TM/CT WT , right: integrin β1 K752E TM/CT . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 02110 . 7554/eLife . 18633 . 022Figure 10 . NMR-monitored titrations of integrin β3 TM/CT by talin . 15N-labeled WT/K716E integrin β3 TM/CT was titrated with unlabeled talin1-F3 . The concentration of all integrins was 100 µM . The samples contained 5% q = 0 . 3 D6PC/POPC/POPS ( POPC:POPS = 2:1 ) bicelles in 25 mM HEPES buffer at pH 7 . 4 . The TROSY NMR spectra were collected at 35°C at 900MHz ( upper panels ) . In both cases six titration points were collected in which the molar ratio between talin1-F3 and the integrin was varied: 0 , 0 . 5 , 1 , 2 , 3 and 4 . The 1H , 15N-TROSY spectra of 15N-labeled integrin from all the titration points were overlaid and plotted . The intensity changes for selected residues , all thought to be at or in the vicinity of the Talin-F3 binding site , were plotted ( bottom panels ) against concentration and the Kd was obtained by a global fit of all the data . Left: integrin β3 TM/CT WT . Right: integrin β3 K716E TM/CT . DOI: http://dx . doi . org/10 . 7554/eLife . 18633 . 022
The goal of this study was to determine whether the canonical mechanism of integrin activation that has emerged from studies of the platelet specific αIIbβ3 integrin extends to the ubiquitously expressed β1 integrins . We show that the structures of the β1 and β3 CT/TM in bicelles are distinct . Also , the highly conserved ‘snorkeling lysine’ that plays a key role in regulating αIIβ3 activation plays an opposite role in β1 integrin function and does not play a role in defining the tilt of either β1 or β3 . Very different affinities exist between β1 and some α subunits relative to those observed between αIIb and β3 . We therefore conclude that the diverse biophysical properties of β1 and β3 integrin TM and CT domains in combination with varying α subunit properties result in distinct mechanisms of integrin activation and function . In this study , we show that , in contrast to integrin αIIbβ3 expressed in a CHO cell , the β1-K752E mutation does not promote integrin activation ( Kim et al . , 2012 ) but actually decreases integrin α1β1 , α2β1 and α5β1 epithelial cell adhesion to their respective ligands , with little effect on cell spreading . Consistent with this , β3 tail mutations that resulted in normal integrin activation and talin binding but decreased cell spreading , were recently described ( Pinon et al . , 2014 ) . A previous study showed that expression of β1-K752L in integrin β1-null fibroblasts did not alter the integrin activation status or the ability of the cells to adhere to fibronectin and laminin-111; however this mutation did inhibit cell spreading and migration ( Armulik et al . , 2004 ) . The mechanism for the altered integrin β1-K752L-mediated cell function was proposed to be due to altered phosphatidylinositol 3-kinase activation that impacted FAK-independent integrin signaling without altering integrin activation ( Armulik et al . , 2004 ) . We did not test the effect of a β1-K752L mutation in CD cells . Nevertheless , a possible reason for the differences in integrin activation in cell systems that are null for β1 and those in CHO cells is that in the former , the mutant integrin is transduced into a β1-null CD cell , while in the CHO cell system the mutated β integrin is expressed into a cell that already has endogenous integrins ( Kim et al . , 2012 ) . Alternatively , it could simply reflect differences in cell type . Another important point is that the only effect of the mutation of the ‘snorkeling lysine’ tested in CHO cells was on integrin activation , and not cell functions that rely on ligand-dependent integrin signaling such as adhesion , migration or spreading . Another key observation is that the β1-K752 and β3-K716 mutations do not seem to play an important role in defining the topology of the TM in the free subunits . β3-K716 was previously believed to stabilize the integrin αIIbβ3 inactive state by forming a salt bridge between its amino side chain and anionic phosphodiester linkages in the phospholipid head groups ( Kim et al . , 2012 ) . This interaction has been proposed to be a key determinant of the TM tilt that is required for optimal outer and inner clasp interactions between the αIIb and β3 subunits , which are present only in the inactive state ( Kim et al . , 2012 ) . This led us to conduct Mn ( II ) -induced paramagnetic relaxation NMR experiments ( PRE ) on WT and β1-K752E to determine whether K752 also promotes tilt of the TM . To our surprise , we observed that β1-WT and β1-K752E integrin exhibit the same PRE patterns , indicating that the tilt of the β1 integrin TM is independent of the charge of residue 752 . Paradoxically , we made the exact same observation for the β3-K716E mutation indicating that neither β1 nor β3 integrins form a lysine/phospholipid ion pair critical for promoting TM tilt . Our studies revealed that previously reported PRE results to the contrary had been subject to an experimental artifact , as detailed in the Results section . Hence , the previous work had mistakenly concluded that the different PRE patterns observed for β3-WT and β3-K716E reflected very different membrane topologies for WT and mutant ( Kim et al . , 2012 ) . Here , we repeated the original experiment , but took pains to eliminate the artifactual interaction , revealing that both WT and the K-to-E mutant forms of both β3 and β1 have membrane topologies that are similar in all four cases . We confirmed these observations by reciprocal NMR PRE experiments using a lipophilic paramagnetic probe . Thus , not only does β1-K752 promote functional output that is opposite to that promoted by β3-K716 , but even in the case of β3 the mechanism by which β3-K716 promotes the inactive signaling state must now again be regarded as unknown . Our results are consistent with another study that used a biotin maleimide chemical modification of engineered Cys side chains to show that the K716P mutation in β3 did not result in any change in membrane topology ( Kurtz et al . , 2012 ) . Given that the functional consequences of mutating β3-K716 and β1-K752 seem to arise from different mechanisms than originally thought , we tested whether the K-to-E mutations alter association of the integrin β tails with talin , an interaction that promotes integrin activation . Consistent with previous work ( Anthis et al . , 2010 ) , we found that the talin F3 domain binds to the WT β3 TM/CT significantly more tightly than to WT β1 TM/CT ( Kd for β3 lower by a factor of 5 ) . Interestingly , the β3-K716E mutation increased Kd for talin binding by a factor of 6 , a finding that that does not correlate well with the proposed activation-promoting nature of this mutation . On the other hand , the β1-K752E mutation alters Kd by less than 2-fold and does so in favor of tighter binding . While possible roles for the F0-F2 talin domains in possible coupling between the TM lysine site and talin binding were not tested in this study , at face value these results suggest that the impact of the TM K-to-E mutation on talin binding is not the basis for how this mutation alters integrin function , once again suggesting a mechanism whereby the TM K-to-E mutation alters integrin function by altering integrin-dependent signaling and not activation . It is interesting that talin binds to both integrin TM/CT domains in bicelles more tightly than to CT-only domains in buffer ( Moore et al . , 2012; Anthis et al . , 2010 ) . Bicelles most likely promote a more native-like conformational state of the CT in the context of its attached and bicelle-anchored TM than for the isolated CT in solution . This work also reveals significant differences in the secondary structures of the β1 and β3 TM/CT , which were here examined under identical bicellar conditions . The β1 TM ends after I758 while the β3 TM ends after I721 , in reasonably good accord with previous NMR , EPR , and biochemical studies . The TM helix of both β1 and β3 was seen to extend into the cytosol and through the juxtamembrane residues that form the inner clasp , albeit with some fraying . However , unlike β1 , where the helix ends with the clasp at K765 , the helix in the β3 extends a full 10 residues further into the subunit , ending only at A737 ( Figure 2C ) . This result is reminiscent of the observations that a continuous β3 TM/CT extends to D740 as part of the αIIbβ3 TM/CT heterodimer in an organic solvent mixture ( Yang et al . , 2009 ) , while in the corresponding complex for the CT-only heterodimer the β3 helix ends at K738 ( Vinogradova et al . , 2002 ) . A previous cross-linking/computational study of the αIIbβ3 complex in cell membranes also indicated the β3 helix extends from the TM far out into the cytosol ( Zhu et al . , 2009 ) . Our results are also generally consistent with a previous NMR study of the β3 TM/CT domain in DPC micelles , where an extended CT helix was observed , but was connected to the TM by a flexible linker ( Li et al . , 2002 ) . Our data also support the presence of hinge motion centered at the TM/CT interface , although our data suggest that hinge formation is transient , not a stable structural feature . Other NMR or EPR structural studies of the β1 or β3 TM in isolation or of the αIIβ3 TM complex did not include the full CT ( Moore et al . , 2012; Lau et al . , 2008a; Yu et al . , 2015 ) . The combined structural and membrane topological studies of our work also offer clear evidence that , on the average , the CT helix for both β1 and β3 TM/CT extends away from the membrane surface into the aqueous phase . Our results are not consistent with the notion that the CT has significant membrane surface affinity , as suggested based on a previous NMR study of crosslinked αIIb and β3 CT in DPC micelles ( Metcalf et al . , 2010 ) . The studies of this work were conducted in model membranes and using excised TM/CT domains rather than full length integrins in cellular membranes . Both of these facts hinder the extrapolation of the observations to full length integrins in native membrane conditions . Nevertheless , they do suggest that some of the functional differences between β1 and β3 are linked to the different intrinsic conformational preferences of their CT , which likely impacts their selectivity and affinity in engaging their cytosolic effector proteins . It is interesting to note that while we observed the β3 helix to extend through site A737 , in an NMR structure of the complex of the β3 CT with the talin F3 domain the helix terminates at amino acid 732 ( Wegener et al . , 2007 ) , suggesting destabilization of the C terminal end of the helix by talin . Conversely , for bicelle-associated β1 the helix was seen to terminate at K765 , while in a crystal structure of the β1 CT with the talin F2F3 domains this helix does not terminate till A773 ( Anthis et al . , 2009 ) . These results suggest that the end of the β3 TM/CT helix is not very stable but is readily disrupted by events such as engagement by talin . This is consistent with the fraying of the CT helix seen in the results of this paper . At the same time the disordered segment C-terminal to the β1 TM/CT helix does have helical propensity that is manifested upon complex formation with talin . The metastability of secondary structure in both β1 and β3 CT seems well suited to enable optimal interactions to cytosolic binding partners . Finally , the data demonstrated that the interactions of different α subunit TM/CT with the β1 TM/CT are characterized by very different affinities , ranging from very weak interactions between α1 or α2 and β1 to much higher affinity interaction between α5 and β1 , similar to that found between αIIb and β3 . Based mostly on studies of the αIIbβ3 integrin it has been widely assumed that the TM/CT of β integrins have an intrinsic affinity for the corresponding domains of their cognate α subunits , such that they will form constitutively inactive heterodimers . Many studies have shown the isolated αIIb and β3 TM associate to form heterodimers in model membranes or as fusion proteins in E . coli or model cell lines ( Lau et al . , 2009; Berger et al . , 2010; Partridge et al . , 2005; Zhu et al . , 2010; Schneider and Engelman , 2004; Schmidt et al . , 2015; Lokappa et al . , 2014; Kim et al . , 2009 ) . We observed similar results for heterodimerization of the α5 and β1 TM/CT , an observation consistent with evidence that this particular β1 integrin is activated according to the canonical model ( Takagi et al . , 2003 ) . In contrast , we found that α1 and β1 as well as α2 and β1 TM/CT interactions were too weak to be quantified in bicelles , even at the high protein concentrations required for NMR spectroscopy . This is surprising in light of studies suggesting that the fusion proteins containing the TM-only domain of these integrin subunits can form heterodimers in E . coli ( Berger et al . , 2010; Schneider and Engelman , 2004 ) . However , these latter studies were conducted in the absence of the β1 , α1 , and α2 CT , which almost certainly profoundly impact heterodimerization ( Briesewitz et al . , 1995; Liu et al . , 2015 ) . Our results suggest that the α1 and α2 CT may actually inhibit formation of α1β1 and α2β1 TM/CT heterodimers , at least in bicelles . The stark contrast between the collagen α1β1 and α2β1 integrins and the fibronectin α5β1 integrin suggests that the role of TM/CT domain heterodimerization in regulating integrin function may vary considerably among different β1 integrins , as previously proposed ( Nissinen et al . , 2012; Abair et al . , 2008b; Bazzoni et al . , 1998; Pepinsky et al . , 2002; Bodeau et al . , 2001 ) . Our results for integrins α1β1 and α2β1 suggest the intriguing possibility that these receptors may remain constitutively in their unclasped α/β-TM/CT-dissociated forms , implying their signaling functions are modulated via mechanisms other than the canonical model of switching between TM/CT-clasped and unclasped forms ( Abair et al . , 2008a; Nissinen et al . , 2012 ) . That some integrins may be constitutively active or unclasped has long been postulated ( Bazzoni and Hemler , 1998 ) . Determining whether this is actually the case will require additional studies . However , even considering that the energetics of heterodimerization of isolated TM/CT will not be the same as local TM/CT heterodimerization in the context of full length integrin subunits , these results indicate that the energetics of TM/CT heterodimerization vary dramatically from integrin to integrin . In conclusion , we present evidence that while integrin αIIbβ3 is found in both active and inactive conformations , a subclass of β1 integrins ( α1β1 and α2β1 ) may adopt a constitutively active conformation . Thus β1 and β3 integrins appear to have distinct mechanisms of action wherein different modes of integrin regulation likely occur within the β1 integrin class based on which α subunit is involved , as well as which cell type . Such cell type-specific mechanisms of integrin function need to be explored if we are to understand how different integrins function in distinct biological settings .
Cell adhesion assays were performed in 96-well plates as previously described ( Chen et al . , 2004 ) . Cells ( 1 × 105 ) were seeded in serum-free medium onto plates containing different concentrations of ECM for 60 min . Adherent cells were fixed , stained with crystal violet , and solubilized , and the optical densities of the cell lysates were read at 570 nm ( OD570 ) . Human placental collagen IV and rat tail collagen I were purchased from Sigma-Aldrich , St Louis , MO . Coverslips were coated with either collagen I ( 0 . 5 μg/ml ) , collagen IV ( 0 . 25 μg/ml ) and fibronectin ( 0 . 5 μg/ml ) and blocked with 2% heat inactivated BSA . Cells were seeded in serum-free medium and incubated for either 15 min , 30 min or 45 min after which they were fixed with 4% paraformaldehyde and stained with Rhodamine labeled Phalloidin . Images were collected with confocal microscopy and the cell area was determined using ImageJ software . The active conformation of integrin β1 on adherent cells was determined using the 12 G10 antibody ( Millipore MAB2247 , Darmstadt , Germany ) that specifically binds to the active conformation of integrin β1 . Total surface expression was determined using AIIB2 antibody . CD cells stably expressing either WT or K752E integrin β1 were allowed to adhere for 1 hr to eight well chamber glass slides ( Millicell EZ slides , Millipore , USA Cat no . PEZGS0816 ) coated with collagen I ( 20 μg/ml ) or fibronectin ( 10 μg/ml ) at 4°C overnight . Adherent cells were fixed with 10% formaldehyde , incubated with primary antibody ( 1:100 ) followed by secondary antibodies ( 1:100 ) and visualized using a Zeiss LSM 510 microscope . Images were taken close to the substrate . The intensities of images were analyzed using ImageJ software ( JACoP ) . Manders' overlap coefficient based on the Pearson's correlation coefficient for average intensity values was in each case quantified and expressed as a percentage of 12 G10 relative to AIIB2 . Statistical plots show the mean values and SD of 30 cells per group . cDNA for the wild type human integrin β1 TM/CT ( resides 719–798 ) , integrin β3 TM/CT ( residues 685–762 ) , integrin β3 transmembrane-only domain ( TM-only ) ( residues 685–727 ) and talin1-F3 domain ( residues 309–405 ) were sub-cloned into a pET16b vector ( Novagen , Darmstadt , Germany ) , which adds an N-terminal His6 purification tag ( MGHHHHHHGM- ) . The single native cysteine present in each construct—C723 in integrin β1 TM/CT , C687 in integrin β3 TM/CT , and C336 in the talin1-F3 domain—were mutated to serine in order to avoid aberrant disulfide formation . No adverse effects of the C687S mutation in β3 or of the C336S mutation in talin1 were reported in previous studies of these proteins ( Lau et al . , 2008a , 2008b; Song et al . , 2012 ) ( Ulmer et al . , 2003 ) . QuikChange site-directed mutagenesis ( Stratagene , La Jolla , California ) was used to produce each of the following variants: K752E integrin β1 TM/CT , K716E integrin β3 TM/CT , and the K716E integrin β3 TM-only . Amino acid-specific isotopic labeling is obtained by using an auxotroph E . coli strain , CT19 , which is incapable of synthesizing Leu/Ile/Val/Ala/Tyr/Phe/Trp/Asp . The cells depend solely on diet to get these amino acids , enabling supplementation of the medium with one of these amino acids in 15N-labeled form , leading to specific labeling . The plasmid encoding the integrin construct was transformed into CT16 cells and plated on LB agar with three antibiotics used for selection: 100 mg/L ampicillin , 100 mg/L kanamycin , and 20 mg/L tetracycline . The plate was incubated overnight at 37°C . As tetracycline is light sensitive , the plate and all the media flasks were covered with aluminum foil . A single colony was used to inoculate 5 ml LB media with those three antibiotics . After overnight growth at 37°C , 1 ml of LB cell culture was used to inoculate 1 L M9 media at room temperature with the addition of 0 . 5 g of each of the eight acids deficient in the CT16 auxotroph . After OD600 reached 0 . 8 the cells were spun down at 3000 rpm for 15 min and transferred to the final medium at room temperature which contained 0 . 2g of the desired 15N-labeled amino acid and 0 . 5 g unlabeled each for other 7 . The cells were then induced by adding IPTG to 1 mM and harvested after overnight growth . The protein was purified following the protocol in the Materials and methods section . This method results in one amino acid type remaining unlabeled in a protein under conditions in which all other amino acids are 15N-labeled . This will lead to 1H , 15N-TROSY or HSQC spectra in which all peaks are seen except for the amino acid that was deliberately not 15N-labeled . Our experience shows this labeling strategy works well for Arg , Lys , and His . The labeling strategy is similar to standard uniform 15N-labeling in 15N-enriched minimal medium with two differences: first , 1 g of the unlabeled amino acid is added to the liquid culture at room temperature immediately before induction and second , cells are harvested 8 hr after induction and continued incubation at room temperature . The protein was purified using the protocol given below . BL21 ( DE3 ) cells were transformed with the pET16b vectors . M9 cultures were grown at room temperature to an OD600 of 1 . 0 prior to induction with 1 mM IPTG . After 24 hr of induction at room temperature , the cells were harvested by centrifugation and stored at −80°C for future use . 15N-NH4Cl , 13C-glucose , and/or D2O were incorporated into the M9 media in order to produce proteins with the desired isotopic labeling for NMR spectroscopy . For protein purification frozen packed E . coli cells were suspended in lysis buffer ( 75 mM Tris-HCl , 300 mM NaCl and 0 . 2 mM EDTA , pH 7 . 7 ) at 20 ml per gram of cells . Next 5 mM Mg ( Ac ) 2 , 0 . 2 mg/ml PMSF , 0 . 02 mg/ml DNase , 0 . 02 mg/ml RNase and 0 . 2 mg/ml lysozyme were added to the cell slurry , which was tumbled at room temperature for 1 hr . The lysate was then sonicated for 10 min at 4°C . Following sonication , the detergent Empigen ( Sigma-Aldrich , St Louis , MO ) was added to 3% ( w/v ) and the mixture was again tumbled at 4°C for 1 hr to solubilize the integrins . Insoluble debris was then removed by centrifugation at 20 , 000 g for 20 min . The supernatant was collected and incubated with Ni2+-NTA resin ( Qiagen , Valencia , CA ) ( 1 ml per liter of cell culture ) at 4°C for 1 hr . The resin was collected by centrifugation at 3700 g for 5 min and packed into a chromatography column , which was washed with buffer A ( 40 mM HEPES , 300 mM CHPNaCl , pH 7 . 5 ) containing 3% Empigen followed by wash buffer ( buffer A plus 40 mM imidazole ( IMD ) and 1 . 5% Empigen ) until the A280 returned to the baseline level . Empigen on the column was then exchanged with the mild detergent dihexanoylphosphatidylcholine ( D6PC , Avanti Polar Lipids , Alabaster , Al , USA ) by washing the column with 8 × 1 column volumes of exchange buffer ( 100 mM NaCl 10 mM IMD , pH 7 . 4 ) containing 2% D6PC , and further exchanged into bicelles by washing the column with two column volumes of exchange buffer containing 2% w/v bicelles and 10% D2O . Integrins were then eluted in 250 mM imidazole ( pH 7 . 4 ) containing 2% bicelles and 10% D2O ( w/w ) . Two different bicelle compositions were used: D6PC/DMPC q = 0 . 3 and POPC/POPS/D6PC ( POPC:POPS = 2:1 ) q = 0 . 3 , where q is the lipid to detergent mole ratio . The eluted protein solution was concentrated 10-fold by using an Amicon Ultra centrifugal filter cartridge with a molecular weight cut-off of 10 kDa and the pH was adjusted to 6 . 5 using acetic acid . 200 μl was then transferred to a 3 mm NMR tube . NMR samples prepared in this manner contained 10% D2O for field frequency locking purposes , 1 mM EDTA , 250 mM IMD , pH 6 . 5 . The final bicelle concentration was 20% ( w/w ) and the integrin concentration was typically 0 . 5 mM . For some samples , IMD was exchanged out as buffer for HEPES . To accomplish this , the concentrated protein solution above ( with 250 mM IMD ) was diluted 10-fold in 25 mM HEPES pH 7 . 4 containing 0 . 7% D6PC ( equal to its critical micelle concentration , CMC ) and 10% D2O , then concentrated again 10-fold . Two additional rounds of dilute/centrifugation buffer exchange were carried out to reduce the final imidazole concentration to <1 mM . The 0 . 7% D6PC present in the exchange buffer was employed to maintain a constant q by offsetting loss of the free population of D6PC ( equal to its CMC ) through the centrifugal filter . Finally , EDTA was added from a 200 mM stock in 25 mM HEPES buffer at pH 7 . 4 to a concentration of 1 mM in the NMR sample . By this method the final NMR samples contained 10% D2O , 1 mM EDTA , 25 mM HEPES , pH 7 . 4 . The final bicelle concentration was 20% ( w/w ) and the integrin concentration was typically 0 . 5 mM , which is 0 . 57% molar percent ( mol% ) relative to total moles of DMPC . Talin1-F3 was purified in a similar manner except that no lipids or detergents were added to any of the purification buffers . Briefly , cells were lysed in an identical manner as described above prior to centrifugation at 20 , 000 g for 20 min in order to remove cellular debris . The lysate was then incubated with Ni2+-NTA resin ( Qiagen , Chatsworth , CA ) ( 1 mL per liter of cell culture ) at 4°C for 1 hr . The resin was then collected by centrifugation at 3700 g for 5 min and packed into a chromatography column . The column was first washed with buffer A followed by buffer A containing 40 mM imidazole . After A280 returned to baseline , the talin1-F3 was eluted in 500 mM IMD ( pH 7 . 4 ) . The talin1-F3 was concentrated to 0 . 8 mM by using an Amicon Ultra centrifugal filter cartridge with molecular weight cut-off of 3 kDa and stored at 4°C for future use . Prior to NMR studies buffer exchange was carried out by 3 rounds of 10-fold dilution/centrifugal concentration with an exchange buffer ( 25 mM HEPES in 10% D2O , 1 mM EDTA , pH 7 . 4 ) . The final talin1-F3 stock concentration was 0 . 8 mM . Protein concentrations were determined from A280 using extinction coefficients calculated by the ‘protparam’ program ( http://web . expasy . org/protparam/ ) . The pH was adjusted using either acetic acid or ammonium hydroxide . The combined TM/CT of integrin α1 ( residues 1132–1179 ) , α2 ( 1125 to 1181 ) , αIIb ( 9189 to 1039 ) , and α5 ( 990 to 1050 ) were expressed and purified as described previously ( Ulmer et al . , 2003; Lu et al . , 2012; Mathew et al . , 2012b ) . These recombinant proteins were eluted from the Ni resin in the presence of 250 mM imidazole and bicelles and prepared for biophysical studies as described above for the integrin beta TM/CT . NMR data were collected for the β1 and β3 TM/CT subunits at 35°C or 45°C on Bruker 900 , 800 and 600 MHz spectrometers equipped with cryogenic triple-resonance probes with z-axis pulsed field gradients . 1H , 15N-TROSY experiments were conducted using the standard Bruker pulse sequence trosyetf3gpsi ( version 12/01/11 ) , which makes use of a sensitivity enhanced , phase-sensitive TROSY pulse sequence and WATERGATE solvent suppression . The integrin β1 TM/CT NMR samples contained 0 . 5 mM ( 0 . 57 mol% relative to moles of DMPC ) uniformly 13C , 15N , 2H-labeled integrin β1 TM/CT , 20% D6PC/DMPC q = 0 . 3 , 1 mM EDTA , 250 mM IMD at pH 6 . 5% and 10% D2O . The integrin β3 TM/CT NMR samples had the same composition as the β1 sample except that the protein was not perdeuterated . The following TROSY-based 3D experiments were collected in order to assign backbone resonances: HNCA , HN ( CO ) CA , HNCACB , HNCO , HN ( CA ) CO , NOESY-TROSY-HSQC . Selective 15N-labeling of specific amino acids was also employed to facilitate resonance assignments . NMR data were processed using NMRPipe ( Delaglio et al . , 1995 ) and analyzed using NMRViewJ ( Johnson , 2004 ) . Backbone chemical shift assignments were analyzed using chemical shift index analysis ( Wishart and Sykes , 1994 ) and TALOS-N ( Shen and Bax , 2013 ) in order to determine the secondary structure . The TROSY peak assignments of the mutant protein or WT protein under pH 7 . 4 D6PC/POPC/POPS bicelles conditions were obtained by comparing spectra to the pH 6 . 5 ‘reference’ assigned WT spectrum . For some mutants/conditions a 3D TROSY-HNCA was collected and analyzed to assist with confirming and completing assignments . The CLEANEX-PM NMR experiment ( Hwang et al . , 1998 ) was carried out using a mixing time of 100 msec to map the extent of hydrogen exchange between backbone amide sites and water occurring over a 100 msec time scale . 15N relaxation experiments on 15N-labeled integrin TM/CT were carried out at 45°C on Bruker Avance NMR spectrometers . A TROSY-based version of the Carr-Purcell-Meiboom-Gill experiment was used to obtain 15N transverse relaxation times ( T2 ) , which are presented as relaxation rates ( R2=T2−1 ) ( Zhu et al . , 2000 ) . The transverse magnetization decay was sampled at nine points: ( 17 , 35 , 52 , 69 , 86 , 104 , 138 , and 173 msec ) . The data were recorded in pseudo-3D mode with a recycle delay of 3 s . All relaxation spectra were recorded with 1024 × 128 complex data points and the spectra widths were 13 by 23 ppm in the 1H and 15N dimensions , respectively . The membrane topologies of the following six proteins were probed: integrin β1 TM/CT , integrin β1 K752E TM/CT , integrin β3 TM/CT , integrin β3 K716E TM/CT , integrin β3 TM-only , and integrin β3 K716E TM-only . For each , two sets of sample conditions were used: one using fully zwitterionic bicelles containing 20% D6PC/DMPC bicelles q = 0 . 3 , with 1 mM EDTA 250 mM IMD at pH 6 . 5 in 10% D2O . The other conditions employ net-negatively charged bicelles composed of 20% D6PC/POPC/POPS ( POPC:POPS = 2:1 ) , q = 0 . 3 , with 25 mM HEPES at pH 7 . 4 in 10% D2O . The protein concentrations were ~0 . 5 mM ( 0 . 57 mol% ) for integrin β1 and ~0 . 3 mM ( 0 . 34 mol% ) for integrin β3 . Two paramagnetic probes were employed; the lipophilic 16-doxyl stearic acid ( 16-DSA ) probe ( Santa Cruz , San Diego , CA , USA ) and the water-soluble Gd-DTPA probe ( Santa Cruz , San Diego , CA , USA ) . For the sake of comparison with Gd-DTPA , in the cases of the integrin β3 TM-only and the β3 K716E TM-only , Mn-EDDA was utilized as a water-soluble probe following the published protocol ( Lau et al . , 2008a ) . The 900 MHz 1H , 15N-TROSY spectrum of each 15N-integrin was monitored to quantitate the line-broadening effects of different paramagnetic probes . For the water-soluble paramagnetic probes , a stock solution ( 50 mM for Mn-EDDA and 500 mM Gd-DTPA in the same buffer and pH as NMR samples ) was made and directly added to NMR samples ( to 1 mM for Mn-EDDA or 10 mM for Gd-DTPA ) . 16-DSA was first dissolved in methanol and the appropriate amount was transferred to an Eppendorf tube and dried under vacuum for 4 hr prior to solubilizing in integrin/bicelle NMR solutions . The final molar concentration of 16-DSA in bicelles was 2 . 5 mM ( 4 mol% of total lipid ) . Matched TROSY spectra of 15N-labeled integrins were collected in the presence and absence of each paramagnetic probe . The intensity ratios for the corresponding peaks in the pairs of spectra were calculated as an indicator of the probe access to the amide site . The spectra were processed using NMRPipe ( Delaglio et al . , 1995 ) and analyzed with NMRViewJ ( Johnson , 2004 ) The formation of heterodimers between different integrin β and α subunits during subunit titration was monitored by 1H , 15N-TROSY spectroscopy at 45°C . The following integrin TM/CT complexes were probed in this study: ( i ) the heterodimers formed by either WT or K752E mutant forms of β1 with WT forms of α1 , α2 and α5 , ( ii ) the heterodimers formed by either WT or K716E forms of β3 with wild type αIIb . All titration points maintained the same model membrane and buffer conditions: 20% ( w/v ) D6PC/POPC/POPS ( POPC:POPS = 2:1 ) , q = 0 . 3 , 50 mM phosphate buffer with 1 mM EDTA in 10% D2O , pH 6 . 5 . All the purified proteins were concentrated by a factor of 10 using centrifugal concentrators ( Amicon Ultra , 10kDal cut-off ) to attain 20% bicelles in the samples . Protein concentration in mol% was calculated as the ( moles of protein X 100 ) / ( total moles lipid ) , where D6PC ( a detergent ) is not considered to be a lipid . Integrin TM/CT complex formation for αIIbβ3 ( WT ) , αIIbβ3 ( K716E ) , α5β1 ( WT ) and α5β1 ( K752E ) were monitored by acquiring 1H-15N-TROSY spectra on a Bruker 900 MHz NMR spectrometer . Rather than titrating a single β subunit sample with multiple aliquots from a stock solution of the alpha subunit , six separate NMR samples were prepared for each titration at mole ratios of α/β = 0 , 0 . 5 , 1 , 2 , 3 , and 4 . The concentrations of WT β3 , WT β1 and the K752E β1 mutant were fixed at 0 . 17 mol% ( 150 µM ) for all the samples , while the concentration for the K716E β3 mutant was fixed at 0 . 10% ( 90 µM ) . For these titrations , no peak shifts were observed , but some β1 1H-15N-TROSY peaks disappeared during the course of the titrations , indicating subunit association/dissociation exchange that is slow on the NMR time scale . Peak intensities were plotted against the concentration of the unlabeled α subunit for each titration . Variations in peak intensities as a function of the α subunit were globally fit by a 1:1 binding model similar to that given in the next paragraph using OriginPro9 . 0 software . Determination of the dissociation constants for integrin α1β1 ( WT ) , α2β1 ( WT ) , α1β1 ( K752E ) and α2K752Eβ1 ( K752E ) TM/CT complexes was based on data from titrations monitored using a Bruker 600MHz NMR spectrometer . Six NMR samples were prepared for each titration , at varying molar ratios of α/β: 0 , 1 . 25 , 2 . 5 , 3 . 75 , 5 , and 7 . 5 . The concentrations of the WT and K752E mutant integrin β1 subunits were fixed at 0 . 23 mol% ( 200 µM ) for all samples . For these titrations on/off binding exchange was rapid on the NMR time scale , such that α1 and α2 subunit-induced chemical shift changes in the spectra were monitored . Hybrid 1H and 15N amide shifts at each point were measured using the following equation ( Ayed et al . , 2001 ) : ( 1 ) Δ ( HN ) =ΔH2WH+2ΔN2WN2 Where WH = 1 and WN = 0 . 154 are weighting factors for the 1H and15N amide shifts , respectively . The hybrid chemical shift changes were plotted against the concentration of unlabeled integrin α1 TM/CT or integrin α2 TM/CT and fit by the 1:1 binding model using OriginPro9 . 0 nonlinear regression: ( 2 ) Δobs=Δmax ( Kd+[L0]+[U0] ) − ( Kd+[L0]+[U0] ) 2− ( 4[U0][L0] ) 2[U]0 where Kd is the dissociation constant , Δobs is the observed hybrid chemical shift change , Δmax is the shift difference between free subunit and saturated complex conditions , and [U0] and [L0] are the mol% concentrations of unlabeled and labeled proteins , respectively ( Anthis et al . , 2010 ) . The affinities by which integrin β1 and β3 wild type and K752E/K716E mutants form heterodimer complexes with α subunits were determined from steady state fluorescence anisotropy measurements . Single cysteine mutants of the α subunit TM/CT ( α1 L1142C , α2 T1132C , α5 E992C and αIIb E960C ) were labeled with the fluorophore IAEDAN and titrated with unlabeled cysteine-free β1 and β3 wild type and K752E/K716E mutants in 2% q = 0 . 3 D6PC/POPC/POPS ( POPC/POPS = 2:1 ) bicelles in 25 mM HEPES buffer pH 7 . 4 at 35°C . Protein preparation and labeling for the above measurements was carried out as follows . His6-tagged mutant single-cysteine α integrin TM/CT were originally eluted from the Ni-NTA column into 0 . 5% DPC micelles in 250 mM imidazole , pH 7 . 8 . The protein concentration was assessed by measuring A280 and then adjusted to ca . 50 μM by diluting with the elution buffer ( 0 . 5% DPC micelles , pH 7 . 8 and 250 mM imidazole ) , protein being quantitated via A280 measurement . EDTA was added from a stock solution to 1 mM and the pH was adjusted to 6 . 5 with acetic acid . DTT was then added to a concentration of 2 . 5 mM and the solution was placed under argon . The solution allowed to mix at room temperature ( RT ) with gentle shaking for 16–20 hr . DTT and imidazole were removed using a PD-10 desalting column pre-equilibrated with 20 mM Tris , 150 mM NaCl , 0 . 5% DPC , pH 7 . 8 . A 10X molar excess of IAEDAN ( from a stock solution 25 mM acetonitrile ) was then added to the protein solution based on the assumption that the protein concentration was ca . 50 μM . The reaction of IAEDAN with the protein TM/CT thiol group was allowed to proceed at room temperature under argon for 3 hr . Excess free label was then removed by desalting over a PD-10 column in 20 mM Tris , 150 mM NaCl , 0 . 5% DPC pH 7 . 8 . Protein concentration and labeling efficiency were calculated using A280 and A373 ( with ε337 = 5600 M−1cm−1 for IAEDAN ) . Labeled proteins were then re-bound to Ni NTA resin via incubation for 1 hr at RT . The labeled proteins on the resin was then washed with 16 column volumes of 0 . 5% DPC micelles in HEPES buffer 25 mM , pH 7 . 2 to further remove unreacted label . The resin-bound integrin was then re-equilibrated with 1% bicelles in HEPES buffer 25 mM , pH 7 . 2 using four column volumes of buffer and then eluted with 250 mM imidazole , 2% bicelles pH 7 . 8 . EDTA was added to 1 mM and the buffer was exchanged via centrifugal filtration with 25 mM HEPES buffer , pH 7 . 4 containing 2% [D6PC/POPC/POPS ( 2:1 ) ] bicelles . The protein concentration was determined using A280 . For use in the titrations unlabeled cysteine-free integrin β1 and β3 wild type and K752E/K716E mutants were eluted from Ni ( II ) -NTA resin using 250 mM imidazole , 2% D6PC/POPC/POPS bicelles pH 7 . 8 . The eluted proteins were then buffer exchanged into 25 mM HEPES buffer , pH 7 . 4 containing 2% ( D6PC/POPC/POPS ) bicelles . The protein concentration was determined by A280 before proceeding with anisotropy experiments . 0 . 2 µM IAEDAN-labeled single cysteine mutants of the α subunit TM/CT ( α1 L1142C , α2 T1132C , α5 E992C and αIIb E960C ) were titrated with unlabeled cysteine-free β1 and β3 wild type and K752E/K716E mutants in 2% D6PC/POPC/POPS bicelles in 25 mM HEPES buffer pH 7 . 4 at 35°C . Each titration point mixture was incubated for 16–20 hr at room temperature before conducting the anisotropy measurements . The mixture was allowed to equilibrate in the cuvette for 5 min before the start of experiments . Anisotropy measurements were carried out using a Horiba Jobin Yvon Fluoromax-3 fluorimeter equipped with an L-format , single cuvette holder with polarizers . Steady state anisotropy was measured using excitation and emission wavelengths of 337 and 487 at 35°C . Increases in anisotropy were measured as a function of the increasing concentration of unlabeled β TM/CT subunit to the IAEDAN-labeled α subunit . Origin 9 . 0 was used to fit a 1:1 stoichiometry binding model to the observed experimental anisotropy values versus the total concentration of unlabeled integrin β1 or β3 , leading to determination of Kd . Concentrations were expressed in mol% ( total moles of protein X100 / total moles of lipid in bicelles ) units , as is appropriate for molecular association involving membrane-associated molecules . The detergent D6PC was not regarded as a lipid in these calculations . Four sets of titrations were carried out involving integrins ( β1 TM/CT , β1 K752E TM/CT , β3 TM/CT , β3 K716E TM/CT ) and the talin1-F3 domain . In all cases , the integrin was 15N-labeled and the concentration was fixed at 100 µM throughout the titration ( using a series of individually prepared samples ) . The NMR spectra were measured in the presence of 0 , 50 , 100 , 200 , 300 , 400 μM of unlabeled talin1-F3 . The total detergent+lipid content was fixed at 5% ( w/v ) at a q-value of 0 . 3 ( POPC/POPS/D6PC ) . The final solutions contained 25 mM HEPES pH 7 . 4% and 10% D2O . The Student's t-test was used for comparisons between two groups , and analysis of variance using Sigma Stat software was used for statistical differences between multiple groups . p<0 . 05 was considered statistically significant . | Proteins called integrins span the membranes of most human cells , and help our cells to interact with their surroundings , enabling them to organise , communicate and to form a variety of structures . Cells in different parts of the body typically produce different integrins so that they can specifically connect with other cells and proteins in their local environment . There are many different kinds of integrin proteins found in cell membranes and they consist of one alpha and one beta subunit . Different integrin pairs can have different effects based on their environment and the other molecules that they encounter . Much of the research into how integrins work has involved one specific integrin found in platelets – cells in blood that aid clotting and wound repair . Yet , it is unknown if all integrins actually operate in the same way as the platelet integrin . Lu , Mathew , Chen et al . studied the part of integrins that are located inside cells ( referred to as the cytoplasmic tail ) and the part that crosses the membrane ( the transmembrane domain ) . Three-dimensional structures of these parts of the proteins showed that they varied between different beta integrin proteins . Further experiments revealed that the strength of the association between different alpha and beta integrins also varied . Finally Lu , Mathew , Chen et al . demonstrated that components shared by several beta integrins actually have different purposes in different contexts . The diversity of structures and interactions within the group of integrin proteins suggests that integrins are likely to behave very differently in different cells . This means that platelet integrins cannot be used to fully understand the activity of all other types of integrin . More work is now needed to understand how the differences between integrins affect the roles that they fulfil and the molecules that they interact with . A deeper understanding of the differences between integrins could ultimately shape the development of strategies to specifically target them to treat a range of diseases – such as cancer and diseases in which there is a build-up of fibrous connective tissue . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2016 | Implications of the differing roles of the β1 and β3 transmembrane and cytoplasmic domains for integrin function |
Mechanical load of the skeleton system is essential for the development , growth , and maintenance of bone . However , the molecular mechanism by which mechanical stimuli are converted into osteogenesis and bone formation remains unclear . Here we report that Piezo1 , a bona fide mechanotransducer that is critical for various biological processes , plays a critical role in bone formation . Knockout of Piezo1 in osteoblast lineage cells disrupts the osteogenesis of osteoblasts and severely impairs bone structure and strength . Bone loss that is induced by mechanical unloading is blunted in knockout mice . Intriguingly , simulated microgravity treatment reduced the function of osteoblasts by suppressing the expression of Piezo1 . Furthermore , osteoporosis patients show reduced expression of Piezo1 , which is closely correlated with osteoblast dysfunction . These data collectively suggest that Piezo1 functions as a key mechanotransducer for conferring mechanosensitivity to osteoblasts and determining mechanical-load-dependent bone formation , and represents a novel therapeutic target for treating osteoporosis or mechanical unloading-induced severe bone loss .
Bone is the vital organ that constantly responds to and adapts to changes in mechanical loads associated with body weight , movement and gravity ( Iwaniec and Turner , 2016 ) . Such a mechanical-load-induced remodeling process is determined through the functional interaction between the bone-forming osteoblasts and the bone-absorbing osteoclasts . Perturbation of this remodeling process can lead to the well-documented bone-loss phenomenon that occurs upon mechanical unloading during long-term confinement in bed or spaceflight ( Nagaraja and Risin , 2013 ) . Impaired osteoblast function might contribute to reduced bone formation ( Carmeliet and Bouillon , 2001 ) . However , the mechanical response properties and the underlying mechanotransduction molecules that are active during bone formation remain poorly understood . The mechanosensitive Piezo1 channel ( Coste et al . , 2010; Coste et al . , 2012; Ge et al . , 2015; Zhao et al . , 2016; Zhao et al . , 2018 ) mediates mechanical responses in various cell types ( Geng et al . , 2017; Murthy et al . , 2017 ) , including vascular and lymphatic endothelial cells ( Choi et al . , 2019; Li et al . , 2014; Nonomura et al . , 2018; Ranade et al . , 2014 ) , smooth muscle cells ( Retailleau et al . , 2015 ) , red blood cells ( Cahalan et al . , 2015 ) , epithelial cells ( Eisenhoffer et al . , 2012; Gudipaty et al . , 2017 ) , neural stem cells ( Pathak et al . , 2014 ) and chondrocytes ( Lee et al . , 2014; Rocio Servin-Vences et al . , 2017 ) . Constitutive knockout of Piezo1 results in embryonic lethality in mice , mainly because of defects in vascular development ( Li et al . , 2014; Ranade et al . , 2014 ) . Piezo1 senses shear stress in vascular endothelial cells and red blood cells , contributing to the regulation of blood pressure ( Rode et al . , 2017; Wang et al . , 2016 ) and of red blood cell volume ( Cahalan et al . , 2015 ) , respectively . Piezo1 can also sense the local cellular environment and thus has a role in epithelial cell homeostasis ( Eisenhoffer et al . , 2012; Gudipaty et al . , 2017 ) , the lineage choice of neural stem cells ( Pathak et al . , 2014 ) , axon growth ( Koser et al . , 2016 ) and axon regeneration ( Song et al . , 2019 ) . Given its widespread function in various cell types and biological processes ( Murthy et al . , 2017 ) , we have reasoned that Piezo1 might play important roles in mechanical-load-dependent bone formation and remodeling . In line with this , expression data obtained from the BioGPS database ( http://biogps . org/#goto=genereport&id=234839 ) indicate that Piezo1 is highly expressed in osteoblasts . Furthermore , a previous study has shown that the expression level of Piezo1 in mesenchymal stem cells ( MSCs ) was correlated to the promotion of osteoblast differentiation and to the inhibition of adipocyte differentiation ( Sugimoto et al . , 2017 ) . Although Piezo1 was proposed to function as a mechanoreceptor that initiates the response to hydrostatic pressure ( HP ) in MSCs , the channel activities and the in vivo contribution to bone formation of Piezo1 expressed in MSCs and MSC-derived cell lineages such as osteoblasts have not been directly examined ( Sugimoto et al . , 2017 ) . In the present study , we set out to address whether Piezo1 might function as a key mechanotransducer that confers mechanosensitivity to the bone-forming osteoblasts and that consequently determines mechanical-load-induced bone formation and remodeling . Toward this goal , we systematically examined the role of Piezo1 in bone formation using Ocn-Cre-dependent Piezo1 knockout mice , mechanical-unloading-induced mouse and cellular models , and human osteoporosis samples . We have found that Piezo1 plays a key role in determining the mechanical response of osteoblasts and the in vivo formation and remodeling of bone in mice and humans .
To explore the role of Piezo1 in bone cell types , we initially examined the mechanical response of the commonly used pre-osteoblast cell line MC3T3-E1 and the pre-osteoclast cell line RAW264 . 7 . To this end , we directly measured the mechanically evoked cationic currents using whole-cell patch clamp electrophysiology coupled with mechanical poking of the cell membrane with a piezo-driven blunt glass pipette . MC3T3-E1 displayed mechanically activated currents in a step-dependent manner with a maximal current of 87 . 7 ± 11 . 6 pA ( Figure 1a , b ) . By contrast , much smaller mechanically activated currents were recorded in cell of the pre-osteoclast cell line RAW264 . 7 ( 24 . 0 ± 5 . 0 pA ) ( Figure 1a , b ) . In line with the recorded mechanically activated currents , the mRNA of Piezo1 was significantly higher in MC3T3-E1 cells than in RAW264 . 7 ( Figure 1c ) . Furthermore , either siRNA-mediated knockdown of Piezo1 ( Figure 1f , g ) or the application of GsMTX4 , a relatively specific blocker of the Piezo channel family ( Bae et al . , 2011 ) , significantly reduced the mechanically activated currents in MC3T3-E1 cells ( from 68 . 6 ± 7 . 0 pA to 17 . 3 ± 2 . 9 pA or 21 . 7 ± 5 . 6 pA , respectively ) ( Figure 1d , e ) . These data suggest that Piezo1 is expressed and mediates the mechanically activated currents in MC3T3-E1 cells . Interestingly , siRNA-mediated knockdown of Piezo1 in MC3T3-E1 cells decreased the expression of the functional marker genes of osteoblasts revealed by QRT-PCR , including alkaline phosphatase ( Alp ) , osteocalcin ( Bglap ) and collagen 1 ( Col1α1 ) ( Figure 1h ) , and reduced the Alp staining ( Figure 1i ) . Furthermore , we found that the expression of Piezo1 was increased in MC3T3-E1 cells cultured with osteogenic medium for 1 day , 3 days and 5 days ( Figure 1—figure supplement 1a , b ) . These data suggest that Piezo1 might mediate the mechanical response and function of differentiated osteoblasts . Thus , we next focused on characterizing the role of Piezo1 in primary osteoblasts and in vivo bone formation . QRT-PCR analysis revealed that Piezo1 was highly expressed in the bone among the various examined mouse tissues ( Figure 2a ) . To investigate the specific expression and function of Piezo1 in the bone-forming osteoblasts , we chose Ocn-Cre mice—which specifically express Cre recombinase under the osteocalcin ( Ocn ) gene promoter in osteoblast-lineage cells , but not in osteoclasts ( Huang et al . , 2016; Zhang et al . , 2002 ) —to generate the conditional Piezo1 KO mice ( Piezo1Ocn/Ocn ) by crossing the Ocn-Cre mice with the Piezo1fl/fl mice ( Cahalan et al . , 2015 ) . In line with the study of MC3T3-E1 cells ( Figure 1 ) , we detected mRNA and protein expression of Piezo1 in the bone tissue and primary osteoblasts derived from littermate Piezo1fl/fl mice ( Figure 2b–e ) . The Piezo1Ocn/Ocn mice show significantly reduced expression of Piezo1 specifically in the bone , but not other tissues ( Figure 2b , c ) , consistent with the specific knockout of Piezo1 in the bone . Furthermore , osteoblasts that were derived from Piezo1Ocn/Ocn mice also had drastically reduced expression of Piezo1 ( Figure 2d , e ) . Notably , the reduction of Piezo1 was more complete in the osteoblasts than in the whole bone tissue , indicating the possible expression of Piezo1 in non-osteoblast-lineage cells ( Figure 2b–e ) , which could not be ablated by using Ocn-Cre mice . We quantitatively compared the mechanically evoked currents of primary osteoblasts isolated from Piezo1fl/fl and Piezo1Ocn/Ocn mice . Consistent with the recordings from MC3T3-E1 cells , Piezo1fl/fl osteoblasts displayed poking-induced currents in a step-dependent manner with a maximal current of 152 . 3 ± 21 . 6 pA ( Figure 2f , g ) . The inactivation Tau of the current is 32 . 8 ± 4 . 1 ms ( Figure 2f , h ) . The maximal currents of the Piezo1Ocn/Ocn osteoblasts were significantly reduced to 66 . 0 ± 14 . 8 pA ( Figure 2f , g ) , while the inactivation Tau was unchanged ( 27 . 8 ± 3 . 9 ms ) ( Figure 2f , h ) . The remaining mechanically evoked currents could be due to incomplete knockout of Piezo1 , as residual Piezo1 proteins were detected in the Piezo1Ocn/Ocn osteoblasts ( Figure 2e ) and the excision rate in osteoblasts of Ocn-Cre mice was estimated to be ~88% ( Zhang et al . , 2002 ) . Alternatively , other mechanosensitive channels that are independent of Piezo1 might account for the remaining mechanically activated currents in the Piezo1Ocn/Ocn osteoblasts . Nevertheless , these data demonstrate that Piezo1 is expressed and functionally mediates mechanically evoked responses in osteoblasts . Piezo1 is a non-selective cation channel that allows Ca2+ influx and initiation of downstream Ca2+ signaling events upon its opening . We therefore assayed whether Piezo1 mediates Ca2+ influx in osteoblasts . Using single-cell Ca2+ imaging with the ratiometric Ca2+ dye Fura2 , we found that Yoda1 , a previously identified Piezo1 chemical activator ( Syeda et al . , 2015 ) , induced a Ca2+ response in WT osteoblasts , which was drastically reduced in Piezo1Ocn/Ocn osteoblasts ( Figure 2i , j ) . Previous studies have shown that Ca2+ influx could lead to phosphorylation of CaMKII and activate the Creb pathway , promoting osteoblast differentiation ( Choi et al . , 2013; Zayzafoon et al . , 2005 ) . A Piezo-dependent Ca2+-CaMKII signaling pathway has been reported in the axon regeneration process ( Song et al . , 2019 ) . In line with these previous findings , we found that the phosphorylation of CaMKII and Creb was apparently reduced in osteoblasts derived from the Piezo1Ocn/Ocn cells ( Figure 2k ) . Furthermore , Runx2 and Atf4 , key transcription factors involved in osteoblast differentiation , were downregulated in the Piezo1Ocn/Ocn cells ( Figure 2k ) . Consistently , the Piezo1Ocn/Ocn osteoblasts showed decreased expression of the differentiation marker genes ( Figure 2l ) and reduced Alp activity ( Figure 2m ) , indicating impaired osteogenesis . To investigate whether Piezo1 plays an autonomous role in osteoblasts instead of in osteoblast precursors , primarily cultured osteoblasts derived from Piezo1fl/fl mice were transfected with the pIRES-EGFP control plasmid ( Ctrl ) or the pCAG-Cre-IRES2-GFP ( Cre ) plasmid to delete Piezo1 . The expression level of Piezo1 in cells transfected with the Cre plasmid was reduced to ~60% of that in cells transfected with the Ctrl plasmid , which is in line with the transfection efficiency ( Figure 2—figure supplement 1a , b ) . The maximal current of cells transfected with the control plasmid was 102 . 6 ± 13 . 97 pA , whereas in the Piezo1fl/fl osteoblasts transfected with the Cre plasmid the maximum current was reduced to 32 . 72 ± 5 . 7 pA ( Figure 2—figure supplement 1c , d ) . Consistent with the results obtained from the Piezo1Ocn/Ocn osteoblasts , the phosphorylation levels of CaMKII and Creb , as well as the levels of Runx2 and Atf4 , were reduced in the Piezo1fl/fl osteoblasts that were transfected with the Cre plasmid ( Figure 2—figure supplement 1e ) . Furthermore , the Piezo1fl/fl osteoblasts that were transfected with the Cre plasmid showed decreased expression of the differentiation marker genes ( Figure 2—figure supplement 1f ) , as well as reduced Alp activity and mineral deposition ( Figure 2—figure supplement 1g , h ) . Given that Ocn-Cre mice express the Cre recombinase under the osteocalcin ( Ocn ) gene promoter in osteoblast-lineage cells , we also examined the expression and function of Piezo1 in primary osteocytes isolated from the Piezo1fl/fl and Piezo1Ocn/Ocn mice . Indeed , osteocytes derived from the Piezo1Ocn/Ocn mice exhibited a reduced level of Piezo1 ( Figure 2—figure supplement 2a , b ) . Piezo1fl/fl osteocytes displayed poking-induced currents in a step-dependent manner with a maximal current of 64 . 9 ± 13 . 7 pA , which was significantly reduced to 34 . 2 ± 4 . 1 pA in the Piezo1Ocn/Ocn osteocytes ( Figure 2—figure supplement 2c , d ) . It has been shown that osteocytes were able to coordinate osteogenesis in response to mechanical stimulation through the regulation of Sost level ( Robling et al . , 2008 ) . We found that Sost expression was upregulated in bone tissue derived from Piezo1Ocn/Ocn mice compared to that from Piezo1fl/fl mice , which is consistent with the reduced osteoblast function observed in Piezo1Ocn/Ocn mice . Collectively , these data suggest that Piezo1 functions as a critical mechanotransduction channel in osteoblast-derived lineage cells in bone , including both osteoblasts and osteocytes . We next examined the in vivo role of Piezo1 in bone formation . Alizarin red and Alcian blue staining revealed that the newborn Piezo1Ocn/Ocn mice had skeletal size similar to that of their Piezo1fl/fl littermates ( Figure 3a ) . However , the Piezo1Ocn/Ocn mice exhibited incomplete closure of the cranial structure ( Figure 3a ) . At 8 weeks of age , the male Piezo1Ocn/Ocn mice showed shorter stature ( Figure 3—figure supplement 1a ) and lower body weight ( Figure 3—figure supplement 1b ) . The length of the femur and tibia of the Piezo1Ocn/Ocn mice was apparently shorter than that of the Piezo1fl/fl control mice ( Figure 3—figure supplement 1c ) . We next carried out micro-CT analysis . The bodyweight-bearing long bones of the Piezo1Ocn/Ocn mice exhibited drastic loss of bone mass , reduced thickness and impaired trabeculation ( Figure 3b , c ) . In the Piezo1Ocn/Ocn mice , bone parameters including the trabecular bone mineral density ( BMD ) , bone volume ( BV/TV ) , trabecular number ( Tb . N ) , trabecular thickness ( Tb . Th ) and cortical bone thickness ( Cort . Th ) were all significantly decreased , whereas the trabecular spacing ( Tb . Sp ) was accordingly increased ( Figure 3d ) . The strength of the long bones of the Piezo1Ocn/Ocn mice was only about half of that of the Piezo1fl/fl control littermates ( Figure 3e ) . Moreover , the rate of bone formation and bone formation rate per area of bone surface were significantly reduced in the Piezo1Ocn/Ocn mice ( Figure 3f ) . Consistent with the bone defects , immunostaining analysis revealed that the expression levels of the osteoblast differentiation markers , including Col1α1 and Ocn , were reduced in the tibias of the Piezo1Ocn/Ocn mice ( Figure 3g ) . Accordingly , the levels of Ocn and PINP ( N-Propeptide of type I Procollagen ) were significantly decreased in serum derived from the Piezo1Ocn/Ocn mice ( Figure 3h ) . Furthermore , the expression of the osteoblast differentiation marker genes was significantly decreased in femurs isolated from the Piezo1Ocn/Ocn mice ( Figure 3i ) . However , the osteoclast activity remained similar in the Piezo1fl/fl and Piezo1Ocn/Ocn mice ( Figure 3—figure supplement 1d–f ) , suggesting an osteoblast-lineage-specific defect in bone formation in the Piezo1Ocn/Ocn mice . As shown in Figure 4 , the Piezo1Ocn/Ocn mice examined at 16 weeks of age showed bone defects that were essentially similar to those of mice at 8 weeks of age ( Figure 3 ) . We also analyzed the in vivo role of Piezo1 in female mice . In the Piezo1Ocn/Ocn female mice , the expression of Piezo1 was reduced ( Figure 3—figure supplement 2a , b ) . The bone parameters including the BMD , BV/TV , Tb . N , Tb . Th and Cort . Th were all significantly decreased , whereas the Tb . Sp was accordingly increased ( Figure 3—figure supplement 2c , d ) . The expression of the osteoblast differentiation marker genes was significantly reduced in femurs isolated from the Piezo1Ocn/Ocn female mice ( Figure 3—figure supplement 2e ) . Thus , the Piezo1Ocn/Ocn female mice showed a phenotype similar to that of male mice . Taken together , these data suggest that Piezo1 deficiency in osteoblast-lineage cells significantly impairs the formation and structural integrity of the bone . To determine whether a Piezo1-mediated mechanical response in the bone is responsible for mechanical-unloading-induced bone loss , we employed the commonly used hindlimb suspension ( HS ) model ( Wang et al . , 2013; Xu et al . , 2017 ) to examine the bone remodeling process in response to the weight-bearing unloading of the Piezo1fl/fl and Piezo1Ocn/Ocn mice . When the Piezo1fl/fl mice were subjected to 28 days of HS , the trabecular bone mass and architecture related parameters , including BMD , BV/TV , Tb . N and Tb . Th , were significantly reduced ( Figure 4a , b ) . These HS-induced phenotypes of the Piezo1fl/fl mice essentially resemble the bone deficits observed in the Piezo1Ocn/Ocn mice without HS treatment ( Figure 4a , b ) . In contrast to the Piezo1fl/fl mice , the Piezo1Ocn/Ocn mice that were subjected to the HS treatment did not show a worsened phenotype in its already impaired bone ( Figure 4a , b ) . Furthermore , HS treatment led to a significant reduction in the bone strength of the hindlimb of the Piezo1fl/fl mice , but not the Piezo1Ocn/Ocn mice ( Figure 4c ) . We observed the corresponding change in the osteoblast function . HS-induced reduction of Ocn and Col1α1 staining in bone tissues ( Figure 4d ) , serum PINP and Ocn levels ( Figure 4e ) , as well as the expression of the differentiation marker genes of osteoblasts ( Figure 4f ) , were specifically observed in the Piezo1fl/fl mice but not in the Piezo1Ocn/Ocn mice . These results demonstrate that the Piezo1fl/fl mice show drastic mechanical-unloading-induced remodeling of the bone , whereas the Piezo1Ocn/Ocn mice are essentially resistant to such remodeling , suggesting that Piezo1 functions as a critical mechanotransducer for the mediation of proper mechanical-load-induced bone remodeling . Given the observation that the HS-treated WT mice essentially recapitulated the defective bone phenotypes and osteoblast dysfunction of the Piezo1Ocn/Ocn mice , we asked whether mechanical unloading might lead to decreased expression of Piezo1 in the bone tissue derived from the Piezo1Ocn/Ocn mice . Indeed , the mRNA and protein levels of Piezo1 were significantly reduced in the Piezo1fl/fl mice , but there was no further reduction in these levels in the Piezo1Ocn/Ocn mice after treatment with HS for 28 days ( Figure 5a , b ) . To examine whether mechanical unloading directly alters Piezo1 expression in osteoblasts , we utilized a cell rotation system to generate a microgravity condition that simulated the effect of mechanical unloading on osteoblasts . Intriguingly , when subjected to the simulated microgravity treatment , the primarily derived osteoblasts had significantly reduced expression of Piezo1 ( Figure 5c , d ) . Furthermore , microgravity-treated osteoblasts showed decreased mechanically activated currents ( Figure 5e , f ) . In line with the importance of Piezo1 in determining osteoblast activities , the simulated microgravity treatment led to significantly reduced expression of osteoblast marker genes ( Figure 5g ) and Alp activity ( Figure 5h ) in the Piezo1fl/fl cells , but not in the Piezo1Ocn/Ocn cells . These data suggest that mechanical unloading can affect the expression of Piezo1 , resulting in dysfunction of osteoblasts and bone formation . To determine whether Piezo1-mediated mechanical response in the bone is responsible for mechanical-loading-induced bone formation , we subjected the Piezo1fl/fl and Piezo1Ocn/Ocn mice to an exercise model involving a treadmill for 21 days ( Wallace et al . , 2007 ) . The mRNA and protein levels of Piezo1 were significantly increased in the Piezo1fl/fl mice , but not in the Piezo1Ocn/Ocn mice ( Figure 6a , b ) . Exercise treatment led to a significant increase in the expression of the osteoblast function marker genes in the Piezo1fl/fl mice , but not in the Piezo1Ocn/Ocn mice ( Figure 6c ) . These data suggest that Piezo1 might respond to mechanical loading in a way that affects bone function . To further examine whether mechanical force directly alters Piezo1 expression in osteoblasts , we utilized fluid shear stress ( FSS ) to stimulate osteoblasts . When subjected to 12 dyn/cm2 FSS treatment for 2 hr , the primary osteoblasts derived from Piezo1fl/fl mice had significantly increased expression of Piezo1 ( Figure 6d , e ) and of the osteoblast marker genes ( Figure 6f ) , as well as increased Alp activity ( Figure 6g ) . By contrast , the FSS-induced effect was not observed in the Piezo1Ocn/Ocn cells . These data suggest that Piezo1 can sense mechanical force and can consequently regulate its own expression , osteoblast function and bone formation . The close relationship between PIEZO1 expression and function in osteoblasts and bone formation prompted us to explore the pathological role of Piezo1 in human osteoporosis . We examined the expression of Piezo1 in femurs from patients with fractures ( Supplementary file 1 ) . Interestingly , the mRNA and protein expression levels of Piezo1 in osteoporosis patients ( T ≤ −2 . 5 ) were significantly lower than those in control patients ( T > −2 . 5 ) ( Figure 7a , b ) . Furthermore , the expression of Piezo1 was positively correlated with the expression of the differential marker genes of osteoblasts , including ALP , BGLAP and COL1A1 , in these human samples ( Figure 7c ) . By contrast , there are no correlations between the expression of Piezo1 and that of the osteoclast marker genes , including Cathepsin K ( CTSK ) , Tartrate-resistant acid phosphatase ( ACP5 ) and Matrix metalloproteinase 9 ( MMP9 ) ( Figure 7d ) . However , there are also no correlations between the expression of Piezo1 and that of the osteocyte marker genes , including Dentin matrix acidic phosphoprotein 1 ( DMP1 ) and Sclerostin ( SOST ) ( Figure 7e ) . These data are consistent with the critical role of Piezo1 in osteogenesis and bone formation observed in mouse models .
Bone undertakes a life-time mechanical-loading-induced remodeling process . However , it remains unclear how the bone tissue directly senses mechanical changes . In the present study , we have focused on characterizing the expression and function of the mechanically activated Piezo1 channel in bone-forming osteoblasts . Piezo1 is characteristically activated by various forms of mechanical stimulation , including poking , stretching , shear stress and substrate deflection ( Coste et al . , 2010; Cox et al . , 2016; Lewis and Grandl , 2015; Poole et al . , 2014 ) , and mediates Ca2+ influx upon opening to initiate downstream Ca2+ signaling , such as the activation of Ca2+-dependent calpain ( Li et al . , 2014 ) , eNOS ( Wang et al . , 2016 ) and CaMKII ( Coste et al . , 2010; Cox et al . , 2016; Lewis and Grandl , 2015; Song et al . , 2019 ) . Piezo1 was directly activated by asymmetric membrane curvature ( Syeda et al . , 2016 ) and lateral membrane tension ( Cox et al . , 2016; Lewis and Grandl , 2015 ) . In line with being an exquisite mechanosensitive cation channel , the full-length 2547-residue mouse Piezo1 forms a remarkable homo-trimeric three-bladed , propeller-like architecture , comprising a central ion-conducting pore module and three highly curved and non-planar blades , each of which is composed of nine transmembrane helical units ( THUs ) each of four transmembrane helices ( TM ) that are connected to the pore by a 9 nm-long beam-like structure ( Ge et al . , 2015; Guo and MacKinnon , 2017; Saotome et al . , 2018; Zhao et al . , 2018 ) . The blade-beam structure might form a lever-like intramolecular transduction pathway for long-distance mechanical gating of the central pore ( Wang et al . , 2018; Zhao et al . , 2018 ) . Bone-residing osteoblasts might constantly experience mechanical forces such as shear stress and gravity changes . Here , we have found that Piezo1 is expressed in primary osteoblasts and osteocytes , mediates mechanically activated cationic currents and Yoda1 ( a Piezo1 chemical activator ) -induced Ca2+ influx , and controls the osteogenesis process via downstream Ca2+ signaling pathways ( Figure 2 ) . Thus , Piezo1 plays an important role in converting mechanical stimuli into osteogenesis of osteoblasts . Strikingly , either loss of Piezo1 in the Piezo1Ocn/Ocn mice or HS-treatment of the Piezo1fl/fl mice led to severely impaired osteogenesis and bone formation , integrity and strength ( Figures 3 and 4 ) , demonstrating the reciprocal relationship between mechanical loading and the Piezo1 channel in determining the mechanotransduction process during bone formation and remodeling . A further highlight of this relationship is the influence of mechanical loading on the expression level of Piezo1 ( Figure 5 and Figure 6 ) . Both microgravity treatment of osteoblasts and mechanical unloading of the mice reduced the expression level of Piezo1 . By contrast , fluid shear stress treatment osteoblasts and mechanical loading of mice increased the expression level of Piezo1 ( Figure 5 and Figure 6 ) . Consistently , previous studies have shown that hydrostatic pressure enhanced the expression of Piezo1 in mesenchymal stem cells and promoted osteogenesis ( Sugimoto et al . , 2017 ) . Furthermore , a stiffer mechanical microenvironment increased the expression of Piezo1 in glioma cells , which in turn enhanced tissue stiffness ( Chen et al . , 2018 ) . Collectively , these data are consistent with a positive feedback loop between the Piezo1 mechanosensor and the mechanical loading experienced by the mechanosensitive cells and organs . Bone is highly sensitive to changes of daily mechanical loading and gravity . It has been documented that bone mineral density decreased at 1% per month at the lumbar spine and 1–1 . 6% per month at the hip in the crew members of the international space station ( LeBlanc et al . , 2000; Vico and Hargens , 2018 ) . Thus , bone loss is one of the most serious problems induced by long-term weightlessness during space flight or in bedridden individuals . The revelation of the positive feedback relationship between Piezo1 and mechanical loading in bone remodeling provides a mechanistic explanation for mechanical-unloading-induced bone loss . On the basis of the results from the simulated microgravity experiments ( Figure 5 ) , the lack of gravitational force or mechanical loading during long-term spaceflight or in bedridden individuals might decrease the expression and function of Piezo1 in osteoblasts , which in turn leads to impaired osteogenesis and bone loss . In line with this , we found an apparent correlation between the expression level of Piezo1 and the bone loss in osteoporosis patients ( Figure 7 ) . Importantly , both the mechanical-unloading-induced bone loss and the mechanical-loading-induced osteogenesis were blunted in Piezo1Ocn/Ocn mice ( Figure 4 ) , suggesting a causal involvement of Piezo1 in bone remodeling . Thus , activating Piezo1 in the bone might represent a novel strategy for preventing or treating mechanical-unloading-induced bone loss . The success in identifying Piezo1 chemical activators such as Yoda1 ( Syeda et al . , 2015 ) , Jedi1 and Jedi2 ( Wang et al . , 2018 ) appears to show the promise to be fulfilled in developing Piezo1-based therapeutics . Given that the Ocn-Cre mice used in the study could also drive Cre expression in osteocytes ( Zhang et al . , 2002 ) , which are derived from osteoblasts and also considered to be mechanosensitive cells in the bone , the observed defects in the bone formation of the Piezo1Ocn/Ocn mice could also be contributed by the expression of Piezo1 in osteocytes . Indeed , we have found that Piezo1 also functions in the osteocytes ( Figure 2—figure supplement 2 ) . The specific contribution of osteocyte-expressing Piezo1 in bone formation remains to be determined by using osteocyte-specific Cre lines in future studies . Nevertheless , we have demonstrated that Piezo1 plays a critical role in controlling the formation and mechanical-loading-dependent remodeling of the bone in mouse models and it is closely related with the occurrence of osteoporosis in human patients .
To delete Piezo1 specifically in osteoblasts , the conditional KO mice were generated by crossing the Piezo1 floxed mice ( Piezo1fl/fl ) ( a generous gift from Dr . Ardem Patapoutian ) ( Cahalan et al . , 2015 ) with the Ocn-Cre transgenic mice ( a generous gift from Dr . Xiaochun Bai ) ( Huang et al . , 2016 ) . We selected Piezo1Ocn/Ocn as experimental mice , Piezo1fl/fl littermates served as controls . The newborn mice were analyzed by polymerase chain reaction genotyping using genomic DNA from the tail . All animal studies were performed according to approved guidelines for the use and care of live animals ( Guideline on Administration of Laboratory Animals released in 1988 , and 2006 Guideline on Humane Treatment of Laboratory Animals from China ) . All the experimental procedures were approved by the Committees of Animal Ethics and Experimental Safety of the China Astronaut Research and Training Center ( Reference number: ACC-IACUC-2017–003 ) . The mouse osteoblast cell line MC3T3-E1 was maintained in minimum essential Eagle’s medium , α modification ( α-MEM ) ( Gibco ) containing 10% fetal bovine serum ( Gibco ) , 100 units/ml penicillin G , and 100 μg / ml streptomycin ( Gibco ) at 37°C with 5% CO2 . Osteogenic medium was prepared by supplementing the maintenance medium with 10 nM dexamethasone ( Sigma ) , 50 µg/ml of ascorbic acid ( Sigma ) and 10 mM β-glycerophosphate ( Sigma ) . The murine osteoclast cell line RAW 264 . 7 was maintained in Dulbecco's modified Eagle's medium ( DMEM ) with 10% fetal bovine serum , 100 units/ml penicillin G , and 100 μg / ml streptomycin . Osteoclast-induced medium was prepared by supplementing the maintenance medium with 50 ng/ml recombinant mouse RANKL protein ( R and D ) . All of the cell lines that were utilized were Mycoplasma-free , as determined by Q-PCR analyses . Cell line identity was validated by the vendors . Primary osteoblasts were isolated from the calvarial bone of newborn ( 1–3 d ) mice by enzymatic digestion in α-MEM with 0 . 1% collagenase and 0 . 2% dispase as described ( Zhao et al . , 2018 ) , and were cultured in α-MEM with 10% FBS . After 2 days , cells were reseeded and cultured in osteogenic medium for the osteoblast differentiation assay . Primary osteocytes were obtained after removal of the bone marrow , and then sequential collagenase and EDTA digestions of the long bone ( Stern et al . , 2012 ) . Tibias and femurs of 7–9-week-old mice were cleaned to remove muscle and connective tissue . Epiphyses were cut , bone marrow was flushed and the bone was cut into 1-mm to 2-mm lengths . These fragments were incubated in 1 mg/ml collagenase solution for 30 min at 37°C and this cell suspension was discarded . This was repeated two more times , for a total of three digestions . The remaining bone fragments were washed with PBS and incubated for 30 min at 37°C with EDTA ( 5 mM , PH 7 . 4 ) in PBS . Cell suspension was again discarded , and bone fragments were washed with PBS . This was repeated two more times for the incubation of bone chips with collagenase and EDTA . Cell suspension was again discarded and the bone fragments were finally incubated with 1 mg/ml collagenase for 30 min at 37°C . Cells were collected , passed through a 70-μm nylon mesh and washed twice . These cells were used for subsequent testing . Cells for RNA interference were transfected with Piezo1 siRNA or NC at 70% confluence using Lipofectamine RNAiMAX in OptiMEM as per the manufacturer’s instructions ( Invitrogen ) . Sequences of the siRNA probes were as follows: Piezo1 negative control siRNA ( NC ) , 5′-UUCUCCGAACGUGUCACGUTT-3′; Piezo1 siRNA , 5′-CACCGGCATCTACG TCAAATA-3′ . pIRES-EGFP and pCAG-Cre-IRES2-GFP plasmid transfection , cells at 70% confluence were transfected with 500 ng pIRES-EGFP ( Ctrl ) and pCAG-Cre-IRES2-GFP ( Cre ) plasmid using Lipofectamine 3000 in OptiMEM as per the manufacturer’s instructions ( Invitrogen ) , cells were analyzed for QRT-PCR , Western Blot and whole cell electrophysiology mechanical stimulation after transfection for 48 hr . Later , cells were analyzed for alp staining after transfection for 5 days and for Alizarin red staining after transfection for 14 days . MC3T3-E1 , RAW 264 . 7 , primary osteoblasts or osteocytes isolated from Piezo1fl/fl and Piezo1Ocn/Ocn mice were cultured on 5 cm2 coverslips . For pCAG-Cre-IRES2-GFP-mediated knockout experiments , the primary osteoblasts derived from Piezo1fl/fl mice were transfected with pIRES-EGFP and pCAG-Cre-IRES2-GFP plasmid for 48 hr , and cells with green fluorescence were isolated for electrophysiological recording . For simulated microgravity treatment , cells were cultured on coverslips in a six-well plate for 24 hr . Next , the coverslips were transferred to a cell cabinet , which was filled up with medium and sealed to prevent air bubbles . The cell cabinet was incubated in the clinostat system , with the clinostat being continuously rotated at 30 rpm/min , at 37°C for 16 hr . The control group was cultured in the same manner without clinorotation . Cells that were grown on coverslips were directly subjected to electrophysiological recording . All experiments were performed at room temperature ( 22–25°C ) . Mechanical stimulation was delivered to the cell during the patch-clamp recording at an angle of 80° using a fire-polished glass pipette ( tip diameter 3–4 μm ) as previously described ( Zhao et al . , 2016 ) . The downward movement of the probe toward the cell was driven by a Clampex controlled piezo-electric crystal micro-stage ( E625 LVPZT Controller/Amplifier; Physik Instrument ) . The probe had a velocity of 1 μm/ms during the downward and upward motion , and the stimulus was maintained for 150 ms . A series of mechanical steps in 1 μm increments was applied every 20 s , and the currents were recorded at a holding potential of −60 mV . GsMTx4 ( Tocris Bioscience ) at a concentration of 4 μM was added to the recording chamber for 30 min before the recording . For siRNA-mediated knockdown experiments , the MC3T3-E1 cells were transfected with 100 nM siRNA using Lipofectamine 3000 ( Invitrogen ) following the manufacturer’s instructions . After 4 to 6 hr , the medium was replaced with fresh α-MEM medium . 48 hr after transfection , the cells were dissociated with trypsin EDTA ( GIBCO ) and triturated in α-MEM medium ( Life Technologies ) supplemented with 10% bovine serum . The resulting cells were placed on coverslips and cultured in a humidified incubator in 5% CO2 at 37°C for 4 hr before the start of recording . Primary osteoblasts derived from WT and Piezo1 cKO mice and grown on coverslips were subjected to single-cell Fura-2 Ca2+ imaging as described previously ( Wang et al . , 2018 ) . The Yoda1-induced amplitude change of the 340/380 fluorescence ratio was calculated by subtracting the baseline ratio prior to Yoda1 application . Yoda1 was solubilized in DMSO as a stock solution of 30 mM and diluted to a final concentration of 30 μM using the Ca2+ imaging buffer . Total RNA from bone tissues or cells was extracted with TRIzol Reagent ( Invitrogen ) according to the manufacturer's instructions . RNA ( 0 . 5 μg ) was reverse transcribed with a PrimeScript RT reagent kit ( TaKaRa ) according to the manufacturer's instructions . cDNA were used to detect mRNA expression by quantitative PCR using TB Green Premix Ex Taq II ( Tli RNaseH Plus ) ( TaKaRa ) . Gapdh was used as a normal control for mRNA . Primers used are listed in Supplementary file 2 . Cells were lysed in lysis buffer ( 50 mM Tris , pH 7 . 5 , 250 mM NaCl , 0 . 1% SDS , 2 mM dithiothreitol ( DTT ) , 0 . 5% NP-40 , 1 mM PMSF and protease inhibitor cocktail ) on ice for 30 min . Bone tissues were ground with a mortar in liquid nitrogen and were lysed in lysis buffer at 4°C for 30 min . Protein fractions were collected by centrifugation at 12 , 000 g , 4°C for 10 min and then 10 μg of lysates were subjected to SDS-PAGE and transferred to polyvinylidene difluoride ( PVDF ) membranes . The membranes were blocked with 5% skimmed milk and incubated with specific antibodies overnight . We used the following antibodies to examine the protein levels in the lysates: Piezo1 ( Proteintech , Cat# 15939–1-AP ) , p-CamkII ( CST , Cat# 12716 s ) , CamkII ( CST , Cat# 4436 s ) , p-Creb ( CST , Cat# 9198 s ) , Creb ( CST , Cat# 9197 s ) , Runx2 ( CST , Cat# 12556 s ) , ATF4 ( CST , Cat# 11815 s ) and Gapdh ( ZSGB-BIO , Cat# TA-08 ) . The ratios of the protein-band intensities relative to that of Gapdh were calculated for each sample using Image J . Alkaline phosphatase staining was monitored using a Vector Blue substrate kit ( procedure number SK-5300 , Vector Laboratories ) . According to the protocol , MC3T3-E1 or primary osteoblasts were incubated with the substrate working solution for 20–30 min . The whole procedure was protected from light . Cells were fixed in 4% paraformaldehyde for 5 min and rinsed with double-distilled H2O . Cells were stained with 40 mM Alizarin red S ( Sigma ) , pH 4 . 0 , for 30 min with gentle agitation . Cells were rinsed five times with double-distilled H2O and then rinsed for 15 min using 1 × PBS with gently agitation . For skeletal preparation , whole-mount skeletal preparations of 5–7-day-old Piezo1fl/fl and Piezo1Ocn/Ocn mice were prepared by removing the skin and internal organs of the mice before immersion in 95% ethanol for 1–3 days . Specimens were stained with 0 . 015% Alcian Blue ( Sigma ) in 80% ethanol with 20% acetic acid . After staining , specimens were washed twice in 95% alcohol for 2 hr , cleared in 1% KOH for 5 hr and stained in 0 . 005% Alizarin red ( Sigma ) in 1% KOH for 1 hr . They were then cleared through 20% , 50% , and 80% glycerine in 1% KOH , then stored in 100% glycerine . For the distal femur , the whole secondary spongiosa of the left distal femur from each mouse was scanned ex vivo using a microCT system ( µCT40 , SCANCO MEDICAL , Switzerland ) . Briefly , 640 slices with a voxel size of 10 µm were scanned in the region of the distal femur beginning at the growth plate and extending proximally along the femur diaphysis . Eighty continuous slices beginning at 0 . 1 mm from the most proximal aspect of the growth plate in which both condyles were no longer visible were selected for analysis . Cortical bone measurements were performed in the diaphyseal region of the femur starting at a distance of 5 . 0 mm from the distal growth plate and extending a further longitudinal distance of 80 slices in the proximal direction . Immediately after the dissection , the femurs were stored in 70% ethanol . Before mechanical testing , the bones were rinsed in PBS . The three-point bending test ( span length , 4 . 0 mm; loading speed 0 . 50 mm/s ) at the mid femur was made using Texture Analyzer Texture Pro CT V1 . 6 Build , Brookfield Engineering Labs Inc . The mice were injected intraperitoneally with calcein green ( 10 mg/kg body weight ) in a time sequence of 10 d and 2 d before euthanasia . The tibias were harvested for undecalcified histology analysis . Unstained 15 µm sections were examined using fluorescence microscopy . Statistical analyses were performed with the Osteomeasure Analysis System . The tibias of mice were fixed with 4% buffered formalin and embedded with paraffin after decalcification with 10% EDTA for 10–15 days , and 5–7 µm sections were prepared on a rotation microtome . Paraffin-embedded sections were deparaffinized in xylene , and rehydrated . Antigen retrieval was performed in citrate buffer ( pH 6 . 0 ) for 15 min at 94–96°C . The sections were incubated with 5% goat serum for 1 hr at room temperature . The samples were stained with Col1α1 ( abcam , Cat# ab64883 ) and Ocn ( proteintech , Cat# 23418–1-AP ) antibody overnight at 4°C . After three washes in PBS , corresponding biotinylated secondary antibodies were then added and incubated for 30 min at room temperature . Negative-control experiments were carried out by omitting the primary antibodies . DAB ( ZSGB-bio ) was used as chromogen , and hematoxylin was used to counterstain . Statistical analyses were performed with the Osteomeasure Analysis System . The analyses were performed according to the manufacturer’s instructions for serum concentrations of PINP ( ELISA , Immunoway ) , Ocn ( ELISA , NOVUS ) and CTX-1 ( ELISA , Sangon Biotech ) . In brief , 50 µl of serum was pipetted in duplicate into the wells of the precoated ELISA plate . Then , 50 µl of antibody solution was added to each well and incubated at room temperature for 90 min on a shaking device . After incubation , the plates were washed three times with wash buffer . The hindlimb-unloading procedure was achieved by tail suspension . Briefly , the 3 month Piezo1fl/fl or Piezo1Ocn/Ocn mice were individually caged and suspended by the tail using a strip of adhesive surgical tape attached to a chain hanging from a pulley . The mice were suspended at a 30° angle to the floor with only the forelimbs touching the floor; this allowed the mice to move and to access food and water freely . The mice were subjected to hindlimb unloading through tail suspension for 28 d . After euthanasia , the whole bone tissues were collected . All animal studies were performed according to approved guidelines for the use and care of live animals ( Guideline on Administration of Laboratory Animals released in 1988 , and 2006 Guideline on Humane Treatment of Laboratory Animals from China ) . All of the experimental procedures were approved by the Committees of Animal Ethics and Experimental Safety of the China Astronaut Research and Training Center ( Reference number: ACC-IACUC-2017–003 ) . To simulate microgravity , we used a 2D clinostat , which was designed and provided by the China Astronaut Research and Training Center ( Beijing , China ) . Rotation causes a gravity vector that is not recognizable by cells . Therefore , the device prevents the cells from feeling gravity . In the present study , cells were seeded at a density of 1 × 106 cells in 25 cm2 cell-culture flasks or 2 × 105 cells on 5 cm2 coverslips adhered to the flask . After cell adhesion , flasks or flasks mounted with coverslips were filled up with culture medium to prevent the presence of air bubbles . The dishes were fixed carefully to the rotating panel of the clinostat system , and rotated at a constant speed of 30 rpm/min for 24 hr to simulate microgravity ( 0 . 01 g ) . For the control , cells were cultured in the same chamber mounted on a stationary clinostat ( 1 g ) . At 4 months of age , mice were divided into four groups: Ctrl- Piezo1fl/fl , Exercise ( Ex ) -Piezo1fl/fl , Ctrl-Piezo1Ocn/Ocn and Exercise ( Ex ) -Piezo1Ocn/Ocn . Each exercise group was subjected to running on a treadmill ( Zhishuduobao , Beijing , China ) at a 5° incline and speed of 12 m/min , 30 min/day for 21 consecutive days . All animal studies were performed according to approved guidelines for the use and care of live animals ( Guideline on Administration of Laboratory Animals released in 1988 , and 2006 Guideline on Humane Treatment of Laboratory Animals from China ) . All of the experimental procedures were approved by the Committees of Animal Ethics and Experimental Safety of the China Astronaut Research and Training Center ( Reference number: ACC-IACUC-2017–003 ) . Fluid flow was applied to cells in a parallel plate flow chamber using a closed flow loop . Cells were plated on 22 × 26 mm glass cover slips and placed into chambers at 80% confluence . After treatment with 12 dyn/cm2 FSS for 2 hr , the apparatus was maintained at 37 °C throughout the duration of the experiment . The correlation between FSS and flow rate was calculated using the equation: τ = 6μQ/bh2 , where Q is the flow rate ( cm3/s ) , µ is the viscosity of the flow media ( 0 . 01 dynes/cm2 ) , h is the height of the channel ( 0 . 05 cm ) , b is the slit width ( 2 . 5 cm ) , and τ is the wall shear stress ( dyne/cm2 ) . The bone tissues of 10 osteoporotic patients and 10 non-osteoporotic people were collected from a clinical setting . The patients were recruited at between 65 and 90 years of age . The classification of the patients into the osteoporotic and non-osteoporotic groups was based on DXA evaluation . We measured the T-score for BMD in the spine of women . A T-score of −2 . 5 or lower qualifies as osteoporosis . Others were control patients ( T > −2 . 5 ) . We obtained informed consent from all participants . The study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki . All clinical procedures were approved by the Committees of Clinical Ethics in the Second Affiliated Hospital of Soochow University ( Reference number: 2016 K-22 ) . All numerical data are expressed as the mean ± SEM from at least three independent samples . Student’s t-test was used for statistical evaluations of two group comparisons . Statistical analysis with more than two groups was performed with one-way analysis of variance ( ANOVA ) . All statistical analyses were performed with Prism software ( Graphpad prism for windows , version 6 . 0 ) . p<0 . 05 was considered statistically significant . | The bones in our skeletons are constantly exposed to mechanical forces , including those exerted by our muscles and also Earth’s gravity . These forces normally help osteoblasts , the cells which build new bone tissue , ensure that bones grow correctly and remain strong . Removing mechanical loads from bones , however , disrupts this process , leading to rapid loss of bone tissue . This is why both astronauts in space ( where gravity is much weaker ) and bed-ridden patients often go on to develop brittle bones . To detect and respond to mechanical forces , cells use specialized sensor proteins . One such ‘mechanosensor’ is a protein called Piezo1 , which is found on the surface of many different types of cells in our bodies . It helps cells respond to touch , pressure , or stretching of the surrounding tissue . For example , Piezo1 in nerve cells underpins our sense of touch , while in the cells lining our blood vessels it senses the force exerted by blood flow . Although osteoblasts clearly respond to mechanical stimuli , exactly how they do so has remained unknown . Sun et al . therefore wanted to find out if Piezo1 also acted as a mechanosensor in osteoblasts , and if so , what role it might play in the loss or formation of bone tissue after changes in the amount of force the bone is exposed to . Experiments using mouse cells grown in the laboratory revealed that Piezo1 was present in osteoblasts and did indeed help the cells respond to mechanical impact of being poked by a microscopic probe . Mice that had been genetically engineered to remove Piezo1 from their osteoblasts did not grow properly , appearing stunted in adulthood . In these mice , the bones supporting most of the body’s weight were also shorter and weaker . Crucially , putting normal bone cells in a low-gravity simulator – therefore mimicking space flight – or exposing mice to conditions mimicking bed-rest was enough to reduce the level of Piezo1 in osteoblasts . In human patients with osteoporosis , where bones become brittle with age , a decrease in levels of Piezo1 is correlated with increasing bone loss . These results show that Piezo1 is required to make healthy bone tissue , and that its loss is probably involved in the increasing fragility that occurs when mechanical forces applied to bones are reduced . This work is an important step towards understanding how our bones are built and maintained . In the future , increasing Piezo1 activity within osteoblasts may lead to treatments for bone loss , whether in hospital patients or astronauts . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2019 | The mechanosensitive Piezo1 channel is required for bone formation |
Obesity produces a chronic inflammatory state involving the NFκB pathway , resulting in persistent elevation of the noncanonical IκB kinases IKKε and TBK1 . In this study , we report that these kinases attenuate β-adrenergic signaling in white adipose tissue . Treatment of 3T3-L1 adipocytes with specific inhibitors of these kinases restored β-adrenergic signaling and lipolysis attenuated by TNFα and Poly ( I:C ) . Conversely , overexpression of the kinases reduced induction of Ucp1 , lipolysis , cAMP levels , and phosphorylation of hormone sensitive lipase in response to isoproterenol or forskolin . Noncanonical IKKs reduce catecholamine sensitivity by phosphorylating and activating the major adipocyte phosphodiesterase PDE3B . In vivo inhibition of these kinases by treatment of obese mice with the drug amlexanox reversed obesity-induced catecholamine resistance , and restored PKA signaling in response to injection of a β-3 adrenergic agonist . These studies suggest that by reducing production of cAMP in adipocytes , IKKε and TBK1 may contribute to the repression of energy expenditure during obesity .
Obesity generates a state of chronic , low-grade inflammation in liver and adipose tissue accompanied by macrophage infiltration and the local secretion of inflammatory cytokines and chemokines that attenuate insulin action , resulting in insulin resistance and the subsequent development of Type 2 diabetes ( Wellen and Hotamisligil , 2005; Hotamisligil , 2006; Lumeng et al . , 2007; Shoelson et al . , 2007 ) . Numerous studies indicate a strong correlation between inflammation and insulin resistance across several populations ( Hotamisligil , 2006 ) . Moreover , genetic ablation or pharmacological inhibition of inflammatory pathways can dissociate obesity from insulin resistance ( Hotamisligil , 2006; Shoelson et al . , 2007 ) , suggesting that local inflammation can be a key step in the generation of insulin resistance . The transcription factor NFκB and its inflammatory program play an important role in the development of insulin resistance in obese liver and adipose tissue ( Yuan et al . , 2001; Arkan et al . , 2005; Wunderlich et al . , 2008; Chiang et al . , 2009 ) . NFκB is activated by the IκB kinase ( IKK ) family , which has four members: IKKα , IKKβ , IKKε , and TBK1 . IKKα and IKKβ act with the scaffolding partner NEMO to activate NFκB ( Hacker and Karin , 2006 ) . Although pharmacologic inhibition or genetic ablation of IKKβ defined a role for this kinase in insulin resistance ( Yuan et al . , 2001; Arkan et al . , 2005 ) , the roles of the noncanonical kinases IKKε and TBK1 are less certain . We recently reported that both mRNA and protein expression levels of IKKε and TBK1 are increased in adipose tissue from mice fed a high fat diet ( Chiang et al . , 2009 ) . Both of these kinases are increased as a consequence of the inflammatory program in obesity ( Reilly et al . , 2013 ) , and contain NFκB regulatory sites in their promoter regions , allowing them to be induced upon NFκB activation ( Kravchenko et al . , 2003 ) . Deletion of the IKKε gene rendered mice partially resistant to some of the deleterious effects of high fat feeding , including weight gain , insulin resistance , hepatic steatosis , and systemic inflammation ( Chiang et al . , 2009 ) . We report , in this study , that IKKε and TBK1 can desensitize lipolytic signaling in white adipose tissue in response to β-adrenergic agonists by phosphorylating and increasing the activity of PDE3B , in the process decreasing cAMP levels . Thus , induction of these noncanonical IκB kinases might contribute to catecholamine resistance during obesity , and blocking their activity has the potential to increase energy expenditure as an anti-obesity and anti-diabetes therapy .
Sympathetic activation of adipose tissue plays a key role in maintaining energy balance by stimulating lipolysis and fat oxidation ( Coppack et al . , 1994; Langin , 2006; Festuccia et al . , 2011 ) . Activation of β-adrenergic signaling by either β-adrenergic agonists or cold exposure in white and brown adipose tissue initiates a cascade of events through cyclic AMP ( cAMP ) , culminating in the transcriptional upregulation of Ucp1 , which results in increased proton leak and energy expenditure ( Himms-Hagen et al . , 2000; Cao et al . , 2004; Yehuda-Shnaidman et al . , 2010 ) . Our previous studies revealed that compared to wild-type ( WT ) controls , IKKε-deficient mice exhibited increased energy expenditure while on a high fat diet ( HFD ) , accompanied by increased expression of Ucp1 in white adipose depots ( Chiang et al . , 2009 ) . Interestingly , increased energy expenditure in IKKε-deficient mice was only seen in HFD-fed mice ( Chiang et al . , 2009 ) , suggesting that upon induction of IKKε during obesity , the kinase might repress an increased adaptive thermogenic response to overnutrition . To explore this possibility , we overexpressed IKKε in 3T3-L1 adipocytes and examined Ucp1 gene expression after treatment with the non-selective β-adrenergic agonist , isoproterenol ( ISO ) , or the β3-adrenergic agonist , CL-316 , 243 . Fold difference in Ucp1 gene expression was calculated by normalization of relative Ucp1 mRNA levels in treated relative to control samples . Treatment of empty vector-expressing cells with ISO or CL-316 , 243 resulted in a 1 . 6-fold or twofold increase in Ucp1 mRNA levels , respectively ( Figure 1A ) . The induction of Ucp1 gene expression in response to ISO or CL-316 , 243 was blunted when WT IKKε was overexpressed in these cells . However , expression of the kinase-inactive mutant of IKKε K38A ( Fitzgerald et al . , 2003 ) was less effective , but still modestly repressed Ucp1 expression . 10 . 7554/eLife . 01119 . 003Figure 1 . IKKε and TBK1 overexpression decrease sensitivity to the β-adrenergic/cAMP pathway in 3T3-L1 adipocytes . ( A ) Fold increase in Ucp1 expression in 3T3-L1 adipocytes expressing empty vector , Flag-IKKε , or Flag-IKKε K38A following treatment with or without 10 μM ISO ( black bars ) or 10 μM CL-316 , 243 ( CL , gray bars ) for 4 hr . **p<0 . 01 . Performed in triplicate . ( B ) Glycerol release from 3T3-L1 adipocytes expressing empty vector ( white bars ) , Flag-IKKε ( black bars ) , or Flag-IKKε K38A ( gray bars ) treated with or without 10 μM ISO or 10 μM CL . *p<0 . 05 and **p<0 . 01 . Performed in triplicate . ( C ) Immunoblots of whole cell lysates from Figure 1B . Results were replicated in triplicate . D . E . stands for dark exposure and L . E . stands for light exposure . ( D ) Immunoblots of whole cell lysates from 3T3-L1 adipocytes expressing empty vector or Flag-IKKε treated with or without 50 μM FSK for 15 min . Results were replicated in multiple experiments . ( E ) cAMP levels from 3T3-L1 adipocytes expressing empty vector , Flag-IKKε , or Flag-IKKε K38A treated with or without 10 μM ISO or 50 μM FSK for 15 min . **p<0 . 0001 and *p<0 . 05 . Performed in triplicate . DOI: http://dx . doi . org/10 . 7554/eLife . 01119 . 00310 . 7554/eLife . 01119 . 004Figure 1—figure supplement 1 . IKKε and TBK1 overexpression decrease sensitivity to the β-adrenergic/cAMP pathway in 3T3-L1 adipocytes . ( A ) Immunoblots of whole cell lysates from 3T3-L1 adipocytes expressing empty vector , Flag-IKKε , or Flag-IKKε K38A treated with or without 10 μM ISO for 15 min . Results were replicated in multiple experiments . ( B ) Immunoblots of whole cell lysates from 3T3-L1 adipocytes expressing increasing amounts of Flag-IKKε or Flag-TBK1 treated with or without 10 μM ISO ( top panel ) or 50 μM FSK ( bottom panel ) for 15 min . Results were replicated in multiple experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 01119 . 004 In addition to increased Ucp1 expression , IKKε knockout mice also exhibited increased lipolysis and fat oxidation ( Chiang et al . , 2009 ) , suggesting that decreased lipolysis in adipose tissue from obese mice might result in part from increased expression of IKKε and TBK1 ( Chiang et al . , 2009 ) . We thus modeled the obesity-dependent increase in the noncanonical IKKs by overexpressing IKKε in 3T3-L1 adipocytes , followed by assay of glycerol release in response to ISO or CL-316 , 243 . Although both isoproterenol and CL-316 , 243 increased lipolysis in empty vector-expressing cells , overexpression of WT IKKε reduced the lipolytic effects of isoproterenol and CL-316 , 243 by greater than 40% , and also reduced basal glycerol release ( Figure 1B ) . The reduction in lipolysis by IKKε overexpression was accompanied by dramatically reduced phosphorylation of HSL and perilipin in response to ISO or CL-316 , 243 ( Figure 1C ) . Expression of the catalytically inactive kinase was less effective in blocking lipolytic signaling , although the levels of protein achieved by overexpression were lower compared to the WT kinase ( Figure 1B , C , Figure 1—figure supplement 1A ) . Overexpression of TBK1 reduced phosphorylation of HSL in response to isoproterenol or the adenylyl cyclase activator , forskolin ( Figure 1—figure supplement 1B ) . Identical results were obtained when IKKε was overexpressed in 3T3-L1 adipocytes stimulated with forskolin ( Figure 1D ) , as detected by western blotting with an anti-phospho-PKA substrate motif antibody . Overexpression of IKKε also repressed the phosphorylation of p38 ( p-p38 ) in response to forskolin ( Figure 1D ) or isoproterenol ( Figure 1—figure supplement 1A ) , whereas overexpression of IKKε K38A was without effect ( Figure 1—figure supplement 1A ) . While glycerol release is likely the result of changes in HSL and perilipin phosphorylation , it is important to note that we have not directly assayed whether re-esterification of glycerol intermediates are also affected . Taken together , these data suggest that similar to what is observed in obesity , overexpression of IKKε or TBK1 can repress lipolytic signaling . The partial effectiveness of the kinase-inactive mutants is puzzling , but may reflect their activation of endogenous IKKε or TBK1 kinases due to dimerization ( Larabi et al . , 2013; Tu et al . , 2013 ) . Since PKA signaling is responsible for Ucp1 induction in response to catecholamines ( Klein et al . , 2000; Cao et al . , 2001 ) , we explored the possibility that both IKKε and TBK1 might reduce β-adrenergic sensitivity of adipocytes by decreasing cAMP levels . IKKε overexpression in 3T3-L1 adipocytes reduced by greater than 80% the increase in cAMP levels produced by both isoproterenol and forskolin , whereas overexpression of IKKε K38A did not ( Figure 1E ) . Previous studies have shown that decreased sensitivity to adrenergic stimuli in adipose tissue can result from reduced β-adrenergic receptors ( Reynisdottir et al . , 1994 ) or increased expression of α2-adrenergic receptors ( Stich et al . , 2002 ) . These studies represent the first demonstration that defects distal to the adrenergic receptor may also contribute to catecholamine resistance , and suggest that IKKε and TBK1 can attenuate the β-adrenergic/cAMP pathway in response to β-adrenergic stimuli in adipocytes in a cell-autonomous manner , and further that induction of these kinases during obesity may account for decreased energy expenditure by reducing sensitivity of adipocytes to β-adrenergic stimulation . Obesity is accompanied by infiltration of proinflammatory macrophages into adipose tissue; these cells secrete inflammatory cytokines , such as TNFα , which generate insulin resistance by stimulating catabolic pathways ( Hotamisligil , 2006; Lumeng et al . , 2007; Ye and Keller , 2010; Ouchi et al . , 2011 ) . Although TNFα is known to increase lipolysis in adipocytes ( Zhang et al . , 2002; Souza et al . , 2003; Green et al . , 2004; Plomgaard et al . , 2008 ) , there is also evidence of a counterinflammatory response in obesity that may serve to repress energy expenditure ( Gregor and Hotamisligil , 2011; Saltiel , 2012; Calay and Hotamisligil , 2013; Reilly et al . , 2013 ) . We thus used TNFα to model the inflammatory milieu of obese adipose tissue in cell culture to determine whether the cytokine might also regulate β-adrenergic signaling in this context . While short-term treatment with TNFα augmented the increase in cAMP produced by forskolin treatment , this effect declined after 12 hr . After 24 hr of exposure , TNFα inhibited the production of the second messenger produced by forskolin ( Figure 2—figure supplement 1A ) . Thus , the catabolic effects of the proinflammatory cytokine TNFα in adipocytes are transient and followed by an inhibitory phase . Our previous studies revealed that treatment of 3T3-L1 adipocytes with TNFα for 24 hr induced the expression of IKKε and increased TBK1 phosphorylation at the active site in a manner that was dependent on the activity of IKKβ and the NFκB pathway ( Reilly et al . , 2013 ) . We thus wondered whether the repression of β-adrenergic sensitivity produced by longer-term treatment with TNFα might be due to increased activity of the noncanonical IKKs . Long-term treatment with TNFα repressed the induction of Ucp1 gene expression in response to β-adrenergic stimuli ( Figure 2—figure supplement 1B ) , whereas the expression of IKKε mRNA ( Ikbke ) was upregulated , as previously reported ( Reilly et al . , 2013 ) . Treatment of 3T3-L1 adipocytes with TNFα for 24 hr decreased glycerol release in response to both isoproterenol and forskolin in a dose-dependent manner ( Figure 2A ) . TNFα treatment also decreased isoproterenol- and forskolin-stimulated cAMP production; an effect that was largely rescued by preincubation of cells with the selective , but structurally unrelated inhibitors of IKKε and TBK1 , amlexanox ( Figure 2B ) ( Reilly et al . , 2013 ) or CAY10576 ( Figure 2C ) ( Bamborough et al . , 2006 ) . 10 . 7554/eLife . 01119 . 005Figure 2 . Prolonged treatment with TNFα decreases the sensitivity of adipocytes to β-adrenergic stimulation in a manner dependent on the activity of IKKε and TBK1 . ( A ) Glycerol release from 3T3-L1 adipocytes treated with or without different concentrations of TNFα as indicated for 24 hr followed by treatment with or without 10 μM ISO or 50 μM FSK . **p<0 . 0001 . Performed in quadruplicate . ( B ) cAMP levels from 3T3-L1 adipocytes treated with or without 100 ng/ml TNFα for 24 hr followed by treatment with or without 10 μM ISO or 50 μM FSK in the presence or absence of pretreatment of 50 μM Amlexanox ( Am ) . **p<0 . 0001 . Performed in quadruplicate . ( C ) cAMP levels from 3T3-L1 adipocytes treated with or without 100 ng/ml TNFα for 24 hr followed by treatment with or without 50 μM FSK in the presence or absence of pretreatment of 1 μM CAY10576 ( CAY ) . **p<0 . 0001 . Performed in triplicate . ( D ) Immunoblots of whole cell lysates from 3T3-L1 adipocytes treated with or without different concentrations of TNFα as same as Figure 2A for 24 hr followed by treatment with or without 10 μM ISO or 50 μM FSK . Results were replicated in multiple experiments . ‘[’ indicates total HSL . ‘n . s . ’ represents non-specific band . Arrow indicates CGI-58 . ( E ) Immunoblots of whole cell lysates from 3T3-L1 adipocytes treated with or without 50 ng/ml TNFα or 100 μg/ml poly ( I:C ) for 24 hr followed by treatment with or without 10 μM ISO for 15 min in the presence or absence of pretreatment with increasing concentrations ( 0 , 10 , 50 , and 200 μM ) of amlexanox for 30 min . Results were replicated in multiple experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 01119 . 00510 . 7554/eLife . 01119 . 006Figure 2—figure supplement 1 . Prolonged exposure of inflammatory cytokines decreases the sensitivity of adipocytes to β-adrenergic stimulation . ( A ) cAMP levels from 3T3-L1 adipocytes treated with or without 100 ng/ml TNFα for the indicated amount of time ( left panel: 0–12 hr , right panel: 0–24 hr ) followed by treatment with or without 25 μM FSK . **p<0 . 01 . Performed in duplicate . ( B ) Relative gene expression ( top panel: Ucp1 , bottom panel: Ikbke ) in 3T3-L1 adipocytes treated with or without 100 ng/ml TNFα for 24 hr followed by treatment with or without 10 μM ISO or 50 μM FSK , or 10 μM CL-316 , 243 for 4 hrs . **P<0 . 01 and **p<0 . 0001 . Performed in tripilicate . ( C ) cAMP levels as measured by Glosensor from 3T3-L1 adipocytes treated with or without 5 μg/ml poly ( I:C ) for 24 hr followed by treatment with or without 50 μM FSK over the course of 75 min . Performed in triplicate . ( D ) Glycerol release from 3T3-L1 adipocytes treated with or without different concentrations of poly ( I:C ) as indicated for 24 hr followed by treatment with or without 10 μM ISO or 50 μM FSK . **p<0 . 0001 . Performed in quadruplicate . ( E ) Immunoblots of whole cell lysates from 3T3-L1 adipocytes treated with or without different concentrations of poly ( I:C ) as same as Figure 2—figure supplement 1D for 24 hr followed by treatment with or without 10 μM ISO or 50 μM FSK . Results were replicated in multiple experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 01119 . 006 Isoproterenol-stimulated β-adrenergic signaling was also decreased by treatment of cells with TNFα ( Figure 2D ) , as manifested by decreased phosphorylation of HSL , perilipin , and other proteins recognized by the PKA substrate motif antibody , whereas IKKε expression was concurrently upregulated and TBK1 phosphorylation was increased by the treatment with TNFα . Pretreatment of 3T3-L1 adipocytes with amlexanox also blocked the inhibitory effect of TNFα on isoproterenol-stimulated β-adrenergic signaling , as determined by western blotting with an anti-phospho-PKA substrate motif antibody , anti-phospho-HSL , and anti-phospho-perilipin antibodies ( Figure 2E ) . Interestingly , phosphorylation of p38 in response to isoproterenol was also dramatically augmented by amlexanox in a dose-dependent manner . Previous studies showed that the toll-like receptor 3 ( TLR3 ) agonist , Poly ( I:C ) , results in the direct activation of IKKε and TBK1 ( Hemmi et al . , 2004; Clark et al . , 2009; Clark et al . , 2011 ) . Similar to TNFα , treatment of 3T3-L1 adipocytes with Poly ( I:C ) simultaneously reduced stimulation of cAMP production , lipolysis and phosphorylation in response to β-adrenergic stimulation ( Figure 2—figure supplement 1C–E ) , and the inhibitory effects of Poly ( I:C ) on the sensitivity to isoproterenol stimulation were partially restored by amlexanox pretreatment , but not to the extent that was observed with TNFα treatment ( Figure 2E ) . It is possible that Poly ( I:C ) -induced desensitization of β-adrenergic pathway engages other pathways that are not directly regulated by IKKε and TBK1 . These results suggest that obesity-associated inflammation leads to the activation of IKKε and TBK1 , which produces reduced sensitivity of adipocytes to β-adrenergic stimulation . cAMP levels can also be regulated by phosphodiesterases , which cleave the second messenger and in the process dampen cAMP-dependent signals . Phosphodiesterase 3B ( PDE3B ) is the major PDE isoform expressed in adipocytes ( Zmuda-Trzebiatowska et al . , 2006 ) . Genetic ablation or pharmacological inhibition of PDE3B in cells and in vivo revealed an important role for the enzyme in lipid and glucose metabolism ( Choi et al . , 2006; Berger et al . , 2009; Degerman et al . , 2011 ) . Phosphorylation and activation of PDE3B by insulin in adipocytes is thought to be mediated by Akt , and cAMP itself acts as a negative feedback regulator of its own levels by promoting PKA-dependent phosphorylation and activation of PDE3B ( Degerman et al . , 2011 ) . Since we observed that cAMP production was impaired in forskolin or isoproterenol-stimulated 3T3-L1 adipocytes overexpressing IKKε ( Figure 1E ) , we examined whether noncanonical IKKs might desensitize adrenergic stimulation through increased activity of PDE3B in adipocytes . Pretreatment with a nonspecific phosphodiesterase inhibitor , IBMX , in 3T3-L1 adipocytes expressing IKKε or TBK1 rescued the full stimulation of cAMP production in response to forskolin ( Figure 3A ) . Interestingly , the selective PDE3B and PDE4 inhibitor , Zardaverine ( Schudt et al . , 1991 ) , also blocked the inhibitory effects of IKKε and TBK1 overexpression on cAMP levels in response to isoproterenol and forskolin in 3T3-L1 adipocytes ( Figure 3B ) , suggesting an important role for PDE3B as a target of the noncanonical IKKs . 10 . 7554/eLife . 01119 . 007Figure 3 . IKKε and TBK1 reduce cAMP levels through activation of PDE3B . ( A ) cAMP levels from 3T3-L1 adipocytes expressing empty vector , Flag-IKKε , or Flag-TBK1 treated with or without 50 μM FSK , 250 μM IBMX , or together for 15 min . *p<0 . 05 and **p<0 . 0001 . Performed in duplicate . ( B ) cAMP levels from 3T3-L1 adipocytes expressing empty vector , Flag-IKKε , or Flag-TBK1 treated with or without 10 μM ISO or 50 μM FSK together with or without 10 μM Zardaverine ( Zarda ) for 15 min . *p<0 . 05 . Performed in duplicate . ( C ) 32P phospho-image of in vitro kinase reaction using either immunoprecipitated HA-PDE3B from HEK293T cells or 1 μg MBP ( myelin basic protein ) as a substrate with recombinant kinases as indicated . Results were replicated in multiple experiments . ( D ) Immunoblots of immunoprecipitation with anti-HA antibodies followed by treatment with or without CIP ( top panel ) and whole cell lysates ( bottom panel ) from Cos-1 cells co-expressing HA-PDE3B with Flag-IKKε/TBK1 or Flag-IKKε/TBK1 K38A . D . E . stands for dark exposure and L . E . stands for light exposure . Results were replicated in multiple experiments . ( E ) Immunoblots of GST-14-3-3 pulldown from HEK293T cells co-expressing HA-PDE3B with Flag-TBK1 or Flag-TBK1 K38A . Ponceau S staining shows the amount of beads used in GST-14-3-3 pulldown . Results were replicated in multiple experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 01119 . 00710 . 7554/eLife . 01119 . 008Figure 3—figure supplement 1 . IKKε and TBK1 interact with PDE3B in a manner dependent on the activity of IKKε and TBK1 . ( A ) Immunoblots of in vitro kinase reaction using either immunoprecipitated HA-PDE3B from HEK293T cells or 1 μg MBP ( myelin basic protein ) as a substrate with recombinant kinases as indicated . Results were replicated in multiple experiments . ( B ) Top panel: 32P phospho-image of in vitro kinase reaction using either immunoprecipitated HA-PDE3B from HEK293T cells or 1 μg MBP as a substrate with recombinant kinases as indicated . Bottom panel: Immunoblots ( IB ) of in vitro kinase reaction . Results were replicated in multiple experiments . ( C ) Immunoblots of in vitro kinase reaction using immunoprecipitated HA-PDE3B from HEK293T cells as a substrate with recombinant MBP-IKKε kinase ( rIKKε ) using increasing amounts of ATP as indicated . Lane 1: Whole cell lysates from HEK293T cells expressing HA-PDE3B , Lane 2: IP flow-through , Lane 3: IP without rIKKε . Results were replicated in multiple experiments . ( D ) Immunoblots of immunoprecipitation with anti-HA antibodies ( top panel ) and whole cell lysates ( WCL , bottom panel ) from HEK293T cells co-expressing HA-PDE3B with Flag-IKKε or Flag-IKKε K38A . Results were replicated in multiple experiments . ( E ) Immunoblots of GST-TBK1 ULD pulldown from 3T3-L1 adipocytes . Ponceau S staining shows the amount of beads used in GST-TBK1 ULD pulldown . Results were replicated in multiple experiments . ( F ) Immunoblots of immunoprecipitation with anti-HA antibodies ( left panel ) and whole cell lysates ( right panel ) from HEK293T cells co-expressing HA-PDE3B with Flag-TBK1 or Flag-TBK1 K38A . Results were replicated in multiple experiments . The schematic model suggests that TBK1 associates with its substrates in an inactive conformation and subsequently dissociate upon phosphorylation in an active conformation . DOI: http://dx . doi . org/10 . 7554/eLife . 01119 . 008 We next examined whether IKKε and TBK1 directly phosphorylate PDE3B to regulate cAMP levels . Recombinant TBK1 , Akt and PKA were incubated in vitro with [γ- 32P]ATP and purified PDE3B as a substrate . Phosphorylation was assessed by SDS-PAGE followed by autoradiography . TBK1 directly catalyzed the phosphorylation of PDE3B; phosphorylation was also produced by incubation with Akt and PKA , as previously reported ( Kitamura et al . , 1999; Palmer et al . , 2007 ) ( Figure 3C ) . IKKε also catalyzed this phosphorylation in vitro ( data not shown ) . This increase in phosphorylation produced by in vitro incubation with TBK1 , IKKε and PKA was also detected when PDE3B was blotted with antibodies that recognize the 14-3-3 binding motif ( Figure 3—figure supplement 1A ) . When purified PDE3B was incubated with the same amount of recombinant TBK1 and canonical IKKβ kinases in vitro , phosphorylation of PDE3B by IKKβ was barely detectable , indicating a level of specificity in which PDE3B is a better target of the noncanonical IKKs ( Figure 3—figure supplement 1B ) . This phosphorylation was dose-dependent with respect to ATP ( Figure 3—figure supplement 1C ) . To determine whether IKKε can phosphorylate PDE3B in cells , we co-expressed IKKε and its inactive mutant K38A with HA-tagged PDE3B in HEK293T cells , followed by immunoprecipitation ( IP ) with anti-HA antibodies . Expression of IKKε in cells caused a shift in electrophoretic mobility of PDE3B , and this shift was not detected when IKKε K38A was expressed ( Figure 3—figure supplement 1D ) . Phosphorylation of PDE3B was also detected after expression of IKKε but not its kinase-inactive mutant K38A in cells , as detected by blotting with antibodies that recognize the 14-3-3 binding motif . To determine whether this molecular shift was dependent on phosphorylation of PDE3B , HA-PDE3B was co-expressed in Cos-1 cells along with IKKε , TBK1 or their kinase inactive mutants , and HA immunoprecipitates were treated with or without calf intestinal phosphatase ( CIP ) . Expression of both of the wild-type kinases reduced the electrophoretic mobility of PDE3B , which could be reversed by treatment with the phosphatase ( Figure 3D , compare lane 3 , 7 to lane 4 , 8 ) . Neither of the kinase-inactive mutants had an effect ( Figure 3D , compare lane 5 , 9 to lane 6 , 10 ) . Previous studies suggested that IKKε and TBK1 bind to their respective substrates through a sequence that includes a ubiquitin-like domain ( ULD ) proximal to their kinase domain . This domain is highly conserved among the IKK family members , and is 49% identical between IKKε and TBK1 ( Ikeda et al . , 2007; May et al . , 2004 ) . To confirm that PDE3B is a bona fide substrate of IKKε and TBK1 , we prepared a GST-ULD domain fusion protein from TBK1 and incubated this fusion protein with 3T3-L1 adipocyte lysates . The fusion protein specifically precipitated endogenous PDE3B from these lysates ( Figure 3—figure supplement 1E ) . To explore further the interaction of these two proteins , we co-expressed WT TBK1 and its K38A mutant with HA-tagged PDE3B in HEK293T cells , and immunoprecipitated the protein with anti-HA antibodies . Kinase-inactive TBK1 was preferentially co-immunoprecipitated with PDE3B , whereas the interaction of PDE3B with WT TBK1 was barely detectable ( Figure 3—figure supplement 1F ) . These data suggest that TBK1 and IKKε associate with substrates such as PDE3B , and subsequently dissociate upon phosphorylation . Next , to test further the role of PDE3B phosphorylation by IKKε and TBK1 in initiating its interaction with 14-3-3β , we prepared a GST-14-3-3β fusion protein which was incubated with lysates from HEK293T cells co-expressing TBK1 with PDE3B . PDE3B was preferentially pulled down by GST-14-3-3β after phosphorylation by TBK1 but not by its inactive K38A mutant , whereas GST beads alone enriched neither PDE3B nor its phosphorylated form ( Figure 3E ) . To evaluate the regulatory role of PDE3B phosphorylation by IKKε and TBK1 , we determined which sites are phosphorylated . HA-PDE3B was co-expressed in Cos-1 cells with IKKε and TBK1 , and phosphorylated PDE3B was enriched by IP with anti-HA antibodies . Phosphorylation sites on human PDE3B were then determined by LC-MS/MS mass spectrometry . This analysis revealed that serines 22 , 299 , 318 , 381 , 463 , 467 , and 503 were phosphorylated by both kinases; there were no differences between the kinases ( Figure 4A ) . Interestingly , the phosphorylation profile of PDE3B matched neither known Akt or PKA profiles ( Lindh et al . , 2007 ) . However , phosphorylation on serine 299 and serine 318 had previously been identified on mouse PDE3B ( residues equivalent to Serine 277 and 296 in mouse PDE3B ) in adipocytes and hepatocytes in response to both insulin and forskolin ( Lindh et al . , 2007 ) . 10 . 7554/eLife . 01119 . 009Figure 4 . IKKε and TBK1 phosphorylate PDE3B at serine 318 , resulting in the binding of 14-3-3β . ( A ) Summary of sites on PDE3B phosphorylated by IKKε or TBK1 ( P-sites ) from mass spectrometry experiments . ( B ) Immunoblots of GST-14-3-3 pulldown from HEK293T cells co-expressing HA-PDE3B or HA-PDE3B S318A with Flag-TBK1 . Ponceau S staining shows the amount of beads used in GST-14-3-3 pulldown . Results were replicated in multiple experiments . ( C ) GST-14-3-3 overlay on nitrocellulose membrane ( top blot ) and an immunoblot ( IB ) of whole cell lysates from HEK293T cells co-expressing HA-PDE3B or HA-PDE3B S318A with Flag-TBK1 ( bottom blot ) . Results were replicated in multiple experiments . ( D ) cAMP levels from 3T3-L1 adipocytes expressing empty vector , HA-PDE3B , or HA-PDE3B S318A treated with or without 100 ng/ml TNFα for 16 hr followed by treatment with or without 25 μM FSK for 15 min . **p<0 . 0001 and **p<0 . 01 . Performed in duplicate . DOI: http://dx . doi . org/10 . 7554/eLife . 01119 . 00910 . 7554/eLife . 01119 . 010Figure 4—figure supplement 1 . Overexpression of PDE3B in 3T3-L1 adipocytes reduces the attenuation of forskolin-stimulated β-adrenergic signaling produced by TNFα . ( A ) Immunoblots of whole cell lysates from Figure 4D . ( B ) Top left panel , pHSL and total HSL signals in whole cell lysates were quantified and normalized to the basal signal ( lane 1 in Figure 4—figure supplement 1A ) . The normalized pHSL/total HSL ratio is presented as the mean ± SEM of two independent experiments . **p<0 . 0001 and *p<0 . 05 . Top right panel , quantification of total IKKε and RalA signals in whole cell lysates was performed as described above . **p<0 . 0001 . Bottom left panel , quantification of phospho-p38 ( p-p38 ) and total p38 signals in whole cell lysates was performed as described above . Bottom right panel , quantification of pTBK1 and total TBK1 signals in whole cell lysated was performed as described above . **p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 01119 . 010 While several serine residues are known to be phosphorylated on PDE3B in response to stimuli , serine 318 ( human ) is the best characterized . This residue resides in a consensus phosphorylation sequence for both Akt and PKA , and also serves as a consensus 14-3-3 binding motif once phosphorylated ( Lindh et al . , 2007; Palmer et al . , 2007 ) . We thus created a Ser318Ala ( S318A ) mutant of PDE3B , and examined its interaction with a GST-14-3-3β fusion protein or by GST-14-3-3 overlay assay . Interestingly , despite incubation with TBK1 , the phospho-defective , S318A mutant of PDE3B , did not specifically interact with GST-14-3-3β , whereas the wild-type protein did ( Figure 4B , C ) . In a GST pull-down assay , the molecular shift of PDE3B S318A was still detected by western blot ( Figure 4B ) , indicating that phosphorylation of PDE3B by TBK1 on other sites still occurred , but were not crucial for 14-3-3β binding . To examine the functional importance of the phosphorylation of PDE3B at Serine 318 , we overexpressed WT PDE3B and its S318A mutant in 3T3-L1 adipocytes , and tested the response of the cells to TNFα . Overexpression of WT PDE3B in cells reduced the attenuation of forskolin-stimulated cAMP production and phosphorylation of HSL produced by TNFα , whereas PDE3B S318A was ineffective ( Figure 4D , Figure 4—figure supplement 1A , B ) . These data suggest that although IKKε and TBK1 can phosphorylate PDE3B on several sites , serine 318 may be particularly important in the regulation of phosphodiesterase function by promoting the interaction between PDE3B and 14-3-3β . More importantly , this residue is the major site mediating the negative effects of IKKε and TBK1 on sensitivity of adipocytes to β-adrenergic stimulation . Disruption of sympathetic activation of lipolysis and fat oxidation may play an important role in the development and maintenance of increased fat storage in obesity . Indeed , while numerous studies have demonstrated catecholamine resistance in obese adipose tissue ( Jensen et al . , 1989; Reynisdottir et al . , 1994; Bougneres et al . , 1997; Arner , 1999; Jocken and Blaak , 2008 ) , the underlying mechanisms remain unclear . To test the functional importance of the noncanonical IKKs in maintaining energy balance in vivo , we investigated whether the administration of a selective inhibitor of IKKε and TBK1 , amlexanox , can reverse diet-induced catecholamine resistance in rodents . We fed mice a high fat or normal diet , treated them with amlexanox by oral gavage for 4 days ( prior to the point when weight loss is seen ) , and then gave a single intraperitoneal ( IP ) injection of the β3-adrenergic agonist CL-316 , 243 . Injection of CL-316 , 243 stimulated a threefold increase in serum FFA and glycerol levels in both vehicle and amlexanox-treated mice on normal diet ( ND ) . The effect of CL-316 , 243 to increase serum FFAs was significantly attenuated in HFD-fed , vehicle-treated mice . However , HFD-fed mice treated with amlexanox responded like normal diet mice , despite the fact that they were weight matched with control HFD-fed mice ( Figure 5A ) . The fold increase in serum glycerol levels was also significantly higher in amlexanox-treated HFD mice , as compared to vehicle-treated HFD-fed mice . In addition , ex vivo pretreatment of white adipose tissues from mice on a HFD with amlexanox enhanced glycerol release ( Figure 5B ) . This effect was more pronounced in the inguinal fat depot , where amlexanox pretreatment increased phosphorylation of HSL , perilipin , and other proteins recognized by the PKA substrate motif antibody in response to CL-316 , 243 treatment compared to vehicle-pretreated tissues ( Figure 5C ) . Amlexanox also concurrently increased the phosphorylation of TBK1 at Ser172 due to the relief of feedback inhibition , as previously reported with other inhibitors ( Clark et al . , 2009; Reilly et al . , 2013 ) . 10 . 7554/eLife . 01119 . 011Figure 5 . The IKKε/TBK1 inhibitor Amlexanox sensitizes β-adrenergic agonist-stimulated lipolysis in white adipose tissue in diet-induced obese mice . ( A ) Fold increase in serum FFA ( left panel ) and glycerol ( right panel ) levels 15 min after CL-316 , 243 injection in ND- or HFD-fed mice treated with amlexanox or vehicle control for 4 days . n = 7 mice per group . *p<0 . 05 . ( B ) Glycerol release from ex vivo epididymal ( left panel ) and inguinal ( right panel ) WATs after 1 hr pretreatment with amlexanox or vehicle . CL-316 , 243 treatment was started at time zero . n = 6 , 3 WAT pieces × 2 mice . *P<0 . 05 . ( C ) Immunoblots in inguinal WAT lysates from Figure 5B after 60 min of CL-316 , 243 treatment . ( D ) cAMP levels in epididymal WAT 20 min after CL-316 , 243 ( CL ) or saline control injection in HFD-fed mice treated with amlexanox or vehicle control for 4 days . n = 2 mice per saline-treated group and n = 3 mice per CL-316 , 324-treated group . *p<0 . 05 . ( E ) Immunoblots in epididymal WAT 5 min after CL-316 , 243 or saline control injection in HFD-fed mice treated with amlexanox or vehicle control for 4 days . ( F ) Relative oxygen comsumption of mice in each treatment group . n = 7 for the vehicle-treated group , n = 5 for the amlexanox-treated group . *P<0 . 05 ( Student’s t test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01119 . 011 To examine whether inhibition of TBK1 and IKKε with amlexanox reverses resistance to catecholamine-induced lipolysis in vivo by increasing stimulation of cAMP production , we measured cAMP levels in epididymal adipose tissue from mice on HFD after CL-316 , 243 IP injection . Interestingly , levels of cAMP were increased after CL-316 , 243 IP injection in mice on HFD pretreated with amlexanox ( Figure 5D ) . Consistent with this , HSL phosphorylation was also increased after CL-316 , 243 IP injection of HFD-fed mice pretreated with amlexanox ( Figure 5E ) . Our previous studies showed that increased expression of Ucp1 in white adipose depots resulted in increased energy expenditure in IKKε-deficient mice ( Chiang et al . , 2009 ) and amlexanox-treated mice ( Reilly et al . , 2013 ) while on a high fat diet but not on a normal diet . To examine whether inhibition of catecholamine resistance in obese adipose tissue by targeting noncanonical IKKs with amlexanox can lead to increase energy expenditure in diet-induced obese mice , we measured oxygen consumption rates of vehicle or amlexanox-treated HFD-fed mice after a single injection of CL-316 , 243 in metabolic cages . The effect of CL-316 , 243 to increase energy expenditure was more pronounced in amlexanox-treated HFD-fed mice , as compared to vehicle-treated HFD-fed mice ( Figure 5F ) . These data suggest that targeting the noncanonical IKKs with the selective inhibitor amlexanox ameliorated catecholamine resistance in obese adipose tissue .
Decreased sympathetic activation of adipose tissue due to impaired catecholamine synthesis or sensitivity has been observed in obese patients ( Reynisdottir et al . , 1994; Stallknecht et al . , 1997; Horowitz and Klein , 2000; Jocken et al . , 2008 ) . Obesity is commonly associated with blunted whole-body catecholamine-induced lipolysis ( Horowitz and Klein , 2000 ) . This is thought to occur through a number of mechanisms , including leptin resistance ( Myers et al . , 2010 ) , as well as the reduced expression of β-adrenergic receptors ( Reynisdottir et al . , 1994 ) or increased expression of α2-adrenergic receptors ( Stich et al . , 2002 ) . White adipose tissue and cultured isolated adipocytes from obese human and mouse models exhibit decreased cAMP-stimulated lipolysis and fat oxidation , due to reduced energy expenditure from decreased mitochondrial uncoupling ( Yehuda-Shnaidman et al . , 2010 ) . This desensitization to adrenergic activation is also a feature of childhood onset obesity ( Bougneres et al . , 1997; Enoksson et al . , 2000 ) , and has been observed in adipocytes from first-degree relatives of obese subjects ( Hellstrom et al . , 1996 ) . We demonstrate here a novel link between obesity and reduced sympathetic activity and β-adrenergic sensitivity , through the inflammation-dependent induction of the noncanonical IκB kinases IKKε and TBK1 . Obesity generates a state of low-grade inflammation in both humans and rodents , which involves activation of the NFκB pathway ( Wellen and Hotamisligil , 2005; Hotamisligil , 2006; Shoelson et al . , 2007 ) . Upon its prolonged activation , NFκB induces the expression of the noncanonical IκB kinases IKKε and TBK1 . The induction of these kinases was blocked by administration of anti-inflammatory agents to mice without producing weight loss , suggesting that they are expressed in response to inflammation rather than obesity per se ( Reilly et al . , 2013 ) . Deletion of the IKKε gene rendered mice partially resistant to weight gain , insulin resistance , steatosis and the long-term inflammation produced by high fat diet ( Chiang et al . , 2009 ) , and administration of the dual specificity IKKε/TBK1 inhibitor amlexanox to diet-induced obese or ob/ob mice produced even more profound effects ( Reilly et al . , 2013 ) . The blockade of these kinases in obese rodents with amlexanox results in increased phosphorylation of PKA substrates in adipose tissue , along with increased expression of Ucp1 , and improved rates of lipolysis and fat oxidation ( Reilly et al . , 2013 ) . Amlexanox was shown to inhibit phosphodiesterase activity of rat peritoneal mast cells via an unknown mechanism ( Makino et al . , 1987 ) . Together these data indicate that IKKε and TBK1 might exert their physiological effects in part by reducing the sensitivity of adipocytes to β-adrenergic stimulation via changes in cAMP . Data presented here suggest that the molecular target of IKKε/TBK1 is the phosphodiesterase PDE3B . Upon increased expression in the obese state , these kinases can phosphorylate PDE3B , causing an increase in the activity of the enzyme that cleaves cAMP , reducing the stimulation of cAMP-dependent phosphorylation of proteins in response to sympathetic activation . These proteins include HSL and perilipin , responsible for β-adrenergic-stimulated lipolysis , and other proteins such as p38 that regulate expression of Ucp1 . The reduced sensitivity to β-adrenergic activation can attenuate lipolysis and fatty acid oxidation , as well as adaptive thermogenesis . Several issues deserve further attention . The first concerns the relative roles of the two noncanonical IKKs in this pathway . Both TBK1 and IKKε are induced in response to obesity-dependent inflammation , and appear to phosphorylate PDE3B on the same residues with equal efficiency . Although there are differences in expression of these kinases in other tissues ( Shimada et al . , 1999 ) , and perhaps differences in the upstream signals that lead to their regulation ( Wunderlich et al . , 2008 ) , their relative roles in controlling this pathway remain uncertain . Additionally , the mechanism by which PDE3B is regulated remains uncertain . While phosphorylation correlates well with decreased levels of cAMP in cells , we have been unable to demonstrate increased catalytic activity of the enzyme due to phosphorylation by IKKε , TBK1 or the other kinases ( Kitamura et al . , 1999; Palmer et al . , 2007 ) thought to regulate the phosphodiesterase . Perhaps the phosphorylation-dependent binding of the enzyme to 14-3-3 exerts changes in its localization and access to its substrate , thus explaining increased activity in cells . How is it that insulin resistance produced by inflammation fails to block continued energy storage ? One possible explanation may lie in the homeostatic response to inflammation itself , typified by the induction of TBK1 and IKKε . Our data confirm previous findings that TNFα and perhaps other inflammatory cytokines can promote lipolytic processes in cells after short-term treatment , but that after longer exposure elicit an inhibitory response that appears to be the result of TBK1 and IKKε induction . Thus , these kinases may be part of a ‘counter-inflammatory’ program that attenuates the extent to which inflammatory signals are effective , and also serves to conserve energy by repressing lipolysis and fatty acid oxidation through activation of PDE3B . Interestingly , PDE3B is also a target of insulin action in adipocytes ( Degerman et al . , 2011 ) . Thus , TBK1 and IKKε appear to co-opt insulin targets to conserve energy during obesity . These insights further suggest that the noncanonical IKKs might be interesting new therapeutic targets for the treatment of obesity and type 2 diabetes .
All chemicals were obtained from Sigma-Aldrich ( Saint Louis , MO ) unless stated otherwise . Anti-Flag antibody was obtained from Sigma , and anti-HA antibody was obtained from Santa Cruz Biotechnology ( Santa Cruz , CA ) . Anti-IKKε , anti-TBK1 , anti-phospho-TBK1 ( Ser172 ) , anti-AKT , anti-phospho-AKT ( Ser473 ) , anti-HSL , anti-phospho-HSL ( Ser660 ) , anti-p38 , anti-phospho-p38 , anti-perilipin , anti-ATGL and anti-PPARγ antibodies were purchased from Cell Signaling Technology ( Danvers , MA ) . Anti-phospho-perilipin ( Ser522 ) was purchased from Vala Sciences Inc ( San Diego , CA ) . Anti-CGI-58 was purchased from Novus Biologicals ( Littleton , CO ) . Anti-RalA antibody was obtained from BD Bioscience ( San Jose , CA ) . Anti-Ucp1 antibody was obtained from Alpha Diagnostics ( San Antonio , TX ) . Anti-PDE3B was provided as a generous gift by the Dr Vince Manganiello ( NHLBI , NIH ) . Enhanced chemiluminescence ( ECL ) reagents were purchased from Thermo Scientific ( Rockford , IL ) . EDTA-free protease inhibitor tablet was purchased from Roche Diagnostics ( Indianapolis , IN ) . Monoclonal anti-HA agarose ( Sigma ) was used for immunoprecipitations , performed using the manufacturer’s protocol . The human PDE3B cDNA was kindly provided by Dr Morris Birnbaum ( University of Pennsylvania ) . The human 14-3-3β cDNA was kindly provided by Dr Ken Inoki ( University of Michigan ) and subcloned into pKH3 ( Chen et al . , 2007 ) and pGEX-4T-1 vectors ( GE Healthcare Life Sciences , MI ) . Amlexanox was purchased from Ontario Chemical Inc . ( Guelph , Ontario , Canada ) . The TBK1/IKKε inhibitor CAY10576 was purchased from Cayman Chemical ( Ann Arbor , MI ) . 3T3-L1 fibroblasts ( American Type Culture Collection , Manassas , VA ) were cultured and differentiated as described previously ( Liu et al . , 2005 ) . The cells were routinely used within 7 days after completion of the differentiation process; only cultures in which >90% of cells displayed adipocyte morphology were used . 3T3-L1 adipocytes were transfected on the second day post FBS using Amaxa Cell Line Nucleofector Kit L ( Lonza , Houston , TX ) according to the manufacturer’s protocol . 3T3-L1 adipocytes were serum starved for 12 hr with 0 . 5% fetal bovine serum ( FBS ) in Dulbecco’s modified eagle medium ( DMEM , Invitrogen , Grand Island , NY ) prior to TNFα treatments ( 50 ng/ml unless otherwise noted ) . 3T3-L1 adipocytes were pre-treated for 1 hr with amlexanox at the given concentrations . Alternatively , 3T3-L1 adipocytes were treated with 50 μM forskolin or 10 μM isoproterenol for 15 min , after a 60 min amlexanox pretreatment . The cells were harvested for total RNA and analyzed by real-time PCR . Cell lysates were resolved on SDS-PAGE and analyzed by immunoblot using the indicated antibodies . HEK293T or Cos-1 cells were cultured to 90% confluence and transfected using Opti-MEM media ( Invitrogen ) and 3 μl Lipofectamine 2000 ( Invitrogen ) per μg DNA according to manufacturer’s protocol . Coexpression of IKKε or TBK1 with HA-PDE3B was done using a 2 μg kinase: 1 μg PDE3B ratio of the expression constructs . Tissues were homogenized in lysis buffer ( 50 mM Tris , pH 7 . 5 , 5 mM EDTA , 250 mM sucrose , 1% NP40 , 2 mM DTT , 1 mM sodium vanadate , 100 mM NaF , 10 mM Na4P2O7 , and freshly added protease inhibitor tablet ) , then incubated for 1 hr at 4°C ( Chiang et al . , 2009 ) . Crude lysates were then centrifuged at 14 , 000 × g for 15 min twice and the protein concentration was determined using BioRad Protein Assay Reagent ( Bio-Rad , Hercules , CA ) . Samples were diluted in sodium dodecyl sulfate ( SDS ) sample buffer and boiled for 5 min at 95°C . Proteins were resolved by SDS-polyacrylamide gel electrophoresis and transferred to nitrocellulose membranes ( Bio-Rad , Hercules , CA ) . Individual proteins were detected with the specific antibodies and visualized on film using horseradish peroxidase-conjugated secondary antibodies ( Bio-Rad , Hercules , CA ) and Western Lightning Enhanced Chemiluminescence ( Perkin Elmer Life Sciences , Waltham , MA ) . Wild-type male C57BL/6 mice were fed a high fat diet consisting of 45% of calories from fat ( D12451 Research Diets Inc . , New Brunswick , NJ ) starting at 8 weeks of age for up to 6 months , while normal diet C57BL/6 controls were maintained on normal chow diet consisting of 4 . 5% fat ( 5002 Lab Diet , St . Louis , MO ) . Animals were housed in a specific pathogen-free facility with a 12-hr light/12-hr dark cycle and given free access to food and water . All animal use was in compliance with the Institute of Laboratory Animal Research Guide for the Care and Use of Laboratory Animals and approved by the University Committee on Use and Care of Animals at the University of Michigan . Total RNA was extracted from differentiated 3T3-L1 adipocytes using the RNeasy Kit ( Qiagen , Valencia , CA ) with a DNase step . The Superscript First-Strand Synthesis System for RT-PCR ( Invitrogen , Grand Island , NY ) was used with random primers for reverse transcription . Real-time PCR amplification of the cDNA was performed on samples in triplicate with Power SYBR Green PCR Master Mix ( Applied Biosystems , Carlsbad , CA ) using the Applied Biosystems 7900HT Fast Real-time PCR System . Adrp was chosen as the internal control for normalization as its expression was not significantly affected by experimental conditions . Sequences of Ucp1 primers used in this study are 5′-AGGCTTCCAGTACCATTAGGT‐3′ and 5′‐CTGAGTGAGGCAAAGCTGATTT‐3′ . Sequences of Ikbke primers used in this study are 5′-ACAAGGCCCGAAACAAGAAAT-3′ and 5′-ACTGCGAATAGCTTCACGATG‐3′ . Data were analyzed using the 2−ΔΔCT method ( Livak and Schmittgen , 2001 ) . Mice were placed on a high fat diet for 6 months , then after 1 week of daily gavage with vehicle , mice were gavaged with either vehicle or amlexanox ( 25 mg/kg ) daily for 4 days . On the fourth day , mice were injected with CL-316 , 243 ( 1 mg/kg ) or saline control . For analysis of blood metabolites , serum samples were collected before and 15 min after the injection , via a submandibular bleed . Mice were euthanized and WAT samples were collected for cAMP measurement and western blot analysis , 15 min or 20 min after injection . Analysis of oxygen consumption was performed in metabolic cages , as previously described ( Reilly et al . , 2013 ) , by the University of Michigan Metabolic Phenotyping Core . Relative oxygen consumption was obtained by normalization of oxygen consumption rates , after the CL-316 , 243 injection , to the oxygen consumption rates on day 3 in the same mouse after saline injection . Both injections were performed at 11 am . For experiments involving CL-316 , 243 , pieces were pre-incubated for 30 min with amlexanox ( 100 μM ) or DMSO vehicle control; then the tissue pieces were transferred to fresh media with and without 10 mM CL-316 , 243 , and media samples were collected every 15 min for 1 hr . To measure glycerol release , 10 μl of supernatant was combined with 200 μl of Free Glycerol Reagent from the Free Glycerol Determination Kit ( Sigma ) and allowed to incubate for 15 min at room temperature . Absorbance at 540 nm was measured to determine glycerol content and was normalized to determine glycerol production per mg of white adipose tissue . 3T3-L1 adipocytes were incubated in DMEM ( Invitrogen ) without phenol red for 2 hr at 37°C . The cells then were incubated for 90 min at 37°C in HBSS-2% fatty acid-free BSA with 10 μM isoproterenol or 10 μM CL-316 , 243 . Free glycerol concentration was measured by reacting 25 μl of conditioned media with 200 μl Free Glycerol Reagent ( Sigma ) and absorbance was measured at 540 nm using the manufacturer’s protocol . Glycerol release was normalized to cellular protein content . 3T3-L1 adipocytes were treated with β-agonists and/or phosphodiesterase inhibitors and allowed to incubate for the indicated amount of time at 37°C . The cells were lysed with 150 μl 0 . 1 M HCl , scraped and spun down . A cAMP Enzyme Immunoassay Kit ( Sigma CA201 ) was used to quantify cAMP levels . 50 μl of cell lysates was combined with 50 μl Assay Buffer 2 in each well and cAMP levels were assayed according to the manufacturer’s protocol . Tissue samples were homogenized in 5% TCA , then extracted with water-saturated ether , and dried before resuspension in Assay Buffer 2 . 80 μl of packed 3T3-L1 cells was electroporated with 3 μg of Glosensor 22-F using Amaxa Cell Line Nucleofector Kit L ( Lonza ) according to the manufacturer’s protocol . Electroporated cells were resuspended in 10 ml of L1-FBS media and 200 μl per well was plated in six columns of an opaque , white , 96-well tissue culture plate ( BD Bioscience , San Jose , CA ) . After 20 hr , media were changed to 100 μl DMEM with 1 . 5 mg/ml luciferin and allowed to equilibrate for 1 hr . The cells were treated with 50 μM forskolin and luminescence was read every 30 s for 75 min . In vitro kinase assays were performed by incubating purified kinase ( IKKε , TBK1 , IKKβ , PKA , or AKT ) in kinase buffer containing 25 mM Tris ( pH 7 . 5 ) , 10 mM MgCl2 , 1 mM DTT , and 10 μM ATP for 30 min at 30°C in the presence of 0 . 5 μCi γ-[32P]-ATP and 1 μg myelin basic protein ( MBP ) per sample as a substrate . IKKε and TBK1 were fused to MBP ( maltose binding protein ) and these fusion proteins were purified from insect SF9 cells by baculovirus expression system by Dr Stuart J Decker ( Life Sciences Institute , University of Michigan ) . IKKβ , PKA , and AKT kinases were purchased from Millipore ( Billerica , MA ) . The kinase reaction was stopped by adding 4X sodium dodecyl sulfate ( SDS ) sample buffer and boiling for 5 min at 95°C . Supernatants were resolved by SDS-polyacrylamide gel electrophoresis , transferred to nitrocellulose , and analyzed by autoradiography using a Typhoon 9410 phosphorimager ( GE Life Sciences , Piscataway , NJ ) . The bands were quantified using ImageQuant . Calf intestinal phosphatase ( CIP ) was obtained from New England Biolabs ( Ipswich , MA ) . Immunoprecipitated PDE3B was incubated for 1 hr at 37°C in a 100 μl reaction containing 50 mM Tris pH 7 . 5 , 150 mM NaCl , 1% NP-40 , EDTA-free protease , and 5 μl CIP . For assays requiring soluble protein , purified GST-14-3-3β protein was eluted from glutathione beads by washing beads with 10 mM glutathione in PBS , pH 8 . 0 . The elution was monitored by A280 readings , and fractions containing protein were pooled and dialyzed overnight against 4 L of ice-cold PBS . The proteins were then concentrated using an Amicon centrifugal filtration unit ( Millipore ) . Concentrated proteins were stored at −80°C in PBS containing 10% glycerol and 10 mM DTT . For GST-14-3-3 pulldowns , cells were washed twice with ice-cold PBS and then lysed in 1 ml of 14-3-3 pulldown buffer ( PD buffer; 15 mM Tris , pH 7 . 5 , 150 mM NaCl , 0 . 5% NP-40 , 1 mM DTT ) supplemented with a protease inhibitor tablet ( Roche ) . Lysates were cleared by centrifuging at 13 , 000×g for 10 min and then were incubated with ∼10 mg of GST or GST-14-3-3β bound to glutathione beads ( GE Healthcare Life Sciences , MI ) for 1 . 5 hr at 4°C . For samples treated with phosphatase , lysates were preincubated with 500 U of calf intestinal phosphatase ( New England Biolabs , Inc . ) at 37°C for 1 hr before adding GST-14-3-3 beads . The beads were washed three times with 1 ml of PD buffer and then resuspended in 2X SDS sample buffer . GST or GST-14-3-3β was labeled with DIG by incubating 25 mg of protein with 2 ml of 5 mM Digoxigenin-3-O-methylcarbonyl-ε-aminocaproic acid-N-hydroxysuccinimide ester ( DIG-NHS; Roche ) in 350 ml of PBS for 15 min at room temperature . The labeling reaction was stopped by adding 100 ml of 1 M Tris-HCl , pH 7 . 4 . Labeled protein was dialyzed against 1 L of 25 mM Tris-HCl , pH 7 . 4 for 1 hr at room temperature , then against 4 L of PBS , pH 7 . 4 for 4 hrs at 4°C , and finally against 4 L of fresh PBS , pH 7 . 4 for 16 hrs at 4°C . Labeled protein was then diluted in 25 ml of TBS ( 50 mM Tris-HCl , pH 7 . 4 , 150 mM NaCl ) containing 2 mg/ml BSA ( Sigma-Aldrich ) and 0 . 01% sodium azide ( Sigma-Aldrich ) . DIG-labeled proteins were stored at 4°C . For overlay assays , PDE3B was immunoprecipitated from HEK293T cells and resolved by SDS-PAGE . Proteins were transferred to a nitrocellulose membrane . The membrane was blocked at room temperature overnight in blocking buffer ( 5% skim milk in TBS-T ) . The membrane was then incubated with DIG-labeled proteins for at least 4 hr at 4°C and then washed three times with TBS-T . The membrane was then incubated with blocking buffer containing anti-DIG HRP antibody ( 1:10 , 000; Roche ) for 2 hr at room temperature , and washed three times with TBS-T . Overlays were visualized by reacting with ECL western blotting substrate ( Perkin Elmer Life Sciences , Waltham , MA ) . In-gel digestion followed by LC-MS/MS analysis was carried out by the mass spectrometry-based proteomics resource in the Department of Pathology , University of Michigan . Briefly , tryptic peptides were resolved on a nano-C18 reverse phase column and sprayed directly onto Orbitrap mass spectrometer ( LTQ-Orbitrap XL , Thermofisher ) . Orbitrap was operated in a data-dependent mode to acquire one full MS spectrum ( resolution of 30 , 000@400 m/z ) followed by MS/MS spectra on six most intense ions ( top 6 ) . Proteins were identified by searching data against human protein database ( Uniprot , rel . 2010-9 ) using X ! Tandem/TPP software suite . Oxidation of Met , carbamidomethylation of Cys and phosphorylation of Ser , Thr , and Tyr were considered as potential modifications ( Maine et al . , 2010 ) . Averaged values are presented as the mean ± SEM . When comparing two groups , we performed Student’s t test to determine statistical significance . When more than two groups and two factors were investigated , we first performed a two-way analysis of variance ( ANOVA ) to establish that not all groups were equal . After a statistically significant ANOVA result , we performed between-group comparisons using the Tukey post hoc analysis for comparisons of all means and Sidak for comparisons of within factor main effect means . ANOVA and Tukey/Sidak tests were performed using GraphPad Prism version 6 . | Obesity is a complex metabolic disorder that is caused by increased food intake and decreased expenditure of energy . Obesity also increases the risk of developing type 2 diabetes , heart disease , stroke , arthritis , and certain cancers . There is considerable evidence to suggest that adipose tissue becomes less sensitive to catecholamines such as adrenaline in states of obesity , and that this reduced sensitivity in turn reduces energy expenditure . However , the details of this process are not fully understood . It is well established that obesity generates a state of chronic , low-grade inflammation in liver and adipose tissue , accompanied by the secretion of signaling proteins that prevent fat cells from responding to insulin , which leads to type 2 diabetes . Activation of the NFκB pathway is thought to have a central role in causing this inflammation . Now Mowers et al . have investigated whether inflammation caused by activation of the NFκB pathway also has a role in producing catecholamine resistance in fat cells . Obesity-dependent activation of the NFκB pathway increases the levels of a pair of enzymes , IKKε and TBK1 . Mowers et al . found that elevated levels of these two enzymes reduced the ability of certain receptors ( called β-adrenergic receptors ) in the fat cells of obese mice to respond to catecholamines . High levels of the two enzymes also resulted in lower levels of a second messenger molecule called cAMP , which increases energy expenditure by elevating fat burning . However , treating the fat cells with drugs that interfere with the two enzymes restored sensitivity to catecholamine , allowing the fat cells to burn energy . Mowers et al . also treated obese mice with amlexanox , a drug that inhibits these enzymes , and found that this treatment made the mice sensitive to a synthetic catecholamine that triggered the release of energy from fat . Mowers et al . suggest , therefore , that IKKε and TBK1 respond to inflammation in the body by reducing catecholamine signaling , thus preventing energy expenditure . Drugs targeting these enzymes may be useful for treating conditions like obesity or type 2 diabetes . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2013 | Inflammation produces catecholamine resistance in obesity via activation of PDE3B by the protein kinases IKKε and TBK1 |
The hemispheric , bi-layered optic cup forms from an oval optic vesicle during early vertebrate eye development through major morphological transformations . The overall basal surface , facing the developing lens , is increasing , while , at the same time , the space basally occupied by individual cells is decreasing . This cannot be explained by the classical view of eye development . Using zebrafish ( Danio rerio ) as a model , we show that the lens-averted epithelium functions as a reservoir that contributes to the growing neuroretina through epithelial flow around the distal rims of the optic cup . We propose that this flow couples morphogenesis and retinal determination . Our 4D data indicate that future stem cells flow from their origin in the lens-averted domain of the optic vesicle to their destination in the ciliary marginal zone . BMP-mediated inhibition of the flow results in ectopic neuroretina in the RPE domain . Ultimately the ventral fissure fails to close resulting in coloboma .
The bi-layered optic vesicles of vertebrates are formed through a bilateral evagination of the late prosencephalon . In teleosts , this process is driven by a migration of single cells that undergo a subsequent intercalation into the epithelium of the expanding optic vesicle ( Rembold et al . , 2006 , England et al . , 2006 , Sinn and Wittbrodt , 2013 , Ivanovitch et al . , 2013 ) . The oval optic vesicle develops into a hemispheric bi-layered optic cup through a process that involves major morphological transformations . A long-held view of this process proposes that the lens-averted epithelium of the optic vesicle differentiates into the retinal pigmented epithelium ( RPE ) , while the epithelium facing the lens gives rise to the neuroretina , which subsequently bends around the developing lens ( Chow and Lang , 2001; Fuhrmann , 2010; Walls , 1942 ) . This neuroepithelial bending is driven by a basal constriction of lens-facing retinal progenitor cells ( RPC ) ( Martinez-Morales et al . , 2009 ) ( Bogdanović et al . , 2012 ) , which ultimately reduces the space occupied by an individual RPC at the basal surface . However , we observed that this is accompanied by a 4 . 7-fold increase in the overall basal optic cup surface area ( Figure 1A–C ) . To identify the cellular origin of this massive increase , we performed in vivo time-lapse microscopy in zebrafish at the corresponding stages , starting at 16 . 5 hpf ( Figure 1D–L , Video 1 ) , in a transgenic line expressing a membrane-coupled GFP in retinal stem and progenitor cells ( Rx2::GFPcaax ) . 10 . 7554/eLife . 05216 . 003Figure 1 . Neuroretinal surface increases during optic cup formation by epithelial flow . ( A ) Scheme showing the orientation of the pictures presented in B–L . ( B ) Basal neuroretinal surface increases from early to late optic cup stage ( dashed yellow lines ) . ( C ) Basal neuroretinal surface was measured in 3D ( superimposed orange lines ) , although RPCs undergo basal constriction during optic cup formation , the surface increases 4 . 7 fold from early to late optic cup stage , ( D–L ) transition from optic vesicle to optic cup over time , shown at a ventral ( D–F ) , a central ( G–I ) , and a dorsal ( J–L ) level . The membrane localized GFP is driven by an rx2 promoter ( rx2::GFPcaax ) , which is active in RPCs . The optic vesicle is bi-layered ( D , G , J ) with a prospective lens-facing ( arrows in D and E ) and a prospective lens-averted ( arrowheads in D , G , J ) epithelium , connected to the forebrain by the optic stalk ( os in D ) , at a ventral level both are connected at the distal site ( circle in D ) , at a central level both are connected distally and proximally ( circles in G ) , notably the morphology of the lens-averted epithelium at a dorsal level is different from central and ventral levels ( arrowhead in J ) . Over time at ventral and central levels ( D–F and G–I , respectively ) , the lens-averted epithelium is being integrated into the forming optic cup ( arrowheads in D , E , G , H , I and arrow in H ) . A patch of cells in the lens-averted domain gives rise to the RPE ( asterisks in H and arrowhead in J , K ) , le: developing lens , os: optic stalk , B and D–L were derived from 4D imaging data starting at 16 . 5 hpf ( D , G , J ) , one focal plane is presented as Video 1 , scalebar in B and C = 50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 00310 . 7554/eLife . 05216 . 004Video 1 . ( related to Figure 1 ) ( control ) Optic vesicle to optic cup transition visualized with rx2::GFPcaax ( orientation as in Figure 1 ) ( imaging starts at 16 . 5 hpf , framerate 1/15 min ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 004 Strikingly , and in contrast to the former model ( Chow and Lang , 2001; Fuhrmann , 2010; Walls , 1942 ) , our analysis shows that almost the entire bi-layered optic vesicle gives rise to the neural retina ( Figure 1D–I ) , with the marked exception of a small lens-averted patch ( see below ) . The majority of the lens-averted epithelium ( Figure 1D , G , between arrowheads ) serves as a neuro-epithelial reservoir , which eventually is fully integrated into the lens-facing neuro-epithelium ( Video 1 ) . This occurs through a sheet-like flow of lens-averted cells into the forming optic cup ( Figure 1E , H ) . This epithelial flow is independent of cell proliferation ( Figure 2—figure supplement 1 , Video 2 ) as demonstrated by aphidicolin treatment . The process is highly reminiscent of gastrulation movements and explains the marked increase of the lens-facing basal neuroretinal surface area . Notably , a small domain of the lens-averted epithelium exhibits a different morphology and behavior . As optic cup formation proceeds , this region flattens , enlarges , exhibits the morphological characteristics of RPE , and eventually ceases expressing RX2 , a marker for retinal stem and early progenitor cells ( Figure 1H , asterisks , Video 3 , in between arrows ) . 10 . 7554/eLife . 05216 . 005Video 2 . ( related to Figure 1 and Figure 2—figure supplement 1 ) Aphidicolin treated embryo . ( imaging started at 17 hpf , framerate 1/10 min ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 00510 . 7554/eLife . 05216 . 006Video 3 . ( related to Figure 1 ) ( control ) Optic vesicle to optic cup transition visualized by H2BGFP RNA into rx2::GFPcaax ( orientation as in Figure 2 ) , data derived from same imaging data as Video 4 , 3D rendered . Arrows mark the border between future RPE and Neuroretina ( imaging starts at 16 . 5 hpf , framerate 1/10 min ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 006 Our data highlight that almost the entire optic vesicle contributes to the formation of the neural retina . This new perspective on optic cup formation raises the question of how the elongated oval optic vesicle is transformed into the hemispheric optic cup . We addressed this by 4D imaging of optic cup formation using a nuclear label ( H2BGFP ) ( Figure 2A ) . We found , concomitant with lens formation , a prominent epithelial flow around the temporal perimeter of the forming optic cup . An involution of cells from the domain of the retinal pigmented epithelium ( RPE ) into the domain of the neuroretina had been proposed ( Li et al . , 2000 ) . Such reorganization of the lens-averted and the lens-facing epithelia , affecting the temporal optic cup , has been subsequently described ( Picker et al . , 2009 ) and confirmed ( Kwan et al . , 2012 ) . It was proposed that such ‘rim movements’ could occur around most of the optic vesicle circumference ( Kwan et al . , 2012 ) . 10 . 7554/eLife . 05216 . 007Figure 2 . Neuroepithelial flow drives morphological changes from optic vesicle to optic cup: the role of the optic fissure and the impact on the forming stem cell domain . ( A ) Dorsal view on optic cup development over time visualized by mosaic nuclear GFP ( H2BGFP ) ( data are derived from 4D imaging data started at 16 . 5 hpf , one optical section is provided as Video 4 ) , while the lens-facing neuroepithelium is starting to engulf the developing lens ( asterisk ) , the lens-averted epithelium is largely integrated into the lens-facing epithelium by flowing around the distal nasal and temporal rims ( arrows ) . A white dotted line marks the border between lens-facing and lens-averted epithelium . ( B ) Scheme showing the key findings of A , the lens ( asterisk ) facing epithelium is marked with red bars . The lens-averted epithelium , which over time is integrated into the lens-facing epithelium is marked with green dots ( except the cells at the edges are additionally marked with a yellow core ) . In between the last cells , which are integrated into the optic cup , the RPE will form in the lens averted domain . ( C ) shows the percentage of movements with a considerable share in dorso-ventral direction for the dorsal , central , and ventral area of the developing eye . In the ventral area of the eye , there is significantly more movement in the dorso-ventral axis , than in the central or dorsal area . ( D ) scheme demonstrating the optic vesicle to optic cup transition ( lateral view ) . Notably , the morphological change from the elongated oval optic vesicle to the hemispheric optic cup is driven mainly by the ventral regions ( arrows mark the orientation of epithelial flow ) ( C and D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 00710 . 7554/eLife . 05216 . 008Figure 2—figure supplement 1 . Epithelial flow is independent of cell division . ( A ) Retinal cell division was inhibited by application of aphidicolin , a well-established DNA polymerase inhibitor . Aphidicolin efficiently inhibited cell proliferation shown by drastically reduced pHH3 positive nuclei ( upper panel , average of 6 pHH3 positive nuclei ) compared to the control ( lower panel , average of 21 pHH3 positive nuclei ) ( the optic cup is encircled with a dotted white line , 21 . 5 hpf ) . ( B ) We addressed the epithelial flow of aphidicolin-treated wild-type embryos , injected with H2BGFP RNA at the one cell stage; please see Figure 2A as control . The embryo was preincubated with aphidicolin 5 hr prior to the start of imaging ( 17 hpf , see also Video 2 ) . The application of aphidicolin did not affect the epithelial flow . As a low level side effect of aphidicolin we observed cell death , in line with previous reports , importantly also not affecting the epithelial flow . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 008 Our data confirm a flow around the temporal perimeter and additionally demonstrate epithelial flow around the nasal perimeter into the forming optic cup . We uncover that the direction of the epithelial flow primarily establishes two distinct neuroretinal domains ( nasal and temporal ) separated by the static dorsal and ventral poles of the forming eye ( Figure 2D , Figure 3A ) . We use these poles as dorsal and ventral reference points throughout the manuscript . Importantly , the prominent rotation of the eye cup only occurs after the epithelial flow has ceased ( 24–36 hpf , Schmitt and Dowling , 1994 ) . 10 . 7554/eLife . 05216 . 009Figure 3 . Development of the CMZ and quantification of the flow towards this domain . ( A ) scheme of optic cup development ( lateral view over time ) including the results of nuclear tracking from the presumptive CMZ back in time to the lens-averted epithelium , remarkably two distinct domains became apparent within the lens-averted epithelium as the source for the presumptive CMZ . ( B ) Establishment of the presumptive CMZ domain ( dorsal view ) , nuclear tracking of cells ( maximum projection ) from the lens-averted domain ( encircled in upper picture ) eventually residing in the forming CMZ ( additionally encircled in lower picture ) , scalebar = 50 µm . ( C ) Scheme showing the optic cup from the lateral side . For quantification four domains were selected , nasal–dorsal , nasal–ventral , temporal–dorsal , and temporal–ventral . Note that the dorsal distal domain is only assembled secondarily and the ventral pole shows the optic fissure . ( D ) Based on differential effective distance , effective speed , and directionality , the migration distance was divided in two phases in the nasal and temporal domain , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 009 The prospective RPE remains in the lens-averted domain and expands in conjunction with the bi-furcated flow of the neuroretina from the lens-averted into the lens-facing domain ( Figure 2A , B , Video 3 ) . To further address the transformation of the elongated , oval optic vesicle into the hemispheric optic cup , we quantified cellular movements along the dorso-ventral axis . We found that the most prominent movements leading to the extension in the dorsal ventral axis occurred in the ventral domain ( Figure 2C ) . A key step in the formation of the ventral neuroretina is the formation of the optic fissure at the ventral pole of the optic vesicle . Lens-averted epithelium flows through this fissure into the forming optic cup to constitute the ventral neuroepithelium ( Figure 2D ) . Taken together , we present a model of optic cup formation , driven by gastrulation-like epithelial flow from the lens-averted into the lens-facing epithelium of the forming optic cup . The epithelium flows in two domains around the temporal and nasal rim , respectively and through the optic fissure of the forming optic cup . Overall , this has far-reaching implications for different aspects of eye development . One is the establishment of the retinal stem cell niche in the ciliary marginal zone ( CMZ ) ( Centanin et al . , 2011 ) , the distal rim of the optic cup/retina . To address whether the CMZ domain originates from a mixed population of progenitor cells that have been ‘set aside’ , or from a predefined coherent domain , we analyzed the transition from optic vesicle to optic cup in 3D over time ( 4D ) ( Video 4 ) . By tracking individual cells , we identified the origin of the distal retinal domain , the future CMZ , as two distinct domains ( nasal and temporal ) within the lens-averted epithelium at the optic vesicle stage ( Figure 3A , B , Video 5 ) . Based on tracking information , we noticed distinct phases during the flow from the lens-averted domain towards the CMZ ( Figure 3D ) . Although cells show high motility in an early phase ( Figure 3D , 1 ) , the directed flow is established only in a later phase ( Figure 3D , 2 ) , in which cells ultimately flow to the rim of the forming optic cup ( Figure 3D ) . 10 . 7554/eLife . 05216 . 010Video 4 . ( related to Figure 2 ) ( control ) Optic vesicle to optic cup transition visualized by H2BGFP RNA into rx2::GFPcaax ( orientation as in Figure 2 ) ( imaging starts at 16 . 5 hpf , framerate 1/10 min ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 01010 . 7554/eLife . 05216 . 011Video 5 . ( related to Figure 2 ) ( control ) Optic vesicle to optic cup transition visualized by H2BGFP RNA into rx2::GFPcaax ( orientation as in Figure 2 ) , data as in video 4 with tracked cells ( maximum projection ) to the presumptive CMZ . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 011 As indicated above , the dorsal pole of the optic vesicle remains static ( Figure 2D ) . Thus , the presumptive dorsal CMZ domain either originates from the lens-facing neuroretina or , alternatively , is established secondarily at a later time point , like the ventral CMZ in the region of the optic fissure . The identification of lens-averted domains as the source of the future nasal and temporal CMZ is consistent with the hypothesis of a distinct origin of retinal stem cells . Our data support a scenario in which the entire optic vesicle is initially composed of stem cells that at the lens-facing side respond to a signal to take a progenitor fate . We propose a tight coupling of morphogenesis with cell determination by inductive signals derived from the surface ectoderm to explain the successive spreading of retinal differentiation from the center to the periphery ( Sinn and Wittbrodt , 2013 ) . Accordingly , lens-averted stem cells might retain their stem cell fate because they are exposed to that signal at the latest point in time . An alternative hypothesis is that stemness might require an active process at the interface to the RPE; it is also possible that both scenarios are involved . Both scenarios are consistent with the expression pattern of rx2 , which is initially found in the entire optic vesicle and subsequently is confined to the CMZ . Strikingly , rx2-positive cells of the CMZ represent multipotent retinal stem cells ( Reinhardt , Centanin et al . , submitted ) . We demonstrated that cell motility and thus tissue fluidity are a prerequisite for neuroretinal flow . These characteristics are likely maintained through signaling , raising the question of which system might be involved . A likely candidate might be BMP , which has been linked to mobility in other tissues during development . In heart jogging , for example , BMP has an ‘antimotogenic’ effect ( Veerkamp et al . , 2013 ) . BMP signaling is important for various aspects of vertebrate eye development such as the enhancement of RPE and the inhibition of optic cup/neuroretina development ( Fuhrmann et al . , 2000; Hyer et al . , 2003; Müller et al . , 2007; Steinfeld et al . , 2013 ) , the formation of the dorso-ventral axis ( Behesti et al . , 2006; Holly et al . , 2014; Koshiba-Takeuchi et al . , 2000; Sasagawa et al . , 2002 ) , and the induction of the optic fissure ( Morcillo et al . , 2006 ) . Specific regions of the eye also seem to depend on the modulation of BMP signaling by the expression of a BMP antagonist ( Sakuta et al . , 2001 , French et al . , 2009 ) . We analyzed BMP signaling activity by assays based on the phosphorylation of the Smads 1/5/8 and the activation of a BMP signaling reporter ( Laux et al . , 2011 ) . BMP signaling is mainly elevated in the temporal domain and to a lesser degree in the nasal domain of the optic vesicle ( 16 . 5hpf , Figure 4A , B , D , E ) . At 21 . 5 hpf BMP signaling is confined to the dorsal pole of the optic cup ( Figure 4C , F ) . The transcriptional BMP sensor is activated with a delay and shows a more confined area of activity ( compare Figure 4A–C to Figure 4D–F ) . 10 . 7554/eLife . 05216 . 012Figure 4 . Analyses of BMP signaling and expression of BMP antagonists during development at 16 . 5 hpf , 19 hpf , and 21 hpf embryos are presented in a lateral view nasal left . ( A–C ) pSmad 1/5/8 immunohistochemistry ( red ) and DAPI nuclear staining . Activated BMP signaling can be appreciated mainly in the temporal domain of the optic vesicle ( arrows ) ( A–B ) and in the dorsal domain of the optic cup ( arrows ) ( C ) . At 16 . 5 hpf , a small domain of activated BMP signaling is visible in the nasal optic vesicle ( arrows ) ( A ) . ( D–F ) Immunohistochemically enhanced BRE::GFP ( green ) and DAPI nuclear staining . Activated BMP signaling can be appreciated in the temporal late optic vesicle ( arrows ) ( E ) and the dorsal optic cup ( arrows ) ( F ) . Hardly any activity can be detected in the optic vesicle at 16 . 5 hpf . Note the delay of activity in comparison to pSmad 1/5/8 . ( G–I ) Whole mount in situ hybridizations with a fsta probe ( Fast Red ) and DAPI nuclear staining . In the optic vesicle as well as in the optic cup two domains ( nasal and temporal ) of fsta expression can be seen ( arrows ) . ( J–L ) Whole mount in situ hybridizations with a bambia probe ( Fast Red ) and DAPI nuclear staining . Bambi expression can be seen in the temporal domain of the optic vesicle ( arrows ) ( J–K ) and in the dorsal domain of the optic cup ( arrows ) ( L ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 012 To address the means by which BMP activity is restricted , we analyzed the activity of prominent BMP antagonists follistatin a ( fsta ) ( Thompson et al . , 2005 ) , and bambi ( bambia ) ( French et al . , 2009 ) . Fsta was expressed in two domains , a nasal and a temporal domain ( Figure 4G–I and Figure 5A ) , whereas bambi was only expressed in the temporal domain of the optic vesicle ( Figure 4J , K ) and the dorsal domain of the optic cup ( Figure 4L ) . The regions of fsta expression correspond to the domains showing neuroretinal flow during optic cup formation . 10 . 7554/eLife . 05216 . 013Figure 5 . BMP antagonism drives neuroepithelial flow during optic cup formation . ( A ) whole mount in situ hybridization for fsta ( NBT/BCIP ) ( 17 . 5 hpf ) . ( B ) GFP expressed in the optic vesicle ( arrows ) of an rx2::GFPcaax zebrafish embryo ( 16 . 5 hpf ) , ( C–D ) GFP driven by the BRE and transmission/brightfield image for orientation . Strong GFP expression can be observed in the eye when BMP is driven under rx2 ( arrows in D ) , whereas only mild GFP can be observed in controls ( arrows in C ) . ( E ) Scheme showing the orientation of the pictures presented in F , ( F ) optic cup development over time of an rx2::BMP4 embryo . Cells are visualized by nuclear GFP ( H2BGFP ) . A dotted line is indicating the border between lens-averted and lens-facing epithelium . Remarkably , the pan-ocular driven BMP resulted in persisting lens-averted domains . The data presented in F are derived from 4D imaging data ( start at 16 . 6 hpf ) one optical section is also presented as Video 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 013 To address the importance of localized BMP signaling in wild-type embryos , we expressed BMP4 in the entire eye using an Rx2 proximal cis regulatory element ( Figure 5B ) , which overrides the localized BMP antagonist in the optic vesicle and optic cup . In BMP reporter fish ( Laux et al . , 2011 ) , we addressed BMP signaling activity under control and experimental conditions . At the optic cup stage , moderate BMP signaling activity was observed in the dorsal retina of control fish ( Figures 4F and 5C ) . The pan-ocular expression of BMP4 resulted in a strong response of the reporter , indicating pan-ocular BMP4 signaling ( Figure 5D ) . Strikingly resembling the BMP dependent ‘antimotogenic’ effect ( Veerkamp et al . , 2013 ) , pan-ocular BMP expression arrested epithelial flow during optic cup formation . Time-lapse in vivo microscopy revealed that cells in the lens-averted part of the future neuroretina remained in the prospective RPE domain and did not contribute to the optic cup ( Video 6 and 7 ) . This ultimately resulted in an apparently ectopic domain of neuroretina that arose from a morphogenetic failure , rather than from a trans-differentiation of RPE ( Figure 6D , Fig . 6—figure supplement 1 , 2 , 3 , Videos 8 , 9 , 10 ) . The severity of the phenotype correlated well with levels of fsta expression in the optic vesicle and was most prominent in the temporal domain of the optic vesicle . These findings highlight the importance of the modulation of BMP signaling for epithelial fluidity during the transformation from optic vesicle to optic cup . We propose that the repression of BMP signaling is crucial to mobilize the lens-averted retinal epithelium , causing it to flow and eventually constitute the neural retina to a large extent . 10 . 7554/eLife . 05216 . 014Video 6 . ( related to Figure 4 ) ( rx2::BMP4 ) Optic vesicle to optic cup transition visualized by H2BGFP RNA ( orientation as in Figure 3 ) ( imaging starts at 16 . 5 hpf , framerate 1/15 min ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 01410 . 7554/eLife . 05216 . 015Video 7 . ( related to Figure 4 ) ( rx2::BMP4 ) Optic vesicle to optic cup transition visualized by rx2::GFPcaax ( orientation as in Figure 3 ) ( imaging starts at 19 hpf , framerate 1/15 min ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 01510 . 7554/eLife . 05216 . 016Figure 6 . Impaired eye gastrulation results in coloboma . ( A–B ) Membrane-localized GFP ( rx2::GFPcaax ) in a developing eye during optic fissure closure ( A = early , B = late ) ( lateral view , derived from imaging data , ( A ) start at 24 hpf , ( B ) after 34 hr imaging at 22°C ) . Rx2 is expressed in retinal stem cells/RPCs ( A ) and after NR differentiation is additionally expressed in photoreceptors and Müller Glia ( B ) while its expression is maintained in retinal stem cells of the CMZ ( Reinhardt and Centanin et al . , submitted ) . The optic fissure margins are still undifferentiated ( arrows in B ) , ( C ) developing eye of rx2::BMP4 fish ( lateral view ) , membrane-localized GFP ( rx2::GFPcaax , anti-GFP immunointensified ) , DAPI nuclear labeling and anti-laminin immunostaining , the optic fissure is visible , noteworthy the temporal retina is mis-shaped and folded into the RPE domain ( best visible in DAPI , arrowheads ) , and located on a basal membrane ( arrowheads in anti-laminin ) , especially the temporal optic fissure margin ( arrowheads in GFPcaax ) is located in the folded part of the temporal retina and not facing the optic fissure ( arrows in GFPcaax ) ( 24 hpf ) ( D–E ) impaired optic fissure closure in rx2::BMP4 embryos over time at a proximal ( E ) and a distal ( D ) level . ( Data obtained from 4D imaging of rx2 ::BMP4/ rx2::GFPcaax started at 21 . 5 hr . Data are also presented as Video 10 . ) Importantly , next to the affected temporal optic cup also the nasal optic cup is mis-folded ( arrowheads in D ) . Remarkably , however , the nasal optic fissure margin extents into the optic fissure ( dashed arrow in E ) but the temporal optic fissure margin does not , likely being the result of the intense mis-bending of the temporal optic cup . This results in a remaining optic fissure ( asterisk in E ) . ( F–G ) Brightfield images of variable phenotype intensities observed in rx2::BMP4 hatchlings . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 01610 . 7554/eLife . 05216 . 017Figure 6—figure supplement 1 . Postembryonic eye development of rx2::BMP4 hatchlings . Although the temporal optic cup is largely malformed and folded it can be seen clearly , that vsx1 as well as vsx2 transgenes ( intensified by wholemount immunohistochemistry ) are expressed in the folded epithelium ( arrows ) . This indicates at least a partial correct differentiation into neuroretinal tissue . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 01710 . 7554/eLife . 05216 . 018Figure 6—figure supplement 2 . Lateral view on optic cup development over time , rx2::GFPcaax ( control ) is compared to rx2::BMP4 at proximal and distal levels . While in controls the lens-averted domain ( yellow dotted line ) is integrated into the developing optic cup it persists in rx2::BMP4 embryos . Note the increasing optic fissure ( arrows ) in rx2::BMP4 embryos . These data were obtained by 4D imaging ( start at 20 hpf ) and are also presented as Video 8 and Video 9 . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 01810 . 7554/eLife . 05216 . 019Figure 6—figure supplement 3 . Dorsal view on optic cup development of an rx2::BMP4 embryo over time at ventral vs central/ dorsal levels . The yellow dotted line indicates the border between the lens-facing and the lens-averted domain . Remarkably , the lens-averted domain is not integrated into the optic cup ( compare to Figures 1 and 2 ) . Notably , an altered morphology of the ventral optic vesicle can be observed showing the domain which is not going to be integrated ( arrows ) . These data are derived from imaging data ( start at 19 hpf ) which is presented in Video 7 . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 01910 . 7554/eLife . 05216 . 020Figure 6—figure supplement 4 . Ventral retinal identity remains in rx2::BMP4 embryos . Whole mount in situ hybridization with a vax2 probe ( NBT/BCIP ) of control ( left ) and rx2::BMP4 embryos ( right ) in a lateral view ( upper pictures ) and a ventral view ( lower pictures ) ( 28 hpf ) . Note that the ventral retinal marker remains expressed in the forming ventral optic cup even if BMP4 is expressed panocularly ( rx2::BMP4 ) . Also note that the vax2 domain in control is broader than in the rx2::BMP4 embryos ( arrows ) in which it is more prominent in the optic stalk region ( arrowhead ) ( the dotted line indicates the optic fissure ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 02010 . 7554/eLife . 05216 . 021Video 8 . Control to video 9 , optic cup development recorded with rx2::GFPcaax ( lateral view ) ( imaging starts at 20 hpf , framerate 1/10 min ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 02110 . 7554/eLife . 05216 . 022Video 9 . ( rx2::BMP4 ) Optic cup development recorded with rx2::GFPcaax ( lateral view ) ( imaging starts at 20 hpf , framerate 1/10 min ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 02210 . 7554/eLife . 05216 . 023Video 10 . Proximal domain of an rx2::BMP4 embryo showing an impaired optic fissure closure ( orientation as in Figure 5E ) ( imaging starts at 21 . 5 hpf , framerate 1/10 min ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05216 . 023 We further investigated the implications of impaired epithelial flow for subsequent steps of eye development ( e . g . , fate of the optic fissure ) . After initiation of neuroretinal differentiation in control embryos , the undifferentiated domains are restricted to the un-fused optic fissure margins and the forming CMZ . Both can be visualized by the expression of Rx2 ( Figure 6A , B ) . The impairment of neuroretinal flow , however , resulted in a mis-organization of the optic fissure . Here , the undifferentiated Rx2-expressing domain was found at the ultimate tip of the lens-averted neuroretinal domain , which failed to flow into the optic cup and persisted in the prospective RPE ( Figure 6C ) . As a result , the temporal optic fissure margin , in particular , failed to extend into the optic fissure ( Figure 6D–E ) . This also holds true , but to a lesser extent , to the nasal optic fissure margin ( Figure 6D ) . As a result , the two fissure margins cannot converge resulting in a persisting optic fissure , a coloboma . Macroscopically , the pan-ocular expression of BMP4 results in phenotypes including a ‘Plattauge’ ( flat-eye ) ( Figure 6G ) , in which the ventral part of the eye is strongly affected and a milder phenotype ( Figure 6F ) , in which the ventral retina develops , but with a persisting optic fissure . It was previously shown that exposing the developing eye to high levels of ectopically applied BMP can cause dorsalization , concomitant with a loss of ventral cell identities ( Behesti et al . , 2006 ) . This is likely the cause for coloboma ( Behesti et al . , 2006; Koshiba-Takeuchi et al . , 2000 , Sasagawa et al . , 2002 ) . Our data based on stable BMP4 expression ( rx2::BMP4 ) in the entire optic vesicle , however , conclusively show that early BMP4 exposure arrests neuroepithelial flow , resulting in a morphologically affected ventral retina . The ventral expression of vax2 in optic cups of rx2::BMP4 embryos indicates the maintenance of ventral retinal fates and argues against early transdifferentiation/dorsalization induced by BMP ( Figure 6—figure supplement 4 ) . Remarkably , the remaining lens-averted domain of those embryos , which was ectopically localized and was not integrated into the optic cup , eventually differentiated into neuroretina ( Figure 6—figure supplement 1 ) , as indicated by the expression of vsx1 ( Kimura et al . , 2008; Shi et al . , 2011; Vitorino et al . , 2009 ) and vsx2 ( formerly Chx10 ) ( Vitorino et al . , 2009 ) . Notably , a localization of neuroretina within the RPE domain might be mistaken for an RPE to neuroretina trans-differentiation , as proposed for other phenotypes ( Araki et al . , 2002; Azuma et al . , 2005; Sakaguchi et al . , 1997 , Bankhead et al . , 2015 ) . Even in amniotes , the histological analyses of consecutive stages of optic cup development are best interpreted as epithelial flow that also enlarges the retinal surface . This can even be appreciated during in vitro optic cup formation using mammalian embryonic stem cells ( Eiraku et al . , 2011 ) . Taken together , our data clearly show that during optic vesicle to optic cup transformation , the lens-averted part of the optic vesicle is largely integrated into the lens-facing optic cup by flowing around the distal rim of the optic cup including the forming optic fissure . Our data have far-reaching implications on the generation of the retinal stem cell niche of teleosts , as the last cells flowing into the optic cup will eventually constitute the CMZ . We identify a part of the lens-averted epithelium as the primary source of the RPE . The arrest of neuroepithelial flow by the ‘antimotogenic’ effect of BMP ( Veerkamp et al . , 2013 ) results in coloboma and thus highlights the importance of the flow through the fissure for the establishment of the ventral optic cup . It is unlikely that the bending of the neuroretina provides the motor for the epithelial flow; in the opo mutant no ectopic neuroretina can be found , indicating that the flow persists , even in the absence of optic cup bending ( Bogdanović et al . , 2012 ) . Consequently , forces established outside the neuroretina are likely to drive the flow . One tissue potentially involved is the mono-layered-forming RPE . We speculate that this tissue contributes to the flow by changing its shape from a columnar to a flat epithelium , massively enlarging its surface ( Figure 1J–K , Video 3 ) . This remains an interesting point , in particular given that epithelial flow is maintained even if cell proliferation is inhibited in both neuroretina and RPE .
BMP4 was cloned via directional Gateway from zebrafish cDNA into a pEntr D-TOPO ( Invitrogen , Germany ) vector with the following primers: forw: 5′ CACCGTCTAGGGATCCCTTGTTCTTTTTGCAGCCGCCACCATGATTCCTGGTAATCGAATGCTG 3′ , rev: 5′ TTAGCGGCA GCCACACCCCTCGACCAC 3′ . The expression construct was assembled via a Gateway reaction using Tol2 destination vector containing a cmlc: GFP ( Kwan et al . , 2007 ) , a 5′Entry vector containing an rx2 promoter ( Martinez-Morales et al . , 2009 ) , the vector containing the BMP4 and a 3′Entry vector containing a pA sequence ( Kwan et al . , 2007 ) . The construct was co-injected with mRNA encoding Tol2 transposase into the cytoplasm of zebrafish eggs at the one cell stage . Stable lines were preselected based on GFP expression in the heart ( cmlc2::GFP ) , raised and validated in F1 and subsequent generations . Lines were maintained as closed stocks and crossed to other lines as indicated in the manuscript . The rx2::GFPcaax construct was assembled with the 5′ and 3′ components described above and GFPcaax in the pEntr D-topo vector via Gateway ( Invitrogen ) and co-injected with mRNA encoding Tol2 transposase into the cytoplasm of zebrafish eggs at the one cell stage . Stable lines were preselected based on GFP expression in the heart ( cmlc2::GFP ) , raised , and validated in F1 and subsequent generations . Lines were maintained as closed stocks and crossed to other lines as indicated in the manuscript . The BRE::GFP zebrafish line ( Laux et al . , 2011 ) was kindly provided by Beth Roman . The Vsx1::GFP zebrafish line ( Kimura et al . , 2008; Shi et al . , 2011; Vitorino et al . , 2009 ) was kindly provided by Lucia Poggi . The Vsx2::RFP zebrafish line ( Vitorino et al . , 2009 ) was kindly provided by the lab of William Harris . Where indicated RNA for H2BGFP ( nuclear localized GFP ) ( 37 ng/µl ) was injected into 1–8 cell staged zebrafish embryos enabling 4D imaging of mosaically nuclear labeled zebrafish . Zebrafish embryos were treated with aphidicolin ( 10 µg/ml , Serva , Germany ) in order to inhibit cell proliferation . 12 embryos were treated with aphidicolin . 4D imaging was performed on one with an aphidicolin pretreatment of 5 hr . The efficacy of the treatment was addressed by analyzing nuclei in mitosis ( positive for the expression of phospho-histone H3 . At 21 . 5 hpf pHH3 positive nuclei were counted in central sections of four control ( untreated embryos from the same batch ) ( average: 21 ) and experimental ( average: 6 ) retinae , respectively . Optic cup surfaces were measured with the help of FIJI ( ImageJ NIH software ) . The mean of the length of the measured lines ( Figure 1C ) of two adjacent optical sections was multiplied by the optical section interval . Confocal data of whole mount immunohistochemical stainings a Leica ( Germany ) SPE microscope was used . Samples were mounted in glass bottom dishes ( MaTek , Ashland , MA ) . Olympus ( Germany ) stereomircoscope was used for recording brightfield images of rx2::BMP4 hatchlings and the overview of the expression of rx2::GFPcaax . For whole mount in situ data acquisition , a Zeiss ( Germany ) microscope was used . Time-lapse imaging was performed with a Leica SP5 setup which was upgraded to a multi photon microscope ( Mai Tai laser , Spectra Physics , Germany ) . It was recorded in single photon modus and multi photon modus . For time-lapse imaging , embryos were embedded in 1% low melting agarose and covered with zebrafish medium , including tricaine for anesthesia . Left and right eyes were used and oriented to fit the standard dorsal view or view from the side . Whole mount in situ hybridization was performed with probes for fsta bambia and vax2 . The probes were selfmade . Sequences were amplified by PCR from zebrafish cDNA and subcloned into pGEMTeasy vector ( Promega , Germany ) . In vitro transcription was performed with Sp6/T7 Polymerase . Hybridization was largely performed according to Quiring et al . ( 2004 ) . The Probe bas visualized with NBT/BCIP ( Roche , Switzerland ) or Fast Red ( Roche ) as indicated . Immunohistochemistry was performed according to a standard whole mount immunohistochemistry protocol . Briefly , embryos/hatchlings were fixed , washed , bleached ( KOH/H2O2 in PTW ) , and blocked ( BSA [1%] , DMSO [1%] , Triton X-100 [0 . 1%] , NGS [4%] , PBS [1×] ) . In case of anti-pSmad 1/5/8 immunohistochemistry embryos were additionally treated with proteinase K ( 10 µg/ml , 16 . 5 hpf: 5 min , 19 hpf and 21 . 5 hpf: 6 min ) . Samples were incubated in primary antibody solution ( anti-laminin , 1:50 , Abcam , Germany ) ( anti-GFP 1:200 , life technologies , Germany ) ( anti-dsRED , Clontech , Germany ) ( anti-pHH3 , 1:100 , Milipore , Germany ) ( anti-pSmad1/5/8 , 1:25 , Cell Signaling , Germany ) in blocking solution . Samples were washed and incubated in secondary antibody solution ( anti-rabbit Dylight , 1:300 , anti-chicken Alexa 488 , 1:300 , Jackson , UK ) with DAPI ( stock: 2 µg/ml , 1:500 ) added . Consecutively , samples were washed and mounted for microscopy . The amount of movement in the dorso-ventral axis was quantified using a supervoxel based Optical Flow algorithm ( Amat et al . , 2013 ) . The pixel wise output was visualized by applying a spherical coordinate system to the eye using a custom made ImageJ plugin ( Source code 1: file plugin ) . The color coding is based on the sign of the polar angle theta and the sign of the azimuth angle phi , as well as on their respective combinations . The quantification was performed by counting the labeled pixels in an ImageJ macro ( Source code 1: file macro ) . Cells were tracked manually using MtrackJ ( Meijering et al . , 2012 ) in Fiji ( ImageJ ) ( Schindelin et al . , 2012 ) back in 4D stacks to their original location or until lost . Only tracks with a significant length were used for the visualizations . Centered on the track cells are represented as spheres . Partially results are presented in a side view where the dorso-ventral axis originally represented as the z-axis has now become the y-axis . A factor of 10 . 5703 is introduced in order to adjust the data of the former z-axis to the other two axes . The color coding is done by choosing colors from an 8 bit lookup table and applying them from the dorsal to the ventral side based on the end of the track . Partially tracking results are presented as tailed spheres . The spheres are based on the tracking data using an average over the last three timepoints . The image is stretched in the z-axis using a factor of 10 . 5703 , to adjust the scale to the x and y axes . Tails are created using a lookup table with 16 different shades per color . The respective shade is defined by the distance and difference in time between the recent position and the position on the tail . | The eye is our most important organ for sensing and recognizing our environment . In humans and other vertebrates , the eye forms from an outgrowth of the brain as the embryo develops . This outgrowth is called the optic vesicle and it is rapidly transformed into a cup-shaped structure known as the optic cup . Defects in this process prevent the optic cup from closing completely , which leads to a severe condition called Coloboma—one of the most frequent causes of blindness in children . The optic cup has two distinct layers: the inside layer—known as the neuroretina—contains light sensitive cells and is surrounded by the other layer called the pigmented epithelium . It is thought that the neural retina is made from cells from the side of the optic vesicle that faces the lens , and the pigmented epithelium is formed by cells from the other side of the vesicle . This is a plausible explanation and is well accepted , but it cannot explain how the neuroretina can become five times larger as the cup forms . Heermann et al . addressed this problem by using four-dimensional in vivo microscopy to follow individual cells as the optic cup forms in living zebrafish embryos . The experiments show that the neuroretina is made of cells from both sides of the optic vesicle . Cells from the back of the optic vesicle ( furthest away from the lens ) join the rest of the cells by moving around the outside rim of the cup . Further experiments found that a signaling molecule called BMP—which is crucial to the normal development of the eye—controls the flow of cells around the developing optic cup . This factor needs to be carefully controlled during the development of the eye; when BMP activity was artificially increased , the flow of cells stopped , resulting in neuroretinal tissue developing in the wrong place ( in the outer layer of the optic cup ) . The experiments also reveal that the stem cells in the retina—which divide to produce new cells throughout the life of the zebrafish—originate from two distinct areas in the optic vesicle . Heermann et al . 's findings challenge the textbook model of eye development by revealing that cells from both sides of the optic vesicle contribute to the neuroretina and that retinal stem cells originate from a specific place in the developing eye . A future challenge will be to understand how the movement of the cells into the neuroretina is coordinated to make a perfectly shaped eye . | [
"Abstract",
"Main",
"Text",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"short",
"report",
"neuroscience"
] | 2015 | Eye morphogenesis driven by epithelial flow into the optic cup facilitated by modulation of bone morphogenetic protein |
The C-cluster of the enzyme carbon monoxide dehydrogenase ( CODH ) is a structurally distinctive Ni-Fe-S cluster employed to catalyze the reduction of CO2 to CO as part of the Wood-Ljungdahl carbon fixation pathway . Using X-ray crystallography , we have observed unprecedented conformational dynamics in the C-cluster of the CODH from Desulfovibrio vulgaris , providing the first view of an oxidized state of the cluster . Combined with supporting spectroscopic data , our structures reveal that this novel , oxidized cluster arrangement plays a role in avoiding irreversible oxidative degradation at the C-cluster . Furthermore , mutagenesis of a conserved cysteine residue that binds the C-cluster in the oxidized state but not in the reduced state suggests that the oxidized conformation could be important for proper cluster assembly , in particular Ni incorporation . Together , these results lay a foundation for future investigations of C-cluster activation and assembly , and contribute to an emerging paradigm of metallocluster plasticity .
Roughly half of all enzymes make use of metal centers to expand their chemical repertoire ( Waldron et al . , 2009 ) . Among the most fascinating of the metallocofactors used for such purposes are Fe-S clusters , which are thought to be the most ancient biological cofactors and which enable chemical transformations ranging from simple electron transfer events to the formation and cleavage of carbon-carbon bonds ( Rees and Howard , 2003; Beinert et al . , 1997 ) . Complex Fe-S clusters , containing alternative metal ions and/or expanded metal frameworks , such as the FeMo-cofactor of nitrogenase , the H-cluster of Fe-Fe hydrogenase , and the C-cluster of Ni-dependent carbon monoxide dehydrogenase ( CODH ) , catalyze fundamental redox conversions that are thought to have enabled early life on Earth ( Rees and Howard , 2003; Rees , 2002 ) . Given their structural complexity and the essential nature of the reactions they catalyze , these clusters have collectively been termed the ‘great clusters’ of biology ( Rees , 2002 ) . The great clusters , and the proteins that house them , have become the focus of extensive mechanistic and structural investigation in the hopes of yielding new applications in clean energy production and bioremediation . In particular , CODH catalyzes the interconversion of the gaseous pollutant CO and the greenhouse gas CO2 , leading to the removal of an estimated 108 tons of CO from the lower atmosphere each year and making it an attractive remediation tool ( Bartholomew and Alexander , 1979 ) . The anaerobic , Ni-dependent CODH has a homodimeric structure containing a total of five metalloclusters , called the B- , C- , and D-clusters . The C-cluster is the site of CO/CO2 interconversion and is composed of a [Ni-3Fe-4S] cubane connected through a linking sulfide ( SL ) to a unique iron site ( Feu ) ( Figure 1 ) ( Drennan et al . , 2001; Dobbek et al . , 2001 ) . Comprehensive spectroscopic analyses have revealed the basic redox states and kinetic properties of this complex metallocluster , and crystal structures with substrates and inhibitors bound have provided snapshots along the reaction pathway ( Figure 2 ) ( Jeoung and Dobbek , 2007; Gong et al . , 2008; Kung et al . , 2009; Jeoung and Dobbek , 2009; Fesseler et al . , 2015; Lindahl et al . , 1990; Kumar et al . , 1993; Anderson and Lindahl , 1994; Anderson and Lindahl , 1996; Seravalli et al . , 1997; Fraser and Lindahl , 1999; Chen et al . , 2003; Seravalli and Ragsdale , 2008; Drennan and Peters , 2003 ) . These studies have revealed the C-cluster to take on four discrete redox states termed Cox , Cred1 , Cint , and Cred2 ( Lindahl et al . , 1990; Kumar et al . , 1993; Anderson and Lindahl , 1994; Anderson and Lindahl , 1996; Seravalli et al . , 1997; Fraser and Lindahl , 1999 ) . The most widely accepted mechanism of CO oxidation involves a one-electron reductive activation of the inactive Cox state to Cred1 followed by a catalytic cycle involving conversion between Cred1 and Cred2 ( Figure 2 ) ( Lindahl et al . , 1990; Kim et al . , 2004; Lindahl , 2008 ) . Despite this relatively unified understanding of CO oxidation activity , there are still many gaps in our understanding of this complicated enzyme that are limiting with regard to both our understanding of CODH biochemistry and potential applications of CODH in industrial settings . In particular , there has been a push to characterize enigmatic redox states and also to probe the effects of molecular oxygen on enzyme activity ( Merrouch et al . , 2015; Wang et al . , 2015; Domnik et al . , 2017 ) . Here , we report the crystal structure of the CODH from Desulfovibrio vulgaris ( DvCODH ) ( Hadj-Saïd et al . , 2015 ) , which reveals a surprising and unprecedented conformational rearrangement of metal ions in the C-cluster and provides the first visualization of the cluster in an oxidized state . Through combined structural and spectroscopic data , we show that conversion between the oxidized and reduced states of the cluster is reversible , consistent with previous electrochemical investigations ( Merrouch et al . , 2015 ) . We further consider the implications of these findings in terms of oxygen sensitivity and cluster assembly and with respect to the other great clusters in biology .
The overall fold and cluster placement of DvCODH is highly similar to other structurally characterized CODHs ( Figure 1—figure supplement 1 ) ( Drennan et al . , 2001; Dobbek et al . , 2001; Domnik et al . , 2017; Doukov et al . , 2002; Darnault et al . , 2003 ) . One noteworthy difference with respect to other CODHs is the identity of the D-cluster , a solvent-exposed Fe-S cluster at the dimer interface that serves as an electron conduit to the surface of the protein . Instead of the expected [4Fe-4S] cluster , ( Drennan et al . , 2001; Dobbek et al . , 2001; Domnik et al . , 2017; Doukov et al . , 2002; Darnault et al . , 2003 ) the electron density is consistent with a [2Fe-2S] cluster ( Figure 1—figure supplement 2 ) as is the placement of cysteine residues in the primary structure . CODH sequence alignments reveal that instead of a C-X7-C D-cluster binding motif , DvCODH , as well as several uncharacterized CODHs , have a C-X2-C motif ( Figure 1—figure supplement 2 ) . This shortened C-X2-C motif appears to constrain the geometry of the ligating cysteine residues such that coordination to a [4Fe-4S] cluster is not possible . Instead , the cysteine positions are ideally suited for coordination of a [2Fe-2S] cluster . In the present work , we determined two structures of as-isolated DvCODH using two independent protein batches and much to our surprise , we observe different conformations of the C-cluster in each structure . A 2 . 50 Å resolution structure of as-isolated DvCODH ( determined using protein batch 1 ) displays the canonical [Ni-3Fe-4S]-Feu C-cluster ( Figure 3a; Figure 3—figure supplement 1a; Supplementary file 1 ) ( Drennan et al . , 2001; Gong et al . , 2008; Doukov et al . , 2002; Darnault et al . , 2003 ) ; the [Ni-3Fe-5S]-Feu state that was observed in structures of Carboxydothermus hydrogenoformans CODH-II ( Dobbek et al . , 2001; Dobbek et al . , 2004 ) is no longer thought to be catalytically relevant ( Jeoung and Dobbek , 2007; Kung et al . , 2009; Drennan and Peters , 2003; Feng and Lindahl , 2004 ) . The ligation of the C-cluster is conserved in DvCODH ( Drennan et al . , 2001; Dobbek et al . , 2001; Gong et al . , 2008; Domnik et al . , 2017; Doukov et al . , 2002; Darnault et al . , 2003 ) with four cysteines ligating the cubane portion of the cluster ( Cys519 is the Ni ligand ) ; one histidine ( His266 ) and one cysteine ( Cys302 ) ligate Feu ( Figure 3a; Figure 3—figure supplement 1a ) . The second , higher-resolution ( 1 . 72 Å ) structure of as-isolated DvCODH ( determined using protein batch 2 ) reveals a novel arrangement of ions within the C-cluster that was confirmed using anomalous diffraction data ( Figure 3b; Supplementary file 1 ) . In this structure , Ni , Feu , and SL are shifted , accompanied by conformational changes of several amino acid side chains , while the positions of the remaining three Fe and three S ions are unchanged . The Ni ion is bound in the site formerly occupied by Feu , coordinated by His266 and Cys302 ( Figure 3b; Figure 3—figure supplement 1b ) . The Ni is additionally ligated by Cys519 and Lys556 ( Figure 3b; Figure 3—figure supplement 1b ) . As mentioned above , Cys519 serves as a ligand to Ni in the canonical C-cluster and here adopts an alternative rotamer conformation such that coordination to Ni is maintained ( Figure 3b; Figure 3—figure supplement 1b ) . The occupancy of the alternative Cys519 conformation correlates with the occupancy of Ni , and both have been refined at an atomic occupancy of 70% , in general agreement with the metal analysis result of 0 . 5 Ni per monomer for the sample that was crystallized ( Supplementary file 2 ) . Lys556 is highly conserved and does not normally coordinate to the C-cluster , but is instead the proposed general base catalyst for deprotonation of water during CO oxidation ( Figure 2 ) ( Drennan et al . , 2001; Dobbek et al . , 2001; Kim et al . , 2004 ) . Here , the lysine amine group comes within 2 . 5 Å of Ni . Together , His266 , Cys302 , Cys519 , and Lys556 ligate the Ni in this altered cluster in a highly distorted tetrahedral coordination geometry that is reminiscent of the geometry of the Ni site in Ni-Fe hydrogenases ( Volbeda et al . , 1995 ) . Concomitant with the shift of the Ni ion , Feu and SL also undergo changes in their coordination environments while remaining associated with the [3Fe-3S] partial cubane ( Figure 3b; Figure 3—figure supplement 1b ) . In addition to interaction with SL , Feu is coordinated by Cys302 , which forms a bridging interaction with Ni , and by a conserved cysteine residue ( Cys301 ) that does not normally serve as a ligand to the C-cluster , resulting in an apparent three-coordinate geometry around Feu ( Figure 3b; Figure 3—figure supplement 1b ) . The shift in the positions of Feu and SL results in the loss of an interaction between SL and a second Fe ion of the cubane ( Fe1 ) , leaving Fe1 with three coordinating ligands , Cys340 and two cubane sulfides . A small peak of residual electron density at a site bridging Feu and Fe1 is present in one of the protein chains in the asymmetric unit ( Figure 3—figure supplement 2 ) . The occupancy of this site is low ( <30% ) precluding identification of the atom/ion . However , if fully occupied , the atom/ion would complete the tetrahedral geometry around these Fe atoms . In addition to changes in Fe coordination , the altered conformation of the C-cluster also involves changes in sulfide coordination state . In particular , the two cubane S ions that coordinate Ni in the canonical cluster are left in a possibly unstable state in which they could be susceptible to protonation or loss as free sulfide ions ( Crack et al . , 2006 ) . In our structure , however , we see no evidence of degradation at these sites , likely due to inaccessibility to solvent or other protective features of the protein environment , in analogy to the case of stable [3Fe-4S] clusters . We next investigated whether this altered C-cluster state is redox dependent . Pre-formed crystals of as-isolated DvCODH ( batch 2 , containing the altered cluster ) were incubated with the reductant sodium dithionite . Strikingly , the resulting crystal structure displays the canonical C-cluster with Ni , Feu , and SL rearranged into their catalytically-relevant positions ( Figure 4a , b; Supplementary file 1 ) . To examine whether this metal rearrangement is reversible upon oxidation , crystals of reduced DvCODH were taken from the anaerobic chamber and incubated under ambient atmospheric conditions . Remarkably , oxidation of the reduced C-cluster by exposure to O2 results in reformation of the unusual cluster architecture ( Figure 4c; Supplementary file 1 ) , whereas both the D- and B-clusters remain intact ( Figure 4—figure supplement 1 ) . Together , these results suggest that this altered cluster is an oxidized form of the C-cluster and that this multi-metal ion rearrangement is reversible ( Figure 4d , Figure 4—video 1 ) . Most likely , a fortuitous oxidation event , affecting DvCODH batch 2 , initially allowed us to obtain the first visualization of an oxidized state of the C-cluster . Given the apparent ability of DvCODH to undergo fully reversible oxidation/reduction events in crystallo , we used electron paramagnetic resonance ( EPR ) spectroscopy to determine the effect of oxidation on the enzyme in solution . First , an EPR spectrum was recorded on a sample of dithionite-reduced DvCODH . This spectrum exhibits resonances characteristic of the one-electron reduced B- and D-clusters ( centered around g ~ 2 ) and of the Cred1 and Cred2 forms of the C-cluster ( gav ~ 1 . 83 and 1 . 87 , respectively; see Figure 2 ) , as has been previously observed for DvCODH ( Hadj-Saïd et al . , 2015 ) ( Figure 4e , black trace ) . A parallel dithionite-reduced sample was incubated under ambient atmospheric conditions to mimic treatment of the DvCODH crystals ( EPR-silent , data not shown ) . The oxygen-exposed sample was then re-reduced with dithionite and the EPR spectrum was recorded , revealing full recovery of the previously observed signals ( Figure 4e , red trace ) . Combined , our crystal structures and EPR data reveal that the metalloclusters of DvCODH are not degraded upon oxidation and that the C-cluster avoids degradation by adopting an alternative , stable , oxidized conformation . Consistent with these results , DvCODH was recently shown to regain activity upon chemical or electrochemical reduction following exposure to molecular oxygen ( Merrouch et al . , 2015 ) . In this respect , the rearranged C-cluster scaffold can be thought of as a ‘safety net’ for retaining cluster ions upon oxygen exposure , and may explain , at least in part , the ability of certain CODHs to recover activity following oxidation in air ( Merrouch et al . , 2015; Wang et al . , 2015; Domnik et al . , 2017 ) . This kind of safety net could improve the ability of an organism to rapidly recover CODH activity after transient exposure to oxic conditions . One of the more noteworthy aspects of the oxidized C-cluster is that cluster ligation involves one residue ( Cys301 ) that is strictly conserved in CODHs but is not a ligand to the canonical C-cluster . Interestingly , previous work on the heterotetrameric CODH/acetyl-CoA synthase from Moorella thermoacetica ( MtCODH/ACS ) showed that mutation of the equivalent cysteine residue ( Cys316 ) to serine resulted in an inactive CODH that appeared to lack an intact C-cluster ( Kim et al . , 2004 ) . To probe the effect of Cys301 in DvCODH , a DvCODH ( C301S ) variant was produced and characterized . Similar to what was observed with MtCODH/ACS ( Kim et al . , 2004 ) , DvCODH ( C301S ) is inactive and does not contain Ni , as assessed by CO oxidation activity assays and inductively coupled plasma optical emission spectroscopy ( ICP-OES ) , respectively ( Supplementary file 2 ) . Unlike wild-type DvCODH , DvCODH ( C301S ) cannot be activated by Ni under reducing conditions , analogous to what has been observed previously for DvCODH samples grown in the absence of the C-cluster maturation factor CooC ( Hadj-Saïd et al . , 2015; Merrouch et al . , 2018 ) . To investigate the architecture of a non-activatable C-cluster , we determined the crystal structure of DvCODH ( C301S ) to 2 . 0 Å resolution ( Supplementary file 1 ) and discovered an intact [3Fe-4S] C-cluster core with an Feu ion that adopts alternative conformations ( Figure 5 ) . Approximately 70% of Feu is in its canonical position coordinated by His266 and Cys302 , whereas approximately 30% is in the Ni-cubane site ( see Methods; Figure 5—figure supplement 1 ) . Thus , in the absence of the non-canonical Feu-ligand Cys301 , the cluster cannot be activated by Ni and Feu appears to be free to occupy multiple sites . Taken together , these data suggest that the novel structure of the C-cluster that we observe here , with Feu coordinated by Cys301 , is relevant to processes beyond oxidation , specifically Ni incorporation . Regardless of its role ( s ) , the dramatic rearrangement observed for the C-cluster adds to a growing appreciation that the ions of great clusters are in fact mobile . The oxidized P-cluster of nitrogenase adopts an open conformation relative to its reduced form through the outward movement of two Fe ions by 1 . 4 and 0 . 9 Å with accompanying loss of S coordination in a rearrangement that is proposed to couple proton transfer to electron transfer ( Figure 6a ) ( Peters et al . , 1997 ) . More recently , Rees and coworkers demonstrated that both the inhibitor CO and an artificially-incorporated Se atom are able to displace a S atom of the cluster and furthermore that , under turnover conditions , the Se atom migrates around the cluster , sampling S positions ( Figure 6b ) ( Spatzal et al . , 2014; Spatzal et al . , 2015 ) . Very recently , it was also revealed that the same S atom of the VFe-cofactor ( containing V in place of Mo ) is displaced by a NH ligand , suggesting that cluster dynamics are likely key in catalysis ( Sippel et al . , 2018 ) . Additionally , an O2-tolerant membrane-bound hydrogenase was shown to have a novel [4Fe-3S] cluster that undergoes redox-dependent structural changes as part of an O2-tolerance mechanism ( Fritsch et al . , 2011; Shomura et al . , 2011 ) ; one Fe ion moves ~1 . 6 Å upon cluster oxidation and becomes coordinated by a protein backbone amide group ( Figure 6c ) ( Shomura et al . , 2011 ) . In the present work , the metal migration of C-cluster atoms is more dramatic than in these other examples , with Ni moving ~3 Å and adopting an entirely new coordination environment , and Feu and SL moving ~1 . 9 and 2 . 6 Å , respectively , with Feu also taking on a new coordination environment . This C-cluster rearrangement from oxidized to reduced appears to involve what amounts to a ‘molecular cartwheel’ with Ni , Feu , and SL following the same trajectory to end up in their canonical positions ( Figure 4d , Figure 4—video 1 ) . In summary , X-ray crystallography has provided views of the ‘great clusters’ of biology , allowing us to marvel at these incredible metallic frameworks that capture and make use of CO , H2 , and N2 gases . We are increasingly finding that these frameworks should not be thought of as rigid scaffolds , but rather as labile assemblies of metal with sulfide . The full nature and significance of this metallocluster lability is just now beginning to emerge and the roles appear to be diverse , including catalysis , electron transfer , protection from oxygen damage , and possibly cluster assembly . The one consistency is that these great clusters continue to surprise us .
DvCODH was expressed in the presence of the C-cluster maturation factor CooC , as described previously ( Hadj-Saïd et al . , 2015 ) . Briefly , the D . vulgaris genes encoding CODH ( cooS ) and the CooC maturase ( cooC ) were cloned into modified pBGF4 shuttle vectors under the control of the promoter of the Desulfovibrio fructosovorans Ni-Fe hydrogenase operon . The CODH construct was N-terminally strep-tagged . A construct encoding DvCODH ( C301S ) was generated from the wild-type sequence by site-directed mutagenesis ( forward primer ACATCAACGTGGCGGGGCTATCCTGCACGGGTAACGAACTGCTC , reverse primer GAGCAGTTCGTTACCCGTGCAGGATAGCCCCGCCACGTTGATGT; mutation underlined ) . Protein was expressed in D . fructosovorans str . MR400 ( Rousset et al . , 1991 ) and purified under anaerobic conditions in a Jacomex anaerobic chamber ( 100% N2 atmosphere ) by affinity chromatography on Strep-Tactin Superflow resin , as described previously ( Hadj-Saïd et al . , 2015 ) . Protein concentrations were determined by amino acid analysis at the Centre for Integrated Structural Biology ( Grenoble , France ) . Metal content ( Supplementary file 2 ) was analyzed by inductively coupled plasma optical emission spectroscopy ( ICP-OES ) . CO oxidation activity was assayed at 37 °C by monitoring the reduction of methyl viologen at 604 nm ( ε = 13 . 6 mM−1·cm−1 ) , as described previously ( Hadj-Saïd et al . , 2015 ) ( Supplementary file 2 ) . All crystals were grown using as-isolated protein samples ( i . e . , samples were not activated with NiCl2 and sodium dithionite prior to crystallization ) . Crystals were grown anaerobically in an N2 atmosphere at 21 °C by hanging drop vapor diffusion in an MBraun anaerobic chamber . Crystals belonging to space group P212121 were obtained as follows: A 1 µL aliquot of as-isolated protein ( 10 mg/mL in 100 mM Tris-HCl pH 8 ) was combined with 1 µL of precipitant solution ( 1 . 0–1 . 1 M ammonium tartrate dibasic pH 7 , 6–9% ( v/v ) glycerol ) on a glass cover slide and sealed over a reservoir containing 500 µL of precipitant solution . Diffraction quality crystals grew in 2–10 d . Crystals were soaked in a cryo-protectant solution containing 1 . 0–1 . 2 M ammonium tartrate dibasic pH 7 , 25% ( v/v ) glycerol and cryo-cooled in liquid nitrogen . Crystals belonging to either space group P21 or P1 were obtained as follows: A 1 µL aliquot of as-isolated protein ( 10 mg/mL in 100 mM Tris-HCl pH 8 ) was combined with 1 µL of precipitant solution ( 150–250 mM MgCl2 , 16–20% ( w/v ) PEG 3350 ) on a glass cover slide and sealed over a reservoir containing 500 µL of precipitant solution . Diffraction quality crystals grew in 1–6 d . Crystals were soaked in a cryo-protectant solution containing 250 mM MgCl2 , 18–20% ( w/v ) PEG 3350 , 9% ( v/v ) glycerol and cryo-cooled in liquid nitrogen . To reduce crystals of as-isolated DvCODH , crystals were transferred into a soaking solution containing 250 mM MgCl2 , 18% ( w/v ) PEG 3350 , 5 mM sodium dithionite and incubated for 30 min . For structures of reduced DvCODH , crystals were transferred to a cryo-protectant solution containing 250 mM MgCl2 , 18% ( w/v ) PEG 3350 , 9% ( v/v ) glycerol and cryo-cooled in liquid nitrogen . For structures of reduced and then oxygen-exposed DvCODH , crystals were transferred into a dithionite-free drop containing 250 mM MgCl2 , 18% ( w/v ) PEG 3350 prior to removal from the anaerobic chamber to avoid reaction of excess dithionite with molecular oxygen . Following removal from the chamber , 0 . 5 µL of aerobically-prepared precipitant solution was added to the drop to initiate equilibration with ambient atmospheric conditions . Crystals were harvested after 2 d as described above . All data were collected at the Advanced Photon Source ( Argonne , IL ) at beamline 24-ID-C at a temperature of 100 K using a Pilatus 6M pixel detector . Where applicable , native , Fe peak , and Ni peak data were collected on the same crystal for a particular sample . Native data were collected at an energy of 12662 eV ( 0 . 9792 Å ) ; Fe peak data at 7130 eV ( 1 . 7389 Å ) ; and Ni peak data at 8360 eV ( 1 . 4831 Å ) . All data were integrated in XDS and scaled in XSCALE ( Kabsch , 2010 ) . Data collection statistics are summarized in Supplementary file 1 . The initial structure of DvCODH was determined to 2 . 50 Å resolution by molecular replacement ( MR ) in the program Phaser ( McCoy et al . , 2007 ) using data from crystals belonging to space group P212121 . The search model for MR was generated from the structure of the CODH from Rhodospirillum rubrum ( 47% sequence identity; PDB ID: 1JQK ) by modification in Sculptor ( Bunkóczi and Read , 2011 ) using the Schwarzenbacher algorithm ( Schwarzenbacher et al . , 2004 ) with a pruning level of 2 to truncate non-identical residues at the Cβ position . Metalloclusters were not included in the search model . A single MR solution was found with an LLG of 311 , TFZ of 21 . 3 , and R-value of 57 . 9 . The model was completed through iterative rounds of model building in Coot ( Emsley et al . , 2010 ) and refinement in Phenix ( Adams et al . , 2010 ) ( see below ) . Subsequent structures were determined by MR in Phaser using the initial DvCODH structure as a search model . Following MR , 10 cycles of simulated annealing refinement were performed in Phenix to eliminate existing model bias . For all structures , refinement of atomic coordinates and atomic displacement parameters ( B-factors ) was carried out in Phenix using noncrystallographic symmetry ( NCS ) restraints . Models were completed by iterative rounds of model building in Coot and refinement in Phenix . In advanced stages of refinement , water molecules were added automatically in Phenix and modified in Coot with placement of additional water molecules until their number was stable . Final stages of refinement included translation , libration , screw ( TLS ) parameterization with one TLS group per monomer ( Painter and Merritt , 2006 ) . For structures determined to less than or equal to 2 Å resolution , NCS restraints were removed in final refinement cycles . In advanced stages of structural refinement of the 1 . 72 Å as-isolated DvCODH structure , it became clear that two conformations of the C-cluster were present . Based on the electron density , the Fe-S scaffold of the oxidized form of the cluster was modeled at an occupancy of 80% , with the Ni ion at 70% . The canonical , reduced form of the cluster without Ni was modeled with an atomic occupancy of 20% . A peak of residual electron density at a position bridging Feu and Fe1 of the oxidized cluster appeared in late stages of refinement . Modeling of a water molecule at this position resulted in a refined occupancy of ~30% . The geometry of this site , however , is not consistent with coordination of H2O/OH− , and given the long Fe-ligand bond distances ( 2 . 4 Å ) , the site is likely occupied by a heavier atom , for example Cl− from the protein buffer . Due to the low occupancy ( <30% ) of an atom heavier than water at this site and the inability to resolve the identity of this ligand crystallographically , this site was left unmodeled in the final structure . The structure of DvCODH ( C301S ) also contained an apparent mixture of cluster types at the C-cluster site . Here , the [3Fe-4S] partial cubane portion of the canonical C-cluster is intact and present at full occupancy ( Figure 5—figure supplement 1 ) ; however , modeling of Feu proved complicated . When modeled and refined as a [3Fe-4S]-Feu cluster at full occupancy , the atomic displacement parameter ( B-factor ) of Feu was higher ( 38 . 7 Å2 ) than the average for the ligating atoms of His266 ( Nε ) and Cys302 ( Sγ ) ( 22 . 9 Å2 ) as well as for the remainder of the cluster ( 24 . 0 Å2 ) , suggesting that Feu may be present at reduced occupancy . Additionally , positive difference electron density ( Fo−Fc ) near the Ni-binding site of the C-cluster was observed , indicating the presence of an atom in this site ( Figure 5—figure supplement 1a ) . Based on the ICP-OES results , DvCODH ( C301S ) does not contain Ni , suggesting that an atom other than Ni occupies this site in the structure . Anomalous difference maps calculated from diffraction data collected at the iron peak wavelength ( 7130 eV ) revealed a shoulder extending from Feu into this site , indicative of the presence of Fe at partial occupancy within the cubane ( Figure 5—figure supplement 1b ) . Together , the native diffraction data , anomalous difference data , and B-factor analysis suggested that there are two different states of the C-cluster in the sample: one with the canonical [3Fe-4S]-Feu scaffold and one in the form of a distorted [4Fe-4S] cubane . Indeed , when Feu is modeled with a split conformation such that at 70% occupancy it is present in its unique binding site and at 30% occupancy it is incorporated into the cubane , the cluster refines well into the electron density and the B-factors of Feu are better matched with those of the surrounding atoms ( Figure 5 ) . Final refinement of each structure yielded models with low free R-factors , excellent stereochemistry , and small root mean square deviations from ideal values for bond lengths and angles . All refinement statistics are summarized in Supplementary file 1 . Side chains without visible electron density were truncated to the last atom with electron density and amino acids without visible electron density were not included in the models . Final models contain the following residues ( of 629 total ) : as-isolated ( batch 1 ) : 4–628 ( chains A and B ) ; as-isolated ( batch 2 ) : 4–629 ( chain A ) , 2–629 ( chain B ) ; reduced ( batch 2 ) : 4–627 ( chain A ) , 4–628 ( chain B ) ; reduced/O2-exposed ( batch 2 ) : 8–63 , 68–286 , 289–629 ( chain A ) , 6–63 , 67–287 , 291–628 ( chain B ) ; DvCODH ( C301S ) : 4–628 ( chain A ) , 4–628 ( chain B ) , 5–627 ( chain C ) , 3–628 ( chain D ) . Models were validated using simulated annealing composite omit maps calculated in Phenix . Model geometry was analyzed using MolProbity ( Chen et al . , 2010 ) . Analysis of Ramachandran statistics indicated that each structure contained the following percentages of residues in the favored , allowed , and disallowed regions , respectively: as-isolated ( batch 1 ) : 96 . 3% , 3 . 4% , 0 . 3%; as-isolated ( batch 2 ) : 96 . 8% , 2 . 9% , 0 . 3%; reduced ( batch 2 ) : 96 . 6% , 3 . 1% , 0 . 3%; reduced/O2-exposed ( batch 2 ) : 96 . 5% , 3 . 2% , 0 . 3%; DvCODH ( C301S ) : 97 . 0% , 2 . 7% , 0 . 3% . Figures were generated in PyMOL ( Schrodinger , 2015 ) . Crystallography packages were compiled by SBGrid ( Morin et al . , 2013 ) . EPR samples were prepared using 57Fe-enriched DvCODH containing 13 . 3 57Fe/monomer and 0 . 5 Ni/monomer , as quantified by ICP-OES . All samples were prepared under oxygen-free conditions in a Coy anaerobic chamber . Samples were incubated with an excess of sodium dithionite ( 30–40 equivalents ) for 20–30 min at 22 °C prior to freezing in liquid N2 under oxygen-free conditions . For the sample of air-exposed and re-reduced DvCODH , an aliquot was removed from the anaerobic chamber and incubated on ice under ambient atmospheric conditions for 50 min to afford full oxidation of the clusters . The sample was then returned to the anaerobic chamber and incubated with 30–40 equivalents of sodium dithionite for 20–30 min . Samples ( 250 µL ) were loaded in Quartz EPR tubes ( QSI Inc , Fairport Harbor , OH ) and frozen in liquid N2 under oxygen-free conditions . EPR spectra were acquired at the Department of Chemistry Instrumentation Facility at MIT on a Bruker EMX Plus continuous wave ( CW ) X-Band spectrometer ( operating at ~9 . 34 GHz ) equipped with a rectangular resonator ( TE101 ) and a cryogen-free system consisting of a Sumitomo RDK-408D2 cold head equipped with a ColdEdge Technologies waveguide cryostat . Spectra were acquired using Bruker Xenon software and were recorded at 10 K at a microwave power of 0 . 2 mW , using a modulation amplitude of 1 mT , a microwave frequency of 9 . 34 GHz , a conversion time of 82 . 07 ms , and a time constant of 81 . 92 ms . Spin quantification was carried out against a Cu2+-EDTA standard containing 200 µM CuSO4 in 10 mM EDTA , under non-saturating conditions . Quantitation of the S = 1/2 [Fe-S] centers amounted to 3 . 2 spins/dimer , similar to our previous report on DvCODH ( Hadj-Saïd et al . , 2015 ) . Quantification of the Cred1 and Cred2 states was carried out on the basis of numerical double-integration of the simulated spectra using the MATLAB-based EasySpin software ( Stoll and Schweiger , 2006 ) ; Cred1 was 0 . 43 spins/dimer and Cred2 was 0 . 35 spins/dimer . | Life relies on countless chemical reactions , almost all of which need to be sped up by enzymes . About half of all enzymes carry metal ions that expand the range of the reactions that they can catalyze . In some enzymes these metal ions assemble with sulfur ions to form so-called metalloclusters . These structures can carry out many different types of reactions , including converting simple forms of elements like nitrogen and carbon into other forms that can be used to make more complicated biological molecules . One enzyme that contains metalloclusters is carbon monoxide dehydrogenase . Known as CODH for short , this enzyme uses a metallocluster called the “C-cluster” to interconvert two gases: the pollutant carbon monoxide and the greenhouse gas carbon dioxide . CODH enzymes are found inside certain bacteria , but they are also of interest for humans , who wish to use them to remove the harmful gases from the environment . But this is not as simple as it may at first seem: CODH enzymes usually become inactive when exposed to air because the metalloclusters fall apart in the presence of oxygen . One CODH enzyme from a widespread bacterium called Desulfovibrio vulgaris , however , is an attractive target for industrial use because it can tolerate oxygen better . Yet , it is still unclear why this enzyme does not get inactivated the way other CODHs do . Wittenborn et al . have now characterized the CODH enzyme from D . vulgaris in more depth via a technique called X-ray crystallography , which can reveal the location of individual atoms within a molecule . By a happy accident , the structures revealed that the C-cluster can adopt a dramatically different arrangement of metal and sulfur ions after being exposed to oxygen . This rearrangement is fully reversible; when oxygen is removed , the metal and sulfur ions move back to their normal positions . This ability to flip between different arrangements appears to protect the metallocluster from losing its metal ions when exposed to oxygen . By providing structural snapshots of how CODH responds to oxygen these results provide a more complete understanding of an enzyme that plays a key role in the global carbon cycle . This understanding could help scientists to develop bioremediation tools to remove carbon monoxide and carbon dioxide from the atmosphere and to engineer bacteria to capture carbon to make biofuels . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2018 | Redox-dependent rearrangements of the NiFeS cluster of carbon monoxide dehydrogenase |
Ankyrins together with their spectrin partners are the master organizers of micron-scale membrane domains in diverse tissues . The 24 ankyrin ( ANK ) repeats of ankyrins bind to numerous membrane proteins , linking them to spectrin-based cytoskeletons at specific membrane microdomains . The accessibility of the target binding groove of ANK repeats must be regulated to achieve spatially defined functions of ankyrins/target complexes in different tissues , though little is known in this regard . Here we systemically investigated the autoinhibition mechanism of ankyrin-B/G by combined biochemical , biophysical and structural biology approaches . We discovered that the entire ANK repeats are inhibited by combinatorial and quasi-independent bindings of multiple disordered segments located in the ankyrin-B/G linkers and tails , suggesting a mechanistic basis for differential regulations of membrane target bindings by ankyrins . In addition to elucidating the autoinhibition mechanisms of ankyrins , our study may also shed light on regulations on target bindings by other long repeat-containing proteins .
Ankyrins are a widely expressed scaffold protein family , which mainly function to link great varieties of functionally related but structurally diverse integral membrane proteins to the spectrin-based cytoskeletons ( Bennett and Baines , 2001; Bennett and Healy , 2009; Bennett and Lorenzo , 2013 ) . Mutations of ankyrins are known to be associated with various diseases in humans such as hereditary spherocytosis ( Eber et al . , 1996; Lux et al . , 1990; Gallagher , 2005 ) , cardiac arrhythmia syndromes and sinus node dysfunction ( Mohler et al . , 2003; Mohler et al . , 2007; Le Scouarnec et al . , 2008; Hashemi et al . , 2009 ) , bipolar disorders ( Baum et al . , 2008; Ferreira et al . , 2008; Schulze et al . , 2009; Scott et al . , 2009; Takata et al . , 2011; Dedman et al . , 2012; Rueckert et al . , 2013 ) , schizophrenia ( Nie et al . , 2015; Ripke et al . , 2011 ) and autism spectrum disorder ( Willsey et al . , 2013; Parikshak et al . , 2013; Shi et al . , 2013; Iqbal et al . , 2013 ) . In vertebrates , the ankyrin family consists of three members: ankyrin-R ( AnkR ) , ankyrin-B ( AnkB ) and ankyrin-G ( AnkG ) . They share similar domain organizations ( Figure 1A ) , but usually locate at different subcellular regions and perform divergent physiological functions ( Mohler et al . , 2002; Abdi et al . , 2006; He et al . , 2013; Bennett and Lorenzo , 2016 ) . Each ankyrin contains an N-terminal membrane binding domain ( MBD ) , which is composed of 24 ANK repeats and responsible for binding to diverse membrane targets . The 24 ANK repeats of ankyrins are among the longest extended repeat-containing domains in living organisms . A trademark of long repeat-containing proteins is their formations of long solenoid structures with high conformational plasticity ( Lee et al . , 2006b; Wang et al . , 2014 ) . Following the MBD is an unstructured linker region of ~130 amino acid residues , followed sequentially by a ZU5-ZU5-UPA supramodule responsible for spectrin binding ( SBD ) ( Mohler et al . , 2004a; Wang et al . , 2012 ) , a death domain ( DD ) , and an intrinsically disordered C-terminal tail ( CT ) with varying lengths ( Figure 1A ) . The folded structural domains ( i . e . MBD , SBD and DD ) in ankyrins share very high sequence similarities ( Figure 1A ) . In contrast , the unstructured regions of ankyrins ( i . e . the linker region and CT ) , although fairly conserved within each isoform , are more divergent among the family members ( Figure 1A ) . Presumably , these isoform-specific and unstructured linker and tail regions are largely responsible for distinct functions of each member of the ankyrin family proteins ( see Abdi et al . , 2006 and He et al . , 2013 for examples ) . However , very little is known regarding how the unique linker and tail regions specifically modulate functions of each ankyrin . Ankyrin is an ancient molecule that appeared as early as in bilaterians over 500 million years ago ( Cai and Zhang , 2006; Bennett and Lorenzo , 2013 ) . The 24 ANK repeats of ankyrins have retained a remarkable level of amino acid sequence conservation throughout the evolution , and different ANK repeats binding membrane proteins with diverse sequences have evolved at different stages of the evolution ( Hill et al . , 2008; Bennett and Lorenzo , 2013; Bennett and Lorenzo , 2016 ) . In an earlier study , we demonstrated that the highly conserved 24 ANK repeats can form multiple semi-independent target binding sites , and each capable of recognizing short and intrinsically disordered segments with variable amino acid sequences distributed on target proteins ( Wang et al . , 2014 ) . As such , ANK repeats are capable of binding to a very diverse array of target proteins with high specificity , a strategy that may be commonly employed by long repeat-containing scaffold proteins ( Wang et al . , 2014; Chook and Blobel , 2001; Conti et al . , 1998; Graham et al . , 2000; Huber and Weis , 2001; Kobe , 1999; Xu et al . , 2010; Zhu et al . , 2011 ) . The target bindings of ankyrin’s ANK repeats must be regulated during the biogenesis and trafficking processes of ankyrins and their targets , as otherwise functionally unrelated membrane targets would be misdirected to the same sites in cellular membranes . Additionally , the formation of each ankyrin/target complex ( e . g . the ankyrin-B/G and sodium channel complexes ) at specific membrane domains also needs to be regulated to fulfill excitation/resting cycles of excitable tissues such as neurons and muscles ( Bréchet et al . , 2008; Garver et al . , 1997; Whittard et al . , 2006 ) . Earlier studies have suggested that ankyrins can adopt autoinhibited conformation ( i . e . by using sequences outside the ANK repeats to directly interact with the repeats ) to regulate their target bindings ( Abdi et al . , 2006; Davis et al . , 1992; He et al . , 2013; Wang et al . , 2014 ) , but the mechanisms of the autoinhibition are unknown . In this study , we systematically investigated the autoinhibition mechanisms of ANK repeats of ankyrin-B and G ( AnkB and G ) . We identified three distinct autoinhibitory segments from the intrinsically disordered linker and CT regions of AnkB , and elucidated the inhibition mechanism of these segments by detailed biochemical and structural investigations . Our study reveals that the three inhibitory segments spread along the entire 24 ANK repeats and bind to the repeats in a quasi-independent manner , so that the inhibitory sites may be combinatorially regulated in response to different binding targets . We further demonstrated that AnkG also adopts a similar overall autoinhibition strategy as AnkB does , but with its own unique binding features . Collectively , the findings of our study not only provide a framework for understanding the interactions and regulations of ankyrins with their membrane targets , but may also be helpful in understanding the regulation of target recognitions by other long repeat-containing scaffold proteins in general .
Ankyrin MBD can be inhibited by its CT regulatory domain in both AnkR and AnkB ( Abdi et al . , 2006; Davis et al . , 1992 ) , and previously we have biochemically and structurally characterized the intramolecular interaction between the 48-residue AnkR CT regulatory domain with all three MBDs from the ankyrin family members ( Wang et al . , 2014 ) . Moreover , our analysis illustrated that the MBDs of AnkR/B/G are highly conserved and share essentially the same binding properties to diverse binding partners ( Wang et al . , 2014 ) . In the structure of AnkB_MBD in complex with AnkR_CT , AnkR_CT adopts an extended conformation lining the inner groove of ANK repeats 1 ~ 14 of AnkB_MBD ( Wang et al . , 2014 ) . Although AnkB and G do not contain AnkR_CT like sequences in their tails , more thorough sequence analysis revealed that both AnkB and G contain a 13-residue segment ( denoted as ‘AI-a’ for the AutoInhibition segment-a , Figure 1A ) that share a similar sequence pattern with the last 13-residue fragment of AnkR_CT and several MBD ‘site-1’ binding targets such as Na+ and K+-channels ( Figure 1B ) . This 13-resdiue AI-a fragment is almost completely conserved in AnkB and AnkG among vertebrates ( alignments not shown ) . Importantly , a Glu residue corresponding to Glu1622 in AnkR_CT and Glu1112 in Nav1 . 2 , which has been shown to be absolutely required for AnkR_CT and Nav1 . 2 to bind to ANK repeats ( Mohler et al . , 2004b; Wang et al . , 2014 ) , is also invariant in AnkB and G ( Figure 1B ) . We found that a peptide encompassing this AnkB_AI-a segment binds to AnkB ANK repeats with a dissociation constant ( Kd ) ~7 . 3 μM based on a fluorescence polarization binding assay ( Figure 1C ) . Point mutations of the positively charged residues in the repeat 1 ( R1 ) of ANK repeats ( 37RAAR40 to 37EAAE40 ) of AnkB significantly decreased its binding to the AI-a peptide ( Figure 1C ) , consistent with our previous structural analysis that the positively charged residues in the ANK repeat R1 play a role in binding to the negatively charged residues from the ‘site-1’-binding peptides . We attempted to elucidate the molecular basis governing the AnkB ANK repeats/AI-a interaction by solving the complex structure using a similar fusion strategy as we used earlier for the Nav1 . 2 peptide ( Wang et al . , 2014 ) ; i . e . by fusing ANK repeats R1-20 to the C-terminal tail of AnkB AI-a . However , this effort was not successful . As an alternative approach , we replaced the corresponding sequence of AnkR_CT with that of AnkB_AI-a and produced an AnkR/B_CT Chimera ( Figure 1D ) . Isothermal titration calorimetry ( ITC ) experiments detected a strong interaction between this chimera peptide and AnkB_MBD , with a Kd value comparable to that of AnkR_CT with AnkB_MBD ( Figure 1E ) . This chimera peptide was fused to the N-terminus of AnkB_MBD for crystallization trials as we have demonstrated for AnkR_CT . With this strategy , we successfully obtained crystals of the AnkR/B_CT Chimera/AnkB_repeats_R1-20 fusion protein , and the crystals were diffracted up to 3 . 3 Å . The structure of the fusion protein was determined by the molecular replacement method using the AnkB_repeats_R1-24 structure as the model ( Wang et al . , 2014 ) ( Table 1 ) . Consistent with the biochemical analysis , the structure showed that the AnkB_AI-a segment binds to ‘site-1’ of AnkB_MBD ( repeats_R1-5 ) using essentially the same binding mode as we observed in the AnkR_CT/AnkB_MBD and Nav1 . 2_ABD/AnkB_R1-9 complex structures ( Wang et al . , 2014 ) ( Figure 1F ) . In particular , Glu1630 from AnkB_AI-a occupies the identical position as Glu1622 in AnkR or Glu1112 in Nav1 . 2 on AnkB_R1-5 , by forming strong hydrogen bonds with Thr94 and Asn98 in the R2-R3 finger loop ( Figure 1F ) . Additionally , hydrophobic interactions between ‘PPIV’ from AI-a and hydrophobic residues ( including the two critical Phe residues , Phe131 and Phe164 , from R4 and R5; see Wang et al . , 2014 ) from R3-R5 also contribute to the binding ( Figure 1F ) . Though the conservation among the three isoforms is limited , the amino acid sequences of the linker connecting MBD and SBD within each isoform of ankyrins are highly conserved ( Figure 1A ) . The crystal structure of AnkR MBD C-terminal 12 ANK repeats ( PDB ID: 1N11 , referred to as AnkR_C12 ) contains a 14-residue fragment of the linker that folds back and binds to the last five repeats ( R20-24 ) of MBD ( Michaely et al . , 2002 ) ( also see below for more details ) . Recently , the linker region of AnkB was reported to directly interact with MBD , thus preventing AnkB from localizing to plasma membranes ( He et al . , 2013 ) . These studies suggest that , besides the AI-a in their CT , the linker regions of ankyrins also play important autoinhibitory roles in regulating functions of MBDs . We chose AnkB to investigate the detailed mechanism governing the linker region-mediated autoinhibition of ankyrins . To probe the linker-mediated autoinhibition , we used three AnkB_MBD binding targets ( Nav1 . 2 , E-cadherin and NF186/L1CAM ) , each with distinct MBD binding mode ( see [Wang et al . , 2014]; and our unpublished data on E-cadherin ) , to test their bindings to three versions of extended MBD: one containing a short 20-residue linker ( aa 28–847 , roughly corresponding to the above-mentioned AnkR_C12 structure but with additional six residues . See Table 2 for a list of key constructs used in the study ) , one containing the entire linker ( aa 28–965 ) , and the third one with the entire linker as well as the entire CT regulatory domain ( i . e . the full-length AnkB with the spectrin binding ZZUD tandem deleted , denoted as AnkB ΔZZUD ) ( Figure 2A ) . Interestingly , extending the linker region from residue 847 to 965 ( AnkB 28–847 vs AnkB 28–965 ) invariably led to weakened or completely loss of interactions between MBD and the three target proteins ( Figure 2B and C ) , indicating that an additional segment within the residues 848–965 can bind to the ANK repeats of AnkB and acts as another autoinhibitory sequence . Including the AI-a fragment in the MBD extension ( AnkB ΔZZUD ) further eliminated the remaining binding of Nav1 . 2 ( Figure 2B and C ) , and this is consistent with our findings that AI-a and Nav1 . 2 bind to R1-5 of MBD in a mutually competitive manner ( Figure 1F; and Wang et al . , 2014 ) . We confirmed the direct interaction between the entire linker ( aa 828–965 ) with the AnkB ANK repeats only ( aa 28–827 ) using an ITC-based binding assay , and data showed a strong binding ( Kd about 0 . 044 μM , Figure 2D ) . The entire linker ( aa 828–965 ) was found to bind to AnkB 28–847 with a Kd ~4 . 2 μM ( Figure 2E ) , further indicating that a fragment within residues 848–965 can directly bind to AnkB_MBD . Another version of extended MBD , AnkB 28–873 interacts with the entire linker ( aa 828–965 ) with a similar affinity as AnkB 28–847 does ( Figure 2F ) , suggesting that residue 848–873 has marginal impact on the AnkB autoinhibition . Taken together , the above biochemical mapping experiments suggested that the AnkB linker region contains two discrete autoinhibitory segments: one segment within the region 848–965 binding to middle repeats of AnkB_MBD ( denoted as AI-b ) , and the other within residues 828–847 overlapping with the short 14-residue segment observed in the AnkR_C12 structure ( denoted as AI-c ) ( Figure 2A; and see below for further detailed mapping ) . To delineate the mechanism governing the AI-b/MBD interaction , we set out to solve the complex structure . First , we mapped the minimal regions of AI-b and the corresponding repeats from MBD responsible for the interaction . Using the scheme described in Figure 2E as the assay , we mapped AI-b to a 32-aa fragment ( aa 865–896 ) from the linker ( Figure 3A ) . Using a similar approach combined with truncations of various repeats , we mapped the AI-b binding regions to R6-14 of MBD ( Figure 3B ) . We determined the crystal structure of AnkB repeats 8–14 fused to the C-terminus of AnkB_AI-b ( aa 857–896 ) at 2 . 35 Å resolution ( Table 1 ) . The AI-b peptide binds to the inner groove formed by repeats 10–13 ( Figure 3C ) , covering a large portion of the target binding ‘site-3’ ( repeats 11–14 ) and a small portion of ‘site-2’ ( repeats 7–10 ) that were defined in our previous study ( Wang et al . , 2014 ) . The structure is also consistent with our biochemical data that the autoinhibitory interaction mediated by AI-b partially inhibits the binding of Nav1 . 2 and NF186 to MBD ( Figure 2B ) , as both targets use the ‘site-3’ of MBD as one of the binding sites ( Wang et al . , 2014 ) . Consistent with this observation , substitutions of hydrophobic residues within the ‘site-3’ with Gln ( Leu366Gln , Phe399Gln , and Leu432Gln ) completely disrupted AI-b’s binding to MBD , and substitutions of hydrophobic residues within the ‘site-2’ with Gln ( Ile267Gln , Leu300Gln ) mildly weakened the AI-b/MBD interaction ( Figure 3E; see Wang et al . , 2014 for the rational of the mutations ) . The electron densities of the MBD_R8-14 bound AI-b only allowed us to trace residues spanning Gly883-Glu892 ( Figure 3C ) . Although residues Glu865-Asp882 could not be traced in the crystal structure , this segment also directly participates in binding to MBD_R8-14 , as deletion of residues from 865 to 873 significantly weakened AI-b’s binding to AnkB_MBD ( Figure 3A ) . The structure of the MBD_R8-14/AI-b complex reveals their binding details: Tyr886 forms a hydrogen bond with His374 and hydrophobic interactions with Phe399 and Ile404; Met884 inserts into a hydrophobic pocket formed by Lys407 , Val437 , Phe440 and Met441; and Arg888 forms a number of hydrogen bonds with the sidechain of Tyr365 and backbone of Leu363 as well as charge-charge interaction with Asp364 ( Figure 3D ) . ITC results showed that single point substitution of Tyr886 from AI-b with Ala significantly decreased its binding to MBD , and Met884Ala , Tyr886Ala double mutations totally abolished the AI-b’s autoinhibition ( Figure 3E ) . It is noted that Met884 in AI-b occupies the same hydrophobic pocket as Ile1588 in AnkR_CT does , although the two ‘site-3’ binding fragments share very little amino acid sequence similarity ( Figure 3F ) , further highlighting the remarkable capacity of ANK repeats in binding to targets with diverse amino acid sequences ( Wang et al . , 2014 ) . The AnkR_C12 structure contains a 14-residue linker region ( aa 832–845 in mouse AnkR ) , and 11 residues ( aa 835–845 ) are defined in the folded back structure ( Michaely et al . , 2002 ) . A longer region of the linker ( i . e . 828–849 in human AnkB ) is highly conserved in all three isoforms of ankyrins ( Figure 6A ) . We speculated that the autoinhibitory AI-c segment of ankyrins may be longer than the 11-residue fragment seen in the structure of AnkR_C12 . To ensure that we do not miss any possible residues forming the AI-c binding segment , we prepared a C-terminal 12 ANK repeats of AnkB followed by a 46-residue linker ( aa 430–873 ) , which covers a small part of AI-b ( Figure 3A ) , and solved the crystal structure of the protein at 1 . 95 Å resolution ( Table 1 ) . The overall structure of our AnkB C-terminal 12 repeats-linker is very similar to that of AnkR_C12 ( RMSD of 1 . 04 Å , Figure 4A ) . However , our structure contains a longer linker folded back binding to the AnkB ANK repeats . A 20-residue linker ( from Thr828 to Asp847 instead of 11 residues in AnkR_C12 ) is clearly defined , and this 20-residue linker extends to the repeat R18 ( Figure 4A ) . Moreover , we observed a prominently positive-charged surface in the inner groove of R16-18 right next to Asp847 , and the five residues following Asp847 are highly negatively charged and conserved ( 847DEEGDD852 ) ( Figure 4B ) . We anticipate that this stretch of negatively charged residues is part of AI-c and involved in binding to MBD ( likely R16-17 based on the length of the segment ) ( Figure 4B ) . This idea is supported by the data from an earlier study showing that substitution of ‘848EEGDDT853’ with ‘NAAIRS’ in AnkB led to increased membrane localization presumably due to the mutation-induced weakened autoinhibition ( He et al . , 2013 ) . In the AI-c folded-back structure , residues from AI-c form extensive hydrogen bonds with the residues in finger loops of ANK repeats ( Figure 4C ) . Perturbations of these hydrogen bonding interactions ( e . g . the Asn834Lys substitution in AI-c or the Asn595Ala mutation in one of the ANK repeats finger loop ) led to a few fold decreases in their bindings ( Figure 4D ) . Hydrophobic interactions also contribute to the AI-c autoinhibition . Met839 is buried in the hydrophobic pocket formed by Leu668 and Leu701; Leu843 interacts with sidechains of Leu597 , Tyr630 , and Ile635 ( Figure 4C ) . Individually mutating these two hydrophobic residues to polar Gln weakened the binding by several folds ( Figure 4D ) . An ideal result would be to obtain structures of MBD of ankyrins binding to two or even all three of these autoinhibitory segments . We have extensively tried such experiments , by fusing the inhibitory fragments combining with the different strategies used in the study . Unfortunately , these extensive efforts have not resulted fruition likely due to the conformational dynamics of the elongated protein complexes . In epithelial cells , AnkG clusters on plasma membranes whereas AnkB largely localizes to intracellular compartments ( Figure 5A and B and He et al . , 2013 ) . Autoinhibition of the MBD by the linker region can modulate AnkB’s membrane vs cytosol distributions in epithelial cells ( He et al . , 2013 ) . We made use of this assay as a functional readout to verify the autoinhibited structures determined in this study and to provide a preliminary glance at the role of AnkB’s autoinhibition . Consistent with the previous report ( He et al . , 2013 ) , WT AnkB mainly localizes in the cytosol , whereas WT AnkG mainly associates with the plasma membranes in polarized MDCK cells ( Figure 5A and B ) . We constructed two mutants in the linker region of AnkB , one weakens the AI-b’s binding to MBD ( the Met884Ala , Tyr886Ala double point mutations , denoted as AnkB 2M ) and the other weakens both AI-b and AI-c’s bindings to MBD ( the Met884Ala , Tyr886Ala , Asn834Lys triple mutations , denoted as AnkB 3M ) , and assayed their membrane vs cytosol distributions in polarized MDCK cells . We observed that the AnkB 2M and 3M mutants show a higher ratio of plasma membrane localization ( Figure 5A and B ) , consistent with releases of the autoinhibition induced by the two mutations . Finally , we deleted essentially the entire linker region encompassing the complete AI-b and AI-c segments ( AnkB ΔLinker , with aa 828–943 deleted ) , and found that this deletion mutant is near completely membrane localized ( Figure 5A and B ) , suggesting that both AI-b and AI-c can regulate AnkB’s membrane localization presumably by modulating its MBD’s binding to membrane-anchored target ( s ) . We have also performed NF186-mediated plasma membrane recruitments of AnkB , and compared the impacts of the 2M and 3M mutations on AnkB’s membrane localizations in HeLa cells which lack endogenous NF186 expression . Co-expression of NF186 partially recruited WT AnkB to the plasma membranes ( Figure 5C ) . The 2M , 3M , and ΔLinker of AnkB mutants displayed sequentially increasing amount of NF186-mediated membrane recruitments in this assay system ( Figure 5C and D ) . Similar as observed in MDCK cells , WT AnkG is better recruited to plasma membranes by NF186 than WT AnkB is ( Figure 5C and D ) . It is not known whether AnkG also contains auto-regulatory segments . Amino acid sequence alignment analysis reveals that residues corresponding to the AnkB AI-c are essentially the same in AnkG , indicating that AnkG also contains an autoinhibitory AI-c segment ( Figure 6A ) . Curiously , the N-terminal half of AnkB AI-b can be nicely aligned with a fragment in the AnkG linker following AI-c , but residues in the C-terminal half of AnkB AI-b that are critical for its binding to MBD ( e . g . Met884 and Tyr886 ) are missing in AnkG ( Figure 6A ) . We hypothesized that AnkG may contain a different autoinhibitory AI-b segment from that of AnkB . Using a similar binding assay developed to discover AI-b in AnkB , we found that elongating the linker from residue 855 to 903 in the extended AnkG MBD ( compare aa 38–855 vs aa 38–903 ) weakened the bindings of Nav1 . 2 , E-cadherin and NF186 to AnkG by 6 ~ 74 folds ( Figure 6B ) , indicating that AnkG linker indeed contains an AI-b segment . We also detected direct interaction between AnkG MBD-AI-c protein ( aa 38–855 , containing the AI-c segment ) and an AnkG’s long linker ( aa 837–920 ) or a shorter version of the linker ( aa 861–920 , lacking AI-c ) ( Figure 6C and D ) . Binding affinity-guided mapping revealed that residues 875 to 903 encompass the complete AI-b of AnkG ( Figure 6E ) . It is noted that AnkG AI-b contains a stretch of highly charged residues in its N-terminal half , and these residues are not found in AnkB , revealing a different binding mechanism of AI-b to MBD in AnkB and G . We found that increasing salt concentrations in the binding buffer significantly weakened the interaction between AnkG_AI-b and AnkG_MBD , supporting the charge-charge interactions between AI-b and AnkG_MBD ( data not shown ) . We substituted three charged residues with neutral or charge reverse residues ( Asp887Ala , Asp889Ala and Lys890Glu , indicated with black triangles in Figure 6A ) from the AnkG linker ( 861-920 ) and found that the mutant completely lost its ability to bind AnkG_MBD ( Figure 6F ) , further supporting the critical roles of the charged residues in AnkG AI-b in its autoinhibition . We attempted to map the AI-b’s binding site on AnkG’s ANK repeats by a similar repeats truncation approach used for AnkB ( Figure 3B ) , but this effort failed due to poor sample behaviors of the AnkG_MBD truncation mutants . Alternatively , we generated the ‘site-1 , -2 , and -3’ target binding mutants of AnkG_MBD based on our earlier work ( Wang et al . , 2014 ) , and tested each of these mutants in binding to the AnkG_AI-b fragment . We found that mutations in ”site-1’ ( Phe141Gln , Phe164Gln ) or in ‘site-2’ ( Ile277Gln , Leu310Gln ) of AnkG_MBD has little effect on AI-b’s binding , whereas mutations in ‘site-3’ ( Leu376Gln , Phe409Gln , Leu442Gln ) completely abolished the interaction between AI-b and ANK repeats ( Figure 6G ) . This result indicates AnkG_AI-b also chiefly binds to the ‘site-3’ ( repeats R11-14 ) of AnkG_MBD . The above result , together with the findings that AnkG also contains the inhibitory AI-a segment ( Figure 1B ) and AI-c segment ( Figure 6A ) , collectively demonstrated that AnkG shares a very similar , three segment-mediated MBD autoinhibitory mechanism as AnkB does , although the AI-b segments in the two ankyrins are different .
Ankyrins are master scaffold proteins assembling very diverse signaling microdomains beneath membrane bilayers . This is achieved by their MBD-mediated bindings to numerous trans-membrane proteins and SBD-mediated anchoring of the protein complex to spectrin-based cytoskeletal meshwork . Formation of a highly elongated and malleable target binding groove by the 24 ANK repeats with multiple semi-independent target binding sites provides a mechanistic explanation to how ankyrin MBD , via combinatorial usages of its target binding sites , can bind to many distinct membrane targets with high specificity ( Figure 7A; and Wang et al . , 2014 ) . A key unanswered question is how MBD-mediated recognitions of membrane targets of ankyrins are regulated . Autoinhibition by C-terminal segment ( s ) to the ANK repeats has been suggested to be one of such regulation mechanisms ( Abdi et al . , 2006 ) . In this study , we characterized the autoinhibition of AnkB and G in detail and discovered that the entire target binding grooves of both AnkB and G MBDs could be inhibited by multiple discrete and intrinsically disordered peptide fragments located in the MBD/SBD linker region and in the tail region right after DD ( Figure 7 ) . Except for the AI-c segment located immediately C-terminal to the ANK repeats , the inhibitory sequences of AI-a and AI-b segments between AnkB and AnkG share limited homology . This is perhaps correlated with the differential potential target binding regulations of the two MBDs . However , the amino acid sequences of all three identified autoinhibitory segments are highly conserved throughout the evolution both for AnkB and for AnkG , implying that the autoinhibition mechanisms discovered here are evolutionary conserved features for the two ankyrins , respectively . Given that the inhibitory segments of both AnkB and G are intrinsically disordered and disordered sequences tend to evolve more rapidly in eukaryotic genomes ( Dyson and Wright , 2005 ) , such high amino acid sequence conservation suggests a strong function-mediated selection against mutational drifts of these inhibitory segments . Expressions of both AnkB and G can be extensively diverse due to their alternative splicing , and therefore the expressed proteins can vary in their sizes dramatically ( Cunha and Mohler , 2009 ) . However , regardless of the extensive alternative splicing , the autoinhibitory segments AI-b and AI-c are always retained in all documented AnkB or AnkG variants according to the Ensembl database ( http://www . ensembl . org ) , suggesting that the autoinhibitory mechanisms characterized here for AI-b and c are common to all isoforms of AnkB and G . Interestingly , the giant AnkG isoforms ( 270 kDa/480 kDa AnkG ) lose the AI-a segment due to alternative splicing; whereas giant AnkB ( 440 kDa AnkB ) keeps this AI-a segment , suggesting the differential regulation of these giant ankyrin isoforms in nervous systems . Our data is also consistent with the findings from super-resolution images ( Leterrier et al . , 2015 ) . In AIS , 270 kDa/480 kDa AnkG’s SBD exhibits a periodic pattern , while their C-terminus part is not periodically arranged and is found ~32 nm radially deeper than the SBD in the axoplasm . AIS AnkG ( 270 kDa/480 kDa isoforms ) lacks AI-a , so the major parts of CT of 270 kDa/480 kDa AnkG are not sequestered by MBD even if AI-b and c are in contact with MBD . Furthermore , the 270 kDa/480 kDa AnkG at AIS/Nodes of Ranvier is likely to adopt an open conformation . As such , the signals of antibodies labeling the C-terminus part of AnkG should not appear near the membrane but is deeper into the axoplasm . Since each target protein normally occupies a few of the total five proposed binding sites along the entire 24 ANK repeats ( Figure 7A ) , the use of multiple semi-independent autoinhibitory segments lining the entire target binding groove of MBD can provide a mechanism for selected release of a few target binding sites ( e . g . sites-1 and 3 for sodium channels ) while keeping other sites closed ( e . g . site-5 ) , thus allowing selected engagements of MBD to certain target proteins at any given membrane microdomains . The combinatorial usage of multiple binding sites by target proteins appears to be a rather common strategy for many long repeat-containing proteins ( Chook and Blobel , 2001; Conti et al . , 1998; Graham et al . , 2000; Huber and Weis , 2001; Kobe , 1999; Xu et al . , 2010; Zhu et al . , 2011 ) . Autoinhibition of an elongated repeats domain by multiple intrinsically disordered segments from the same protein are rather uncommon , and only a few cases have been reported . Kap60p , a karyopherin family protein , adopts a somewhat similar autoinhibition mode as ankyrins do . In this case , two NLS recognition sites on the armadillo repeats were shown to be autoinhibited by two consecutive segments from Kap60p’s N-terminal unstructured region ( Matsuura and Stewart , 2004 ) . In another case , two NLS binding sites of importin-α are inhibited by the same fragment from its N-terminal unstructured region at a 1:2 stoichiometry instead of multiple segments with different sequences as found in ankyrins and Kap60p ( Catimel et al . , 2001 ) . Other autoinhibition mechanisms found in long repeat-containing proteins include the HEAT repeats of karyopherin-β2 occupied by an internal long loop from the middle region of the HEAT repeats ( Chook and Blobel , 1999; Lee et al . , 2006a ) , the HEAT repeats of exportin chromosome region maintenance 1 ( CRM1 ) autoinhibited by a C-terminal α helix immediately following the HEAT repeats ( Saito and Matsuura , 2013 ) , and the armadillo repeats of Diaphanous-related formins ( Drfs ) autoinhibited by a short α helix from its distal C-terminal region ( Lammers et al . , 2005 ) . It will be interesting to investigate whether the autoinhibition by multiple discrete segments as observed in ankyrins here might also be used by other long repeat-containing proteins capable of binding to diverse targets in the future . Ankyrin MBDs are versatile membrane target binders and many of these ankyrin binding membrane proteins are located in specific membrane microdomains ( e . g . ion-channels in the axon initial segments of neurons as well as specifically patterned membrane domains in cardiomyocytes ) playing vital physiological roles . Defects of interactions between ankyrins and their membrane targets are frequently linked to human diseases including hereditary spherocytosis , cardiac arrhythmia , and several types of psychiatric disorders ( Eber et al . , 1996; Mohler et al . , 2004b; Van Camp et al . , 1996 ) . In view of the vital physiological roles of the interactions between ankyrins and these membrane targets , a priori assumption is that such ankyrin-mediated membrane target bindings must be regulated . For example , the membrane target binding groove of MBDs is likely to be closed , via the autoinhibition mechanism elucidated in this study , during biogenesis and trafficking processes of ankyrins . Upon reaching each specific membrane domain , certain mechanisms are likely existing to release a few selected target binding sites for specific assembly of ankyrin/membrane target complexes . The existence of multiple semi-independent inhibitory sequences in the tail of ankyrins suggests that these sequence may be regulated in a combinatorial fashion so that ankyrins can scaffold many different target proteins in different cellular settings and in different tissues . The positioning of the spectrin binding ZZUD domains with respect to the auto-inhibitory sites-a and b suggests a possible spectrin/SBD interaction-induced release of ankyrin autoinhibition . The spectrin binding ZZUD tandem are flanked by the AI-a and AI-b segments both in 220 kDa AnkB and 190 kDa AnkG , and the connection sequences between the ZZUD tandem and the two inhibitory segments are not very long ( Figures 1A and 7B ) . Once ankyrins and spectrins meet at specific membrane microdomains , both the bulky ankyrin SBD/spectrin complex as well as the perceived ‘pulling force’ exerted by the spectrin-bound SBD may dislodge the two inhibitory segments , AI-a and AI-b , flanking the ZZUD tandem and thus lead to the release of the autoinhibition . Accompanied by the spectrin binding-induced release of the AI-a and AI-b segments , the target binding ‘sites-1 , 2 , 3’ are opened and the corresponding membrane proteins can bind to the ankyrin MBDs . It should be noted that such coordinated membrane microdomain targeting and spectrin/actin network binding-induced release of ankyrin MBD autoinhibition model , if it is indeed employed in cells , is likely used together with other regulatory mechanisms in regulating the conformational opening of ankyrins . Ankyrins at AIS/nodes of Ranvier are chiefly 270 kDa/480 kDa AnkG isoforms . Interestingly , these two larger AnkG isoforms do not contain the ‘AI-a’ sequence in their CTs due to alternative splicing . Thus , the 270 kDa/480 kDa AnkG seem to have different autoinhibition mechanisms with respect to other ankyrins ( e . g . 190 kDa AnkG and 220 kDa AnkB ) . The lack of the ‘AI-a’ site in 270 kDa/480 kDa AnkG means that these two AnkG isoforms can still bind to Nav channels or NF186 ( See Figure 2B , AnkB 28–965; and also Hedstrom et al . , 2008 ) , albeit with much lower affinities ( i . e . a form of semi-open conformation ) . Spectrin binding to the AnkG SBD may promote further opening of ANK repeats and thus leading to full engagements of the membrane targets of AnkG . Such mutual reinforcement of AnkG-mediated membrane microdomain assembly model is consistent with the findings showing that membrane targets and βIV spectrin mutually stabilize each other through AnkG at AIS or nodes of Ranvier ( Yang et al . , 2004; Patzke et al . , 2016 ) . Additionally , post-translational modifications ( e . g . phosphorylation by different protein kinases ) can provide another level of differential activations of ankyrins ( see below ) . Although AnkB and AnkG share the similar overall autoinhibition modes , the detailed amino acid sequences in their autoinhibitory segments are quite different ( Figures 6A and 7A ) . In parallel with this finding , AnkB and AnkG are often targeted to different subdomains in the same tissues including neurons and cardiomyocytes ( Galiano et al . , 2012; Lowe et al . , 2008; Mohler et al . , 2005 ) . It is possible that the unique sequences of their inhibitory segments provide means for AnkB and G to be differentially regulated . One such possible mean is via post translational modifications such as phosphorylations ( Lu et al . , 1985; Cianci et al . , 1988 ) . We surveyed the PhosphoSitePlus database ( http://www . phosphosite . org/ ) and found that Ser855 in AI-c , Thr888 , Tyr891 in AI-b of AnkG and Thr838 , Ser846 in AI-c , Tyr889 , Ser890 in AI-b of AnkB can be phosphorylated . These residues are positioned within the autoinhibitory segments . It is possible that phosphorylation on some of these residues may lead to differential releases of the MBD autoinhibition of AnkB or AnkG . Again , the disordered nature of the autoinhibitory segments of ankyrins is favorable for their accesses to different protein kinases . It is further noted that different ankyrin-enriched membrane microdomains ( for example AIS ) are often selectively enriched with certain protein kinases ( Bréchet et al . , 2008; Hund et al . , 2010; Tapia et al . , 2013 ) . Post translational modifications on the intrinsically disordered inhibitory segments of ankyrins as a mean of their target recognition regulations may be a fertile ground to explore in the future . In summary , we systemically studied the autoinhibition mechanisms of the 24 ANK repeats of AnkB and G by their own respective MBD/SBD linker and proximal tail segments . We biochemically characterized these intramolecular interactions , solved the representative complex structures and quantitatively evaluated the inhibitory effects of these autoinhibitory segments on MBD’s bindings to different physiological membrane targets . We demonstrated that the combinatorial uses of multiple semi-independent , intrinsically disordered autoinhibitory segments can provide graded regulations of targets binding by ankyrin MBDs . Our findings on the autoinhibition as well as target bindings of the 24 ANK repeats of ankyrins may also shed light on target recognitions and regulations by other long repeat-containing scaffold proteins that are quite abundant in the mammalian proteomes .
The coding sequences of the AnkR constructs were PCR amplified from a mouse muscle cDNA library . The coding sequences of AnkB and AnkG constructs were PCR amplified from the full-length human 220 kDa AnkB template or the full-length rat 270 kDa AnkG template respectively ( both templates as well as the HA tagged full length NF186 construct are generous gifts from Dr . Vann Bennett ) . E-cadherin ( aa 734–884 ) , NF186 ( aa 1187–1214 ) , L1CAM ( aa 1206–1233 ) and Nav1 . 2 ( aa 1035–1129 ) coding sequences were PCR amplified from mouse brain or muscle cDNA libraries . All of the constructs that used for protein expression were cloned onto a home-modified pET32a vector . All truncation mutations of ANK repeats constructs were made with the same strategy as described in our previous study ( Wang et al . , 2014 ) . The fusion constructs of AnkR/B_CT Chimera/AnkB_repeats_R1-20 and AnkB_AI-b/AnkB_R8-M14 were made by standard two-step PCR with a coding sequence of ‘GSLVPRGSGS’ as the flexible linker ( M14 means replacing the αB of R14 with a capping sequence corresponding to the αB of R24 for protein stabilization as we used earlier ) . The same strategy was used in making other fusion constructs described in this study . All point mutations were created using the Quick Change site-directed mutagenesis kit and confirmed by DNA sequencing . Protein expression and purification protocols are the same as previously described ( Wang et al . , 2012; 2014 ) . Recombinant proteins were expressed in BL21 ( DE3 ) Escherichia coli cells with induction of 0 . 25 mM IPTG at 16°C . The N-terminal Trx-his6-tagged proteins were purified using Ni2+-NTA agarose affinity column followed by size-exclusion chromatography ( Superdex 200 column from GE Healthcare , Little Chalfont , UK ) in the final buffer containing 50 mM Tris-HCl , 1 mM DTT , and 1 mM EDTA , pH 7 . 8 with either 100 mM NaCl or 500 mM NaCl as required . For simplicity , we use human 220 kDa AnkB ( NM_020977 . 3 ) , rat AnkG ( MBD-UPA: NM_001033984 . 1 and DD-CT: NM_031805 . 1 ) and mouse AnkR ( NM_001110783 . 3 ) for the amino acid numbering throughout the manuscript . The various constructs of ankyrins used in this study are listed in Table 2 . Isothermal titration calorimetry ( ITC ) assays and fluorescence assays were carried out with the same protocol as described earlier ( Wang et al . , 2014 ) . Briefly , isothermal titration calorimetry assays were performed on a VP-ITC MicroCal calorimeter ( MicroCal , Northampton , MA ) at 25°C and data were analyzed and fitted using the program Origin7 . 0 ( Microcal ) . Fluorescence-based binding assays were performed on a PerkinElmer LS-55 fluorimeter equipped with an automated polarizer at 25°C . The Kd values were obtained by fitting the titration curves with the classical one-site binding model . All crystals were obtained by hanging drop or sitting drop vapor diffusion methods at 16°C . Crystals of AnkR/B_CT Chimera/AnkB_repeats_R1-20 were grown in solution containing 4 M ammonium acetate and 0 . 1 M Bis-Tris propane ( pH 7 . 0 ) . Crystals of AnkB_AI-b/AnkB_R8-M14 were grown in solution containing 0 . 1 M HEPES ( pH 7 . 0 ) , 1 M ammonium sulfate and 0 . 5% w/v PEG 8 , 000 . Crystals of AnkB_AI-c/AnkB_R13-24 were grown in solution containing 0 . 2 M CaCl2 , 0 . 1 M HEPES ( pH 7 . 5 ) and 28% v/v PEG 400 . Crystals were soaked in crystallization solution containing additional 20% glycerol for cryoprotection . All datasets were collected at the Shanghai Synchrotron Radiation Facility at 100 K . Data were processed and scaled using HKL2000 ( Otwinowski and Minor , 1997 ) . Structures were solved by molecular replacement using PHASER ( McCoy et al . , 2007 ) with fragments of the entire 24 ANK repeats ( PDB: 4RLV ) as the searching models . Peptides were manually built according to Fo-Fc difference maps in COOT ( Emsley et al . , 2010 ) . Further manual model adjustment and refinement were completed iteratively using COOT ( Emsley et al . , 2010 ) and PHENIX ( Adams et al . , 2010 ) . The final models were validated by MolProbity ( Chen et al . , 2010 ) and statistics are summarized in Table 1 . All structure figures were prepared by PyMOL ( http://www . pymol . org ) . The coordinates of the structures reported in this work have been deposited to PDB under the accession codes of 5Y4D , 5Y4E and 5Y4F for the RB-Chimera/AnkB_R1-20 , AI-b/AnkB_R8-M14 and AI-c/AnkB_R13-24 structures , respectively . MDCK cells were seeded on 35 mm dishes with 10 mm diameter uncoated glass bottom ( MatTek , Ashland , MA ) and grown in 10% FBS supplemented DMEM at 37°C incubator with 5% CO2 . After around 20 hr , when the confluency reached 30 ~ 40% , cells were transfected with 300 ng plasmids using Lipofectamine 2000 transfection reagent ( Invitrogen , Carsbad , CA ) following the protocol suggested by the manufacturer . After transfection , MDCK cells were grown in 10% FBS supplemented DMEM until they were fully polarized . HeLa cells were cultured in the same media and culture condition as used for MDCK cells . HeLa cells were co-transfected with 400 ng HA tagged NF186 plasmids and 500 ng ankyrin constructs using Viafect transfection reagent ( Promega , Madison , WI ) when the confluency reached 20 ~ 30% . Then cells were grown in 10% FBS supplemented DMEM for 24 hr and fixed . HeLa ( RRID: CVCL_0030 ) and MDCK ( RRID: CVCL_0422 ) cells were originated from ATCC . These cells were not individually authenticated and not found to be on the list of commonly misidentified cell lines ( International Cell Line Authentication Committee ) . Cells were tested negative for mycoplasma contamination by cytoplasmic DAPI staining . The MDCK or HeLa cells were fixed with 4% paraformaldehyde at room temperature for 15 min , and permeabilized with 0 . 2% Triton X-100 at room temperature for 15 min followed by a 60 min blocking in PBS buffer containing 5% bovine serum albumin . For immunostaining , cells were then incubated with primary antibodies ( goat anti GFP , ab6658 , Lot: GR206330-6 , RRID: AB_305631; Rabbit anti HA , Sigma , H6908 , RRID: AB_260070 ) at 4°C overnight . The next day , cells were washed with PBS buffer three times and then incubated with fluorescence-conjugated secondary antibodies ( Alexa Fluor 488 or 594 ) at room temperature for 2 hr , followed by incubating with 500 nM DAPI for 5 min to stain nucleus . Then cells were washed with PBS before imaging . All the cell culture images were captured by a Zeiss LSM 880 laser-scanning confocal microscope . The MDCK cell and HeLa cell images were captured using a 40 × 1 . 4 oil objective with pinhole setting to 1 Airy unit . Fluorescence intensity were analyzed with ImageJ software ( https://imagej . nih . gov/ij/ ) and statistically analyzed with GraphPad Prism five using one-way ANOVA followed by Dunnett’s multiple comparisons test . | The membrane that surrounds the cells of animals is organized into regions known as microdomains . Different microdomains have different roles; for example , one microdomain may strengthen the cell while another may contain the apparatus used by cells to signal to each other . A family of proteins called the ankyrins plays a central role in forming microdomains , and mutations to these proteins have been linked to many diseases , including heart rhythm abnormalities and bipolar disorder . Ankyrins connect certain proteins in the cell membrane to each other and to the network of microfilaments that organize the cell interior . A region called the membrane binding domain at one end of the ankyrin interacts with the membrane proteins . The other end , or “tail” , of the ankyrin is thought to regulate this interaction , although little is known about how it does so . The members of the ankyrin family are similar but tend to localize to different parts of the cell and play different roles . Chen , Li et al . have now used biochemical and structural techniques to analyze the sequences of two ankyrins , called Ankyrin-B and Ankyrin-G . The results show that the tail regions of both of these proteins contain three “autoinhibition” segments that restrict the activity of the ankyrins . Each segment independently interacts with different sites on the ankyrin’s membrane binding domain , and this interaction joins the ends of the protein together into a ‘head-to-tail’ conformation that prevents it binding to membrane proteins . The autoinhibition segments of Ankyrin-B and Ankyrin-G are quite different , which Chen , Li et al . suggest could explain why these proteins have different activities . Ankyrins play a vital role in many tissues , and so the findings presented by Chen , Li et al . will be of interest to scientists working in a wide range of basic biology and medical research fields . An immediate question for further investigation is how the autoinhibition of ankyrins is regulated to allow ankyrin-organized microdomains to form and work properly . Results from such studies will help researchers to understand the mechanisms that underlie diseases caused by mutations of ankyrins or the proteins that they bind to , and to develop potential therapies for these diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2017 | Autoinhibition of ankyrin-B/G membrane target bindings by intrinsically disordered segments from the tail regions |
Shifting attention among visual stimuli at different locations modulates neuronal responses in heterogeneous ways , depending on where those stimuli lie within the receptive fields of neurons . Yet how attention interacts with the receptive-field structure of cortical neurons remains unclear . We measured neuronal responses in area V4 while monkeys shifted their attention among stimuli placed in different locations within and around neuronal receptive fields . We found that attention interacts uniformly with the spatially-varying excitation and suppression associated with the receptive field . This interaction explained the large variability in attention modulation across neurons , and a non-additive relationship among stimulus selectivity , stimulus-induced suppression and attention modulation that has not been previously described . A spatially-tuned normalization model precisely accounted for all observed attention modulations and for the spatial summation properties of neurons . These results provide a unified account of spatial summation and attention-related modulation across both the classical receptive field and the surround .
Our eyes are constantly bombarded by a welter of visual stimuli , only a small fraction of which can be processed thoroughly ( Chun et al . , 2011; Kastner and Ungerleider , 2000 ) . Spatial attention sifts through the plethora of stimuli – enhancing perception at behaviorally-relevant locations – but the underlying neural principles of this process are not fully understood ( Chun et al . , 2011; Kastner and Ungerleider , 2000; Posner , 1980; Carrasco , 2011; Roelfsema et al . , 1998; Anton-Erxleben and Carrasco , 2013 ) . Neuronal responses modulate as attention shifts among stimuli at different receptive-field locations ( Moran and Desimone , 1985; Treue and Maunsell , 1996; Reynolds et al . , 1999; Martínez-Trujillo and Treue , 2002; Ghose and Maunsell , 2008; Lee and Maunsell , 2010; Ni et al . , 2012; Recanzone and Wurtz , 2000; Luck et al . , 1997; Motter , 1993; Chelazzi et al . , 1998; Zénon and Krauzlis , 2012 ) . These response modulations can be complicated . For example , depending on the stimulus configuration , attending to a non-preferred stimulus can either increase or suppress activity ( e . g . Treue and Maunsell , 1996 ) . Normalization models of attention provide a succinct framework in which these complex response modulations can be understood ( Reynolds et al . , 1999; Ghose , 2009; Reynolds and Heeger , 2009; Lee , 2009; Boynton , 2009 ) . However , only a few studies have directly tested these models against the responses of individual neurons to various stimulus configurations ( Ni et al . , 2012; Lee , 2009; Sanayei et al . , 2015; Xiao et al . , 2014 ) . Normalization models of attention assume that attention acts on stimulus-induced excitation and suppression to modulate neuronal responses . Importantly , both excitation and suppression vary spatially within the receptive field: excitation is largely restricted to the classical receptive field ( cRF ) , while suppression extends far beyond into the surround ( Cavanaugh et al . , 2002a , 2002b; Sceniak et al . , 1999; Desimone and Schein , 1987; Carandini et al . , 1997 ) . Crucially , how attention interacts with the receptive field structure of neurons remains unclear . For example , the way that attention acts on neuronal responses when shifted among stimuli inside the cRF versus when shifted to stimuli inside the surround has not been compared directly ( Motter , 1993; Sanayei et al . , 2015; Sundberg et al . , 2009 ) . Differences may occur because feedforward- , feedback- and intracortical circuitries are thought to contribute differentially to the suppressive and excitatory inputs associated with stimuli in either the cRF or the surround ( Angelucci et al . , 2014 ) , and because the cRF and the surround presumably serve different functional roles ( Angelucci et al . , 2014; Schwartz and Simoncelli , 2001; Vinje and Gallant , 2000 ) . More generally , it remains unknown if and how attention operates on the spatially-varying excitation and suppression of a neuron's receptive field . This is a pivotal open question because , as we will show below , the interaction between attention and the receptive field structure determines which neurons are most affected by attention and consequently are most likely to influence attentional behavior . We measured how attention affects neuronal responses to various stimulus configurations both inside and outside the cRF of V4 neurons , and fitted normalization models to the responses of individual neurons . We find that the principles that drive attention modulation are remarkably similar within the classical receptive field and the surround . We show that stimuli induce excitation and suppression that varies spatially , and that attention interacts with this spatially-varying excitation and suppression . This interaction explained the large differences in attention modulations across neurons , and a non-additive relationship among stimulus selectivity , stimulus-induced suppression and attention modulation . A spatially-tuned normalization model , wherein attention multiplies both the excitatory and spatially-varying normalization term , precisely accounted for all neuronal responses to either single or multiple stimuli , either attended or unattended , presented inside either the cRF or the surround . The model relates stimulus selectivity , stimulus-induced suppression and attention-related modulation to each other , and unifies spatial summation and attention-related modulation across different regions of the receptive field .
We trained two rhesus monkeys to perform a visual-detection task in which spatial attention was controlled and measured . In each trial , a sequence of stimuli was presented at four locations equidistant from the fixation point ( Figure 1A ) . Stimuli were full-contrast static Gabor stimuli with one of two orthogonal orientations . The monkey's task was to detect a faint white spot ( target; Figure 1A right ) that appeared in the center of one Gabor during a randomly selected stimulus presentation . We manipulated attention in blocks of trials by cueing the monkey at the start of each block as to which stimulus location was most likely to contain the target ( Materials and methods ) . In 91% of trials the target was presented at the cued location ( valid cue; position of the black circle Figure 1A ) . On the remaining 9% of trials the target appeared with equal probability at one of the three uncued locations: either next to the cued location ( invalid near; position of the yellow circle Figure 1A ) , or at one of two locations contralateral to the cued location ( invalid far; position of the blue circles Figure 1A ) . 10 . 7554/eLife . 17256 . 003Figure 1 . Task and performance . ( A ) Every trial consisted of a sequence of stimulus presentations . On each stimulus presentation ( 200 ms duration; 200–1020 inter-stimulus interval ) , Gabor stimuli of two orthogonal orientations could be presented at four possible stimulus locations . The monkey was rewarded for detecting a faint white spot ( target ) in the center of one Gabor during one stimulus presentation . For 91% of trials the target was presented at the cued location ( location of the black circle; valid trials ) . On the remaining 9% of trials the target was presented at one of three uncued locations: adjacent to the cued location ( location of the yellow circle; invalid near ) , or at one of two locations on the opposite side of the fixation point ( location of the blue circles; invalid far ) . Colored circles in ( A ) are shown for illustrative purposes , never presented during the task . ( B ) Average performance across recording sessions for monkey M1 . Proportion correct ( ± SEM based on N = 52 sessions; proportion correct at equal target strength: Valid: 0 . 79; Invalid near: 0 . 42; Invalid far: 0 . 30 ) as a function of target strength for trials in which the target occurred at the cued ( gray: valid ) or uncued ( yellow: invalid near; blue: invalid far ) location . Target strength is defined as the opacity of the target . The pictograms below the target-strength axis illustrate the nature of the target-strength manipulation but do not represent actual target-strength values used during the recordings . ( C ) Average performance across recording sessions for monkey M2 ( N = 78 sessions; proportion correct at equal target strength: Valid: 0 . 56; Invalid near: 0 . 24; Invalid far: 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17256 . 003 The attention cue considerably affected behavioral performance in the task: targets were much more likely detected at a cued location than at an uncued location , even when the uncued location was adjacent to the cued location ( Figure 1B , C; valid vs . invalid near: monkey M1: p=8 × 10−27 , M2: p=1 × 10−26; valid vs . invalid far: M1: p=1 × 10−36 , M2: p=2 × 10−58; paired t-test on the average proportion correct across sessions; M1: N = 52; M2: N = 78 ) . The improved performance indicates that the monkeys preferentially attended to the cued stimulus location , which allowed us to compare neuronal responses among conditions in which attention was directed to different stimulus locations within neurons' cRF or surround . We examined the principles by which attention affects neuronal responses to stimuli inside the classical receptive field ( cRF ) or within the surround ( sRF ) . Using chronically implanted microelectrode arrays , we recorded from 728 neurons in visual area V4 in the left hemisphere of two monkeys ( monkey M1: 264; M2: 464 ) while they performed the visual-detection task in which spatial attention was controlled . All results presented here are based on the activity of these 728 single neurons , but all findings were confirmed in the responses of 12 , 067 multi-unit clusters ( M1: 4709; M2: 7358 ) . During each session we simultaneously measured the activity of multiple neurons , and optimized the orientation and position of stimuli for a randomly selected unit . The neurons' receptive field centers were located in the lower right visual field ( black dots in Figure 2A for an example session ) . 10 . 7554/eLife . 17256 . 004Figure 2 . Stimulus conditions . ( A ) Neurons' receptive field centers were located in the lower right visual field: black dots indicate receptive-field centers of 16 simultaneously recorded neurons from one recording session . White circles ( 1 , 2 , 3 ) indicate the three stimulus locations near the neurons' receptive field for this example session . Within a block of trials , only two stimulus locations were used: locations 1+2 or 1+3 . ( B ) Nine possible stimulus combinations resulting from two stimulus locations and two orthogonal orientations . ( C ) Two receptive-field configurations: cRF-cRF stimulus configuration with two stimuli inside the neuron's classical receptive field . White dotted circle illustrates the cRF . ( D ) sRF-cRF stimulus configuration with one stimulus inside a neuron's cRF and an adjacent stimulus in its surround . Each stimulus location near the neurons' receptive fields ( stimulus location 1 , 2 , 3 in 2A ) had a corresponding stimulus location on the opposite side of the fixation point ( stimuli near Away in C , D; see also Figure 2—figure supplement 1 ) . ( E ) Pictograms illustrate for one Gabor pair the stimulus configurations used to calculate all indices . Cyan circles indicate the preferred Gabor ( P ) , orange circles the non-preferred Gabor ( N ) . Solid circles represent task conditions wherein attention was directed toward a stimulus location near the neurons' receptive field ( PAttN , PNAtt ) . Dashed circles indicate that the stimulus was unattended and attention was directed toward another location . DOI: http://dx . doi . org/10 . 7554/eLife . 17256 . 00410 . 7554/eLife . 17256 . 005Figure 2—figure supplement 1 . Average PSTH for individual Gabor stimuli presented inside the classical receptive field ( cRF ) and within the surround . Shown are the average responses from the same V4 neurons to a Gabor stimulus placed either inside the cRF ( dashed line ) or within the surround ( solid line ) . Surround stimuli on average slightly suppressed the baseline response . Black vertical line indicates stimulus onset . Shading over the lines indicates ± SEM . Based on the responses from 558 neurons for which a surround position was examined . DOI: http://dx . doi . org/10 . 7554/eLife . 17256 . 005 In different blocks of trials , we measured neuronal responses to stimuli presented at three different receptive-field locations ( stimulus locations 1 , 2 , 3 in Figure 2A ) . Within a block of trials , only two of these stimulus locations were used: e . g . location 1+2 or 1+3 in Figure 2A ( Materials and methods ) . During each stimulus presentation within a trial , we presented one , two , or no stimuli at the two stimulus locations near the receptive field ( Figure 2B ) . Depending on the location of each neuron's receptive field , stimuli fell either inside the cRF or within the surround . We distinguished between two receptive-field configurations: one in which the two stimulus locations both lay inside the neuron's cRF ( cRF-cRF , Figure 2C ) , and another in which one stimulus location was positioned inside the neuron's cRF while the other stimulus location was positioned inside its surround ( sRF-cRF , Figure 2D ) . Because we tested the responses to stimuli shown at two locations pairings ( e . g . locations 1+2 in Figure 2C vs . 1+3 in Figure 2D ) , 309 neurons were tested in both a cRF-cRF and an sRF-cRF configuration ( M1: 97; M2: 212 ) . We classified locations as belonging to the cRF or sRF using stimulus presentations that included only one Gabor ( Figure 2B; Materials and methods ) . Locations where either stimulus orientation generated a response were considered to lie within the cRF . Those where neither stimulus orientation generated a response were considered to lie within the surround ( Figure 2—figure supplement 1 ) . In different blocks of trials , the monkeys directed their attention toward all possible stimulus locations , one attended location per block of trials , each time ignoring the other stimulus locations . Attention was directed toward stimulus locations near the neurons' receptive fields ( e . g . locations 1 , 2 , or 3 in Figure 2A ) , or toward stimulus locations away from the receptive fields ( 'Away' in Figure 2C , D; attend away ) , i . e . to stimulus locations on the opposite side of the fixation point from the neuron's receptive field . We quantified the stimulus selectivity of the neurons separately for each stimulus configuration . For each of four Gabor pairs ( Figure 2B ) at each pair of stimulus locations ( i . e . location pairings 1+2 or 1+3 , Figure 2A ) , we used a selectivity index ( 'Selectivity' , Figure 2E ) : ( P−N ) / ( P+N ) , that ranges from zero ( unselective ) to one ( completely selective ) . Here P is the response to the component Gabor of a Gabor pair that generated the stronger average response when presented alone ( preferred ) , and N is the response to other component Gabor that generated the weaker average response when presented alone ( non-preferred ) . Note that the preferred and non-preferred Gabor within a pair were presented at two different receptive-field locations , and could have the same or a different orientation ( Figure 2B ) . Thus stimulus selectivity between members of a Gabor pair could arise from a neuron's orientation selectivity and from its preference for spatial locations . In subsequent analyses , we will show that the relationship between attention modulation and stimulus selectivity does not depend on whether the stimulus feature is space or orientation . What is critical for attention modulation is a differential response to the component stimuli of a compound stimulus . For both the cRF-cRF and sRF-cRF condition , we measured each neuron's stimulus-induced suppression for each Gabor pair at each pair of stimulus locations using a stimulus-induced suppression index: ( P−PN ) / ( P+PN ) ( middle right pictogram Figure 2E ) . PN is the response to the Gabor pair ( P and N defined as before ) . This index is negative when the neuronal response increases when a non-preferred stimulus is added to the preferred stimulus , and positive when the neuronal response is suppressed by the addition of a non-preferred stimulus to the preferred stimulus . By definition , neurons do not respond to a stimulus when it appears alone inside the surround , so the surround stimulus is invariably assigned as non-preferred ( N ) . For both the selectivity index and the stimulus-induced suppression index , the responses to the preferred ( P ) , non-preferred ( N ) , and their combined presentation ( PN ) were measured in the same attention state: when attention was directed away from the neuron's receptive field ( attend away ) . These responses are shown in the bar-plot insets in Figure 3A–D . 10 . 7554/eLife . 17256 . 006Figure 3 . Example attention modulations . Responses of four different neurons to a selected Gabor pair are shown ( measured in different sessions ) . ( A ) Example 1: cRF-cRF configuration . Left panel shows this neuron's receptive-field map with the two stimulus locations at which the Gabors were presented overlaid ( white-gray dots ) . Right panel PSTHs show the neuronal responses to the Gabor pair when attention was directed toward the preferred Gabor ( cyan line; PAttN ) , the non-preferred Gabor ( orange line; PNAtt ) , or a stimulus on the opposite side of the fixation point ( green dashed line; PN; attend away ) . Bar-plot inset shows the responses of this neuron to a Gabor pair ( PN ) and its component Gabors ( P , N ) , all measured in the attend away condition . This neuron's response was selective to the component Gabors of the Gabor pair ( P vs . N ) , suppressed by the addition of a non-preferred Gabor to a preferred Gabor ( P vs . PN ) , and strongly modulated when attention was shifted between the two component Gabors of the Gabor pair ( PAttN vs . PNAtt ) . ( B ) Example 2: another neuron in the cRF-cRF configuration . This neuron showed weak selectivity , hardly any suppression , and little attention modulation . ( C ) Example 3: sRF-cRF configuration with one Gabor inside the neuron's cRF , and one Gabor inside its surround . By definition , the cRF Gabor is preferred ( P ) and the silent surround Gabor is non-preferred ( N ) . The neuron responded highly selectively to the cRF and the surround Gabor when presented alone ( P vs . N ) , showed surround suppression ( P vs . PN ) , and was modulated by attention ( PAttN vs . PNAtt ) . ( D ) Example 4: another neuron in the sRF-cRF configuration . This neuron was highly selective to the component Gabors of the Gabor pair , but only weakly suppressed by the surround Gabor , and showed little attention modulation . The insets show the average waveforms of the recorded neurons ( blue ) plus that of the multi-unit activity measured at the same electrode ( grey ) . Shading around the mean represents ± 2 median absolute deviation ( MAD ) . Scale bars indicate 50 μV and 0 . 1 ms . The receptive-field maps were normalized to the maximum response for each neuron during receptive-field mapping ( RF max ) , dark blue shows the baseline response . Error bars represent ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 17256 . 006 Figure 3A–D shows examples of attention-related response modulations of four different neurons to one selected Gabor pair: two neurons in the cRF-cRF configuration ( A , B ) and two neurons in the sRF-cRF configuration ( C , D ) . The neuron in Figure 3A responded selectively to the two component Gabors of the Gabor pair shown inside the neuron's cRF ( inset: P vs . N; selectivity index=0 . 44 ) . Its response to the preferred Gabor was suppressed when the non-preferred Gabor was added to it ( inset: P vs . PN; suppression index=0 . 22 ) . The position of attention profoundly affected this neuron's responses: Compared to when attention was directed away from the neuron's receptive field ( dashed green line; PN; attend away ) , attention to the preferred Gabor increased this neuron's response ( cyan line; PAttN ) , whereas attention to the non-preferred Gabor suppressed its response ( orange line; PNAtt ) . Attention-related modulation was quantified using an attention-modulation index: ( PAtt N−PNAtt ) / ( PAtt N + PNAtt ) ( lower pictogram Figure 2E ) , which is positive when the neuronal response increases when attention is directed toward the preferred Gabor , compared to when attention is directed toward the non-preferred Gabor . The attention-modulation index for example 1 was 0 . 48 . In contrast to example 1 , the response of the neuron in Figure 3B was poorly selective to the component Gabors of the Gabor pair ( P vs . N; selectivity index=0 . 066 ) , showed little suppression when a non-preferred Gabor was placed alongside a preferred Gabor ( P vs . PN; suppression index=0 . 04 ) , and was only weakly modulated when attention shifted between the preferred and the non-preferred Gabor within the cRF ( cyan vs . orange line; attention-modulation index=0 . 04 ) . Figure 3C shows the responses of a neuron to a Gabor pair in another stimulus configuration , in which one Gabor was placed inside the neuron's cRF and another Gabor inside its surround ( sRF-cRF ) . As expected , the neuron responded much more to the cRF Gabor than to the surround Gabor ( P vs . N; selectivity=0 . 963 ) . When the surround Gabor was placed alongside the cRF Gabor , the neuron's response was greatly reduced , the hallmark of surround suppression ( P vs . PN; suppression index=0 . 336 ) . The neuron showed strong attention-related modulation: Compared to when attention was removed from both the cRF and the surround Gabor ( dashed green line; attend away ) , attention to the cRF Gabor increased this neuron's response ( cyan line ) , while attention to the surround Gabor sharply decreased its response ( orange line; attention-modulation index=0 . 58 ) . The response of the fourth example neuron in Figure 3D was highly selective to the component Gabors of the Gabor pair ( P vs . N ) , only slightly suppressed by the surround Gabor ( P vs . PN ) , and its firing rate was barely modulated by attention ( cyan vs . orange line; selectivity index=0 . 9; suppression index=0 . 05; attention-modulation index=0 . 08 ) . These examples illustrate the diverse stimulus selectivities , stimulus interactions ( i . e . stimulus-induced suppression ) and attention-related modulations in the neuronal responses in visual cortex . Next , we asked how variability in stimulus selectivity and stimulus-induced suppression relates to variability in attention modulation within the cRF and the surround across the sample of recorded neurons . We first examined the relationship between selectivity and attention modulation . Shifting attention between two Gabors inside the cRF was associated with larger response changes for neurons with more selective responses to the component Gabors of the Gabor pair ( Figure 4A; cRF-cRF configuration; p=4 × 10−109 for a non-zero slope; linear regression ) ( Reynolds et al . , 1999 ) . Attention-related modulation was also stronger for neurons that responded more selectively to the cRF and surround stimulus ( Figure 4B; sRF-cRF configuration; p=3 × 10−76 for a non-zero slope; linear regression ) . Low selectivity can occur in the sRF-cRF configuration when the cRF stimulus produces little response because it has a non-preferred orientation or is positioned at a weakly responsive cRF location . Comparing Figure 4A and B shows that attention-related modulation increases more with selectivity in the cRF-cRF than in the sRF-cRF configuration ( p=5 × 10−4 for different slopes in each receptive-field configuration; general linear model ) . 10 . 7554/eLife . 17256 . 007Figure 4 . First-order analyses suggest that attention modulation follows different principles for stimuli inside the cRF and the surround . ( A , B ) Average attention modulation as a function of the stimulus selectivity in the cRF-cRF and sRF-cRF configuration respectively . Low selectivity occurs in the sRF-cRF configuration when neurons respond weakly to the cRF stimulus , e . g . because of a non-preferred orientation or a weakly responsive cRF location , and have a baseline response to the surround stimulus . ( C , D ) Histogram of all stimulus-induced suppression indices measured in the cRF-cRF and sRF-cRF configuration respectively . The suppression index is negative when neurons increase their response when a non-preferred stimulus is added to the preferred stimulus ( enhancing ) , and positive when neurons decrease their response when a non-preferred stimulus is added to the preferred stimulus ( suppressing ) . Black bars indicate indices associated with Gabor pairs for which the suppression index differed significantly from zero ( p<0 . 01; permutation t-test; see also Figure 4—figure supplement 1 ) . Triangle points to the mean suppression index . ( E , F ) Average attention modulation as a function of stimulus-induced suppression in the cRF-cRF and sRF-cRF configuration respectively . Error bars represent ± SEM . ( G , H ) Stimulus-induced suppression versus stimulus selectivity for all Gabor pairs in the cRF-cRF ( N = 1769 ) and sRF-cRF ( N = 1768 ) configuration respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 17256 . 00710 . 7554/eLife . 17256 . 008Figure 4—figure supplement 1 . Example neurons with strong surround suppression . A , B , C three neurons with significant surround suppression ( p<0 . 01 ) . The left panels show the average responses of the neurons to single stimuli presented either inside the cRF ( black; cRF ) , the surround ( light grey; Surround ) and the responses to the combined presentation of both the cRF and the surround stimulus ( grey; cRF + surround ) . The insets show the average waveforms of the recorded neurons ( blue ) plus that of the multi-unit activity measured at the same electrode ( grey ) . Shading around the mean represents ± 2 median absolute deviation ( MAD ) . The right panels show each neuron's receptive-field map with the two stimulus locations at which the Gabors were presented overlaid ( white-gray dots ) . The receptive-field maps were normalized to the maximum response for each neuron during receptive-field mapping ( RF max ) , dark blue shows the baseline response . See Figure 3C for another example . DOI: http://dx . doi . org/10 . 7554/eLife . 17256 . 008 We next examined stimulus-induced suppression . V4 neuronal responses on average decrease when a non-preferred stimulus is added to a preferred stimulus inside their cRF ( Figure 4C; average suppression index = 0 . 08 , p=4 × 10−104; t-test for a difference from zero ) ( Reynolds et al . , 1999 ) . Similarly , stimulating the surround decreases the average responses of V4 neurons ( Figure 4D; average suppression index = 0 . 04 , p=2 × 10−28; t-test ) ( Schein and Desimone , 1990 ) . However , stimuli inside the surround suppressed the neuronal responses less than stimuli inside the cRF: the average suppression index for the surround condition ( sRF-cRF ) was significantly smaller than the average suppression index for the cRF condition ( M1: p=9 × 10−6; M2: p=4 × 10−15; t-test; see below and Figure 7 for further discussion ) . The black bars in Figure 4C and D represent neurons that were significantly ( p<0 . 01 ) suppressed by the non-preferred ( surround ) stimulus . See Figure 4—figure supplement 1 for some example neurons with significant surround suppression ( see also Figure 3C ) . Surround suppression was also weaker than cRF suppression when comparing only suppression indices that differed significantly from zero ( p<0 . 001 ) . Extending previous findings in area MT ( Ni et al . , 2012; Lee , 2009 ) , we find that V4 neurons with stronger stimulus-induced suppression by cRF stimuli also showed stronger attention modulation ( Figure 4E; p=1 × 10−31 for a non-zero slope; linear regression ) . Furthermore , and consistent with a previous study ( Sundberg et al . , 2009 ) , attention modulation was also stronger for neurons whose responses were more suppressed by a surround stimulus ( Figure 4F; p=5 × 10−4 for a non-zero slope; linear regression ) . However , comparing Figure 4E and F shows that attention-related modulation increases more with stimulus-induced suppression in the sRF-cRF than in the cRF-cRF configuration ( p=0 . 005 for different slopes in each receptive-field configuration; general linear model ) . A previous study in V4 examining the relationship among stimulus selectivity , sensory interaction ( akin to stimulus-induced suppression ) and attention modulation , found a strong correlation between stimulus selectivity and sensory interaction ( Reynolds et al . , 1999 ) . In the present study , however , stimulus selectivity and stimulus-induced suppression were not significantly correlated with each other across neurons ( Figure 4G , H Pearson correlation = 0 . 02 , p=0 . 32; see Discussion for further comments on the difference between studies ) . This finding shows that the correlations between stimulus selectivity and attention modulation , and that between stimulus-induced suppression and attention modulation , are not explained by an underlying correlation between selectivity and suppression . Furthermore , and in contrast to previous studies , the lack of a correlation between both indices allowed us to examine the separate contributions of selectivity and suppression to the magnitude of attention modulation . The above-mentioned different relationships between stimulus selectivity , stimulus-induced suppression and attention-related modulation in the cRF-cRF and the sRF-cRF configuration , suggest that the rules that govern attention modulation differ within the cRF and the surround . Next we falsify this suggestion and show how these different relationships in the two receptive-field configurations can be explained by a common rule . We used multiple linear regression to examine if attention-related modulation depends on the joint magnitude of stimulus selectivity and stimulus-induced suppression . For both receptive-field configurations , the regression model included a main effect of selectivity and a main effect of stimulus-induced suppression . Importantly , in each RF configuration the regression model also included an interactive product term , which measured the dependency of attention-related modulation on both selectivity and stimulus-induced suppression , i . e . this term measures whether the relationship among selectivity , suppression and attention modulation is non-additive ( see Materials and methods for further information ) . Figure 5 shows how attention modulation varies with selectivity and stimulus-induced suppression ( 5A; cRF-cRF , 5B; sRF-cRF ) . For both configurations , the relationship is non-additive . Specifically , Figure 5A and B show that when stimulus-induced suppression is low , attention modulation will be weak , even when attention is shifted between a strong and a weak stimulus ( upper left corner in Figure 5A , B ) . That is , the plots show that the effect of selectivity near zero stimulus-induced suppression is weak , although significant ( main effect of selectivity at zero stimulus-induced suppression: cRF-cRF: p=2 × 10−64; sRF-cRF: p=2 × 10−60; M1: p=7 × 10−136 across RF configurations; M2: p=5 × 10−30 across RF configurations ) . 10 . 7554/eLife . 17256 . 009Figure 5 . Selectivity and stimulus-induced suppression interact to control attention modulation . ( A , B ) Average attention modulation as a function of stimulus-induced suppression ( x-axis ) and stimulus selectivity ( y-axis ) in the cRF-cRF and sRF-cRF configuration respectively . The magnitude of attention modulation is indicated by color ( red = strong , blue = weak ) . Note that , although the data covered most of this space ( see Figure 4G , H ) , few regions , e . g . the lower right corner in ( B ) , were not well sampled . ( C ) Model schematic . Every stimulus contributes an excitatory drive ( L1 and L2 ) to the neuron's response ( R1 , 2att ) to a Gabor pair . Each stimulated receptive-field location , either inside the cRF or inside the surround , contributes divisive suppression ( α1 and α2 ) to the neuron's response . The divisive suppression is fixed for each receptive-field location , independent of the stimulus presented at that location . A small amount of baseline suppression is further added ( σ parameter; not shown ) . Directing attention toward a stimulus location has a multiplicative effect ( β ) on the parameters ( L2 and α2 ) corresponding to the attended receptive-field location ( location 2 in the schematic ) . ( D , E ) Average model-predicted attention modulation as a function of the observed stimulus-induced suppression ( x-axis ) and the observed stimulus selectivity ( y-axis ) in the cRF-cRF and sRF-cRF configuration respectively ( See also Figure 5—figure supplement 1 ) . Same conventions as in ( A , B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17256 . 00910 . 7554/eLife . 17256 . 010Figure 5—figure supplement 1 . Example single-neuron responses and their corresponding model fits . ( A ) Neuron with a preferred ( P ) and non-preferred ( N ) stimulus presented inside the cRF ( cRF-cRF condition ) . Black: observed responses . Grey: modeled responses . P , Patt , N , Natt show the responses to the individually-presented preferred and non-preferred stimulus with attention away ( P , N ) , or attention directed to the stimulus ( Patt , Natt ) . PN shows the condition in which both stimuli were presented simultaneously with attention away ( PN ) , attention directed toward the preferred stimulus ( PattN ) , or directed toward the non-preferred stimulus ( PNatt ) . The values of the model parameters for each example neuron are shown on the right . Note that these parameter values correspond to spike counts in a 250 ms window and should be multiplied by four to obtain spikes/s . This neuron's response is suppressed when a non-preferred stimulus is added to a preferred stimulus ( P vs . PN ) . The model accounts for this difference because the non-preferred stimulus induces few excitation ( small L2 ) but large enough suppression ( α2 ) . So suppression dominates over excitation . The model also captures the strong attention modulation ( PattN vs . PNatt ) through the β parameter , which multiplies the excitatory drive ( L ) and suppressive drive ( α ) of the attended stimulus . By increasing the weight of both drives , attention effectively focuses on the inputs related to the attended stimuli , as if the inputs from other stimuli were attenuated . So attention to a weak stimulus decreases the response , while attention to a strong stimulus increases the response ( i . e . attention modulation ) . ( B ) Neuron with a cRF ( P ) and surround ( N ) stimulus ( sRF-cRF condition ) . The model accounts for the observed suppression and attention modulation , which is similar to that of the neuron in A ( cRF-cRF condition ) . ( C ) Neuron with a cRF ( P ) and surround ( N ) stimulus ( sRF-cRF condition ) . The surround stimulus induces no surround suppression ( low α2 and L2 value ) . As a result , shifting attention between the cRF and the surround stimulus leads to virtually no attention modulation . DOI: http://dx . doi . org/10 . 7554/eLife . 17256 . 010 Conversely , when selectivity is low , attention modulation will also be weak , even when attention is shifted between stimuli that strongly suppress each other’s response ( bottom right corner in Figure 5A , B; effect of stimulus-induced suppression at zero selectivity: cRF-cRF: p=0 . 5; sRF-cRF: p=0 . 6; M1: p=0 . 4 across RF configurations; M2: p=0 . 3 across RF configurations ) . Strong attention-related modulation occurs only when selectivity and suppression are both large , and this was true for both RF configurations ( upper right corner in Figure 5A , B; interaction between stimulus-induced suppression and selectivity: cRF-cRF: p=2 × 10−9; sRF-cRF: p=3 × 10−6; M1: p=1 × 10−4 across RF configurations; M2: p=3 × 10−20 across RF configurations ) . The interaction between selectivity and stimulus-induced suppression did not differ significantly between the cRF-cRF and sRF-cRF configuration ( M1: p=0 . 6; M2: p=0 . 85; 3-way interaction ) , nor did any other interaction with RF configuration . Because a non-significant effect does not indicate the absence of an effect , we performed a Bayesian regression analysis ( Materials and methods ) . This analysis showed that the observed data are 347 times more likely to agree with a regression model that does not distinguish between the cRF-cRF and sRF-cRF configurations than with a model that does include RF-configuration as a predictor . Thus attention modulation is driven by similar mechanisms within the cRF and the surround . To examine response modulations associated with shifting attention between two receptive-field stimuli , previous studies used two different stimuli ( e . g . stimuli of different orientations , colors , directions ) , each presented at a different but approximately equally-responsive cRF position ( Moran and Desimone , 1985; Reynolds et al . , 1999; Ghose and Maunsell , 2008; Lee and Maunsell , 2010; Ni et al . , 2012 ) . Similar to these previous studies , we reproduced the above findings using the data from conditions with low spatial selectivity . When attention shifted between stimuli at two approximately-equally responsive cRF positions ( less than 2 spike/s response difference when each of two cRF positions is stimulated with an identical single stimulus ) , similar effects were observed ( main effect of orientation selectivity: p<0 . 001; main effect suppression: p=0 . 7; interaction between feature selectivity and suppression: p=0 . 02; multiple linear regression ) . Next , we examined whether the converse situation , i . e . same stimuli at unequally-responsive cRF positions , would produce attention modulations comparable to those described earlier . We found similar attention-related modulations using Gabor pairs consisting of identical Gabor stimuli presented at various cRF positions . Note that in this situation selectivity to the component Gabors of a Gabor pair originates solely from a neuron's spatial preferences ( receptive field ) , because the Gabors are identical . All of the above findings were replicated using only the data obtained with Gabor pairs consisting of identical Gabors ( main effect spatial selectivity: p=4 × 10−22; main effect suppression: p=0 . 29; interaction between spatial selectivity and suppression: p=6 × 10−6; multiple linear regression ) . This indicates that attention-related modulation depends on a differential response to the component stimuli of a compound stimulus , regardless of the origin of the response difference ( feature or spatial ) . Accordingly , at their most abstract level , models of neuronal attention modulation only need to account for the responses arising from different component stimuli , whether they arise from preferences for specific stimulus features , preferences for certain parts of the receptive field , or both . Our findings reveal a striking uniformity in the rules that govern attention modulation inside the cRF and within the surround: the interaction between stimulus selectivity and stimulus-induced suppression strongly influences how much attention modulates neuronal responses . Hence , any model of neuronal attention modulation needs to embody this relationship . We found that a spatially-tuned normalization model can readily capture this interaction ( Materials and methods ) . We used a spatially-tuned normalization model , described as follows ( Figure 5C ) : ( 1 ) R1 , 2=L1+L2α1+α2+σ where R1 , 2 is the neuronal response to a Gabor pair consisting of Gabors 1 and 2 . L1 and L2 are the excitatory drives associated with each component Gabor . The α1 and α2 parameters control the suppressive drive of each stimulated cRF or surround location . In this model , α1 and α2 are each associated with one receptive-field location , and do not vary with the orientation of the stimuli shown at those locations . Because the suppression , or normalization , is free to vary across receptive-field locations , the normalization is spatially tuned . In fitting the data , we fixed α1 at one to constrain the model . The α parameter adds baseline suppression . Directing attention toward the first ( R1att , 2; Equation ( 2 ) ) or second ( R1 , 2att; Equation ( 3 ) ) receptive-field location has a multiplicative effect on the parameters corresponding to the attended receptive-field location . This is described by the β parameter in Equations ( 2 ) and ( 3 ) : ( 2 ) R1att , 2=βL1+L2β+α2+σ ( 3 ) R1 , 2att=L1+βL21+βα2+σ The model was fit to each neuron's responses in all stimulus conditions: including conditions with one stimulus or two stimuli near the receptive field , and conditions with attention directed toward stimulus locations near the receptive field or away from it ( Materials and methods ) . The spatially-tuned normalization model provided an accurate account of the neuronal data , giving a median two-fold cross-validated explained variance of 87% across neurons ( M1: 86%; M2: 88% ) . For the 309 neurons ( M1: 97; M2: 212 ) that were tested in both a cRF-cRF and an sRF-cRF configuration the responses were equally well explained ( M1: 86%; M2: 87% ) . The model captures the way attention modulates neuronal responses to stimuli inside the cRF or the surround . Figure 4A , B , E , F ( light grey points ) show that the model precisely accounts for attention modulation across the full range of observed stimulus selectivity and stimulus-induced suppression values , within both the cRF-cRF and sRF-cRF configuration . Figure 5D , E shows the average model predictions based on the model fits from all neurons in the cRF-cRF ( D ) and sRF-cRF ( E ) configurations ( see Figure 5—figure supplement 1 for response fits of individual neurons ) . The model reproduces the way stimulus selectivity and stimulus-induced suppression interacted in both the cRF-cRF and the sRF-cRF configurations: predicting large attention modulation when both selectivity and suppression are strong , but little attention modulation when either selectivity or suppression is low . Thus this single model describes how attention modulates responses to stimuli inside the cRF or the surround . The previous analyses were based on stimulus configurations with two stimuli inside the neurons' receptive field . Importantly , the model also accounts for the neuronal effects of attention to single stimuli at different receptive field locations . This is shown in Figure 6 , which shows how the effect of attention varies when attention is directed to single stimuli at various distances from the receptive field center: for the observed ( A ) and the modeled ( B ) responses . Attention modulation to single stimuli is typically small compared to attention modulation with multiple stimuli . This is because in single stimulus conditions there are no suppressive influences from a flanking stimulus , and we have shown that these suppressive influences are necessary to induce strong attention modulation ( Figure 5 ) . The model accounts for these smaller attention modulations with single stimuli . 10 . 7554/eLife . 17256 . 011Figure 6 . The spatially-tuned normalization model captures how attention modulates responses to single stimuli presented at various receptive field locations . ( A ) Observed responses . Average response as a function of the distance between the single stimulus and the receptive field center , when the stimulus is attended ( white ) or unattended ( black ) . ( B ) Same as ( A ) but for the modeled responses . The responses of each neuron were normalized to the maximum response across conditions in which a single stimulus was presented inside the receptive field . The receptive field distance is given by the Mahalanobis distance from the Gaussian receptive-field center . The Mahalanobis distance is akin to the number of standard deviations ( σ ) away from the receptive-field center . Only neurons whose receptive fields were well fitted with a two-dimensional Gaussian profile ( >80% explained variance; 306 neurons; M1: 95; M2: 211 ) were included . Error bars represent ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 17256 . 011 What determines the spatial tuning of suppression ? For each neuron we sorted the value of the suppression-parameter α , associated with each of the three measured receptive-field locations ( Figure 2A ) , as a function of the distance of each receptive-field location from the neuron's receptive-field center ( Figure 7A ) . Within neurons , locations closest to the receptive-field center induced on average greater suppression than locations furthest away from the receptive-field center ( average α-parameter values of closest vs . furthest location: p=2 × 10−23; M1: p=7 × 10−5; M2: p=0 . 005; paired permutation t-test ) . This is further illustrated in Figure 7B , C ( gray ) , which shows the average normalized α-parameter value as a function of the distance from the receptive-field center . Lower α-values at greater distances are consistent with the observation that surround suppression is significantly weaker than cRF suppression ( Figure 4C , D; see above ) . 10 . 7554/eLife . 17256 . 012Figure 7 . Spatially-tuned excitation and suppression decrease with distance from the receptive-field center , but at different rates . ( A ) Each recording session we measured neuronal responses to stimuli presented at three different receptive-field locations ( Figure 2A ) . The responses of each neuron were fitted with the spatially-tuned normalization model . The value of the suppression parameters α associated with each of the three measured receptive-field locations were ranked according to the proximity of those receptive-field locations to the neuron's receptive-field center: 1 being closest , and 3 being furthest away from the receptive-field center . The suppression parameter values were then normalized by the maximum α-value for each neuron . For each ranking number , the normalized suppression parameter values were subsequently averaged across neurons . Stimulus locations closest to the receptive-field center contributed more suppression to the neurons' response than those furthest away . ( B ) Average normalized suppressive drive ( α , gray ) and excitatory drive ( L , black ) as a function of the distance ( in visual degrees ) of its corresponding receptive-field location from the receptive-field center . The value of the excitatory drive parameter L for stimuli of different orientations were averaged per receptive-field location , and normalized by the maximum excitatory drive across the three measured receptive-field locations of a neuron . ( C ) Same as ( B ) but with an alternative distance measure , namely the Mahalanobis distance , which is akin to the number of standard deviations ( σ ) away from the receptive-field center . In ( B ) and ( C ) , each excitatory-drive value L ( black ) has a corresponding suppressive-drive value α ( gray ) . Error bars represent ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 17256 . 012 In addition to suppression , Figure 7B , C also shows the strength of the excitatory drive ( measured by the average L-parameter values ) as a function of the distance from the receptive-field center ( black ) . Comparing the curves for the excitatory ( black ) and the suppressive drive ( gray ) reveals a striking similarity between the receptive field structure of V4 neurons and that previously observed in primary visual cortex ( V1 ) ( Cavanaugh et al . , 2002b; Sceniak et al . , 1999; DeAngelis et al . , 1994 ) : both excitation and suppression are maximal near the receptive-field center , but excitation is more spatially concentrated , with suppression stretching over larger distances . The spatial pattern of suppression suggests that suppression in the cRF and the surround are continuous extensions of one another . These findings indicate that attention operates uniformly on the spatially-continuous excitation and suppression of a neuron's receptive field . Is a variable top-down attention signal necessary ? We fixed the attention parameter β at a constant value for all neurons , i . e . as the mean β value across neurons when estimated as a free parameter . This constrained model accounted almost as well for the data as the model in which β was free to vary ( median two-fold cross-validated percentage explained variance 86% , compared to 87% for the unconstrained model ) . This finding suggests that differences in attention modulation between neurons are only weakly related to differences in the top-down attention signal across neurons ( see also [Ni et al . , 2012] ) . In contrast , spatial tuning is important . When we instead kept the α terms constant and allowed β to vary across neurons , the model's performance decreased significantly ( 77%; p=0 . 004 , Sequential F-test ) , especially for neurons that were tested in both a cRF-cRF and an sRF-cRF configuration ( 73%; p=0 . 002 ) . Hence when a fixed stimulus is shown , differences in attention modulation across neurons appear to arise when a relatively uniform top-down attention signal interacts with the different amounts of excitation and suppression elicited by that stimulus in each neuron .
We measured the dependency of neuronal attention modulation on stimulus selectivity and stimulus-induced suppression throughout different receptive field regions , including the surround , of V4 neurons . We found that stimulus selectivity and stimulus-induced suppression strongly interact to determine the magnitude of attention modulation in neurons . This interaction determined attention modulation within both the classical receptive field and the surround , indicating that remarkably similar principles drive attention modulation inside the center and surrounding regions of the receptive field . A spatially-tuned normalization model , fitted to the responses of individual neurons , captured the dependency of attention modulation on both stimulus selectivity and suppression , and provided an excellent account of how attention operates across different regions of the receptive field , with either single or multiple stimuli shown inside . Each stimulus configuration induced variable amounts of excitation and suppression in different neurons , depending on the receptive field position of the stimuli . Attention operates on this variable excitation and suppression , thereby explaining why the magnitude of attention-related modulations varies so widely across neurons . Reynolds et al . ( 1999 ) observed a strong correlation between stimulus selectivity and their index of stimulus-induced suppression . It is important to note that this strong correlation is not a general property of spatial summation in visual cortex . Instead , the correlation they observed likely arose from the experimental design used in that study . Specifically , for each neuron Reynolds and colleagues presented different stimulus pairs that fell on two locations that were always chosen to lie well within the cRF at similar distances from the receptive field center . The authors kept the stimulus at one location fixed ( reference stimulus ) , but varied the orientation and color of the stimulus at the second location . But varying the orientation and color of a stimulus , but not its location , varies predominantly the excitation and not the suppression to the neuron ( see Appendix; we found no evidence for orientation-tuned suppression in our data ) . Consequently , both the selectivity and the stimulus-induced suppression index varied with a single variable , the strength of the second stimulus relative to the strength of the fixed stimulus . This explains the strong correlation between stimulus selectivity and suppression in that study . In contrast , we presented stimuli at locations across both the classical and surrounding receptive field , allowing both stimulus-induced excitation and suppression to vary across stimulus configurations . This is the primary reason for the difference between the studies . The variable reference stimulus across neurons in Reynolds et al . , and the slightly different indices used in the two studies will have further amplified the differences in findings in both studies . Hence , under more general stimulus conditions wherein stimuli can fall on any receptive field location , stimulus selectivity and stimulus-induced suppression are not correlated . Importantly , by avoiding a correlation between selectivity and suppression , we were able to examine their separate contributions to the magnitude of attention modulation . Shifting attention between a strong and a weak stimulus , each presented at different receptive-field locations , changes which neuronal inputs are emphasized , thereby causing neuronal-response modulations . However , we show that such clear-cut differential processing only occurs if the weaker stimulus also induces strong suppression: without suppression , attention has no leverage to amplify input differences . The spatially-tuned normalization model captures the dependency of attention-related modulation on both selectivity and suppression , and does so for neurons with stimuli in either the cRF or the surround . Both Sundberg et al . ( 2009 ) and Sanayei et al . ( 2015 ) used conditions with one stimulus inside the classical receptive field and at least one other stimulus in the surround ( sRF-cRF ) . However , neither of these studies used a condition with both stimuli positioned inside the cRF ( cRF-cRF ) . Hence , a direct comparison of attention modulation within the cRF and the surround could not be performed in these studies . This comparison is crucial to determine whether the neuronal effects of attention differ between the cRF and the surround . Importantly , we found that seeing the similarity between attentional effects in both receptive field regions requires examination of the combined relationship , i . e . interaction , between attention modulation and both stimulus selectivity and stimulus-induced suppression ( Figure 4 versus 5 ) . The interaction between these variables was similar in both receptive field configurations . Sanayei et al . fit different ( normalization ) models , but never measured the neuronal effects of attention when attention was directed to the surround stimulus; the authors only compared the effects of attention to the cRF stimulus , with or without surround stimulation , versus attention to a distant stimulus . Attention was never directed to the surround stimuli . Thus , Sanayei et al . lacked the crucial information needed to examine how surround suppression affects attention modulation and to test the efficacy of normalization models . We tested whether a single model could fit both the cRF-cRF and sRF-cRF data . The good fits of the spatially-tuned normalization model to the data obtained in both receptive field configurations provided further evidence that attention acts similarly inside the cRF and the surround . Suppression and excitation may rely on distinct mechanisms in different regions of the receptive field ( Angelucci et al . , 2014 ) . Our data do not pertain to these different mechanisms and we may have missed some small differences in attention modulation associated with these distinct mechanisms . Nonetheless , our findings show that the way attention interacts with excitation and suppression across different regions of the receptive field is remarkably similar . In recent years , several normalization models of attention have been proposed ( Reynolds et al . , 1999; Ghose and Maunsell , 2008; Ni et al . , 2012; Ghose , 2009; Reynolds and Heeger , 2009; Lee , 2009; Boynton , 2009; Lee et al . , 1999 ) . Two of the more elaborate models explicitly assumed that attention acts on a specific receptive-field structure , namely one that encompasses a relatively narrow excitatory field in addition to a wider suppressive field ( Ghose , 2009; Reynolds and Heeger , 2009 ) . This receptive-field structure is based on findings from primary visual cortex ( V1 ) ( Cavanaugh et al . , 2002b; Sceniak et al . , 1999; DeAngelis et al . , 1994 ) . It is important to note , however , that none of these studies empirically tested if attention actually operates on the spatially-varying excitation and suppression implied by such a receptive field structure . We started with a spatially-tuned normalization model that made no assumptions about the structure of excitation and suppression in the receptive field of V4 neurons . Furthermore , and in contrast to the earlier-mentioned studies ( Ghose , 2009; Reynolds and Heeger , 2009 ) , we explicitly fitted models to the responses of individual neurons to test relationships with the underlying receptive field structure . Interestingly , this naive model reveals that the receptive field organization of V4 neurons strongly resembles that of V1 neurons ( Cavanaugh et al . , 2002b; Sceniak et al . , 1999; DeAngelis et al . , 1994 ) : both excitation and suppression are maximal near the receptive-field center , but excitation is more spatially concentrated , while suppression stretches over larger distances . These findings suggest that similar receptive field organizations can be found throughout different stages of visual cortex . Importantly , our findings show that attention operates uniformly across the spatially-varying excitation and suppression of a receptive field: throughout the receptive field , including the surround , attention-related modulations of neuronal responses is governed by very similar normalization rules . The finding that the rules of neuronal attention modulation are similar across different regions of the receptive field simplifies our view of attentional operations in visual cortex , and provides strong support for normalization models of attention ( Ghose , 2009; Reynolds and Heeger , 2009 ) . We also show that the origin of the stimulus-induced excitation is not important for determining the magnitude of attention modulation: we found no distinction between excitation related to a neuron's feature tuning ( i . e . orientation tuning ) or spatial tuning ( i . e . receptive field ) . What matters for neuronal attention modulation is stimulus-induced excitation , regardless of its origin , in conjunction with spatially-tuned suppression . It follows that when a particular stimulus configuration induces variable amounts of excitation and suppression in different neurons , attention-related modulations will vary across these neurons . The fact that attention operates on the spatially-varying excitation and suppression of a receptive field has important implications , as it determines which neurons will be most influenced by attention . For instance , with a given number of stimuli presented inside the receptive field , attention to a preferred stimulus shown inside the center of the receptive field typically has the greatest potential to elevate neuronal responses . This is true not only because stimuli near the receptive field center generally elicit most excitation , but also because such stimuli are most likely to induce the greatest suppression . The elevated suppression by center stimuli gives them more weight in normalization mechanisms as it allows them to better discount the suppressive influences from other simultaneously presented stimuli . Similarly , attention to a weak stimulus inside the receptive field center will in general reduce responses more than attention to a weak stimulus elsewhere in the receptive field , including the surround . This does not mean that stimuli in the surround , which induce relatively less suppression , have little impact on attention modulation . Indeed , because the surround is so much larger than the cRF it can contribute considerable suppression . Such strong surround suppression likely occurs under natural viewing conditions where stimuli are shown throughout the visual field , many of them covering the surround ( Vinje and Gallant , 2000; Ozeki et al . , 2009; Haider et al . , 2010; Coen-Cagli et al . , 2015 ) . The normalization model predicts that such strong surround suppression may robustly amplify attention modulation , much beyond the attention modulation observed without surround suppression . This effect is illustrated in Figure 8 in which spatial attention was applied to model neurons with ( upper panels Figure 8 ) or without a surround ( lower panels Figure 8 ) . The model neurons with a surround strongly modulated their responses by attention , but those without a surround much less ( Figure 8F upper vs . lower panel ) . Hence , although the precise role of the surround is still unknown ( Schwartz and Simoncelli , 2001; Vinje and Gallant , 2000; Sachdev et al . , 2012 ) , an important contribution of the surround may lie in its ability to amplify attention-related response modulations . 10 . 7554/eLife . 17256 . 013Figure 8 . The surround may amplify spatial attention under natural viewing conditions . ( A ) Original image . ( B ) Model neurons tiled the image . Each pixel contained one model neuron with its receptive-field centered on that pixel . An example cRFs ( solid red ) and surround ( red dashed ) for one neuron are shown . The radius of each neuron's surround was approximately five times larger than the radius of its cRF . The model neurons computed local contrast within the excitatory and suppressive component of their receptive field . The response maps show each neuron's response: neurons near high-contrast regions responded most as indicated by the luminance of the pixels . ( C ) Original image scaled according to the response map in ( B ) . ( D ) Attention was directed to the left eye . Attention weighed the excitatory and suppressive inputs with its Gaussian kernel , resulting in stronger responses of the neurons with receptive fields near the attended location relative to neurons with receptive fields outside the locus of attention . ( E ) Original image scaled according to the response map in ( D ) , illustrating the way attention changes the visual representation . ( F ) Attention modulation of each neuron , defined as ( responseAtt - response ) / ( responseAtt + response ) . Here , response is the response map without attention as in ( B ) , while responseAtt is the response map with attention as in ( D ) . Upper panels ( B–F ) are based on model neurons with a suppressive surround . Lower panels ( B–F ) are based on model neurons without a suppressive surround , but with the same amount of suppression inside the cRF as the neurons with a suppressive surround . DOI: http://dx . doi . org/10 . 7554/eLife . 17256 . 013
Two male rhesus monkeys M1 and M2 ( Macaca mulatta , both 9 kg ) were trained to perform a spatial attention task . Monkeys were pair housed in standard 12:12 light-dark cycle and given food ad libitum . Before training , each animal was implanted with a head post . After completion of the behavioral training ( ~7 months ) , we implanted a 10 × 10 array of microelectrodes into area V4 of the left cerebral hemisphere , between the lunate sulcus and the superior temporal sulcus . Before surgery , animals were given buprenorphine ( 0 . 005 mg/kg , intramuscular ) and flunixin ( 1 . 0 mg/kg , intramuscular ) as analgesics , and a prophylactic dose of an antibiotic ( Baytril , 5 mg/kg , intramuscular ) . For surgery , animals were sedated with ketamine ( 15 mg/kg , intramuscular ) and xylazine ( 2 mg/kg , intramuscular ) and given atropine ( 0 . 05 mg/kg , intramuscular ) to reduce salivation . Anesthesia was maintained with 1–2% isoflurane . Antibiotic was administered again 1 . 5 hr into surgery; buprenorphine and flunixin were given for 48 hr post-operatively . All procedures were approved by the Institutional Animal Care and Use Committee of Harvard Medical School ( Boston , MA; protocol #04214 ) . Stimuli were presented on a gamma-corrected cathode-ray tube ( CRT ) display with a 100 Hz frame rate and a resolution of 1024 × 768 pixels . Monkeys were seated 57 cm from the center of the screen . Stimuli consisted of full-contrast achromatic odd-symmetric static Gabor stimuli ( 0 . 6–2 . 2 cycles per degree; one spatial frequency per daily session ) presented on a gray background ( 42 cd/m [Kastner and Ungerleider , 2000] ) and were rendered online using custom-written software ( https://github . com/MaunsellLab/Lablib-Public-05-July-2016 . git ) . The Gabor stimuli were truncated at three SD from their center . We trained monkeys to perform a visual detection task in which spatial attention was manipulated ( Figure 1A ) . The trial started when the monkey fixated a small spot in a virtual 1 . 5° square fixation window in the center of the video display for 240–700 ms . Eye movements were tracked using an infrared eye-tracking camera ( EyeLink 1000 ) sampling binocularly at 500 Hz . The duration of the fixation period was randomly drawn from a uniform distribution . Following fixation a sequence of stimuli was presented , in which each stimulus presentation lasted 200 ms and was separated from other stimuli by 200–1020 ms interstimulus intervals ( Figure 1B ) . The durations of the interstimulus intervals were randomly drawn from an exponential distribution ( τ = 200 ms ) . During the interstimulus interval only a gray screen with the fixation dot was shown . The stimulus presentations were short to prevent animals from adjusting their attention within a stimulus presentation in response to the number of stimuli presented ( Lee and Maunsell , 2010; Ni et al . , 2012; Lee , 2009; Williford and Maunsell , 2006 ) . On each trial , stimuli appeared at two locations near the receptive fields of neurons , but the two locations differed between blocks of trials . One stimulus location ( the middle location: location 1 in Figure 2A ) never varied , but in different blocks of trials the second stimulus location was shifted either clockwise ( location 2 in Figure 2A ) or counterclockwise ( location 3 in Figure 2A ) . For the example session in Figure 2A , the possible stimulus-location pairings were 1+2 and 1+3 . All stimulus locations were equidistant from the fixation point , and stimulus locations 2 and 3 were equidistant from stimulus location 1 . The two different pairs of stimulus locations assured that many neurons were tested in both receptive-field configurations ( cRF-cRF and sRF-cRF ) . On each stimulus presentation within a trial , we presented one , two , or no stimuli at the two stimulus locations near the neurons' receptive fields . The stimuli could be of one of two orthogonal orientations . Each session , the stimulus orientation was optimized for a randomly selected unit , so that different orientations were used across sessions . A representative set of nine possible stimulus combinations ( for a particular orientation pair ) is shown in Figure 2B . Using these different stimulus combinations we could measure stimulus selectivity , stimulus-induced suppression and attention modulation . Each stimulus location near the neurons' receptive fields ( stimulus location 1 , 2 , 3 in Figure 2A ) had a corresponding and equally eccentric stimulus location on the opposite side of the fixation point ( e . g . stimuli near Away in Figure 2C , D ) . As outlined below , we instructed monkeys to direct their attention to one stimulus location , either near or away from the receptive fields . This way we could measure not only how attention modulated neuronal responses when directed to different stimuli near the neurons' receptive fields , but also measure stimulus selectivity and stimulus-induced suppression with attention directed away from the neurons' receptive fields . On each stimulus presentation ( of multiple in a trial ) , each stimulus location was equally likely to contain one orientation , the other orientation or no stimulus . When Gabor pairs were presented near the neurons' receptive fields , their centers were separated by a median of 2 . 3° ( range: 1 . 6–4 . 8° ) , and always separated by at least six Gabor standard deviations ( mean Gabor σ: 0 . 45°; range: 0 . 17–0 . 5° ) . With such inter-stimulus distances , two stimuli can be presented within the receptive fields of V4 neurons . Subjects were required to detect a faint white spot , labeled Target in Figure 1A , B . The target appeared at one of the four stimulus locations ( two near the neurons' receptive field and two counterparts on the opposite side of the fixation point; see above ) during one stimulus presentation within a trial . The target never appeared on the first stimulus presentation of a trial , but could occur with equal probability on any other stimulus presentation ( range: 2–8 ) . Two to five percent of the trials contained no target and the monkey was rewarded for maintaining fixation . Targets were presented in the center of Gabor stimuli to encourage the monkeys to confine their attention to a restricted part of visual space , near the cued stimulus location . Task difficulty was manipulated by varying the target strength , defined as the opacity of the target ( range of alpha-transparency values: 0 . 06–0 . 28 ) . Each session we used six different target strengths ( Figure 1B , C ) . The monkey was rewarded with a drop of juice for making a saccade to the target location within 350 ms of its appearance . Attention was cued to one location in blocks of ~150 trials . Before the start of each block the monkey performed three to five instruction trials in which stimuli were presented at a single ( cued ) location . The instruction trials cued the monkey to attend to that location during subsequent trials in which stimuli could occur at all four locations . Within a block of trials , the target appeared at the cued location in 91% of the trials ( valid trials; position of the black circle in Figure 1A ) . In the remaining 9% of the trials ( invalid trials ) the target appeared at one of the three other ( uncued ) stimulus locations , with equal probability ( position of the yellow and blue circles in Figure 1A ) . We used a single target strength for the invalid trials , as this allowed us to obtain reliable estimates of behavior at the unattended locations despite the small number of invalid trials ( Figure 1B , C ) ( Cohen and Maunsell , 2009 ) . Using invalid trials , we could compare performance between attended and unattended locations . We recorded neuronal activity using a 10 × 10 array of microelectrodes ( Blackrock Microsystems; impedances: 0 . 3–1 . 2 MΩ at 1 kHz; 1 mm long electrodes; 0 . 4 mm between adjacent electrodes ) , chronically implanted into area V4 of the left cerebral hemisphere of each monkey . The data presented here are from 130 daily sessions of recording ( Monkey M1: 52; Monkey M2: 78 ) . At the beginning of each recording session , we mapped the receptive fields and optimized stimulus parameters ( position , orientation ) for a randomly selected unit . We first measured the orientation-tuning curve of each neuron using a large Gabor that covered the lower right visual field . Orientation tuning was measured using Gabors of 8 different orientations spanning 180° . We then mapped the spatial receptive field of each neuron using a Gabor with the preferred orientation of the selected unit , and a Gabor with an orientation orthogonal to that neuron's preferred orientation . Using two orthogonal orientations assured that most neurons were responsive to at least one stimulus . The same two orientations were subsequently used during the attention task . Receptive-field mapping was computer controlled and used a full-contrast static Gabor with the two orthogonal orientations ( spatial frequency: 1 . 1 cycles per degree; Gabor sigma: 0 . 3° ) on an 8 by 8 grid of positions ( ~azimuth range: −1 to 8°; ~elevation range: 2 . 5 to −8° ) . The center-of-mass of the receptive field ( unfitted data ) was defined as the receptive-field center . The receptive field plots in Figure 3 are based on the linearly-interpolated spike counts measured at each grid location . The spike counts were obtained within a 200 ms window starting 50 ms after stimulus onset . Action potential waveforms were sorted offline using spike-sorting software ( Offline Sorter , Plexon Inc ) . Waveforms for which the first two principal-component scores formed a well-defined cluster , separate from other waveforms , were classified as single units . The receptive fields of the units were located in the lower right quadrant at an average eccentricity of 3° for monkey M1 and 4° for monkey M2 . We included only neuronal data from stimulus presentations from correct , validly-cued trials . We excluded incorrect trials , invalidly-cued trials , instruction trials , trials with no target , the first stimulus presentation of a trial ( on which no target could occur ) , and stimulus presentations with a target . Neurons were included in the analyses if they responded significantly above baseline to any single Gabor presented at any stimulus location in the attend away condition ( ANOVA; α=0 . 05 ) . Responses in the attend away condition were obtained by averaging the firing rates from the conditions in which attention was directed to either of the two stimulus locations furthest away from the receptive-field center of the neuron ( Away in Figure 2C , D ) . The small subset of neurons whose responses where significantly suppressed below baseline by all stimuli ( N=13 ) was not further analyzed . Neuronal responses were computed based on the spikes in the interval from 50 ms to 300 ms after stimulus onset . Similar results were obtained using different intervals . A stimulus location was considered within the classical receptive field ( cRF ) if the neuron responded significantly to any single stimulus ( of either orientation ) presented at that location , measured with attention away from the neuron's receptive field ( attend away ) . The median distance from the receptive-field center of a stimulus inside the cRF was 1 . 7° ( interquartile range 1 . 2° to 2 . 5° or 0 . 7–1 . 5σ , where σ is the Mahalanobis distance from those neurons whose receptive fields were well fitted with a bivariate Gaussian function: >80% explained variance , N=306 neurons ) . A stimulus location was considered to be within the surround of a neuron if the neuron did not increase its firing rate significantly to any single stimulus ( of either orientation; N>36 trials per stimulus ) presented at that location , measured with attention away from the neuron's receptive field ( attend away ) . Note that neurons for which a surround location was measured did respond significantly to at least one of the stimuli when it was presented inside the cRF instead of the surround ( Figure 2—figure supplement 1 ) . The median distance of a surround location from the receptive-field center was 3 . 5° ( interquartile range 2 . 7° to 4 . 3° , or 1 . 9σ to 3 . 1σ ) . We obtained similar results when we additionally required surround positions to lie at more than 2 . 5σ from the receptive field center . We also recorded from neurons with two stimuli inside their surround , i . e . sRF-sRF configuration . These data were not further analyzed due to a lack of responses . The peristimulus time histograms ( PSTHs ) in Figure 3A–D were computed by counting the number of spikes in 1 ms bins and smoothing with a Gaussian filter of σ = 5 ms . A selectivity index was computed based on the responses to the component Gabors of each Gabor pair ( four pairs in Figure 2B ) . Selectivity indices were computed for each Gabor pair presented at each pair of stimulus locations ( stimulus locations 1+2 or 1+3 in Figure 2A ) . We thus obtained eight selectivity indices per neuron . The selectivity index is defined as ( P - N ) / ( P + N ) . Here , P ( preferred ) and N ( non-preferred ) are the neuronal responses to the strongest and weakest component Gabor of a Gabor pair when presented alone . P and N were measured with attention away from the neurons' receptive fields ( attend away ) . The upper pictogram in Figure 2E illustrates the computation of the selectivity index for one Gabor pair . By definition the neuron does not respond to the stimulus when it appears alone inside the surround . It follows that for the sRF-cRF receptive-field configuration , the surround Gabor is always assigned as non-preferred ( N ) and the cRF Gabor as preferred ( P ) . Stimulus-suppression indices were similarly computed for each of eight possible Gabor pairs as ( P - PN ) / ( P + PN ) , where P is the response to the preferred Gabor of a Gabor pair as described before , and PN is the response to the Gabor pair . Both P and PN were measured with attention away from the neurons' receptive fields ( attend away ) . The middle pictogram in Figure 2E illustrates the computation of the stimulus-induced suppression index for one Gabor pair . We obtained similar results when defining a suppression index as ( P+N-PN ) / ( P+N+PN ) . Note that the stimulus-induced suppression index is distinct from the α terms in the model . This is because the stimulus-induced suppression index is based on the observed neuronal responses , which comprise both an α and L term ( i . e . response = L/ ( α+σ ) ) . In terms of the model parameters , the stimulus-induced suppression index is given by:Stimulus−induced\ suppression index=responseP−responseP+NresponseP+responseP+N=LPαP+σ−LP+LNαP+αN+σLPαP+σ+LP+LNαP+αN+σ , where LP is the excitatory drive from the preferred component Gabor of a Gabor pair , LN is the excitatory drive from the non-preferred component Gabor , αP is the suppressive drive from the preferred component Gabor , and αN is the suppressive drive from the non-preferred component Gabor . So the stimulus-induced suppression index depends on both the excitatory and suppressive drive from the stimulus . Attention-modulation indices were computed for each of eight possible Gabor pairs as ( PAttN - PNAtt ) / ( PAttN + PNAtt ) , where PAttN is the neuronal response to the Gabor pair with attention directed to P , PNAtt is the neuronal response to the Gabor pair with attention directed to N . The lower pictogram in Figure 2E illustrates the computation of the attention-modulation index for one Gabor pair . All Gabor pairs for which a neuron responded on average with at least 1 spike ( in the 250 ms window ) in the attend away condition were further analyzed , but similar results were obtained using other criteria . This way neuronal data from 728 neurons were analyzed ( monkey M1: 264; M2: 464 ) . In Figures 4 , 5 selectivity and stimulus-induced suppression indices are computed for each neuron and all Gabor pairs , so neurons contribute more than one index . Due to the chronic nature of our recordings , it is likely that some neurons were resampled across days . Because we adjusted the values of the stimulus orientations and locations each day for a randomly selected unit , resampling rarely involved identical stimulus configurations . Similar results were obtained for both monkeys ( see Results ) . We used multiple linear regression to examine if attention-related modulation depends on the interaction between stimulus selectivity and stimulus-induced suppression . For both RF configurations , the model included a main effect of selectivity , supplemented with a main effect of stimulus-induced suppression . The model also included an interactive product term , which measured the dependency of attention-related modulation on both selectivity and stimulus-induced suppression . The regression model is given by:attentionmodulation=selectivity⋅β1+suppression⋅β2+selectivity⋅suppression⋅β3+error In this model , the main effect of selectivity measures the contribution of selectivity to attention modulation given that suppression is zero . Similarly , the main effect of suppression measures the contribution of suppression to attention modulation given that selectivity is zero . For example , if suppression is zero , the suppression and the interaction term ( selectivity × suppression × β3 ) both go to zero , leaving only the selectivity term β1 , which specifies the contribution of selectivity to attention modulation . Conversely , if selectivity is zero , the only non-zero term is the β2 term , which specifies the contribution of suppression to attention modulation . Thus the main effects are not estimated from a particular selection or a subset of the dataset , but follow mathematically from the linear regression model with interaction term . For the Bayesian regression analysis , we compared the marginal likelihood of the data given a regression model that does not include receptive field configuration as a factor to the marginal likelihood of the data given a regression model that does include receptive field configuration as a factor ( i . e . the Bayes factor , using the lmBF function from the BayesFactor package in R [Rouder et al . , 2012] ) . The plots in Figure 5A , B , D , E were obtained using regularized bilinear interpolation on the observed or modeled attention-modulation indices from all Gabor pairs and all neurons ( Surface Fitting using gridfit ( http://www . mathworks . com/matlabcentral/fileexchange/8998 ) , MATLAB Central File Exchange ) . Tuned normalization has been applied before ( Carandini et al . , 1997; Schwartz and Simoncelli , 2001; Lee et al . , 1999; Rust et al . , 2006 ) and has been used to explain neuronal-response modulation when attention is shifted between two stimuli with different motion directions inside the cRF of MT neurons ( Ni et al . , 2012 ) . The spatially-tuned normalization model is described by Equation ( 1 ) . This spatially-tuned normalization model was fitted to the neuronal responses of all 728 neurons used in the analyses . The model parameters are: L11 , L12 , L21 , L22 , L31 , L32 , α2 , α3 , σ , β . Specifically , Lij ( adopted from linear response ) is the excitatory drive from a stimulus of orientation j ( j = 1 , 2 ) at receptive-field location i ( i = 1 , 2 , 3 ) . αp is the suppression parameter associated with stimulated receptive-field location p=1 , 2 , 3 ) . β adds attention to the model and is multiplied with the parameters associated with the attended location ( L and α; see Results and Figure 5C ) . In the conditions with attention directed away from the receptive fields ( attend away ) β = 1 ( see Equation ( 1 ) ) . σ is the semi-saturation constant that is fixed across conditions and serves as a baseline suppression parameter . The σ parameter was introduced in Heeger's ( Heeger , 1992 ) original divisive normalization model to model the shape of contrast-response functions of neurons in primary visual cortex . It also stabilized the response when stimuli of low ( or zero ) contrast are presented by preventing division by zero . For our data , which involve only high-contrast stimuli , it represents baseline suppression , which may arise from spontaneously active inhibitory neurons or suppression caused by constant stimuli ( e . g . the edge of the stimulus display ) visible to the monkey . It is an important parameter to accommodate the neuronal effects of attention to single stimuli inside the receptive field ( see Figure 5—figure supplement 1A vs . B: P vs . Patt ) . The median value of the sigma parameter was 0 . 06 ( median absolute deviation ( MAD ) : 0 . 26 ) . A model with no sigma term performs significantly worse at explaining responses to isolated attended stimuli ( median two-fold cross-validated percentage explained variance 84% , compared to 87% for the model with sigma term; p=0 . 006; sequential F-test ) . We also fit for each neuron a model with only one free excitatory ( L ) term capturing excitation across all stimulus conditions . This model with one L term performed significantly worse at explaining neuronal responses than the full model with all L-terms ( median two-fold cross-validated percentage explained variance 48% , compared to 87% for the model with all L terms; p<0 . 0001; sequential F-test ) . We tested two pairs of receptive-field locations ( see Spatial attention task and Figure 2A ) . α1 is the suppression parameter related to the receptive-field location common to both of the receptive-field location pairs ( stimulus location 1 for the example session in Figure 2A ) , and is set to one to constrain the model . All parameters were constrained to be nonnegative . For each neuron , the model was fitted to the neuronal responses of 36 distinct attention and stimulus combinations by minimizing the sum of squared error using a simplex optimization algorithm ( MATLAB fminsearch . m; MathWorks ) . The goodness-of-fit of the model was calculated for each neuron as the percentage explained variance , which was determined by taking the square of the correlation coefficient between the model-predicted responses and the observed neuronal responses across all stimulus conditions . The explained variance was calculated using the average neuronal responses from trials not used to fit the model . For this purpose , we employed two-fold cross-validation , fitting the model based on half of the randomly-chosen trials , and using the remaining data to measure the goodness of fit of the model . This procedure was repeated five times , each time using a different random draw , and subsequently averaged across all cross-validations to produce the reported goodness-of-fit . In Figure 8 , each image pixel contained one model neuron with its receptive field centered on that pixel . The neurons' receptive fields consisted of an excitatory and suppressive receptive field . These receptive fields were modeled as a circular two-dimensional Gaussian with a standard deviation of eight pixels for the excitatory field and 40 pixels for the suppressive field . The model neurons computed local contrast within their excitatory and suppressive receptive field . The excitatory input was divisively normalized by the suppressive input to generate the model neuron's response . Model neurons without a surround experienced no suppression from stimuli positioned outside their classical receptive field . The classical receptive field was defined as all pixels within 16 pixels , i . e . two standard deviations from the excitatory receptive field , from the receptive-field center . Attention was modeled as a circular two-dimensional Gaussian kernel with a standard deviation of five pixels and amplitude equal to six . Attention weighed the excitatory and suppressive inputs with its kernel , resulting in stronger responses of the model neurons near the locus of attention relative to model neurons outside the locus of attention . | At any moment , our brain receives an enormous amount of information from our senses . However , we are not aware of all of this information; only the information we decide to focus on is perceived in detail . This ability to focus our attention is important for survival . The neurons involved in vision respond best to information that comes from a small ‘window’ in what is being seen . When something appears in this window ( known as the neuron’s receptive field ) , the activity of the neuron either increases or decreases . How does focusing attention on an object change the neuron’s response ? Verhoef and Maunsell investigated this question by recording electrical activity in an area of the brain called V4 in monkeys as they focused their attention on objects in different locations of the neuron’s receptive field . The recordings show that a single rule determines when attention influences a neuron’s activity . If an object inside the neuron’s receptive field decreases the activity of the neuron , then attention can change that neuron’s activity . Attention then changes the activity of the neuron by either removing or further boosting the influence of these objects . Verhoef and Maunsell then developed a mathematical model based on these results , and found that the model could explain why the activity of a neuron changes when attention is focused on objects at different locations in its receptive field . The next step is to understand exactly how the brain works to either remove or boost the influence of an object that causes a neuron’s activity to decrease . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | Attention operates uniformly throughout the classical receptive field and the surround |
The process of cell fate determination has been depicted intuitively as cells travelling and resting on a rugged landscape , which has been probed by various theoretical studies . However , few studies have experimentally demonstrated how underlying gene regulatory networks shape the landscape and hence orchestrate cellular decision-making in the presence of both signal and noise . Here we tested different topologies and verified a synthetic gene circuit with mutual inhibition and auto-activations to be quadrastable , which enables direct study of quadruple cell fate determination on an engineered landscape . We show that cells indeed gravitate towards local minima and signal inductions dictate cell fates through modulating the shape of the multistable landscape . Experiments , guided by model predictions , reveal that sequential inductions generate distinct cell fates by changing landscape in sequence and hence navigating cells to different final states . This work provides a synthetic biology framework to approach cell fate determination and suggests a landscape-based explanation of fixed induction sequences for targeted differentiation .
Multistability is a mechanism that cells use to achieve a discrete number of mutually exclusive states in response to environmental inputs , such as the lysis/lysogeny switch of phage lambda ( Arkin et al , 1998; Oppenheim et al . , 2005 ) and sporulation/competence in Bacillus subtilis ( Süel et al . , 2006; Schultz et al . , 2009 ) . In multicellular organisms , multistable switches are also common in the cellular decision-making including the regulation of cell-cycle oscillator during cell mitosis ( Pomerening et al . , 2003 ) , Epithelial-to-Mesenchymal transition and cancer metastasis ( Jolly et al . , 2016; Lee et al . , 2014a ) , and the well-known cell differentiation process , which is a manifestation of cellular state determination in a multistable system ( Laurent and Kellershohn , 1999; Guantes and Poyatos , 2008 ) . However , loss of multistability can drive cells to acquire metastatic characteristics and stabilize highly proliferative , pathogenic cellular states in cancer ( Lee et al . , 2014b ) . C . H . Waddington hypothesized the ‘epigenetic landscape’ to explain canalization and fate determination mechanism during cell differentiation ( Waddington , 1957 ) . In this hypothesis , differentiation is depicted as a marble rolling down a landscape with multiple bifurcating valleys and eventually settles at one of the local minima , corresponding to terminally differentiated cells . More recent theoretical studies further proposed the local minima to be modeled as steady states or attractors of dynamical systems , which can be mathematically described using differential equations ( Zhang and Wolynes , 2014; Li and Wang , 2013a ) . As such , cell differentiation can be interpreted as a state transition process on a multistable dynamic system . A myriad of theoretical analysis have investigated the functioning of such systems and quantified the Waddington landscape and developmental paths through computation of the probability landscape for the underlying gene regulatory networks ( Li and Wang , 2013a; Wang et al . , 2011; Li and Wang , 2013b; Ferrell , 2012; Bhattacharya et al . , 2011; Macarthur et al . , 2009; Huang et al . , 2007 ) . Recent studies also revealed that the potential landscape and the corresponding curl flux are crucial for determining the robustness and global dynamics of non-equilibrium biological networks ( Wang , 2015; Xu et al . , 2014; Wang et al . , 2008 ) . Furthermore , the multiple stable steady states have been predicted beyond the bistable switches with or without epigenetic effects , which is reflected in slow timescales ( Wang , 2015; Xu et al . , 2014; Li and Wang , 2013b; Feng and Wang , 2012; Wang et al . , 2011; Feng et al . , 2011 ) . Experimental researches , however , mostly focus on bistable switches , involving transitions between only two states . And demonstrations , from a combination of experiments and computational modeling , for the existence and operation of such a landscape in a higher dimensional multistable system are still lacking . Moreover , it remains unknown how gene regulatory networks ( GRNs ) , gene expression noise , and signal induction together shape the attractor landscape and determine a cell’s developmental trajectory to its final fates ( Schmiedel et al . , 2015; Tanouchi et al . , 2015; Prindle et al . , 2014; Chalancon et al . , 2012; Murphy et al . , 2010; Balázsi et al . , 2011; Kramer and Fussenegger , 2005; Bennett et al . , 2008; Maamar et al . , 2007 ) . Complex contextual connections of GRNs have impeded experimentally establishing the shape and function of the cell fate landscape . Rationally designed and tunable synthetic multistable gene networks in E . coli , however , could form well-characterized attractor landscapes to enable close experimental investigations of general principles of GRN regulated cellular state transitions . Since the functioning of these principles only requires the most fundamental aspects of gene expression regulation , they would also be applicable for cell differentiation regulations in mammalian cells . Here , we combine mathematical theory , numerical simulations , and synthetic biology to probe all possible sub-networks of mutually inhibitory network with positive autoregulations ( MINPA , Figure 1A ) , which has been hypothesized to have multistability potentials ( Guantes and Poyatos , 2008; Huang et al . , 2007 ) . Moreover , MINPA and its sub-networks are recurring motifs enriched in GRNs regulating hematopoietic development ( Gata1-Pu . 1 , [Graf and Enver , 2009] ) , trophectoderm differentiation ( Oct3/4-Cdx2 , [Niwa et al . , 2005] ) , endoderm formation ( Gata6-Nanog , [Bessonnard et al . , 2014; Li and Wang , 2013a] ) , and bone , cartilage , and fat differentiation ( RUNX2-SOX9-PPAR-γ , [MacArthur et al . , 2008; Rabajante and Babierra , 2015] ) . 10 . 7554/eLife . 23702 . 003Figure 1 . Conceptual and experimental design of MINPA and its sub-networks . ( A ) Abstract diagram of MINPA topology , where X and Y mutually inhibit ( T-bars ) each other and auto-activate ( arrowheads ) itself . Four inducers to regulate the four color-coded regulatory edges are also listed . ( B ) Molecular implementation of the MINPA network . Para/lac ( purple arrow ) is activated by AraC ( yellow ) and repressed by LacI ( light green ) , while Plux/tet ( cyan arrow ) is activated by LuxR ( blue ) and repressed by TetR ( red ) . Arabinose and AHL ( oval ) can induce AraC and LuxR activation , respectively . IPTG and aTc ( hexagon ) can respectively relieve LacI and TetR inhibition . GFP and mCherry serve as the readout of Para/lac and Plux/tet , respectively . Therefore , TetR and AraC collectively form the node X in ( A ) , color-coded as purple rectangle . Similarly , LuxR and LacI collectively form the node Y in ( A ) , color-coded as cyan rectangle . Genes , promoters and regulations are color-coded corresponding to the topology in ( A ) . ( C ) List of MINPA and its 14 sub-networks . Numbering of indices is converted from topologies’ binary name ( see Figure 1—figure supplement 1E for more details ) . T represents ‘topology’ . R represents ‘repression’ , and A represents ‘autoactivation’ . Superscript is used to describe the number of such types of edges . Topologies with shaded background were later constructed and analyzed experimentally . ( D–E ) Dynamic responses for Para/lac ( D ) and Plux/tet ( E ) through induction with Arabinose ( Ara ) and IPTG , and AHL and aTc , respectively . Presented data was the mean value of three replicates . mCherry and GFP serves as the readout of the two promoters . DOI: http://dx . doi . org/10 . 7554/eLife . 23702 . 00310 . 7554/eLife . 23702 . 004Figure 1—figure supplement 1 . Experimental design , topological hierarchy and multistability probability analysis of MINPA sub-networks . ( A ) Abstract diagrams and molecular implementation of the eight MINPA sub-network topologies , including tunable positive feedbacks ( T6 and T9 ) , mutual inhibition ( T5 ) , dual-positive feedbacks ( T10 ) , and their combinations ( T7 , T11 , T13 , and T14 ) . Genes , promoters , and regulations are color-coded corresponding to the topology on the left side . ( B–C ) Biological devices for testing promoter Para/lac ( B ) and Plux/tet ( C ) , respectively . Fluorescence was measured by flow cytometry at 12 hr and 24 hr ( not shown ) after adding the inducers . All the data points were averaged from three repeated experiments . Grey arrows represent constitutive promoters ( BBa_K176009 ) . ( D ) Principle components and optimal Hill functions fitted for hybrid promoter Para/lac and Plux/tet . ( E ) List of MINPA and its 14 sub-networks , which are named , and color-coded , as binary numbers based on the existence of regulatory edges . ( F ) Topological hierarchy of the 15 motifs . ( G ) Probability of multistability ( tristability and quadrastability ) for the nine experimentally constructed networks . SSS: stable steady states . The range for activation/repression strength is [0 . 3 , 0 . 8] , and the probability was calculated from the number of parameter sets giving rise to multistability over 2000-repeated computational simulations . ( H ) Histograms for parameter combination that have multistability in sub-networks of T10 , T13 , T14 and T15 ( MINPA ) . For each sub-network , the network activation and inhibition strength Sa , Su , St and Sl are randomly selected from given ranges and the number of corresponding SSS is recorded . We repeat this procedure for all sub-networks for 2000 times and calculate the probabilities of bistability ( blue lines ) , tristability ( black lines ) and quadrastability ( red lines ) . ( I ) Scatter plots for all parameter combinations that have multistability in MINPA . We generate parameters using the same approach as ( H ) and select the parameter combinations that can generate three or four SSS in MINPA . DOI: http://dx . doi . org/10 . 7554/eLife . 23702 . 004
Engineered circuits of MINPA ( Figure 1B ) and its sub-networks ( Figure 1—figure supplement 1A ) are designed to use two hybrid promoters , Para/lac and Plux/tet , which are characterized experimentally to show small leakage and high nonlinearity ( Figure 1D–E and Figure 1—figure supplement 1B–D ) . For MINPA topology , hybrid promoter Para/lac drives AraC and TetR expression , representing the node X in Figure 1A , whereas Plux/tet controls LuxR and LacI transcription , representing the node Y . AraC and LuxR activate Para/lac and Plux/tet in the presence of Arabinose and AHL ( 3oxo-C6-HSL ) respectively , forming positive autoregulations . IPTG inhibits the repressive effect of LacI on TetR expression , while aTc counteracts TetR repression on LacI . Hence , the two nodes form the topology presented in the conceptual design shown in Figure 1A . Green fluorescent protein ( GFP ) and mCherry serve as the corresponding readouts of Plux/tet and Para/lac activities in living cells ( Figure 1B ) . Topologies of MINPA and all its subnetworks can be divided into four layers , from one- to four-dimensional networks based on the number of regulatory edges ( Figure 1C and Figure 1—figure supplement 1E–F ) and further categorized into nine groups based on the configurations of activation and inhibition . By computationally searching a large parameter range for each of the nontrivial networks ( Faucon et al . , 2014 ) , we found that networks with two auto-activations , including A2 , RA2 , R2A2 , have high probability of tristability or quadrastability ( Figure 1—figure supplement 1G ) , defined as having three or four stable steady sates ( SSS ) under a common induction condition . However , MINPA has broader parameter distributions than the other two ( Figure 1—figure supplement 1H–I ) , which suggests it is more resistant to parameter change and thus likely to achieve multistability in experimental settings . In order to experimentally evaluate dynamic properties of these networks , we constructed nine circuits including tunable positive feedbacks ( T6 and T9 ) , mutual inhibition ( T5 ) , dual-positive feedbacks ( T10 ) , and their combinations ( T7 , T11 , T13 , T14 and T15 ) . One-dimensional networks ( T1 , T4 , T2 and T8 ) and trivial two-dimensional networks ( T3 and T12 ) are excluded for their low multistability probability . All motifs were constructed using the same set of components ( Figure 1 ) . Probing a circuit’s multistability typically requires thorough hysteresis experiments covering wide ranges of doses for all inducers ( Acar et al . , 2005; Angeli et al . , 2004; Gardner et al . , 2000 ) , which becomes infeasible for nine complex networks with four inducers . To improve the efficiency of probing multistability and tunability , we designed a ‘sequential induction’ method to accelerate exploration of unknown high dimensional bifurcation spaces ( see Appendix text for details ) , instead of conventional ‘back and forth’ hysteresis on one parameter dimension . The main concept relies on the fact that multistable gene networks could exhibit discontinuous jump from one state to another in response to changing parameter ( inducer ) combinations . Taking the classic ‘toggle switch’ as an example ( Gardner et al . , 2000 ) , the circuit can be tuned by two external inducers and its two-parameter bifurcation diagram has a stretched S shape ( Figure 2A ) . Initialized at an arbitrary state A , the cells could reach State C in the bistable region directly when induced with both inducers simultaneously . If the cells are first induced by Inducer I to go to state B , they will also reach State C after Inducer II is added . However , if the same dose of Inducer II is applied first , cells will cross the bifurcation plane to state D on the low-Response surface and then reach state E with addition of Inducer I ( Figure 2A ) . State C and E are two different steady states with the same induction dosages , illustrating hysteresis and verifying multistability . 10 . 7554/eLife . 23702 . 005Figure 2 . Sequential induction of MINPA and its sub-networks . ( A ) Schematic illustration of rationale for sequential induction . This two-parameter bifurcation diagram of a bistable toggle-switch depicts all steady state values of response ( Z-axis ) with combinations of inducer I and II ( X and Y axes ) . Arrows illustrate order and direction of inductions and consequent steady state value changes . Solid lines on the X-Y plane are the boundaries of bistability . Dashed lines on the X-Y plane are projections of solid white arrowheads . ( B ) Arabinose ( Ara ) and IPTG were sequentially ( left and middle columns ) or simultaneously ( right column ) applied to induce T9 , T13 , T11 , and T15 . T: topology . The concentration of Arabinose and IPTG is 2 . 5*10−5m/v , and 5*10−5 M , respectively . To indicate the effects of inducers , we used the same color for applied inducers and its regulated connections , which were also shown in bold lines . The other non-regulated connections are represented by thin lines . ( C ) AHL and aTc were sequentially ( left and middle ) or simultaneously ( right ) applied to induce T6 , T7 , T14 , and T15 . The concentration of AHL and aTc is 1*10−4 M , and 200 ng/ml , respectively . ( D ) Ara and AHL were sequentially ( left and middle ) or simultaneously ( right ) applied to induce T10 , T14 , T11 , and T15 . The concentration of Arabinose and AHL is 2 . 5*10−5m/v , and 1*10−8 M , respectively . Samples were treated with the first inducer till OD600 is about 0 . 15 and then the second inducer was added . Cells were grown for another 24 hr before measured by flow cytometry . The experiments were performed in triplicate and repeated two times , and representative results are presented . The inducers are color-coded as visual assistance to indicate which edge of inset diagram it regulates . DOI: http://dx . doi . org/10 . 7554/eLife . 23702 . 00510 . 7554/eLife . 23702 . 006Figure 2—figure supplement 1 . Experimental design and validation of sequential induction strategy in a synthetic toggle switch circuit . ( A ) Abstract diagram and molecular implementation of the toggle switch circuit . TetR ( R ) and LacI ( I ) mutually inhibit each other through binding to Ptet and Plac promoter , respectively . IPTG and aTc ( hexagon ) can respectively relieve LacI and TetR inhibition . GFP serves as the readout of Ptet . ( B ) Time course results of the sequential induction . The y-axis represents forward scatter ( FSC-A ) , and the x-axis indicates GFP fluorescence . IPTG and aTc were sequentially ( left and middle columns ) or simultaneously ( right column ) applied to induce the toggle circuit . The first inducer was added to the media for 5 hr , and then the second inducer was added . Fluorescence was measured by flow cytometry at 0 hr , 5 hr , 12 hr , and 24 hr after the second inducer was added into the cultures . The concentration of IPTG and aTc is 8*10−5 M , and 100 ng/ml , respectively . Experiments were repeated for at least three times , and representative results were shown . DOI: http://dx . doi . org/10 . 7554/eLife . 23702 . 00610 . 7554/eLife . 23702 . 007Figure 2—figure supplement 2 . Time course results of sequential induction for the MINPA ( T15 ) circuit . ( A ) Arabinose ( Ara ) and IPTG were sequentially ( left and middle columns ) or simultaneously ( right column ) applied to induce T15 . The first inducer was applied for 5 hr , and then the second inducer was added into the culture . Fluorescence was measured by flow cytometry at 0 hr , 12 hr , and 24 hr after the second inducer was added into the culture . The concentration of Ara and IPTG is 2 . 5*10−5m/v , and 5*10−5 M , respectively . ( B ) AHL and aTc were sequentially ( left and middle ) or simultaneously ( right ) applied to induce T15 . The first inducer was applied for 6 . 5 hr , and then the second inducer was added into the culture . The concentration of AHL and aTc is 1*10−4 M , and 200 ng/ml , respectively . ( C ) Ara and AHL were sequentially ( left and middle ) or simultaneously ( right ) applied to induce T15 . The first inducer was applied for 5 hr , and then the second inducer was added into the culture . The concentration of Arabinose and AHL is 2 . 5*10−5m/v , and 1*10−8 M , respectively . ( D ) IPTG and aTc were sequentially ( left and middle ) or simultaneously ( right ) applied to induce T15 . The first inducer was applied for 6 . 5 hr , and then the second inducer was added into the culture . The concentration of IPTG and aTc is 1*10−4 M , and 200 ng/ml , respectively . Fluorescence was measured by flow cytometry at 0 hr , 12 hr , and 24 hr after the second inducer was added into the culture . The inducers are color-coded as visual assistance to indicate which edge of inset diagram it regulates . DOI: http://dx . doi . org/10 . 7554/eLife . 23702 . 00710 . 7554/eLife . 23702 . 008Figure 2—figure supplement 3 . Sequential induction for circuits T5 , T7 , T13 , and T15 with inducers IPTG and aTc . Left: IPTG was first applied to induce the circuits , and then aTc was added; Middle: aTc was first applied to induce the circuits , and then IPTG was added; Right: IPTG and aTc were added simultaneously into the medium . The concentration of IPTG and aTc is 1*10−4 M and 200 ng/ml , respectively . Samples were treated with the first inducer for 6 . 5 hr and then the second inducer was added . Fluorescence was measured at 24 hr by flow cytometry . DOI: http://dx . doi . org/10 . 7554/eLife . 23702 . 008 To test our theoretical analysis , a synthetic toggle switch circuit was constructed ( Figure 2—figure supplement 1A ) . Following experimental design principles ( see Appendix text for details ) , we designed a protocol to show the sequential induction effects . We first employed IPTG to induce the circuit for 5 hr , and then aTc was added . Time course results showed that cells stayed at low-GFP state till 24 hr ( Figure 2—figure supplement 1B ) . However , cells induced with aTc first , and then IPTG mainly stayed at high-GFP state , another stable steady state under this condition . Simultaneous aTc and IPTG induction produced similar cell distributions . These results show that sequential induction can be used as a strategy to quickly explore a multistable potential landscape for complex non-equilibrium systems . Without knowing the exact bifurcation range beforehand , such ordered sequential inductions could help quickly explore the irregular bifurcation space to reveal multistability for systems with complicated bifurcations , which is typically caused by interfering parameters . Similar sequential induction techniques have been shown to enable access of otherwise hard-to-reach cell death states in breast cancer cells ( Lee et al . , 2012 ) . This strategy has also been widely employed in directed differentiation of stem cells to specific lineages ( Paşca et al . , 2015; Pagliuca et al . , 2014; Kroon et al . , 2008 ) and reprogramming somatic cells to induced pluripotent stem cells ( Liu et al . , 2013 ) . Although specific inducer concentrations are required to observe the effects of this strategy in synthetic circuits , sequential induction with pre-selected inducer combinations can help perform a coarse-grained exploration from different directions in the parameter space . Furthermore , stochastic gene expression of the circuits also contributes to cellular population distribution thus leads to pronounced sequential induction effects , given experimentally feasible amount of time , when the system is entering its multistable region from different directions . Therefore , distinct final states , or even different population distributions , under sequential induction strongly suggests the existence of nonlinear dynamics , including multistability ( see Appendix text for details ) . Using the sequential induction approach , we tested the nine circuits using flow cytometry . Cells were first induced by inducer I , inducer II was then added into the media for another 24 hr . Depending on the network configuration , four different dual-inducer combinations were used . For example , Arabinose and IPTG were applied sequentially and simultaneously to T9 , T13 , T11 and T15 , respectively ( Figure 2B ) . It can be seen only T15 exhibits significant expression difference between three induction patterns , while the others show little change ( Figure 2B and Figure 2—figure supplement 2A ) . It should be noted that T15 also exhibits tri-modality of fluorescence expression , suggesting multistability given the presence of gene expression noise , which is partially consistent with our computational predictions . Similarly , AHL and aTc were applied to T6 , T7 , T14 , and T15 , respectively ( Figure 2C and Figure 2—figure supplement 2B ) . Results show that only T15 exhibits significant fluorescence pattern change with different inductions , whereas T6 and T7 exhibit minor uniform shifts of expression . T14 , although exhibiting bimodality , only shows a ratio change of two populations between three inductions and no sign of bifurcation . Sequential induction by Arabinose and AHL combinations has little effect on T10 , T14 and T11 , but T15 displays three notable populations for AHL-then-Arabinose induction ( Figure 2D and Figure 2—figure supplement 2C ) . IPTG and aTc were also tested on T5 , T7 , T13 and T15 , but no notable dynamics were observed ( Figure 2—figure supplement 2D and Figure 2—figure supplement 3 ) . Taken together , T15 , the full MINPA topology , shows the most variety and complexity in population heterogeneity under sequential inductions , suggesting this circuit has the highest potential to generate complex multistability within our induction range and hence enable us to approach the Waddington landscape . Next , operating principles and full tunability of T15 ( MINPA ) were further examined by using four inducers ( Arabinose , AHL , aTc , and IPTG ) to fine tune the strength of regulations and perturb the system ( Figure 3A ) . Uninduced cells showed low GFP and low mCherry expression ( low-low state , LL ) . In the presence of AHL and aTc , high GFP and low mCherry ( GFP state ) is observed; low GFP and high mCherry ( mCherry state ) emerged with induction of Arabinose; and high GFP and high mCherry ( high-high state , HH ) was achieved when induced with Arabinose and AHL . These results verify that our engineered MINPA circuit is functioning as designed and fully controllable with four distinct states reachable through appropriate inductions , respectively . 10 . 7554/eLife . 23702 . 009Figure 3 . Bifurcation analysis and hysteresis of MINPA . ( A ) Engineered MINPA is tunable to reach four individual states: low-low , GFP , mCherry , and high-high , under no induction , 1*10−4 M AHL and 100 ng/ml aTc , 2 . 5*10−5 ( m/v ) Arabinose , 1*10−4 M AHL and 2 . 5*10−3 ( m/v ) Arabinose , and respectively . To indicate the effects of inducers , we used the same color for applied inducer and its regulated connection ( bolder lines ) in the MINPA topology . The other non-regulated connections are represented by thin lines . ( B ) 3-D bifurcation diagram of MINPA . AR/AL is a lumped parameter composed of increasing concentrations of Arabinose and AHL , but the ratio of Arabinose and AHL is fixed , i . e . , [Arabinose]/[AHL] is a constant . GFP and mCherry represent the states of node X and Y . Blue lines represent stable steady states , while red ones are unstable steady states . Grey , green , rose , and golden spheres represent low-low , GFP , mCherry , and high-high state , respectively . And the size of spheres correlates with the attractiveness of each state . C1 , C2 , C3 , and C4 are four increasing concentrations of Arabinose and AHL used for experimental probing . ( C–D ) Hysteresis results of MINPA under induction of AR/AL . C1LL-C4LL: cells with low-low initial state ( C ) are induced with AR/AL at C1 to C4; C1HH-C4HH: cells with high-high initial state ( D ) are induced with AR/AL for 24 hr at C1 to C4 . C1: no inducers; C2: 2 . 5*10−6m/v Arabinose and 1*10−7 M AHL; C3: 2 . 5*10−5m/v Arabinose and 1*10−6 M AHL; C4: 2 . 5*10−3m/v Arabinose and 1*10−4 M AHL . Arabinose and AHL were added at the same time to induce the cells . 100 , 000 cells were recorded for each sample by flow cytometry . DOI: http://dx . doi . org/10 . 7554/eLife . 23702 . 00910 . 7554/eLife . 23702 . 010Figure 3—figure supplement 1 . Another view of the 3-D bifurcation diagram of MINPA at C2 . The circuit’s quadrastability is illustrated as four similar-sized colored spheres on the same gray plane , which represents the low-low , GFP , mCherry , and high-high state , respectively . Blue lines represent stable steady states , while red ones are unstable steady states . Grey , green , rose , and golden spheres represent low-low , GFP , mCherry , and high-high state , respectively . And the size of spheres correlates with the attractiveness of each state . C2: 2 . 5*10−6m/v Arabinose and 1*10−7 M AHL . DOI: http://dx . doi . org/10 . 7554/eLife . 23702 . 01010 . 7554/eLife . 23702 . 011Figure 3—figure supplement 2 . Bifurcation analysis for and hysteresis of MINPA with induction of Arabinose and AHL . ( A–D ) One-dimensional bifurcation analysis for all four inducers in MINPA . We perform bifurcation analysis for each inducer while setting the concentration of other inducers to be very small ( 10−10 ) . The blue curves are branches of stable steady states ( SSS ) , while the red curves are branches of unstable steady states ( USS ) . The bifurcation analyses are performed using Matlab . ( E–H ) Dual induction bifurcation analysis for Arabinose and AHL in MINPA . We perform bifurcation analysis for dual induction of different β ( see Appendix for details ) . The blue curves are branches of SSS , while the red curves are branches of unstable steady states . ( I ) Hysteresis of MINPA for initial low-low state cells with induction of Arabinose and AHL . Cells with initial low-low state were induced with a series of concentrations of Arabinose ( from 2 . 5*10−6m/v to 2 . 5*10−5m/v to 2 . 5*10−3m/v ) and AHL ( from 1*10−7 M to 1*10−6 M to 1*10−4 M ) . Data circled by red rectangles are shown in Figure 3C . Cells were grown for 24 hr before measured by flow cytometry . 10 , 000 events were recorded . ( J ) Hysteresis of MINPA for initial high-high state cells with induction of Arabinose and AHL . Initial high-high state cells were collected from the initial low-low state cells induced with 2 . 5*10−3m/v Arabinose and 1*10−4 M AHL for 12 hr . Cellular states were then monitored by flow cytometry to ensure its high-high state profile . The high-high state cells were washed and then inoculated into fresh medium with the same concentrations of Arabinose and AHL ( from 2 . 5*10−6m/v to 2 . 5*10−5m/v to 2 . 5*10−3m/v ) and AHL ( from 1*10−7 M to 1*10−6 M to 1*10−4 M ) . Data circled by red rectangles are shown in Figure 3D . Cells were grown for 24 hr before measured by flow cytometry . 10 , 000 events were recorded . DOI: http://dx . doi . org/10 . 7554/eLife . 23702 . 011 To help design experiments to further investigate the circuit’s quadrastability , a detailed mathematical model was developed to describe the system ( see Appendix for details ) . Using parameters derived from hybrid promoter testing experiments , bifurcation analysis was carried out to systematically quantify MINPA’s dynamic behavior ( Figure 3B , Figure 3—figure supplement 1 and Figure 3—figure supplement 2A–H ) . Figure 3B is the three-dimensional bifurcation diagram , where levels of GFP and mCherry represent the states of node X and Y , and ‘AR/AL’ is a lumped parameter composed of a fixed ratio of the concentrations of Arabinose and AHL . Overall , it can be seen that the system , initialized without induction , is predicted to be quadrastable ( shown as four colored spheres , representing LL ( grey ) , GFP ( green ) , mCherry ( rose ) , and HH ( golden ) state , respectively ) but with the low-low state to have dominant attractiveness ( shown as the big gray sphere ) when AR/AL is low ( C1 ) . However , when AR/AL level is within an intermediate range , relative stabilities between different states become comparable . When AR/AL level increased from C1 to C2 , the circuit’s quadrastability becomes well pronounced , illustrated as four similar-sized colored spheres on the same gray plane , which represents the low-low , GFP , mCherry , and high-high state , respectively ( Figure 3—figure supplement 1 ) . As AR/AL continues to increase from C2 to C3 , while the other three SSS remain stable , the stability of the GFP branch disappears . Further increase of AR/AL results in only one stable state-the high-high state , shown as the orange sphere with biggest size . To establish MINPA’s quadrastability and tristability as predicted , hysteresis , a hallmark of multistability ( Acar et al . , 2005; Wu et al . , 2014 , 2013 ) , of the network was tested . Initialized at the low-low state , cells were induced by increasing doses of AR/AL corresponding to C1 to C4 and measured by flow cytometry ( Figure 3C and Figure 3—figure supplement 2I ) . As predicted , C1LL ( cells with initial Low-Low state grown at C1 condition ) experiment demonstrates uniform low-low fluorescence profile , due to the low-low state’s dominant attractiveness , and C4LL shows a uniform high-high profile . Interestingly , C3LL indeed illustrates tri-modality , which is the result of predicted tristability . C2LL experiment , on the other hand , exhibits enough heterogeneity to signal high-high , low-low , and mCherry state , but does not illustrate significant trace of GFP state . Given that GFP state is achieved through combinational induction of AHL and aTc ( Figure 3A ) , we hypothesize that the GFP state here is not easily accessible with AHL induction only . Next , cells initialized at high-high states were collected and diluted into fresh media with the same concentrations of AR/AL ( Figure 3D and Figure 3—figure supplement 2J ) . As predicted , these cells keep high-high expression profile even with inductions as low as C1 , another demonstration that the system is already multistable at C1 . Taken together , the two sets of experiments demonstrated clear hysteresis and verified the existence of three of the four predicted SSS . To further investigate what determines the accessibility of certain SSS in this quadrastable system and how cells navigate this attractor landscape , we take into account gene expression stochasticity ( Wu et al . , 2013 ) to sketch out MINPA’s quasi-potential attractor landscape ( Figure 4A and Appendix ) , which is calculated as the negative logarithmic function of stationary distribution density in the phase space of GFP and mCherry . Using the weighted ensemble random walk algorithm ( Appendix ) , the stationary density distribution can be efficiently calculated from the initial uniform distribution . It can be seen that when there is no inducer , MINPA is already quadrastable with four local minima , which is consistent with bifurcation analysis for C1 condition . Furthermore , the much stronger stability of the low-low state ( deepest well , Top landscape ) and high state-transition barrier explain homogeneous low-low population ( C1 experiment in Figure 3C ) when cells were initialized with no inductions . 10 . 7554/eLife . 23702 . 012Figure 4 . Model-guided quadrastability of MINPA through triple induction . ( A ) Dynamic evolution of computed energy landscapes of MINPA under sequential/simultaneous inductions of Arabinose , and/or AHL and aTc . Center route: simultaneous induction with three inducers; Left route: sequential induction with AHL and aTc first , and then Arabinose . Right route: sequential induction with Arabinose , and then AHL and aTc . Deeper wells represent the higher stability of corresponding states . For each three-dimensional landscape , corresponding two-dimensional state-potential plots were also shown . Red line sketches the potentials from mCherry state to high-high to GFP state while green one represents the potentials from mCherry state to low-low to GFP states . mC: mCherry; HH: high-high; LL: low-low . GFP* and mCherry* is the computed GFP and mCherry abundance from the model . To indicate the effects of inducers , we used the same color for applied inducers and its regulated connections , which were also shown in bolder lines . ( B–D ) Experimental validations of model-predicted quadrastability using flow cytometry . Quadrastable steady states were observed when Arabinose , AHL , and aTc were simultaneously added into the media ( B ) , corresponding to the Center route in A ) . Four populations were also observed when AHL and aTc were first added to growth media for 6 . 5 hr and then Arabinose was added , and cells were grown for another 24 hr before measurement ( C ) , corresponding to the Left route in A ) . Bimodality ( low-low and mCherry states ) was generated when Arabinose was first applied and then AHL and aTc were added ( D ) , corresponding to the Right route in A ) . Concentrations for Arabinose , AHL and aTc are 2 . 5*10−5m/v , 1*10−4 M , and 400 ng/ml , respectively . Representative results from three replicates are showed and 100 , 000 cells were recorded for each sample by flow cytometry . DOI: http://dx . doi . org/10 . 7554/eLife . 23702 . 01210 . 7554/eLife . 23702 . 013Figure 4—figure supplement 1 . Cells’ states under induction with the first inducer , microfludic results to demonstrate quadrastability with IPTG and aTc induction , and time course of sequential induction of AHL , aTc and Ara . ( A ) Up left: Flow cytometry result for cells simultaneously induced with 2 . 5*10−5m/v Arabinose , 1*10−4 M AHL and 200 ng/ml aTc for 6 . 5 hr . Up right: Cells were first induced with 1*10−4 M and 400 ng/ml aTc for 6 . 5 hr , and then measured by flow cytometry . About 12% cells were moving from low-low state to GFP state at 6 . 5 hr . Bottom left: Cells were first induced with 2 . 5*10−5m/v Arabinose and no obvious state transition was observed at 6 . 5 hr . However , at 9 . 5 hr , most cells ( 84 . 6% ) were transitioned to mCherry state ( Bottom right ) . ( B ) Microfluidic setup and device design ( adopted from Dr . Hasty lab ( Ferry et al . , 2011 ) . ( C ) Images showing E . coli growing in the device . White arrows indicate the flow direction . ( D ) Time course of the cells growing and fluorescence state change with 2*10−4 M IPTG and 200 ng/ml aTc induction in the trap . The red flow is medium without inducer for 6 hr , and then cells switch to medium with inducers for 18 hr . Small white arrows show single cells with state change from GFP to mCherry . Magnification: 40x . ( E ) Time-course sequential induction with AHL , aTc first and then Arabinose ( corresponding to the Left route in Figure 4A ) . The indicated time point is the time after Arabinose added into the culture . 10 , 000 events were recorded . The low-low state cells changed from 21 . 2% ( 12 h ) to 24 . 6% ( 24 h ) to 30 . 6% ( 36 h ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23702 . 013 Since Arabinose and AHL combination is not sufficient to enable the cells to reach all four SSS , we chose to add aTc to the mix to further facilitate cell transitions among these four SSS . Using our expanded model , we simulated simultaneous and sequential inductions and computed corresponding quasi-potential landscape ( Figure 4A ) , showing cells harboring the same MINPA network exhibiting distinct landscapes under different inductions . AHL and aTc promote a more stable GFP state ( Left center ) , while Arabinose induction modulates the landscape to be biased toward mCherry state ( Right center ) . When the three inducers were applied simultaneously , the landscape changes and the four states show comparable stabilities ( Bottom ) , suggesting a higher possibility of quadramodal cell population experimentally . Experimental validation is shown as flow cytometry measurements of cells treated with Arabinose , AHL , and aTc simultaneously for 24 hr ( Figure 4B , and Figure 4—figure supplement 1A ) . Such a hybrid induction greatly facilitates the cells’ transition from low-low state to the other three states so that a quadramodal distribution emerges . Single-cell time lapse microscopy results also showed that the initial low-low state cells could differentiate into GFP , mCherry and high-high state cells ( Figure 4—figure supplement 1B–D and Appendix 1—Video 1 ) . This also finally verifies predicted quadrastability of MINPA . There are two other strategies to reach this condition: sequential inductions with AHL-and-aTc and then Arabinose ( Figure 4A , Left route ) or Arabinose and then AHL-and-aTc ( Right route ) . Even though the initial and final landscapes are the same , the dynamics for each route are quite different , which could lead to distinct outcomes . By comparing state barrier heights ( Figure 4A ) , we hypothesize that cells walking through the left route would start transitioning from low-low state to GFP state upon induction of AHL and aTc . Following Arabinose induction would then make the mCherry state accessible . So some cells with GFP state would transition to high-high state while some low-low state cells transition to mCherry state , resulting in cells in all four states . Experimental testing indeed shows four stable populations ( Figure 4C ) . At 6 . 5 hrs of AHL and aTc induction , about 12% cells were moving to GFP state while the rest of them still stay ‘undecided’ at low-low state ( Figure 4—figure supplement 1A and E ) . This is consistent with the simulated landscape as these two states are more stable and accessible to each other ( Figure 4A , Left ) . Arabinose induction promoted some cells to transition into mCherry state while some cells continued moving into GFP state , of which some further transitioned to high-high state . Interestingly , the right route is predicted to generate different results . When first induced with Arabinose , the mCherry valley is so deep that it would be difficult for cells to jump out to high-high state , and low-low state cells are also hardly transit to GFP state due to its low attractiveness , and thus most cells would stay at mCherry and low-low state even with AHL and aTc inductions ( Figure 4A , Right ) . Experimental testing of the right route indeed only produces two populations with low-low and mCherry state ( Figure 4D ) . With 5 hrs of Arabinose induction , most cells still stay at low-low state because of slow transition to the mCherry state ( Figure 4—figure supplement 1A ) , but 84 . 6% cells transitioned to mCherry state with 15 . 3% cells at low-low state at 9 . 5 hr ( Figure 4—figure supplement 1A ) . This is consistent with our model predictions . The high barrier between the mCherry state and high-high state blocks the transition from mCherry state to high-high state , while the low attractiveness and relatively high barrier of the GFP state also decreases the probability of cells transitioning from low-low to GFP state . Hence , when AHL and aTc are applied , cells are predominantly in the mCherry state with a small portion in low-low state with low probability of transitioning out , resulting in a bimodal distribution .
Multistability and the resulting landscape has long been proposed as an underlying mechanism that cells use to maintain pluripotency and guide differentiation ( Kauffman , 1993; Laurent and Kellershohn , 1999; Huang et al . , 2007; Guantes and Poyatos , 2008; Palani and Sarkar , 2009; Narula et al . , 2010; Faucon et al . , 2014 ) . Theoretical frameworks have also been established to quantify the Waddington landscape and biological paths for cell development ( Li and Wang , 2013a , 2013b; Wang et al . , 2011 ) . Experimental validation of this hypothesis and a full understanding of this mechanism will help reveal differentiation dynamics and routes for all cell types , which remains an outstanding problem in biology . In this study , we engineered the quadrastable MINPA circuit and show that it can guide cell fate choices , represented by fluorescence expression , through shaping the potential landscape . MINPA represents one of the most complicated two-node network topologies and includes four genes to implement a web of regulations . Biological complexity correlates with the number of regulatory connections ( Szathmáry et al . , 2001 ) , not the number of genes . Hence , dense connectivity and complex dynamics of MINPA may provide a framework to understand similarly densely connected gene regulatory networks . Combining mathematical modeling and experimental investigation , this study serves as a proof-of-principle demonstration of the Waddington landscape . Furthermore , we used this circuit to demonstrate how different sequential inductions can change the landscape in a specific order and navigate cells to different final states . Such illustrations suggest mechanistic explanations of the need for fixed induction sequences for targeted differentiation to desired cell lineage . Overall , this study helps reveal fundamental mechanisms of cell-fate determination and provide a theoretical foundation for systematic understanding of the cell differentiation process , which will lead to development of new strategies to program cell fate .
All the molecular cloning experiments were performed in E . coli DH10B ( Invitrogen , USA ) , and measurements of MINPA and sub-networks were conducted in E . coli K-12 MG1655ΔlacIΔaraCBAD strain as previously described ( from Dr . Collins Lab [Litcofsky et al . , 2012] ) . The sequential induction for the toggle circuit was conducted in E . coli MG1655ΔlacI strain as previously described ( Litcofsky et al . , 2012 ) . Cells were grown at 37°C in liquid and/or solid Luria-Bertani broth medium with 100 µg/mL ampicillin or kanamycin . Chemicals AHL ( 3oxo-C6-HSL , Sigma-Aldrich ) , Arabinose ( Sigma-Aldrich , USA ) , isopropyl β-D-1-thiogalactopyranoside ( IPTG , Sigma-Aldrich ) , and anhydrotetracycline ( aTc , Sigma-Aldrich ) were dissolved in ddH2O and diluted into indicated working concentrations . Chemical aTc solution was stocked in brown vials , and experiments involving aTc were performed in cabinet without light , and cell cultures were grown in darken incubator at 37°C . Cultures were shaken in 5 mL and/or 15 mL tubes at 220 rotations per minute ( r . p . m ) . All the plasmids ( MINPA and its nine sub-networks ) in this study were constructed using standard molecular cloning protocols and assembled by standardized BioBricks methods based on primary modules ( Table 1 ) from the iGEM Registry ( www . parts . igem . org ) . Hybrid promoter Para/lac was from Dr . Collins lab and amplified using forward primer: CGGAATTCGCTTCTAGAGAATTGTGAGCGGATAAC; and reverse primer: CGCTGCAGGCACTAGTTTGTGTGAAATTGTTATCCG . PCR product was purified using GenElute PCR Clean-Up Kit ( Sigma-Aldrich ) , and then cut by restriction enzymes EcoRI and PstI . The purified product was inserted into pSB1K3 backbone , and finally verified by DNA sequencing . The MINPA circuit was constructed from promoter Para/lac and nine other Biobrick standard biological parts: BBa_B0034 ( ribosome binding site , RBS ) , BBa_C0080 ( AraC gene ) , BBa_C0040 ( tetR gene ) , BBa_K176000 ( Plux/tet hybrid promoter ) , BBa_C0062 ( luxR gene ) , BBa_C0012 ( lacI gene ) , BBa_B0015 ( transcriptional terminator ) , BBa_E0240 ( GFP generator ) , and BBa_J06702 ( mCherry generator ) . The fragment and vector were separated by gel electrophoresis ( 1% TAE agarose ) and purified using GenElute Gel Extraction Kit ( Sigma-Aldrich ) . Then , fragment and vector were ligated together using T4 DNA ligase , and the ligation products were transformed into E . coli DH10B and clones were screened by plating on 100 μg/ml ampicillin LB agar plates . Finally , their plasmids were extracted and verified by double digestion ( EcoRI and PstI ) . The detailed procedures of assembling DNA constructs were described in our previous study ( Wu et al . , 2014 ) . Restriction enzymes ( EcoRI , XbaI , SpeI , and PstI ) and T4 DNA ligase were purchased from New England Biolabs . All the constructs were inserted into high copy number plasmid pSB1A3 and pSB1K3 . All the constructs were verified by DNA sequencing ( Biodesign sequencing lab in ASU ) step by step . 10 . 7554/eLife . 23702 . 014Table 1 . Components from the Registry of standard biological partsDOI: http://dx . doi . org/10 . 7554/eLife . 23702 . 014Biobrick numberAbbreviation in the paperDescriptionBBa_C0080AraCAraC arabinose operon regulatory protein from E . coliBBa_C0040TetRTetracycline repressor from transposon Tn10BBa_C0062LuxRLuxR activator from Aliivibrio fischeriBBa_C0012LacILacI repressor from E . coliBBa_E0240GFPGFP generatorBBa_J06702mCherryRFP generatorBBa_K176002Plux/tetHybrid promoter with LuxR/HSL- and TetR-binding sitesBBa_B0034RBSRibosome binding siteBBa_B0015TerminatorTranscriptional terminator ( double ) BBa_K176009CPConstitutive promoterpSB1K3pSB1K3High copy BioBrick assembly plasmid with kanamycin resistancepSB1A3pSB1A3High copy BioBrick assembly plasmid with ampicillin resistance All the samples were analyzed at the indicated time points on an Accuri C6 flow cytometer ( Becton Dickinson , USA ) with excitation/emission filters ( 488/530 nm for GFP , and 610 LP for mCherry ) . The data were collected in a linear scale and non-cellular low-scatter noise was removed by thresholding . All measurements of gene expression were obtained from at least three independent experiments . For each culture , 100 , 000 events were collected at a slow flow rate . Data files were analyzed using MATLAB ( MathWorks ) . For sequential induction , initially uninduced overnight cell culture was diluted into fresh media without or with inducer I , grown at 37°C and 220 r . p . m till OD600 is 0 . 15 ~ 0 . 25 ( the time usually takes 5 ~ 6 . 5 hr , depends on the inducers and concentrations ) . For samples induced individually by Ara , or AHL , or IPTG , it is ~5 hr; for samples induced with aTc , it takes ~6 . 5 hr . According to our experience , gene ( GFP ) is starting to be partially expressed while steady states are not yet stable . Then inducer II was added into the culture , and grown for another 24 hr . Flow cytometry was performed at 0 hr , 12 hr , and 24 hr after the second inducer was added into the culture . For each set of sequential induction , the first scenario: add inducer I first , then add inducer II; the second scenario: add inducer II first , then add inducer I; the third scenario: add inducers I and II at the same time . As a control , cells without any inducer were also prepared and measured . Inducer I and II were the two of four commercial chemicals: AHL , Arabinose , IPTG , and aTc . All the experiments were repeated for at least three times and only representative results were showed . For hysteresis experiments , initially uninduced cells were diluted into fresh media and distributed into new 5 ml tubes . Various amounts of Arabinose and AHL ( 3oxo-C6-HSL ) were added into the media , and cells were then grown at 37°C shaker . The initially high-high state cells induced with 2 . 5 *10−3 m/v Arabinose and 1*10−4 M AHL were collected with low-speed centrifugation , washed twice , resuspended with fresh medium , and at last inoculated into fresh medium at a 1:100 ratio with the same series of inducer ( Arabinose and AHL ) concentrations . C1 , C2 , C3 , and C4 ( Figure 3B–D ) are four increasing concentrations of Arabinose and AHL used for experimental probing , but the ratio of Arabinose and AHL is fixed . Specifically , cells were induced with the Arabinose and AHL at the same time ( the third scenario ) , at concentrations from C1 to C4 . C1: no inducers; C2: 2 . 5*10−6m/v Arabinose and 1*10−7 M AHL; C3: 2 . 5*10−5m/v Arabinose and 1*10−6 M AHL; C4: 2 . 5*10−3m/v Arabinose and 1*10−4 M AHL . Flow cytometry analyses were performed at 12 hr and 24 hr to monitor the fluorescence levels . Experiments were repeated two times with three replicates . Cells with MINPA circuit were grown overnight , which was then re-diluted into 5 mL fresh LB medium with Kanamycin the next day . When OD600 of the cells reached about 0 . 2 , cells were spun down with low speed and resuspended in 5 ml of fresh medium and loaded into the device . Detailed description of chip design and device setup could be found from Hasty Lab ( Ferry et al . , 2011 ) . Two media were prepared: one with inducers and the other without . Cells in the trap were first supplied by the medium without inducer for 6 hr , and then switched to medium with inducers for anther 18 hr , which was controlled by adjusting the heights of the medium syringes relative to one another . Images were taken by using Nikon Eclipse Ti inverted microscope ( Nikon , Japan ) equipped with an LED-based Lumencor SOLA SE . Phase and fluorescence images were taken every 5 min for 24 hr in total under the magnification 40x . Perfect focus was maintained automatically using Nikon Elements software . Experimental detail can also be found in Appendix . Ordinary differential equation models were developed to describe and analyze the MINPA system . Details are provided in the Appendix . | Cells in animals use a process called differentiation to specialize into specific cell types such as skin cells and liver cells . Proteins called transcription factors drive particular steps in differentiation by controlling the activity of specific genes . Many transcription factors interact with each other to form complex networks that regulate gene activity to determine the fate of a cell and control the whole differentiation process . Some individual gene networks can program cells to become any one of several different cell fates , a feature known as multistability . In the 1950s , a scientist called Conrad Waddington proposed the concept of an “epigenetic landscape” to describe how the fate of a cell is decided as an animal develops . The cell , depicted as a ball , rolls down a rugged landscape and has the option of taking several different routes . Each route will eventually lead to a distinct cell fate . As the ball moves down the hill , the choice of routes and final destinations becomes more limited . Theoretical approaches have been used to understand how gene regulatory networks shape the epigenetic landscape of an animal . However , few studies have experimentally tested the findings of the theoretical approaches and it is not clear how environmental inputs help to determine which path a cell will take . Although bacteria cells do not generally specialize into particular cell types , bacteria cells can use multistability in transcription factor networks to switch between different behaviors or “states” in response to cues from the environment . Wu et al . used a bacterium called E . coli as a model to investigate how a gene network called MINPA from mammals , which is involved in differentiation and is believed to show multistability , can guide cells to adopt different states . The work combined experimental and mathematical approaches to design , construct and test an artificial version of the MINPA gene network in E . coli . The experiments showed that MINPA could direct the cells to adopt four different stable states in which the cells produced fluorescent proteins of different colors . With the help of mathematical modeling , Wu et al . charted how the landscape of cell states changed when external chemical cues were applied . Exposing the cells to several cues in particular orders guided the cells to different final states . The findings of Wu et al . shed new light on how the fate of a cell is determined and provide a theoretical framework for understanding the complex networks that control cell differentiation . This could help develop new ways of directing cell fate that could ultimately be used to generate cells to treat human diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology"
] | 2017 | Engineering of a synthetic quadrastable gene network to approach Waddington landscape and cell fate determination |
A major challenge in biology is identifying distinct cell classes and mapping their interactions in vivo . Tissue-dissociative technologies enable deep single cell molecular profiling but do not provide spatial information . We developed a proximity ligation in situ hybridization technology ( PLISH ) with exceptional signal strength , specificity , and sensitivity in tissue . Multiplexed data sets can be acquired using barcoded probes and rapid label-image-erase cycles , with automated calculation of single cell profiles , enabling clustering and anatomical re-mapping of cells . We apply PLISH to expression profile ~2900 cells in intact mouse lung , which identifies and localizes known cell types , including rare ones . Unsupervised classification of the cells indicates differential expression of ‘housekeeping’ genes between cell types , and re-mapping of two sub-classes of Club cells highlights their segregated spatial domains in terminal airways . By enabling single cell profiling of various RNA species in situ , PLISH can impact many areas of basic and medical research .
In parallel with the development of single-cell RNA sequencing ( scRNA-seq ) , there have been rapid advances in single-molecule in situ hybridization ( smISH ) techniques that localize RNAs of interest directly in fixed cells ( Shah et al . , 2016; Huss et al . , 2015; Chen et al . , 2015; Wang et al . , 2012; Larsson et al . , 2010; Raj et al . , 2008; Femino et al . , 1998 ) . These smISH techniques involve hybridization of fluorescently-labeled oligonucleotide probes , typically 24–96 per gene , to mark individual RNA molecules with a discrete , diffraction-limited punctum that can be quantitatively analyzed by fluorescence microscopy . smISH has been used in cultured cells to study the subcellular distribution of RNAs ( reviewed in [Buxbaum et al . , 2015] ) , the consequences of stochastic noise on gene expression ( Raj et al . , 2010; Raj et al . , 2006 ) , and the impact of cell shape and environment on expression programs ( Moffitt et al . , 2016a; Battich et al . , 2015 ) . An increasingly important application for smISH is the simultaneous localization of customized panels of transcripts in tissue , which is used to validate putative cell subtypes identified by scRNA-seq studies ( Grün and van Oudenaarden , 2015 ) . Performing smISH in intact tissue can also reveal the spatial relationship between the cells expressing secreted signaling factors and the cells expressing the corresponding receptors , information that current scRNA-seq approaches cannot resolve because they require tissue dissociation with irretrievable loss of spatial context . Finally , when applied on a genome-wide scale in tissues , smISH has the potential to entirely bypass scRNA-seq as an upfront discovery tool . The development of multiplexed smISH for use in tissue has been challenging due to autofluorescent background and light scattering ( Shah et al . , 2016; Sylwestrak et al . , 2016; Moffitt et al . , 2016b; Chen et al . , 2016; Choi et al . , 2014; Lyubimova et al . , 2013 ) . One strategy for addressing this problem is to amplify probe signals by the hybridization chain reaction ( HCR , reviewed in ( Choi et al . , 2016 ) ; see also ( Wang et al . , 2012 ) for branched-DNA amplification ) , which provides up to five orthogonal detection channels . Higher levels of multiplexing can be achieved by repeated cycles of RNA in situ hybridization followed by a re-amplification step ( Shah et al . , 2016 ) , but because a single round of probe hybridization in tissue sections takes hours , multiplexing with HCR is laborious . Unamplified smISH techniques have the practical advantage that hundreds of endogenous RNA species can be barcoded in a single reaction , and then read out with rapid label-image-erase cycles ( Moffitt et al . , 2016b; Moffitt et al . , 2016c ) , but these do not provide adequate signal in tissues . Ideally , a technique for high-throughput profiling would combine all of the RNA probe hybridization and signal amplification steps into a single reaction . Previously , Nilsson and colleagues presented an elegant enzymatic solution to this problem ( Larsson et al . , 2010; Ke et al . , 2013 ) . They used barcoded padlock probes to label cDNA molecules in cells and tissues , and rolling-circle amplification ( RCA ) to transform the circularized probes into long tandem repeats . The approach worked in tissues and handled an unbounded number of orthogonal amplification channels . The only limitations were that the RNA-detection efficiency was capped at about 15% ( each transcript could only be probed at a single site because the 3' end of the cDNA served as the replication primer ) , and that the approach required an in situ reverse transcription step with specialized and costly locked nucleic-acid primers . Here , we report an in situ hybridization technique with performance characteristics that enable rapid and scalable single-cell expression profiling in tissue . Our approach is a simplified variant of the padlock/RCA technique which replaces padlock probes with RNA-templated proximity ligation ( Söderberg et al . , 2006; Frei et al . , 2016 ) at Holliday junctions ( Labib et al . , 2013 ) ; hence , we term it proximity ligation in situ hybridization ( PLISH ) . As demonstrated below , PLISH generates data of exceptionally high signal-to-noise . Multiplexed hybridization and signal amplification of all target RNA species is carried out in a single parallel reaction , and the RNAs are then localized with rapid label-image-erase cycles . PLISH exhibits high detection efficiency because it probes multiple sites in each target RNA , and high specificity because of the proximity ligation mechanism . PLISH utilizes only commodity reagents , so it can be scaled up inexpensively to cover many genes . It works well on conventional formalin-fixed tissues that have been cryo- or paraffin-embedded , and can be performed concurrently with immunostaining , making it extremely versatile . Using the murine lung as a characterized model tissue , we show that multiplexed PLISH can rediscover and spatially map the distinct cell types of a tissue in an automated and unsupervised fashion . An unexpected discovery from this experiment is that murine Club cells separate into two populations that differ molecularly and segregate anatomically . PLISH constitutes a novel , single cell spatial-profiling technology that combines high performance , versatility and low cost . Because of its technical simplicity , it will be accessible to a broad scientific community .
Proximity ligation at Holliday junctions offers a simple mechanism for the amplified detection of RNA ( Labib et al . , 2013 ) . First , a transcript is targeted with a pair of oligonucleotide 'H' probes designed to hybridize at adjacent positions along its sequence ( Figure 1A ) . The left H probe includes a single-stranded 5' overhang while the right probe includes a 3' overhang . Importantly , target RNAs can be tiled with H probe pairs at multiple sites , which is critical for efficient detection of low abundance transcripts ( Figure 1B ) . The overhangs are then hybridized to 'bridge' and linear 'circle' oligonucleotides with embedded barcode sequences to form a Holliday junction structure , after which ligation at the nick sites creates a closed circle . Finally , the 3' end of the right H probe primes rolling-circle replication , which generates a long single-stranded amplicon of tandem repeats . Addition of fluorescently-labeled 'imager' oligonucleotides complementary to the barcodes generates an extremely bright punctum at the site of each labeled transcript . Because each barcode sequence is unique , the puncta derived from different target RNAs can be labeled with different colors ( Figure 1C ) . To implement PLISH , we adapted protocols for antibody-based proximity ligation ( Söderberg et al . , 2006 ) . The technique utilizes conventional oligonucleotides , two commercially available enzymes , and procedures familiar to molecular biologists . The ligase and polymerase enzymes are less than half the size of an immunoglobulin G , and they diffuse at least as rapidly as the 60mer DNA hairpins used for HCR amplification ( Choi et al . , 2014; Joubert et al . , 2003; Lapham et al . , 1997; Modrich et al . , 1973 ) . Our initial studies produced bright puncta that were absent if any of the oligonucleotide or enzyme reagents was withheld . The signal from the individual RCA amplicons exceeded cellular and tissue fluorescence background by more than 30-fold , rendering autofluorescence inconsequential ( Figure 1—figure supplements 1A–B and [Jarvius et al . , 2006; Blab et al . , 2004] ) . Histograms of puncta intensities fit to a negative binomial distribution , as expected for a DNA replication process that terminates stochastically and irreversibly ( Figure 1—figure supplement 1C–F ) . The coefficients of variation for the puncta intensity distributions were typically between one and two . The requirement for coincident hybridization of two probes at adjacent sites in an RNA transcript should make PLISH highly specific . To evaluate this , we performed several experiments . First , we used PLISH to detect the transcription factor SRY-box 4 ( SOX4 ) in cultured HCT116 cells . A pool of ten H probesets exhibited much higher RNA detection efficiency than a single H probeset , as expected ( Figure 1D ) . However , when the RNA-recognition sequence of either the left or right H probe in each set was scrambled , there were no detectable puncta . Thus , both H probes had to be correctly targeted to generate a signal . Second , we tested the sequence-specificity of the PLISH signal in tissue by pre-incubating samples with antisense 'blocking' oligonucleotides complementary to the target RNA at the H probe hybridization sites . For these experiments , we stained mouse lung sections for secretoglobin 1a1 ( Scgb1a1 ) , a marker of airway Club cells . Antisense oligonucleotides drastically attenuated the number of PLISH puncta , whereas scrambled blocking oligonucleotides of the same length had no apparent effect ( Figure 1E ) . Third , we analyzed murine lung sections for the co-localization of the mRNA transcript and protein product of surfactant protein C ( Sftpc ) , which is expressed in alveolar epithelial type II ( AT2 ) cells . Of the cells that were positive for PLISH signal , 98 . 5% were also positive for antibody staining ( n = 184 , Figure 1—figure supplement 2 ) . This level of specificity is excellent relative to HCR-amplified smISH , where off-target binding of hybridization probes can account for a quarter of the observed puncta ( Shah et al . , 2016 ) . To quantify the sensitivity and accuracy of RNA detection , we benchmarked PLISH measurements against a reference-standard dataset of single-cell , quantitative reverse transcription polymerase chain reaction ( qPCR ) and RNA-seq measurements on HCT116 cells ( Wu et al . , 2014 ) . For genes with fragment-per-kilobase-per-million-read ( FPKM ) values greater than one , the single-cell qPCR technique detected mRNA in >90% of the cells ( Figure 1F ) . However , the fraction of transcript-positive cells dropped quickly between FPKM values of 1 and 0 . 1 . A fit of the qPCR data to a Poisson sampling model suggested that an FPKM value of one corresponded to 2 . 5 copies per cell ( see also [Marinov et al . , 2014; Battich et al . , 2013] ) . The PLISH technique detected RNA transcripts with a sensitivity comparable to single-cell qPCR . For example , Caspase-9 ( CASP9 ) has an FPKM value of 2 , and it was observed in 100% of the cells by PLISH . We detected an average of 8 puncta per cell , which is consistent with the prediction of 5 copies per cell from the fit to the qPCR data ( Figure 1F , inset ) . For a set of ten genes covering the full spectrum of expression levels in HCT116 cells , the number of PLISH puncta per cell correlated with bulk FPKM values ( Figure 1G ) . To quantify RNA-detection efficiency in tissue , we marked a set of axin 2 ( Axin2 ) transcripts in mouse lung sections using an HCR-amplified smISH procedure ( Choi et al . , 2014 ) and then determined the fraction of the marked transcripts that could be identified by PLISH . We chose the Axin2 gene because of its low expression level in the lung . HCR detected a sparse population of cells with one to two puncta each ( the HCR detection efficiency was low because we used a single HCR probe rather than 24 ) . PLISH puncta generated with a pool of four H probe pairs co-localized with 32% of the HCR puncta ( Figure 1—figure supplement 3 ) . Thus , the four PLISH probesets detected Axin2 transcripts with a composite efficiency of 32% and an average per-site efficiency of 9% . This probe efficiency matches or exceeds that of other smISH techniques . The PLISH detection efficiency can be tuned on a per gene basis by altering the number of H probe pairs . Decreasing the number of probesets pro-rates the number of puncta from highly-expressed genes , while increasing the number of probesets can facilitate sensitive detection of very low-abundance transcripts . We next characterized the performance of PLISH for low-plex RNA localization in tissues . This experimental format uses a disposable hybridization chamber that is sealed to a coverslip or slide surrounding a tissue section ( Figure 2A ) . PLISH detection of up to 5 RNA species is accomplished by stepwise application of reagents through the inlet and outlet ports of the chamber . The puncta from each RNA species are then labeled in a unique color by hybridization to 'imager' oligonucleotides with spectrally-distinct fluorophores . After imaging , the fluorescence micrographs are interpreted by direct visual inspection . We aimed to test whether PLISH provides single-molecule and single-cell resolution in tissues , whether it robustly detects low-abundance RNA species , whether the spatial distribution of RNA is consistent with prior knowledge , whether PLISH is compatible with simultaneous immunostaining , and whether it is compatible with formalin-fixed , paraffin-embedded ( FFPE ) samples . First , we analyzed murine lung sections for RNA expression of the ciliated-cell marker Forkhead box J1 ( Foxj1 ) , and the Club-cell marker Scgb1a1 . Foxj1 is a low-abundance transcript with an FPKM value of 10 in ciliated cells , as measured by scRNA-seq ( Treutlein et al . , 2014 ) . We observed single cells with multiple discrete Foxj1 puncta in the terminal bronchiolar epithelium , surrounded by numerous strongly Scgb1a1 positive cells ( Figure 2B ) . These data establish PLISH's single-molecule and single-cell resolution in tissues , and its ability to detect low-abundance transcripts . Second , we analyzed human lung FFPE sections for RNA expression of SCGB1A1 , and for protein expression of the basal cell marker , Keratin 5 ( KRT5 ) . To do this , we appended two antibody incubation steps to the standard PLISH protocol . Strongly SCGB1A1 positive cells were localized to the lumen of the airways , overlying KRT5 positive cells ( Figure 2C ) , matching the known anatomical distribution of Club and basal cells , respectively . These data establish PLISH's compatibility with simultaneous immunostaining , and with FFPE samples . Third , we analyzed murine lung sections for RNA expression of three genes: the AT2 cell marker Sftpc , the macrophage-enriched marker Lysozyme 2 ( Lyz2 ) , and Scgb1a1 . Overlays of the three channels provided a striking visual depiction of the different cell types . Macrophages were bright in the Lyz2 channel , but absent in the other channels ( Figure 2D ) . AT2 cells were bright in the Sftpc channel , moderately bright in the Lyz2 channel and absent in the Scgb1a1 channel ( Figure 2D , white cells in the overlay ) . Club cells were very bright in the Scgb1a1 channel , but otherwise absent ( Figure 2E ) . Finally , putative bronchioalveolar stem cells ( Kim et al . , 2005 ) ( BASCs ) were bright in the Sftpc channel with a weak punctate signal in the Scgb1a1 channel ( Figure 2E ) . Thus , raw PLISH data can be interpreted without any computational processing , made possible by PLISH's exceptional signal-to-noise in tissues . We also evaluated how PLISH performs in primary samples of diseased human tissue , to assess whether it will be useful for molecular analysis of the many human diseases that cannot be accurately modeled in animals . One example is idiopathic pulmonary fibrosis ( IPF ) , a fatal lung disease of unknown pathogenesis ( Travis et al . , 2013 ) . The diagnosis of IPF is based on the presence of specific histological features , including clusters of spindle-shaped fibroblasts , stereotyped 'honeycomb' cysts , and epithelial cell hyperplasia . In this regard , single-cell profiling approaches that operate on dissociated tissue ( Xu et al . , 2016 ) are intrinsically limited because they cannot correlate molecular data with cytologic and spatial features . As a preliminary test , we used PLISH to analyze RNA expression of the AT2 cell marker SFTPC in resected lung tissue from control and IPF patients . In contrast to the uniformly cuboidal SFTPC-expressing AT2 cells distributed throughout alveoli of non-IPF lungs ( Figure 2F ) , we observed clusters of SFTPCHi cells of heterogeneous size and varying degrees of flattening lining the airspace lumen of IPF lungs . Surprisingly , many cells that did not appear to be epithelial ( i . e . , they were not lining an airway lumen ) expressed SFTPC at low levels . Based on this pilot experiment , PLISH should be a suitable tool for building atlases of RNA expression in human disease . The PLISH data can be overlaid with monoclonal antibody staining patterns that are the mainstay of pathologic diagnosis and classification . Highly multiplexed measurement of different RNA species requires iterated data collection cycles , since conventional fluorescence microscopy only provides up to five channels ( Figure 3A ) . The data collection cycles include fluorescent labeling of a subset of the 'barcodes' ( i . e . , unique nucleotide sequences complementary to fluorescently labeled 'imager' oligonucleotides ) in a sample , imaging of the labeled transcripts , and erasure of the fluorescent signal . Ideally , the cycles should be fast , and the erasure should not cause any mechanical or chemical damage to the sample . Consistent with prior work , we found that PLISH puncta could be imaged in the presence of a 3 nM background of freely-diffusing imager oligonucleotides ( Figure 3—figure supplements 1A and [Blab et al . , 2004] ) . This allowed us to streamline data collection by eliminating a wash step , and also presented a simple erasure strategy . By using short imager oligonucleotides that equilibrate rapidly on and off of RCA amplicons ( Jungmann et al . , 2014 ) , we could erase fluorescence from a previous cycle by a simple buffer exchange ( Figure 3—figure supplement 1B ) . We also established an erasure method based on uracil-containing imager oligonucleotides , which were removed with a 15 min enzymatic digestion ( Figure 3—figure supplement 1C ) . Thus , we could image PLISH puncta in five different color channels with spectrally-distinct fluorophores , and we were able to complete cycles in as little as 20 min , which approaches the cycle time of an Illumina MiSeq instrument ( https://support . illumina . com ) . To demonstrate and validate the multiplexing capacity of PLISH , we co-localized the mRNA of eight selected genes in ~2900 single cells from an adult mouse lung ( Figure 3B ) . Our panel included four commonly used lung cell type markers that have been previously characterized , and four ubiquitously expressed genes . The targeted transcripts were Sftpc ( AT2 cells ) , advanced glycosylation end product-specific receptor ( Ager , AT1 cells ) , Scgb1a1 ( Club cells ) , Lyz2 ( macrophage and AT2 cell subset ) , ferritin light polypeptide 1 ( Ftl1 ) , beta actin ( Actb ) , inactive X specific transcripts ( Xist ) , and glyceraldehyde-3-phosphate dehydrogenase ( Gapdh ) . The eight RNA species were barcoded in a single PLISH reaction , and the data were collected with a pair of label-image-erase cycles using the enzymatic erasure approach described above ( Figure 3A ) . A nuclear counterstain ( DAPI ) and transmitted light micrograph were also obtained . To quantify the expression of all eight genes on a per-cell basis , we created a PLISH-specific pipeline in CellProfiler , an open-source software package ( Kamentsky et al . , 2011 ) . The pipeline first identified nuclei in the DAPI channel , which were used as anchor points for expansion to full-cell assignments . Fortuitously , the bulk of the detected mRNAs in AT1 cells , which have an extremely flat and broad morphology , were clustered around the nuclei . We summed the PLISH signal for each gene in the nuclear and peri-nuclear regions of each cell , and saved the results as single-cell expression profiles indexed on anatomical location . We also created a utility to pseudocolor cells in a transmitted light micrograph according to their inferred cell type ( see below ) , so that we could visualize the relationship between cellular gene expression and anatomical localization . An important scientific challenge is to identify and map all of the molecularly distinct cell types that make up complex tissues , and in situ single-cell profiling should be a powerful tool for working towards this goal . As a proof-of-concept for this , we asked whether known lung cell types could be rediscovered by an automated and unsupervised analysis of our multiplexed PLISH data set . We used two standard data analysis tools , K-means clustering ( Figure 3C ) and t-distributed stochastic neighbor embedding ( van der Maaten and Hinton , 2008 ) ( t-SNE , Figure 3D ) , to classify and visualize the entire population of cells . The automated analysis identified ten cell classes , four of which were labeled 'other' because they were defined primarily by 'signature' profiles of ubiquitously-expressed genes . The remaining six classes were associated with a known lung cell type based on marker-gene expression . The Sftpc and Scgb1a1 positive cell classes were labeled as AT2 and Club , respectively , while the Lyz2 positive class was labeled as macrophage ( one of the two AT2 cell classes was also Lyz2 positive as previously reported in ( Desai et al . , 2014 ) , Figure 4—figure supplement 1A ) . The cell class with the highest Ager expression was labeled as AT1 , but Ager mRNA was also detected in a subset of AT2 and Club cells , and in one of the four 'other' cell classes , indicating it is not particularly specific for AT1 cells . We validated the PLISH results by indirect immunohistochemistry ( Figure 4—figure supplement 1B ) and by comparison with previously published scRNA-seq data ( Figure 4—figure supplement 1C ) , which confirmed the low specificity of Ager for AT1 cells . We also analyzed the RNA expression of Akap5 , another transcript that is highly-enriched in AT1 cells ( Treutlein et al . , 2014 ) , and found that its localization correlated closely with Ager's ( Figure 4—figure supplement 1D ) . For a higher-resolution analysis of cellular gene expression , we examined the expression pattern of individual genes in re-colored t-SNE plots ( Figure 4A ) . We found a small cloud of cells between the Club and AT2 clusters that expressed both Sftpc and Scgb1a1 . On the basis of this dual expression , we assigned them as the BASC type ( Kim et al . , 2005 ) . We also noted that Lyz2 expression partitioned the AT2 cells into two classes designated Lyz2+ and Lyz2- , while Actb segregated Club cells into two classes designated ActbHi and ActbLo . Gapdh was the most uniformly expressed transcript , consistent with its role as a 'housekeeping' gene ( Figure 4B ) . Ftl1 expression was highest in alveolar macrophages , as expected , where it is believed to play a role in processing iron from ingested red blood cells ( McGowan and Henley , 1988 ) . Unexpectedly , Ftl1 was also highly expressed in Club cells . Actb expression was highest in macrophages , presumably because of its functional role in motility , and in AT1 cells , which must maintain a flat morphology and expansive cytoskeleton ( Foster et al . , 2010 ) . To validate the PLISH results , we pseudocolored the cells in transmitted-light images according to their class ( Figure 4C–D ) . Importantly , no spatial information was included in the k-means clustering . Several observations confirmed the accuracy of the automated classification . First , the Club cell class mapped perfectly onto the bronchial epithelium , while cells from the AT1 and AT2 classes were distributed throughout the alveolar compartment . The rare BASCs also localized precisely to the bronchioalveolar junctions , where they have been shown to reside by immunostaining ( Kim et al . , 2005 ) ( Figure 4C and Figure 4—figure supplement 1E ) . The macrophage class was primarily found inside the alveolar lumen , and many exhibited a characteristic rounded cell shape . The Otherd class of cells was enriched in pulmonary arteries , and therefore might represent endothelial or perivascular cells . We further observed a striking spatial segregation of the two Club cell classes . ActbHi Club cells clustered together at the bronchial terminus , while ActbLo Club cells populated more proximal domains ( Figure 4D–E ) . While the significance of this pattern is not immediately obvious , it emphasizes how PLISH can readily integrate molecular and spatial features of single cells to generate insights that would be missed with either piece of information alone .
PLISH represents a practical technology for multiplexed expression profiling in tissues . It combines high performance in four key areas: specificity , detection efficiency , signal-to-noise and speed . The specificity derives from coincidence detection , which requires two probes to hybridize next to one another for signal generation . Efficient detection of low-abundance transcripts is accomplished by targeting multiple sites along the RNA sequence . Enzymatic amplification produces extremely bright puncta , and allows many different RNA transcripts to be marked with unique barcodes in one step . The different RNA transcripts can then be iteratively detected to rapidly generate high dimensional data . While low-plex PLISH on a handful of different genes can be valuable , the PLISH technology is also scalable , without requiring specialized microscopes ( or other equipment ) , software , or computational expertise . The oligonucleotides and enzymes are inexpensive and commercially available from multiple vendors . The H probes are the cost-limiting reagent , but can be synthesized in pools ( Murgha et al . , 2014; Beliveau et al . , 2012 ) . Assuming five pairs of H probes for each target RNA species , and 20 cents for a 40mer oligonucleotide , the cost of PLISH reagents amounts to $3 per gene . It should therefore be practical to simultaneously interrogate entire molecular systems , such as signaling pathways or super-families of adhesion receptors . The high specificity and signal-to-noise of PLISH will be advantageous for deep profiling , where non-specific background increases with increasingly complex mixtures of hybridization probes ( Moffitt et al . , 2016c ) . Our initial studies demonstrate PLISH's capacity for rapid , automated and unbiased cell-type classification , and illustrate how it can complement single-cell RNA sequencing ( sc-RNAseq ) . Sc-RNAseq offers greater gene depth than in situ hybridization approaches , but it is less sensitive , fails to capture spatial information , and induces artefactual changes in gene expression during tissue dissociation ( van den Brink et al . , 2017; Lee et al . , 2015 ) . PLISH provides the missing cytological and spatial information , and it is applied to intact tissues . Going forward , sequencing can be used to nominate putative cell types and molecular states based on the coordinate expression of 'signature genes' , and multiplexed PLISH can be used to distinguish true biological variation from technical noise and experimentally-induced perturbations . Importantly , multiplexed PLISH provides the tissue context of distinct cell populations , which is essential for understanding the higher-order organization of intact systems like solid tumors and developing organs . In diseases like IPF where morphology and gene expression are severely deranged ( Xu et al . , 2016 ) , histological , cytological and spatial features may even be essential for making biological sense of sequencing data . Currently , efforts are underway to more deeply characterize cellular states by integrating diverse types of molecular information . We have already demonstrated the combined application of PLISH with conventional immunostaining . Going one step further , oligonucleotide-antibody conjugates make it possible to mix and match protein and RNA targets in a multiplexed format ( Weibrecht et al . , 2013 ) . The generation of comprehensive , multidimensional molecular maps of intact tissues , in both healthy and diseased states , will have a fundamental impact on basic science and medicine .
Unless otherwise specified , all reagents were from Thermo-Fisher and Sigma-Aldrich . Oligonucleotides were purchased from Integrated DNA Technologies . T4 polynucleotide kinase , T4 ligase , USER enzyme and their respective buffers were purchased from New England Biolabs . Nxgen phi29 polymerase and its buffer were purchased from Lucigen . Abbreviations: BSA , bovine serum albumin; DAPI , 4 , 6-diamidino-2-phenylindole; DEPC , diethylpyrocarbonate; EDTA , ethylenediaminetetraacetic acid; min , minutes; PBS , phosphate buffered saline; PFA , paraformaldehyde; RCA , rolling circle amplification; RT , room temperature . All oligonucleotide sequences are listed in Supplementary file 1 . HCT116 cells ( ATCC; CCL-247 ) were authenticated by HLA typing and confirmed negative for Mycoplasma contamination using PCR . Cells were grown on poly-lysine coated #1 . 5 coverslips ( Fisherbrand 12–544 G ) using standard cell culture protocols until they reached the desired confluency . The cells were rinsed in 1X PBS and fixed in 3 . 7% formaldehyde with 0 . 1% DEPC at RT for 20 min . The fixed cells were treated with 10 mM citrate buffer ( pH 6 . 0 ) at 70°C for 30 min , dehydrated in an ethanol series , then enclosed by application of a seal chamber ( Grace Biolabs 621505 ) to the coverslip . Lungs were collected from adult B6 mice ( Jackson Labs ) and fixed by immersion in 4% PFA as previously described ( Desai et al . , 2014 ) . Non-IPF human lung tissue was obtained from a surgical resection , and IPF tissue from an explant . All mouse and human research were approved by the Institutional Animal Care and Use Committee and Internal Review Board , respectively , at Stanford University . The tissues were fixed by immersion in 10% neutral buffered formalin in PBS at 4°C overnight under gentle rocking , cryoprotected in 30% sucrose at 4°C overnight , submerged in OCT ( Tissue Tek ) in an embedding mold , frozen on dry ice , and stored at −80°C . 20 μm sections were cut on a cryostat ( LeicaCM 3050S ) and collected on either poly-lysine coated #1 . 5 coverslips or glass slides ( Fisherbrand Superfrost ) , air dried for 10 min , and post-fixed with 4% PFA at RT for 20 min . The human lung tissue in Figure 2A was formalin-fixed and paraffin-embedded ( FFPE ) according to standard protocols , and 20 μm sections were cut on a microtome and collected on glass slides . The FFPE sections were deparaffinized by immersion in Histoclear ( National Diagnostics , HS-200 ) for 3 × 5 min , then dehydrated in an ethanol series and post-fixed with 4% PFA at RT for 20 min . Tissue sections were treated with 10 mM citrate buffer ( pH 6 . 0 ) containing 0 . 05% lithium dodecyl sulfate at 70°C for 30 min , or in some experiments , with 0 . 1 mg/ml Pepsin in 0 . 1M HCl for 8 min at 37°C and dehydrated in an ethanol series . Following treatment , sections were air dried for 10 min and enclosed by application of a seal chamber . Target RNAs were probed at ~40 nucleotide detection sites , with 1 to 10 sites per RNA species depending on expression level . NCBI BLAST searches were used to eliminate detection sites that shared 10 or more contiguous nucleotides with a non-target RNA . The detection sites were also selected to minimize self-complementarity as indicated by the IDT oligo analyzer . Each detection site was targeted with a pair of H probes designated HL ( left H probe ) and HR ( right H probe ) . The HL and HR probes included ~20 nucleotide binding sequences that were complementary respectively to the 5' and 3' halves of the detection site . The binding sequences were chosen so that the 5' end of the HL binding sequence and the 3' end of the HR binding sequence would abut at a 5’-AG-3’ or a 5’-TA-3’ dinucleotide in the target RNA . The lengths of the binding sequences were adjusted so that the melting temperature of the corresponding DNA duplex would fall between 45–65°C as computed by IDT Oligo analyzer using default settings of 0 . 25 μM oligo concentration and 50 mM salt concentration . To generate H probes , suitable HL and HR binding sequences were catenated at their respective 5' and 3' ends with overhang sequences taken from one of eight modular design templates ( Supplementary file 1 ) . The left and right overhang sequences in each design template were complementary to a specific bridge ( B ) and circle ( C ) oligonucleotide , which directed a desired fluorescent readout . The design templates reported here utilized a common 31 base oligonucleotide for the bridge . Following previous work ( Söderberg et al . , 2006 ) , the circle oligonucleotides were ~60 bases long with 11 base regions of complementarity to cognate H probes on either end . The circle sequences were chosen to minimize self-complementarity . Each imager oligonucleotide was complementary to a barcode embedded in one of the C oligonucleotides , allowing unique detection of the corresponding RCA amplicon . The H-probe oligonucleotides were ordered on a 25 nanomole scale with standard desalting . The B and C oligonucleotides were ordered on a 100 nanomole scale with HPLC purification , and phosphorylated with T4 polynucleotide kinase according to the manufacturer recommendations . Imager oligonucleotides were purchased either as HPLC-purified fluorophore conjugates ( A488 , Texas Red , Cy3 , Cy5 ) , or as amine-modified oligonucleotides that were subsequently coupled to Pacific Blue-NHS ester according to the manufacturer recommendations . Six buffers were used for PLISH barcoding: H-probe buffer ( 1M sodium trichloroacetate , 50 mM Tris pH 7 . 4 , 5 mM EDTA , 0 . 2 mg/mL Heparin ) , bridge-circle buffer ( 2% BSA , 0 . 2 mg/mL heparin , 0 . 05% Tween-20 , 1X T4 ligase buffer in RNAse-free water ) , PBST ( PBS + 0 . 1% Tween-20 ) , ligation buffer ( 10 CEU/μl T4 DNA ligase , 2% BSA , 1X T4 ligase buffer , 1% RNaseOUT and 0 . 05% Tween-20 in RNAse-free water ) , labeling buffer ( 2x SSC/20% formamide in RNAse-free water ) , and RCA buffer ( 1 U/μl Nxgen phi29 polymerase , 1X Nxgen phi29 polymerase buffer , 2%BSA , 5% glycerol , 10 mM dNTPs , 1% RNaseOUT in RNAse-free water ) . An H cocktail was prepared by mixing H probes in H-probe buffer at a final concentration of 100 nM each . If an RNA was targeted with more than five probe sets , the concentrations of the H probes for that RNA were pro-rated so that their sum did not exceed 1000 nM . A BC cocktail was also prepared by mixing B and C oligonucleotides in bridge-circle buffer at a final concentration of 6 μM each . Single-step barcoding was performed in sealed chambers . The workflow consisted of three steps: ( i ) The sample was incubated in the H cocktail at 37°C for 2 hr . The sample was then washed 4 × 5 min with H-probe buffer at RT , and incubated in the BC cocktail at 37°C for 1 hr . ( ii ) Following a 5 min wash with PBST at RT , the sample was incubated in ligation buffer at 37°C for 1 hr . ( iii ) The sample was washed 2 × 5 min with labeling buffer at RT , and washed with 1X Nxgen phi29 polymerase buffer at RT for 5 min . The sample was then incubated in RCA buffer at 37°C for 2 hr ( typical for cultured cells ) to overnight ( typical for tissue ) . Finally , the sample was washed 2 × 5 min with labeling buffer . Barcoded PLISH samples were fluorescently labeled by two different procedures , designated 'washout' and 'fast' . In the washout procedure , the sample was incubated with imager oligonucleotides in imager buffer ( labeling buffer with 0 . 2 mg/mL heparin ) at a final concentration of 100 nM each for 30 min , and then washed 2 × 5 min with PBST at RT . In the fast procedure , the sample was incubated for 5 min with imager oligonucleotides in imager buffer at a final concentration of 3 nM each , and then imaged immediately . Samples that did not require label-image-erase cycles were stained with DAPI ( stock 1 mg/ml; final concentration - 1:1000 in PBS ) for 5 min and mounted in H-1000 Vectashield mounting medium ( Vector ) . Data were collected by confocal microscopy ( Leica Sp8 and Zeiss LSM 800 ) using a 40X oil immersion or a 25X water immersion objective lens . 20 μm z-stacks were scanned , and maximum projection images were saved for analysis . For 5-color experiments , DAPI was added after the Pacific Blue channel had been imaged , and the Texas Red and Cy3 channels were linearly unmixed using Zeiss software . Transmitted light images were acquired on a Leica Sp8 confocal microscope using the 488 nm Argon laser and the appropriate PMT-TL detector . Images from serial rounds of data collection were aligned using the nuclear stain from each round as a fiducial marker . Unless otherwise stated , imaging data of cells and mouse lung tissue are representative of three independent experiments with ≥4 fields of view each . Imaging data of human lung tissue are representative of two independent experiments with ≥4 fields of view each . HCR was performed following a published protocol ( Choi et al . , 2014 ) with probes that targeted two sites covering nucleotides 621–670 and 1159–1208 in the mouse Axin2 transcript , and AlexaFluor 488-/AlexaFluor 647-labeled amplifier oligonucleotides . The samples were then processed for PLISH with H probes targeting four sites covering nucleotides 347–386 , 1878–1917 , 2412–2451 and 2956–2995 in the Axin2 transcript , and imaged using a Cy3-labeled imager oligonucleotide . PLISH barcoding was performed as described above . Subsequently , the sample was washed 3 × 5 min with PBST at RT , and incubated in blocking solution ( 50 μl/ml [5%] normal goat serum , 1 μl/ml [0 . 1%] Triton X-100 , 5 mM EDTA and 0 . 03 g/ml [3%] BSA in PBS ) at RT for 1 hr . The sample was then incubated with primary antibody ( Rabbit anti-pro-Sftpc , Millipore , 1:500 or Rabbit anti-Cytokeratin 5 , Abcam Ab193895 , 1:400 ) in blocking solution at 37°C for 2 hr under gentle rocking , washed 4 × 5 min with PBST at RT , and incubated with secondary antibody ( Goat anti-Rabbit-Cy5 , Jackson Lab , 1:250 ) and DAPI ( 1:1000 ) in blocking solution at RT for 1 hr . The sample was washed 3 × 5 min in PBST at RT and mounted in H-1000 Vectashield . Mouse lung tissue cryosections were collected on slides , post-fixed and processed as described above . The samples were incubated with a 60-base oligonucleotide complementary to nucleotides 219–278 in the Scgb1a1 mRNA , or with a scrambled 60-base oligonucleotide , at 100 nM final concentration in H-probe buffer at 37°C for 2 hr . The samples were then washed 2 × 5 min with H-probe buffer at RT , and processed for PLISH using H probes that targeted nucleotides 229–268 in the Scgb1a1 transcript . To perform enzymatic erasure , 15–20 base imager oligonucleotides were ordered with the dT nucleotides replaced by dU nucleotides . Following imaging , the signal was erased by incubating the sample with 0 . 1 U/μL USER enzyme in 1X USER enzyme buffer at 37°C for 20 min , followed by washing 2 × 3 min with PBST at RT . To perform rapid erasure , short 10–11 base oligonucleotides were ordered . Following imaging , the signal was erased by incubating the sample with PBST at 37°C for 15 min . Lungs collected from B6 and the Lyz2+/EGFP mouse strains ( Faust et al . , 2000 ) were fixed and immunostained as whole mounts as previously described ( Desai et al . , 2014 ) . Primary antibodies were chicken anti-GFP ( Abcam ab13970 ) , rat anti-Ecad/Cdh1 ( Invitrogen ECCD-2 ) , goat anti-Scgb1a1 ( gift from Barry Stripp ) , rabbit anti-pro-Sftpc ( Chemicon AB3786 ) , and rat anti-Ager ( R and D MAB1179 ) . Fluorophore-conjugated secondary antibodies raised in Goat ( Invitrogen ) or Donkey ( Jackson Labs ) were used at 1:250 and DAPI at 1:1000 . FIJI was used to pseudocolor unprocessed micrographs for display as three-color overlays . A custom CellProfiler ( Kamentsky et al . , 2011 ) pipeline ( Source code 1 ) was created to measure RNA signal intensities at the single-cell level . Briefly , the centers of cell nuclei were first identified as maxima in a filtered DAPI image , and associated with a numerical index . Nuclear boundaries were assigned by a propagation algorithm , and then expanded by ~1 micron to define sampling areas . The following data were then recorded: ( i ) average pixel intensities for each data channel over each sampling area; ( ii ) the coordinates of the sampling areas; ( iii ) shape metrics for the corresponding nuclei; and ( iv ) an image with the boundary pixels of each nucleus set equal to the associated index value . For each RNA species , the PLISH data were first normalized onto a 0:10 scale by dividing through by the largest value observed in any cell over all of the fields of view , and then multiplying by ten . The data were then log-transformed onto a −1:1 scale by the operation: transformed_data = log ( 0 . 1 + normalized_data ) . Custom Matlab scripts were used to perform hierarchical clustering of the log-transformed single-cell expression profiles , to generate heatmaps , and to create images with the boundary pixels of each nucleus colored according to a cluster assignment ( Source code 1 ) . Custom R scripts were used for k-means clustering and to make t-SNE projection plots ( Source code 1 ) . | The human body contains several hundred types of specialized cells that have different roles . The cells form tissues , and each tissue can only work if it has the right cells and if they are correctly organized and distributed to build working structures . This is how the same body parts , e . g . lungs , brains , hearts , end up looking and working the same way in almost everyone . Cells organize themselves into tissues by exchanging short-range messages between nearby cells . Understanding how cells communicate to form , maintain and repair tissues is a challenge for biologists . Finding ways to examine different signals at the same time would improve our understanding of these important processes . Now , Nagendran , Riordan et al . developed a microscopy technique that can tackle this issue . The cells can be stained and tagged with already established dyes and markers to measure the location and signals of all cell groups at the same time . By calculating how these cells are distributed in space it is then possible to estimate how they interact with each other . Based on this , Nagendran , Riordan et al . then successfully tested their tool in lung tissues from mice and humans . This cheap , high-speed technology works on tissue samples from any animal , including humans , and can be easily combined with existing technologies and so be adapted for a wide range of uses . A deeper knowledge of how combinations of signals guide tissue formation and maintenance could help us to better understand what causes developmental diseases , organ failures and cancer . Tools like this could even help to identify key targets for new treatments . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"tools",
"and",
"resources"
] | 2018 | Automated cell-type classification in intact tissues by single-cell molecular profiling |
Active flight requires the ability to efficiently fuel bursts of costly locomotion while maximizing energy conservation during non-flying times . We took a multi-faceted approach to estimate how fruit-eating bats ( Uroderma bilobatum ) manage a high-energy lifestyle fueled primarily by fig juice . Miniaturized heart rate telemetry shows that they use a novel , cyclic , bradycardic state that reduces daily energetic expenditure by 10% and counteracts heart rates as high as 900 bpm during flight . Uroderma bilobatum support flight with some of the fastest metabolic incorporation rates and dynamic circulating cortisol in vertebrates . These bats will exchange fat reserves within 24 hr , meaning that they must survive on the food of the day and are at daily risk of starvation . Energetic flexibly in U . bilobatum highlights the fundamental role of ecological pressures on integrative energetic networks and the still poorly understood energetic strategies of animals in the tropics .
Energy intake , incorporation and expenditure are fundamental to animal behavior and evolution ( Brown et al . , 2004; Weiner , 1992 ) . Animals must balance between generating enough metabolic power to find and acquire food and maintaining sufficient reserves to sustain daily maintenance , and repair and reproduce . This basic requirement of life can drive the foraging strategies of entire clades ( Williams et al . , 2014 ) and extensive links among various behavioral and physiological strategies have evolved in response to single ecological pressures including diet and pathogen environments ( Cohen et al . , 2012 ) . This is largely a consequence of the sequential and linear process of energetic input ( feeding ) , and that energy expenditure is additive across parallel aspects of physiology ( Weiner , 1992 ) . Energetic networks then link across physiological systems from mitochondrial oxidation to digestion , to respond to changes in resource availability and maintain physiological integrity . Well-adapted energy metabolisms must then both be able to conserve reserves and deliver enormous energetic power outputs in an efficient and effective manner . However , few animal models currently allow us to follow energy from intake to delivery of energetic currency to fuel metabolism , and finally to the countermeasures taken to slow down energetic expenditure when it is not needed . Furthermore , accumulating evidence shows that data collected in laboratory settings may not reflect the full range of strategies animals employ to deal with this energetic dilemma ( Bishop et al . , 2015; Bowlin et al . , 2005; Calisi and Bentley , 2009; Geiser et al . , 2007; Ward et al . , 2002 ) . This makes quantitative data from naturally behaving animals in the wild even more important to test the balance and integration of physiological adaptations to energetic limitations . Flying vertebrates are an excellent example of this balance . While flight is one of the most efficient modes of locomotion per unit distance traveled , it is costlier per unit time than any other mode of locomotion ( Norberg , 1990; Schmidt-Nielsen , 1979 ) . To fulfill the exceptional demands of powered flight , both birds and bats have undergone dramatic physiological reorganization that emphasizes the need to supply fuel to large flight muscles ( Maina , 2000; Norberg , 1990 ) . Bat flight in particular is an extreme case of vertebrate locomotor energetics . In comparison to those of non-flying mammals of comparable size , hearts and lungs of bats are larger and have higher blood oxygen transport potential , delivering more oxygen per heart beat than non-flying terrestrial mammals ( Neuweiler , 2000 ) . Bats use some of the highest mass-specific metabolic rates during flight; 3–5 times greater than any other mammals and maximum increases of 15–16 times minimum resting metabolic rates ( Speakman and Thomas , 2003 ) . This may place bats at their energetic ceiling , and integrated physiological networks that allow them to maintain high metabolic rates at or near their limits over extended periods of time may be under equally strong selection to reduce resting energetic expenditure below what is commonly found in mammals . Bats launch themselves directly into energy-demanding flight at the onset of their activity period and on an empty stomach , fueling flight by limited fat reserves ( Voigt et al . , 2010 ) . They must then efficiently find and ingest food , and make energy available to their metabolism rapidly , as high metabolic rates and small body size place them at risk of starvation if sufficient food is not found . This risk is enhanced in the many species that specialize on ephemeral food sources . One strategy to cope with this energetic vulnerability is through daily reduction of metabolic rate ( torpor ) found in small-bodied bat species especially from the temperate zone . By entering a distinct low-energy state characterized by low body temperature , some bats reduce metabolic rates by 99% during torpor when ambient temperatures are lower than their thermoneutral zone ( Geiser and Stawski , 2011; Ruf and Geiser , 2015 ) . In tropical and sub-tropical regions where ambient temperatures are high , it may be impossible to lower body temperature beyond these critical minimum temperatures to save energy , therefore reductions in heart rate may reflect reductions in cellular respiration rates and gene expression in multiple pathways and be an effective measure of energetic conservation ( Dechmann et al . , 2011; Dzal et al . , 2015; McNab , 1969; Storey and Storey , 2004 ) . This may be particularly important in those that feed on sugar dense foods as they are at the highest risk of starvation ( McNab , 1969; Voigt and Speakman , 2007 ) . Heart rate has a quadratic relationship with metabolic oxygen consumption ( Bishop and Spivey , 2013; Grubb , 1982 ) , and by measuring it directly it is possible to gain insight into energetic expenditure at high temporal resolution . Heart rates in bats may more than double in the transition from rest to flight , reflecting enormous flight power requirements ( Thomas , 1975 ) . Controlled experiments in wind tunnels and laboratory conditions have yielded incredible insight into the regulation of metabolism and energy consumption across a wide variety of activities and physiological states . However , heart rates of exercising animals in nature are unpredictable and metabolic rates measured during wind tunnel flight may not indicate the full scope of natural behavior . In a tropical insectivorous bat , heart rate increases from 129 bpm in the roost to 847 bpm during flight , a six-fold increase that is larger than predicted from other captive bats in wind tunnels ( Dechmann et al . , 2011 ) . Alternatively , heart rates of free-flying animals may be much lower than expected . For example , bar-headed geese traverse the Himalayas with heart rates of 250–475 bpm ( Bishop et al . , 2015 ) , 20% lower than what is expected from captive measures ( Ward et al . , 2002 ) , and during migration , heart rates in Swainson’s thrushes are 10% lower than comparable long flights in wind tunnels ( Bowlin et al . , 2005 ) . This indicates that sustainable metabolic rates possible during exercise may differ greatly from maximal rates or extrapolations in captive studies and we have only been able to get an initial glimpse into the heart rates used by flying bats . Once they begin to feed , bats fuel their enormous demand for power by directly and rapidly metabolizing ingested food , but this can lead to high risk of starvation via rapid fat turnover ( Caviedes-Vidal et al . , 2008; Voigt and Speakman , 2007 ) . One mechanism that may help animals to adjust the timing and intensity of shifts in metabolic scopes are glucocorticoids . They are key integrators between the environment and energy balance that ensure rapid response to changes in energetic needs ( Cohen et al . , 2012 ) . Elevated levels of glucocorticoid hormones in blood plasma suppress glycogen formation and promote gluconeogenesis ( Haase et al . , 2016 ) , fat oxidation ( Brillon et al . , 1995 ) , and play a primary role in energy balance ( Nieuwenhuizen and Rutters , 2008 ) . Most bats that have been studied show high baseline glucocorticoid concentrations ( Reeder et al . , 2004; 2006 ) , which indicates that they are in a ready state to rapidly mobilize glucose and glycogen reserves . By manipulating circulating levels of glucocorticoids or those tied to binding globulin , individual use of energy reserves can be modulated ( Schneider , 2004 ) . Bats are then faced with an energetic dilemma where they must rapidly power flight , but quickly switch to conserving energetic stores gained during foraging . To better understand the interplay of energy expenditure and conservation , we describe the daily energetic life of Peters’ tent-making bat ( Uroderma bilobatum , family Phyllostomidae ) in Gamboa , Panamá . These bats are central-place foragers that leave a stable roost location to feed primarily on juice extracted from ripe figs ( Ficus spp ) . We hypothesized that they would not use torpor during their regular daily life and that their energy intake and turnover rates would be high . Daily energy intake and expenditure should then be closely matched , resulting in a specialized life-style at the energetic edge . Testing our hypotheses was made possible by newly miniaturized heart rate transmitters to describe both the activity patterns of the species and their instantaneous energetic expenditure throughout the day , including the first flying heart rates of free-ranging individuals . We also tested how these bats fuel their metabolism through measurement of metabolic incorporation rates and fat turnover from stable isotope ratios in their breath in short-term captivity ( Voigt and Speakman , 2007 ) . In combination with an estimate of energy mobilization potential via elevated circulating cortisol , this allows us a more complete view into how these small-bodied , high-metabolic frugivores meet daily energetic demands .
We tracked the heart rates ( Figure 1—figure supplement 1 ) of four bats for 13 . 6 ± 4 . 9 hr ( mean ± SD ) each day for two to four days ( 13 days total ) . This included 4 . 03 ± 0 . 05 hr of activity outside of the roost at night and the approximately 12 hr that bats spend in their roost during the day for 350 hr of total recording time . All bats left their roosts between 18:00 – 18:30 and flew three to seven minutes to their initial foraging sites . Bats executed multiple short flights of 1–2 min each ( mean ± SD: 1 ± 1 . 5 min ) that were consistent with flying to a fruiting tree , selecting a fruit , and carrying it to a separate feeding perch . During our tracking , flight accounted for 13 ± 6% ( 30 . 6 ± 15 . 6 min ) of the time outside of the roost . We were able to locate several food trees , all of which were Ficus insipida , but all bats also fed for short periods across the Panama Canal at sites inaccessible during tracking . Bats returned to their day roosts between 22:30 – 06:00 . When bats returned early in the night , they left for an additional one to two hours later in the morning . The minimum time that we tracked a bat foraging , including short bouts away from the roosts was two hours and the maximum total time outside of the roost was 10 hr . All bats returned to their home roost each night where they remained for the rest of the day . Uroderma bilobatum used a large range in heart rates ( fH ) across the day , ranging from 173 to 1066 bpm ( Figure 1—figure supplement 2 ) . Analysis of activity-specific fH shows that bats expend 4 . 9 ± 0 . 8 kJ h−1 during flight ( mean ± SD; fH: 766 ± 56 bpm , Figure 1 ) . Amplitude fluctuations of the fH radio signals show that minimum fH of flying bats was 750 bpm . Maximum recorded flying fH was 1066 bpm . Uroderma bilobatum then needed to generate a minimum of 0 . 98 W to fly ( 3 . 5 kJ h−1 ) , but flight typically had higher costs of 1 . 36 ± 0 . 23 W with a maximum recorded output of 2 . 3 W . This is a mass-specific metabolic power of 75 . 89 ± 11 . 9 W kg−1 and a maximum mass specific power of 145 . 6 W kg−1 . Nightly non-flight activity when bats were stationary required 2 . 2 ± 1 . 1 kJ h−1 ( fH: 492 ± 128 bpm , Figure 1 ) . Surprisingly , U . bilobatum periodically lower fH to 200–250 bpm from a mean fH of 374 ± 112 bpm throughout their daily resting periods where they remain relatively inactive in their roosts , and during which they consume 1 . 2 ± 0 . 8 kJ h−1 ( 0 . 33 — 0 . 23 W or 18 . 84 ± 13 . 59 W kg−1 , Figure 2 ) . During these periods bats are typically sitting quietly , although bats can be alert during these times and engage in bouts of agonism , grooming , and may fly from the roost due to disturbances around the roosting sites . Bats suppress fH by 30% 2–3 times per hour ( mean: 1 . 54 ± 1 . 18 sd times per hour ) for 5–7 min throughout the day ( Figure 2 ) . This cyclic bradycardia is a yet undescribed strategy that was only detectable through complete sampling of daily heart rate recordings . These lowered heart rates were followed by a return to the more stable rates between 300–400 bpm , or often to a brief arousal state with elevated heart rates above resting rates . All bats employed these reduced heart rates but one individual only used them on two of the four days it was observed . This lowered heart rate resulted in a median resting metabolic rate ( RMR ) of 0 . 54 ± 0 . 01 kJ h−1 compared to RMR 0 . 75 ± 0 . 04 kJ h−1 at higher mean fH . Using the mean energetic expenditure by each bat on each night it was tracked ( Supplementary file 1 ) we can estimate typical field metabolic rate ( FMR ) of 45 . 79 kJ if a bat spends 2 hr in flight and executes daily cyclic bradycardia ( Figure 3 ) . Two hours may be an over-estimate of time flying in the resource dense region where we tracked bats , but likely reflects areas with more dispersed fruit trees . Based on median values for each individual mean metabolic scope was 5 . 39 ± 1 . 80 . This short , cyclic bradycardia then saved U . bilobatum 0 . 3–0 . 5 kJ h−1 or 3 . 5–6 kJ total over the 12 hr resting phase which is 10% ( 7 . 6–13 . 1% ) of their total FMR . We used a diet switching experiment that transitioned bats from a natural diet , dominated by figs with low δ13C values , to an experimental diet with high δ13C values ( agave sugar ) to model the speed at which ingested sugar enters metabolism by measuring the changes in the δ13C composition of exhaled CO2 . After a baseline sample , bats ( n = 8 ) were fed a solution of agave nectar . Their exhaled breath was rapidly enriched in 13C and reached an asymptotic value of −16 . 5 ± 2 . 0 ‰ 50 min after initial feeding ( Figure 4A , Supplementary file 2 ) which is lower than the δ13C value of the diet and indicates that fat or glycogen stores continued to be metabolized ( δ13Cdiet = −12 . 0 ± 0 . 1 ‰ , t = −12 . 6 , df = 32 , p<0 . 001 ) . Overall δ13C breath enrichment followed a mean single pool incorporation model of δ13C breath ( t ) =−16 . 575–12 . 841e-0 . 081t , with 50% of metabolism fueled by ingested food after only 8 min ( t50 = 8 . 1 ± 15 . 6 min ) . The large standard deviation in t50 is due to one distinctive bat ( Individual A ) that showed a nearly linear enrichment curve with no asymptote ( Supplementary file 2 ) . If this bat is excluded , t50 drops to 7 . 6 min and an incorporation curve of δ13C breath ( t ) =−16 . 497–13 . 138e-0 . 091t . Bats fed with ripe Ficus indica ( n = 6 ) did not show any change in δ13Cbreath over the course of the following 90 min ( Figure 4A , F1 , 24 = 2 . 614 , p=0 . 113 ) . Over the course of the next three days , bats kept in captivity and fed on agave nectar showed increasingly 13C enriched baseline δ13C breath values at the beginning of the night ( Figure 4B ) and after not eating for the entire day , which is typical of feeding patterns of these bats . We estimated a mean single-pool exponential model of δ13Cbreath ( t ) =−16 . 412–12 . 901e-0 . 801t , with a t50 = 13 . 2 ± 4 . 6 hr , and by the third night bats approached an asymptotic starting value of −17 . 06 ± 1 . 27 ‰ which is not different from the asymptotic value of the initial feeding experiment ( t13 = 0 . 91 , p=0 . 38 ) . This indicates that fifty percent of an individual’s fat reserves are then exchanged after 13–17 hr , and a carbon atom has a residency period of 1–2 days , with a full exchange of fat after 3 days . Bats captured at their roosts ( 15 F , 6 M ) showed low baseline values of circulating cortisol concentrations ( ng ml−1 ) that did not differ by sex ( F: 64 . 81 ± 158 . 81 ng ml−1 , M: 57 . 66 ± 137 . 07 ng ml−1 , F1 , 19 = 0 . 009 , p=0 . 92 ) . When restrained in a cloth bag for one hour they showed a strongly sex biased response: restraint-induced values were 10–15 times baseline conditions ( Figure 5 ) , and were two-fold greater in females than in males ( F: 989 . 50 ± 450 . 78 ng ml−1 , M: 428 . 34 ± 94 . 45 ng ml−1 , F1 , 19 = 0 . 8 . 89 , p=0 . 008 ) .
We tracked the heart rates free-ranging bats throughout the 24 hr period , including foraging , to estimate total energetic costs . As hypothesized , we found that heart rate derived energy expenditure of U . bilobatum during flight is high and this is achieved through rapid incorporation of ingested food into their metabolism . We found flying heart rates that were 4–5 times higher than resting rates during the day and twice the heart rates of bats roosting at night . These bats replace nearly half of their fat reserves within a single day , resulting in short potential starvation times . Uroderma bilobatum counter this high energetic expenditure by spending relatively little time in flight and they have exceptionally low circulating cortisol values at rest during the day . These low basal values promote conservation of glucose reserves , but can be elevated up to 15 times , at least in response to stress , and could be used to generate the high metabolic power needed for flight . Most surprising , we found that by cyclically lowering heart rates during the day , they save 10% of their energy budget . This cyclic bradycardia is a novel strategy that minimizes energetic expenditure at relatively high ambient temperatures and allows U . bilobatum to maintain a FMR expected for their size . Only by completely sampling these high-resolution data from naturally behaving bats were we able to detect these lowered heart rates and quantify their effect on bat energetic strategies . Using the energetic expenditure derived from median heart rates of resting bats in their natural roosts during the day ( 0 . 54 kJ h−1 or 0 . 27 W ) we can estimate a RMR of 13 . 0 kJ day−1 , which closely approximates previous measures of BMR ( 12 . 8 kJ day−1 [McNab , 1969] ) . After commuting to the foraging patch , figs are collected during short flights of 1–2 mins and most of the remaining time is spent more or less at rest in their night feeding roosts resulting in only 30 mins per night in actual flight . Although this may differ among sites or during periods of less favorable food availability , this perch-resting with short fruit collection flights is an important part of their energy saving strategy . The subsequently low heart rates and activity patterns estimate an estimated FMR of ca . 46 kJ day−1 ( Bishop and Spivey , 2013 ) , which is within the general predictions for FMR based on body mass from a broad taxonomic sampling of studies using doubly-labeled water ( Speakman , 2005 ) . While energetic expenditure met estimates for 16–19 g bats , this was only possible due to U . bilobatum restriction of total active flying time to less than about two hours per night ( Figure 3 ) and the cyclic suppression of heart rates while resting during the day . The cyclic bradycardia during daytime rest in our study is unprecedented . It is possible that these cycles are linked to REM and pre-REM sleep , but when humans and cats sleep their heart rate slows immediately prior to the elevation of heart rates during REM cycles ( Taylor et al . , 1985; Verrier et al . , 1998 ) . In both taxa the change in heart rate lasts only seconds and amounts to a total change of 3–5% from the resting heart rate as compared to the minutes-long 30% reduction in U . bilboatum . A reversed pattern in heart rate is found in hibernating ground squirrels that show irregular heart rates that speed up for 30–50 s before slowing again to a steady rate ( Milsom et al . , 1993; Milsom et al . , 1999 ) . The regular occurrence of this cyclic bradycardia suggests that it is a standard and regular aspect of the way that U . bilobatum rests and is likely a further extension of the energy conservation of sleep ( Benington and Heller , 1995; Kilduff et al . , 1993; Schmidt , 2014; Walker et al . , 1979 ) . Bradycardia is common aspect of the dive response where diving mammals slow their heart rates to conserve oxygen when submerged for long periods ( Noren et al . , 2012 ) . Mammals in torpor are also bradycardic ( Currie et al . , 2014; Dechmann et al . , 2011; Elvert and Heldmaier , 2005; Heldmaier et al . , 2004 ) , but the cyclic and varying nature of the heart rate depressions we find in U . bilobatum are not characteristic of any of these physiological states . Animals enter torpor and hibernation through controlled reductions of heart rate via increased inter-beat interval and skipped beats ( Elvert and Heldmaier , 2005; Milsom et al . , 1999 ) . It may be that the slowed heart rates in U . bilobatum reflect the initial descent into short and shallow torpor events with a decrease in heart rate preceding the shift to torpor . Further investigation into the nervous control of bradycardic states in U . bilobatum would clarify both how these reductions are executed and any similarity to a sleep-torpor transition ( Milsom et al . , 1999 ) . Thus far , bats and hummingbirds in torpor and at rest have showed low and constant heart rates without any indication of the cycling we observe ( Currie et al . , 2014; Dechmann et al . , 2011; Schaub and Prinzinger , 1999 ) . Species that are capable of daily torpor generally lower their body temperatures to maximize energetic savings , particularly when exposed to cold temperatures ( Ruf and Geiser , 2015 ) , and this commonly occurs in tropical and sub-tropical mammals at temperatures below 24°C ( Canale et al . , 2012; Geiser and Stawski , 2011 ) . However , a similar torpor response is not possible for the tropical U . bilobatum . Instead , they actively defend their body temperatures when exposed to cold , and more than triple metabolic rates to maintain a body temperature of 36°C at ambient temperatures of 10°C vs 30°C ( McNab , 1969 ) . In our respirometry calibrations , U . bilobatum maintained a constant body temperature around 37°C across the measured range of heart rates of 300–800 bpm . In fact , they may alter heart rate dynamics independently of body temperature ( Tøien et al . , 2011 ) . The low heart rates we observed are similar to the minimum resting heart rates of small bats ( i . e . , 200–400 bpm ) in thermonetural conditions ( Currie et al . , 2015; Kulzer , 1967; Leitner , 1966; Leitner and Nelson , 1967 ) . The thermoneutral zone for U . bilobatum is reported to be 29–35°C ( McNab , 1969; Rodríguez-Herrera et al . , 2016 ) , which is still slightly higher than the ambient temperature of our field site ( mean: 25 . 87 ± 1 . 21°C , range: 23 . 38–28 . 24°C . It is unclear if the frequency and intensity of the cycling we observed are in response to ecological and energetic interactions , such as lowered foraging success , that decouples resting metabolic rates from overall FMR ( Nilsson , 2002; Welcker et al . , 2015 ) . Decoupling resting metabolic rates from total energetic expenditure is hypothesized to be found in animals that live near their energetic ceilings ( Welcker et al . , 2015 ) with high metabolic rates , like U . bilobatum . We found the number of cycles per hour varied both within and among individuals , but we do not yet have enough information on the relationship between total nightly energy expenditure , energy intake , and the lowering of heart rates . However , it is unclear why these bats move between two apparently stable low energy states at rest . Further investigation into the potential relationship between this heart rate change and more commonly perceived torpor states would help understand energetic adaptations in tropical environments . There are relatively few data on the heart rates of free-flying animals ( Green , 2011 ) all of which are larger than U . bilobatum . Furthermore , accurately estimating energy consumption during flight under controlled conditions has often been unpractical or impossible due to the conditions needed in wind tunnels and mask respirometry . Our bats' heart rates and metabolic power are surprisingly low and variable when compared to flight tunnel studies . The heart rate derived estimate for cost of flight in U . bilobatum ( 1 . 36 ± 0 . 23 W , range 0 . 98–2 . 3 W ) was slightly lower than mass loss estimates for bats of similar sizes and wing shapes ( 1 . 96–2 . 45 W: ( von Busse et al . , 2013; Winter and von Helversen , 1998 ) . However , U . bilobatum show a mass specific power of 76 W kg−1 with a maximum output of 145 W kg−1 , which is within the power requirements of bats that are up to 44 times larger ( Carpenter , 1986; Thomas , 1975 ) . While our estimates of energy consumption were not directly calibrated with flying bats , they provide the best potential estimates available based on broad patterns of the relationships among heart rate , stroke volume , and oxygen consumption during exercise ( Bishop and Spivey , 2013 ) , and must be interpreted with some caution . The large changes in heart rate among activity states remain lower than would be expected based on body size and reinforce the emerging pattern of lower energy consumption by free-flying animals versus those in controlled laboratory conditions ( Bishop et al . , 2015; Bowlin et al . , 2005; Ward et al . , 2002 ) . Metabolic power of bat flight may be difficult to predict as a function of body size , but more likely the context in which the animals fly plays a strong role in determining the energy used . Laboratory experiments have been our best window into animals' physiological possibilities , but it is increasingly important to study energetics in relevant ecological settings to understand how these physiological mechanisms evolve . The ability to rapidly fuel metabolism through ingested food seems to be a common adaptation among hummingbirds and bats with diets of simple carbohydrates and the fastest incorporation rates measured for vertebrates ( Welch et al . , 2016 ) . Nectarivorous hummingbirds ( 3–5 g ) and bats ( 10 g ) fuel 50% of their metabolism within 3–9 min of feeding ( Suarez et al . , 2011; Voigt and Speakman , 2007; Welch et al . , 2008 ) . At three times their body size , U . bilobatum shows similar fractional incorporation rates . In contrast , other fruit-eating bats use incorporation rates of 10–12 min regardless of body size ( Amitai et al . , 2010 ) . These comparative data indicate strong pressure on all flying frugivores , regardless of size , to mobilize ingested food to power flight and this is mediated through paracellular absorption ( Caviedes-Vidal et al . , 2008; Price et al . , 2014 ) . While initiating flight on stored energy , U . bilobatum and other sugar-focused bats rely heavily on ingested carbohydrates to supplement rapidly depleted glycogen at the onset of flight , further taxing the sugar oxidation cascade to push energy to muscle as quickly as possible ( Suarez et al . , 2011; Welch et al . , 2016 ) . Frugivorous bats deplete the large glycogen stores in their liver within 24 hr ( Pinheiro et al . , 2006 ) . Our fat turnover experiments also showed that half of fat and sugar storage is mobilized within a single day . Specialization on foods rich in simple but rapidly incorporated carbohydrates seems to come with high risks that necessitate additional physiological and behavioral strategies to ensure energetic stability . Uroderma bilobatum further control energetic incorporation and conservation by maintaining exceptionally low baseline cortisol levels that then are elevated to some of the highest recorded naturally induced values for mammals ( Sapolsky et al . , 2000 ) . Basal glucocorticoid values of other bat species are especially high for mammals of their size ( 100–800 ng ml−1; ( Reeder et al . , 2004; 2006 ) and show large potential maximal output when challenged with ACTH ( Lewanzik et al . , 2012 ) . However , the difference between baseline and restraint-induced circulating cortisol especially in female U . bilobatum is more similar to the extremes found in lemmings ( Lemmus trimucronatus ) that seasonally elevate their baseline corticosterone values by 10–80 times to concentrations of over 4000 ng ml−1 ( Romero et al . , 2008 ) or in flying squirrels ( Glaucomys sp ) that elevate cortisol values 38–40% above already high baseline values ( Desantis et al . , 2016 ) . The low baselines we found may be a consequence of capturing resting or sleeping animals in their day roosts at least 4 hr after sunrise when circulating glucocorticoids were at their lowest ( Sapolsky et al . , 2000 ) . However , this cannot explain peak values 1 . 5x greater than those observed in other mammals . We suggest that rapid increases in circulating cortisol levels during the acute stress response act in concert to mobilize energy stores , but more importantly , by suppressing glucocorticoid secretion during rest these bats are able to further minimize energetic expenditure and lower their metabolic rates ( Haase et al . , 2016; Nieuwenhuizen and Rutters , 2008; Palme et al . , 2005 ) and minimize additional fat oxidation ( Brillon et al . , 1995 ) . Unpredictable fruit availability can have dramatic effects on survival and some bats , including U . bilobatum take advantage of their roosts to leverage social information and identify newly available food items ( O'Mara et al . , 2014a; Ramakers et al . , 2016 ) . Furthermore , the potential for rapid declines in food availability has likely shaped conservative physiological strategies in these bats to minimize energy expenditure while allowing for rapid resource mobilization needed for powered flight . These dynamic energetic strategies likely contribute to the success and diversity of the over 1300 bat species throughout the world ( Simmons , 2005 ) .
We captured 4 adult Uroderma bilobatum ( 2f/2m , 18 . 1 ± 1 . 5 g body mass ) from their day roosts in Gamboa , Panamá in December 2014 . Bats were fitted with a heart rate transmitter ( ca 0 . 8 g; SP2000 HR Sparrow Systems , Fisher , IL USA ) that emitted a continuous long-wave signal modulated by cardiac muscle potentials ( Bowlin et al . , 2005 ) . This added 4 . 5 ± 0 . 04% of body mass and is within the range of the additional loading ( 5% ) that should have minimal impact on behavior and physiology of bats and birds ( Aldridge and Brigham , 1988; Barron et al . , 2010; Elliott , 2016; O'Mara et al . , 2014b ) , particularly broad-winged understory foragers like U . bilobatum . We trimmed the dorsal fur below the shoulder blades . A topical analgesic was then applied ( Xylocaine gel , Astra Zeneca , Wedel Germany ) and after disinfecting the electrodes and back of the bats with 70% EtOH , the transmitter’s two copper plated gold electrodes were inserted ca . 3 mm through a puncture made with a 23 G sterile needle . The transmitters were mounted on thin , flexible cloth and glued over the electrode insertion points using a silicone-based skin adhesive ( Sauer Hautkleber , Manfred Sauer , Germany ) . The electrodes are flexible and do not appear to disturb the animals , and we expect superficial healing of the small punctures within one hour . While behavioral responses may not directly reflect physiological stress ( Ditmer et al . , 2015 ) , our radio tracking data show typical behavior for this species , and both the large variation and temporal consistent heart rate data we collect do not indicate that bats are either under excessive stress or that habituation was needed to accommodate the added load of the transmitter ( O'Mara et al . , 2014b ) . After calibration of heart rate versus oxygen consumption ( below ) animals were tracked for 2–6 days ( mean: 3 . 75 d ) . We recaptured three of the four bats and removed their transmitters . Bats lost 0 . 0–0 . 5 g ( 0 . 17 ± 0 . 29 g ) which is within the daily mass fluctuations ( 1–2 g ) observed in this species ( O’Mara , unpublished data ) . We measured rates of oxygen consumption ( V˙O2 ) carbon dioxide production ( V˙CO2 ) , heart rate ( fH ) , and body temperature ( Tb ) of these four bats with an open-flow , push-through respirometry system . External air ( >75% relative humidity , ~26°C ) was dried with Drierite ( WH Hammond Driertie Co , Ltd , Xenia , OH , USA ) and pumped through a mass flow controller ( FB8 , Sable Systems International , Las Vegas , NV , USA ) into a 1 L respirometry chamber fitted with a thermocouple within a 20 L insulated cooler that was dark and temperature controlled ( PELT5 , Sable Systems ) . Flow rate was 600 ml min−1 , chamber temperature was maintained at 28–29°C , and relative humidity and vapor production were measured with a RH-300 ( Sable Systems ) , and an additional empty chamber served as a reference to the animal chamber . After drying the air leaving the chamber with Drierite we measured CO2 concentration , and after scrubbing the air of CO2 with Ascarite ( Thomas Scientific , Swedesboro NJ , USA ) we determined O2 concentrations ( FOXBOX , Sable Systems ) . Chamber temperature , CO2 , O2 , and relative humidity were recorded directly with Expedata via the UI-2 data acquisition interface ( Sable Systems ) . V˙O2 and V˙CO2 were then calculated across five minute intervals ( Lighton , 2008 ) . Bat Tb was monitored with a temperature sensitive PIT-tag ( BioThermo13 , Biomark Inc , Boise ID , USA ) injected dorsally under the skin and recorded every minute ( Stockmaier et al . , 2015 ) . Bats remained normothermic throughout the experiment with Tb = 36 . 9 ± 1 . 6°C ( Figure 1—figure supplement 3 ) . Heartbeat of bats in the respirometry chamber was recorded as a sound file ( see below ) , and fH was averaged over the one minute preceding each Tb measurement . This gave five Tb and fH measures for each measurement of V˙O2 and V˙CO2 . After three hours bats were released at their roosts . Respirometry measures were taken between 19:00 – 04:00 hr . Heart rate provided a better fit in a single factor generalized linear mixed effect model ( bat identity as a random effect ) of energy consumption than body temperature ( R2adjusted = 0 . 758 vs R2adjusted = 0 . 145 , respectively ) , and the inclusion of Tb in a two-factor model did not improve the model’s predictive ability . In the best-fit model , energy expenditure was related to fH as ( kJ h−1 ) =0 . 004 * fh - 0 . 3228 ( Figure 1—figure supplement 4 ) . We recorded fH of the four free-ranging bats during 2–6 days and nights using telemetry receivers ( AR8000 , AOR Ltd ) connected to 3-element Yagi antennae ( Sparrow Systems ) . This was then recorded via mini-dv output to a wave file ( 44 . 1–48 kHz ) on a digital recorder ( Tascam DR-05 ) . Receivers were placed under roosts to record fH during the full inactive cycle during daylight hours . One to two people then followed the bats at emergence ( ca 18:00 ) for 4–8 hr during the night’s activity and continuously recorded estimated activity ( flight , inactivity , grooming ) via fluctuations in the amplitude of the transmitted signal . Transmitter signal could be detected within 70–100 meters in the forest and up to a kilometer over open space ( the Panama Canal ) . This gave us 18–20 hr of heart rate recordings per individual per day for a total of 350 hr . Daytime mean ambient temperature was 25 . 87 ± 1 . 21°C ( mean daytime minimum to mean maximum: 23 . 38–28 . 24 ) , and mean nightly ambient temperature was 23 . 74 ± 0 . 50°C ( mean nightly minimum to mean maximum: 22 . 74–24 . 78°C ) . Ambient temperature was recorded by the Autoridad del Canal de Panamá for Gamboa and provided by the Smithsonian Tropical Research Institute’s Physical Monitoring Program . Heart rate from radio transmitters was scored previously by visually measuring the interval needed to encompass 5–10 heart beats at sampling intervals of 0 . 5–10 min apart ( Barske et al . , 2014; Bowlin et al . , 2005; Dechmann et al . , 2011; Sapir et al . , 2010; Steiger et al . , 2009 ) . We fully sampled the recorded data using an automated approach in R 3 . 2 ( Core Team , 2016 ) to identify and count all heartbeats ( Figure 1—figure supplement 1 ) . We used a finite impulse response filter in seewave with a window length of 1500–2000 samples to select the carrier frequency of the transmitter . We counted individual heartbeats by applying a timer function in seewave that ran over non-overlapping windows of 500 samples . This created a resolution of 88–98 sampling windows per second . We then applied a kernel density filter in KernSmooth to further eliminate noise that was outside of the 90% quantile . This approach is conservative in that it may have eliminated some heart rate outliers , but the autocorrelated nature of heart rate allowed us to filter out errors likely induced by static or other interference in the recordings . Automated samples were inspected periodically to validate the filtering method , particularly in periods with high variation . We then estimated total energy consumption in two ways . First , we used the five minute V˙CO2 production from the respirometry chamber estimate total energy consumption using a conversion of 1 mL CO2 ≅ 26 J and matched this to an average of the preceding one minute Tb and fH measurements . While we attempted to get a range of fH within the respirometry chamber , we could not attain the high heart rates typical of U . bilobatum during flight or the very low fH we observed during day rest . Furthermore , high fH of animals due to factors other than exercise , such as our respirometry chamber , may under-estimate energy consumption caused by changes in stroke volume and oxygen extraction efficiency during exercise ( Bishop and Spivey , 2013 ) . However , we can use the relationship between heart mass ( Mh , a proxy for stroke volume ) , and body mass ( Mb ) to model oxygen consumption as function of fH as V˙O2 = 0 . 0402 Mb0 . 328±0 . 05Mh0 . 913±0 . 045fh2 . 065±0 . 03 ( Bishop and Spivey , 2013 ) . This estimate is based on the exercise response of 24 species of endotherms across 5 orders of magnitude of body size . This model is able to accurately estimate energy consumption during the primary mode of locomotion ( Ward et al . , 2002 ) , which has been the major shortfall of experimental calibration of heart rate against V˙O2 in respirometry conditions where locomotion is restricted . We estimated individual heart mass as 1% of body mass at capture ( Canals et al . , 2005 ) . Because the bulk of U . bilobatum diet is carbohydrate we then converted V˙O2 estimates to energy by assuming that 1 ml O2 ≅ 21 . 11 J . We used a feeding experiment to measure the change in δ13C values in exhaled CO2 and estimate the time needed for ingested food to enter metabolic processes and exit as waste CO2 . Uroderma bilobatum feed on figs with a low enrichment of 13C , typical for a C3 plant . By feeding an enriched 13C source from a CAM plant ( agave nectar ) we could measure how quickly sugar entered metabolism ( McCue and Welch , 2016; Voigt and Speakman , 2007 ) . Bats were captured from their day roosts and housed individually in mesh-lined cages . At time zero , bats were removed from their cage and immobilized by gently wrapping them in cotton gauze , excluding their heads and feet . They were then placed into a 6 × 6 × 4 cm plastic container with an 18 G needle hermetically attached . After sealing the container , ambient air was washed of CO2 using NaOH and flushed through the plastic container at a flow rate of 700 mL min−1 . The flushing gas exited the container through the attached needle . The pump was turned off 2 min prior to collection to allow breath to accumulate in the plastic box . To collect accumulated CO2 we pierced the teflon membrane of an exetainer ( LabCo Exetainer Buckinghamshire , UK ) with the needle tip attached to the plastic container . This vacuumed approximately 4 . 5 ml of headspace into the vacutainer . After the initial sample collection ( time 0 ) , bats were removed from the container and fed either freshly-collected Ficus insipida , or approximately 1 . 5 ml of a solution of 20% ( w/w ) agave nectar ( Organic Blue Agave Nectar , Wholesome Sweeteners , Sugar Land Texas , USA ) , 2% ( w/w ) Nutri-Cal ( Vétoquinol Prolab Inc , Princeville , Québec , Canada ) and water using a transfer pipette . Breath samples were collected at 0 , 10 , 20 , 30 , 40 , 60 , and 90 min after the initial feeding . Bats fed figs ( n = 6 ) were placed back in their home cage and allowed to feed ad libitum after sample collection at 40 mins and 60 mins . Bats fed agave nectar ( n = 8 ) were fed an additional 0 . 5 ml of the agave nectar solution after sample collection at 30 and 60 min to ensure that the bats’ breath was equilibrated isotopically with the new diet . Additional samples at 50 and 70 mins post initial feeding were collected during the agave feeding experiments . Bats fed on figs were returned to their capture site after the last sample collection . To measure fat turnover following ( Voigt and Speakman , 2007 ) , bats fed agave nectar were given an additional 1 ml of agave solution after the final sample collection and returned to their home cages . They were maintained on the agave nectar solution supplemented with Nutrical ( tomlyn , Fort Worth USA; δ13C Agave + Nutri-Cal: 12 . 023 ± 0 . 11 ‰ ) as their only source of food for the following three nights . Bats were offered agave nectar and water ad libitum and were also fed by hand every 3 hr to ensure that they were feeding consistently . Food was removed during the day and bats were fasted for at least 10 hr prior to sample collection . At the beginning of each night were removed from their holding cages to collect a single breath sample as in the previous experiment to measure their baseline δ13C . Following breath collection , bats were fed with the agave nectar solution and returned to their holding cages . Body mass and body condition were monitored throughout the experiment to ensure animal optimum health , and one animal was released after night 2 because of weight loss . Breath samples were then shipped to the stable isotope laboratory of the Leibniz Institute for Zoo and Wildlife Research where δ13CO2 was analyzed in a blind protocol using a GasBench ( Thermo Scientific , Bremen Germany ) connected to a stable isotope ratio mass spectrometer ( Delta V Advantage Thermo Scientific , Bremen Germany ) . Samples were analyzed together with a laboratory standard gas that was previously calibrated with the international 13C reference materials NBS 19 and L-SVEC . Ratios of 13C and 12C were expressed relative to the international standard ( Vienna-PeeDee Belemnite ) using the δ notation in parts per mill ( ‰ ) : δ13CV-PDB = Rsample/Rstandard-1 ) x 103 where Rsample/Rstandard is the ratio of heavy and light carbon isotopes ( 13C/12C ) in the sample and the standard . Precision was always better than ±0 . 06‰ ( 1 SD ) . To measure the isotopic composition of the agave nectar solution a sample was dried in a drying oven until constant mass and 3 subsamples were then separated , weighed , and loaded in a tin capsule . Samples were analysed together with laboratory standards of known stable carbon isotope ratios using an elemental analyser ( Flash elemental analyser , Thermo Scientific , Bremen , Germany ) connected in continuous mode via a Conflo III to a stable isotope mass spectrometer ( Delta V Advantage; Thermo Scientific , Bremen , Germany ) . Samples were combusted under chemically pure helium gas in the analyser and resulting gases were then routed to the IRMS for the analysis of stable carbon isotope ratios . The analytical precision was always better than 0 . 13 per mille ( one standard deviation ) . We estimated the fractional rate of isotopic incorporation ( k ) using a one-pool model for each individual bat as δ13Cbreath ( t ) = δ13Cbreath ( ∞ ) + [δ13Cbreath ( 0 ) – δ13Cbreath ( ∞ ) ] e-kt; where δ13Cbreath ( t ) is the isotope composition , δ13Cbreath ( ∞ ) is the asymptotic equilibrium isotope composition , and k is the fractional rate of isotope incorporation . The time at which 50% of carbon isotopes are exchanged in the animal’s breath is calculated as t50=-ln ( 0 . 5 ) /k . The reciprocal of the fractional incorporation rate ( k−1 ) estimates the average residence time of a carbon atom in fat reserves . We used a one-compartment model as this typically reflects isotopic incorporation into breath better than models with more complicated dynamics ( Martínez Del Rio and Anderson-Sprecher , 2008 ) . Non-linear least-squares models based on one-pool dynamics were fit to individual bats . We sampled circulating cortisol values from bats ( 15 F , 6 M ) captured from their natural day roosts under the roofs of houses in Gamboa , Panama using a hoop net between 10–12 hr in November 2013 when females were not palpably pregnant or with dependent young . Baseline cortisol samples were collected by puncturing the antebrachial or cephalic vessels with a sterile 23 G needle and collecting ca . 70 uL of blood in heparin-coated hematocrit tubes . Blood samples were collected within 3 min of capture and placed on ice . Bats were placed in a soft cloth bag for 60 min and then a second blood sample of equal volume was collected . After the second blood collection , bats were fed 30% sugar water and released at the capture site . Blood samples were spun in a centrifuge for seven minutes at 7000 g , the plasma removed and snap frozen in liquid nitrogen before storage at −30°C . Samples were then mailed on dry ice to Trent University where total plasma cortisol was measured in duplicate using a commercially available radioimmunoassay ( MP Biomedicals ImmuChem Coated Tube Cortisol 125I RIA Kit; MP Biomedicals , LLC , Diagnostic Division , Orangeburg , NY , USA ) . This kit was validated for parallelism with plasma from U . bilobatum . Tests for differences between slopes on log-transformed data showed that the serially diluted plasma curve was parallel to the assay standard curve ( F1 , 9 = 0 . 21 , p=0 . 66 ) . The intra-assay coefficient of variation ( CV ) was 2 . 4% and all samples were run in a single assay . Seven generalized linear mixed effects models in lme4 were used to evaluate circulating cortisol concentrations ( ng ml−1 ) for effects of sex and time point sampled ( baseline or restraint ) using animal identity as a random effect . | To survive , all animals have to balance how much energy they take in and how much they use . They must find enough food to fuel the chemical processes that keep them alive – known as their metabolism – and store leftover fuel to use when food is not available . Bats , for example , have a fast metabolism and powerful flight muscles , which require a lot of energy . Some bat species , such as the tent-making bats , survive on fruit juice , and their food sources are often far apart and difficult to find . These bats are likely to starve if they go without food for more than 24 hours , and therefore need to conserve energy while they are resting . To deal with potential food shortages , bats and other animals can enter a low-energy resting state called torpor . In this state , animals lower their body temperature and slow down their heart rate and metabolism so that they need less energy to stay alive . However , many animals that live in tropical regions , including tent-making bats , cannot enter a state of torpor , as it is too hot to sufficiently lower their body temperature . Until now , scientists did not fully understand how these bats control how much energy they use . Now , O’Mara et al . studied tent-making bats in the wild by attaching small heart rate transmitters to four wild bats , and measured their heartbeats over several days . Since each heartbeat delivers oxygen and fuel to the rest of the body , measuring the bats’ heart rate indicates how much energy they are using . The experiments revealed for the first time that tent-making bats periodically lower their heart rates while resting ( to around 200 beats per minute ) . This reduces the amount of energy they use each day by up to 10% , and helps counteract heart rates that can reach 900 beats per minute when the bats are flying . Overall , these findings show that animals have evolved in various ways to control their use of energy . Future research should use similar technology to continue uncovering how wild animals have adapted to survive in different conditions . This knowledge will help us to understand how life has become so diverse in the tropics and the strategies that animals may use as climates change . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology"
] | 2017 | Cyclic bouts of extreme bradycardia counteract the high metabolism of frugivorous bats |
Serum amyloid A ( SAA ) is an evolutionally conserved enigmatic biomarker of inflammation . In acute inflammation , SAA plasma levels increase ~1 , 000 fold , suggesting that this protein family has a vital beneficial role . SAA increases simultaneously with secretory phospholipase A2 ( sPLA2 ) , compelling us to determine how SAA influences sPLA2 hydrolysis of lipoproteins . SAA solubilized phospholipid bilayers to form lipoproteins that provided substrates for sPLA2 . Moreover , SAA sequestered free fatty acids and lysophospholipids to form stable proteolysis-resistant complexes . Unlike albumin , SAA effectively removed free fatty acids under acidic conditions , which characterize inflammation sites . Therefore , SAA solubilized lipid bilayers to generate substrates for sPLA2 and removed its bioactive products . Consequently , SAA and sPLA2 can act synergistically to remove cellular membrane debris from injured sites , which is a prerequisite for tissue healing . We postulate that the removal of lipids and their degradation products constitutes a vital primordial role of SAA in innate immunity; this role remains to be tested in vivo .
The serum amyloid A ( SAA ) family consists of 12-kDa proteins that have been highly evolutionally conserved at least since the Cambrian period , from sea cucumber to human ( Uhlar et al . , 1994; Sun et al . , 2016 ) . SAA is an enigmatic biomarker of inflammation that is better known as a protein precursor of systemic amyloid A ( AA ) amyloidosis , a life-threatening complication of chronic inflammation , than for its beneficial action ( Westermark et al . , 2015; Papa and Lachmann , 2018 ) . Inducible human SAA is produced mainly by the liver under the control of pro-inflammatory cytokines , is secreted into blood , and binds its major plasma carrier , high-density lipoprotein ( HDL ) ( Benditt and Eriksen , 1977 ) . SAA is also secreted locally at inflammation sites and is implicated in cytokine production and immune cell recruitment to these sites ( De Buck et al . , 2016; Eklund et al . , 2012; Ye and Sun , 2015 ) . During the acute-phase response , which is a complex systemic response to severe inflammation , infection or injury ( Gabay and Kushner , 1999 ) , human SAA isoforms 1 , 2 and 3 are upregulated ( Uhlar and Whitehead , 1999 ) , while isoform 4 is constitutively expressed at much lower levels ( reviewed in Sun et al . , 2016 and in De Buck et al . , 2016 ) . Plasma levels of inducible SAA are elevated in infections such as tuberculosis , in autoimmune disorders such as rheumatoid arthritis , lupus , and Crohn’s disease , and in certain cancers ( De Buck et al . , 2016; Eklund et al . , 2012; Ye and Sun , 2015; Sack , 2018 ) . Although chronically elevated SAA is deleterious as a protein precursor of amyloidosis and as a causal risk factor for atherosclerosis ( Eklund et al . , 2012; Getz et al . , 2016; Thompson et al . , 2018 ) , the beneficial action of SAA is less clear . In fact , SAA has been reported to be pro- or anti-inflammatory in various studies , and its functions in acute and chronic inflammation remain enigmatic ( reviewed in Eklund et al . , 2012 , Ye and Sun , 2015 , Sack , 2018 , and Kisilevsky and Manley , 2012 ) . Remarkably , in acute inflammation , during infection , after injury or following surgery , plasma levels of SAA increase swiftly more than 1 , 000-fold , reaching up to 3 mg/ml in 24–48 hr , and then the levels drop ( Sun et al . , 2016; De Buck et al . , 2016; Uhlar and Whitehead , 1999; Sack , 2018 ) . The advantage for survival of this dramatic but transient increase is unclear . However , high sequence conservation in this ancient protein family ( Uhlar et al . , 1994; Sun et al . , 2016 ) and a major and rapid commitment of liver and local tissues to SAA biosynthesis suggest that SAA is vital for survival . One potential beneficial role of SAA is its ability to mobilize HDL cholesterol for cell repair . In the acute-phase response , SAA becomes a major HDL protein that can reroute the transport of HDL cholesterol by interacting with several cellular scavenger receptors that bind SAA-modified HDL ( reviewed in Kisilevsky and Manley , 2012 ) . However , HDL undergoes additional modifications during the acute-phase response ( Jahangiri , 2010; Tall and Yvan-Charvet , 2015 ) , and the role of SAA in the homeostasis of these modified HDLs can be relatively minor ( de Beer et al . , 2013 ) . Moreover , rerouting HDL cholesterol transport cannot explain rapid and massive secretion of SAA over a period of hours following the onset of acute inflammation in various organisms , including those lacking HDL . Hence , the key primordial function of SAA must be different from HDL homeostasis . Although most circulating SAA is bound to HDL , like other HDL proteins , SAA is an exchangeable apolipoprotein that can transiently dissociate in a labile ‘free’ form ( Wilson et al . , 2018 ) . Free SAA can bind a range of other apolar ligands , including cholesterol ( Liang et al . , 1996 ) , retinol ( Derebe et al . , 2014 ) , phospholipids , lysophospholipids , and free fatty acids ( FFA ) ( Takase et al . , 2014; Jayaraman et al . , 2015; Frame and Gursky , 2016; Tanaka et al . , 2017; Jayaraman et al . , 2017a; Frame et al . , 2017; Jayaraman et al . , 2018 ) . Our in vitro studies showed that SAA binds various phospholipid vesicles and spontaneously solubilizes them to form HDL-sized particles de novo ( Frame et al . , 2017; Jayaraman et al . , 2018 ) . We proposed that this ability hinges upon the binding of diverse apolar ligands at a large concave apolar face of the SAA molecule ( Frame and Gursky , 2016 ) . This face , formed by two amphipathic α-helices , was observed in the atomic-resolution x-ray crystal structures of human SAA1 . 1 and murine SAA3 ( Derebe et al . , 2014; Lu et al . , 2014 ) . The shape of this apolar face , whose key features are conserved in the SAA family ( Frame and Gursky , 2016 ) , helps to explain the preferential binding of SAA to highly curved apolar surfaces , a property that is essential for HDL binding and lipid sequestration ( Frame and Gursky , 2016; Frame et al . , 2017 ) . These findings compelled us to propose that SAA’s ability to solubilize phospholipid bilayers and to form lipoprotein nanoparticles de novo reflects the primordial role of this Cambrian protein in the removal of cell membrane debris from injured sites , a function that pre-dates SAA binding to HDL ( Frame et al . , 2017 ) . Here , we consider a functional link between SAA and another ancient lipophilic plasma protein , phospholipase A2 ( PLA2 ) . PLA2 is a superfamily of diverse enzymes that hydrolyze phospholipids in the sn2 position ( Burke and Dennis , 2009 ) . The reaction products , lysophospholipids and FFA , are bioactive lipids that are precursors of signaling molecules in many vital processes ( Burke and Dennis , 2009 ) . Secretory PLA2 ( sPLA2 ) is a family of pro-inflammatory enzymes that are involved in the immune response ( Boyanovsky and Webb , 2009; Murakami et al . , 2016 ) , which is especially relevant to SAA . For example , sPLA2 group-IIa ( sPLA2-IIa ) is an antimicrobial acute-phase reactant whose concentration in plasma and at inflammation sites can increase several hundred-fold simultaneously with that of SAA ( Pruzanski et al . , 1993 ) . Notably , sPLA2 is co-expressed with SAA and is induced by the same group of pro-inflammatory cytokines ( Vadas et al . , 1993 ) . Moreover , SAA stimulates smooth muscle cells to express sPLA2-IIa ( Sullivan et al . , 2010 ) . Clinical studies have reported a direct link between the plasma levels of SAA and the enhanced activity of sPLA2 during the early stages of inflammation , whereas in vitro studies have shown that SAA enhances the remodeling of sPLA2-induced lipoproteins via an unknown mechanism ( Pruzanski et al . , 1995; Pruzanski et al . , 1998 ) . Furthermore , sPLA2 hydrolyzes highly curved micelle-like surfaces in lipoproteins such as HDL ( diameter 8–12 nm ) , but not intact planar bilayers ( Høyrup et al . , 2004; Halperin and Mouritsen , 2005 ) , whereas SAA preferentially binds to such highly curved surfaces or forms them de novo by solubilizing lipid bilayers ( Jayaraman et al . , 2018; Lu et al . , 2014 ) . Taken together , these findings compel us to postulate not only a spatiotemporal overlap between SAA and sPLA2 at inflammation sites in vivo , but also their potential synergy in lipid clearance ( Jayaraman et al . , 2016 ) . This study explores this synergy and its mechanism .
The murine and human SAA isoform 1 . 1 ( mSAA1 . 1 and hSAA1 . 1 ) proteins used in this study are major isoforms that bind HDL and form amyloid in vivo ( Westermark et al . , 2015 ) . Recombinant mSAA1 . 1 ( hereafter termed SAA for brevity ) was used in most experiments . To determine whether and how SAA influences phospholipid hydrolysis by sPLA2 , model and plasma lipoproteins that differed in size and composition were used as substrates for either sPLA2 group-III ( sPLA2-III ) , which preferentially hydrolyses phosphatidylcholine ( PC ) , or sPLA2-IIa , which preferentially hydrolyses phosphatidylethanolamine but also acts on PC ( Burke and Dennis , 2009; Boyanovsky and Webb , 2009 ) . Unless otherwise stated , both sPLA2-III and sPLA2-IIa enzymes are collectively termed sPLA2 . We first probed how the lipid surface curvature imposed by SAA influences the sPLA2 reaction . SAA was incubated with multilamellar vesicles ( MLV ) of a model phospholipid palmitoyl-oleoyl-PC ( POPC ) . The results presented in Figure 1 show that SAA solubilized POPC MLV ( diameter circa 200 nm ) to form smaller SAA-POPC complexes ( ~8 nm ) . The time course of this microsolubilization was monitored by turbidity , and its products were observed by non-denaturing PAGE ( Figure 1A , B ) . Increasing the SAA to POPC molar ratio from 1:100 to 1:10 increased the rate of solubilization ( Figure 1A ) and resulted in the generation of slightly smaller SAA-POPC particles; for all ratios explored , the particle size was 7 . 5–8 . 5 nm ( Figure 1B ) . These small particles , along with POPC MLV , were used as substrates for sPLA2 , whose enzymatic activity was assessed by measuring free fatty acid products . In the absence of sPLA2 , no significant hydrolysis of SAA-POPC particles was detected , but in the presence of sPLA2 , extensive hydrolysis of the SAA-POPC particles was observed ( Figure 1C ) . By contrast , MLV were not hydrolyzed by sPLA2: the levels of FFA were below the detection limit of our assay , and thin-layer chromatography showed only the presence of PC ( Figure 1D ) . Consequently , SAA readily solubilizes model phospholipid bilayers such as POPC MLV and converts them into small HDL-size particles that provide excellent substrates for sPLA2 . This ability distinguishes SAA from other major HDL proteins , such as apoA-I and apoA-II , which cannot spontaneously solubilize POPC MLV ( Figure 1A ) . Next , we tested the effects of SAA on the lipolysis of plasma HDL by sPLA2 . Human HDL that contained various amounts of bound SAA ( up to 27% of the total protein mass ) , termed SAA-HDL , were prepared by incubation of HDL with SAA using 1:1 or 4:1 protein weight ratio of exogenous SAA to endogenous apoA-I as described in the 'Materials and methods' ( Figure 2—figure supplement 1A–C ) . The lipoprotein fraction containing only HDL-bound proteins was isolated by size-exclusion chromatography ( SEC ) from the total incubation mixture ( marked SEC Fr and total in Figure 2—figure supplement 1B ) , and was hydrolyzed by sPLA2 . A progressive increase in activity with an increasing amount of bound SAA was observed ( Figure 2A , B ) . As the particle curvature was similar in these experiments , this increased activity must have stemmed from the presence of bound SAA . To determine whether unbound ( free ) SAA also enhanced the enzymatic activity of sPLA2 , we used sPLA2 to hydrolyze total incubation mixtures that contained HDL-bound and free proteins ( Figure 2—figure supplement 1A ) . For the same amount of SAA , the enhancementof lipolytic activity was comparable in the presence of HDL-bound SAA or a mixture of HDL-bound and free SAA ( compare 4:1 SAA-HDL Sec Fr in Figure 2A , B with SAA-HDL total in Figure 2C , D ) . To directly probe the role of free SAA , we incubated it with normal human LDL that does not bind SAA; the mixture ( SAA +LDL , Figure 2—figure supplement 1D ) was hydrolyzed with sPLA2 . Increased lipolysis was observed upon addition of free SAA to LDL ( Figure 2C , D ) . Consequently , free SAA augmented the lipolysis by sPLA2 of various model and plasma lipoproteins . In summary , the results in Figure 1 and Figure 2—figure supplement 1D show that SAA enhances the hydrolysis by sPLA2-III or sPLA2-IIa of diverse substrates , including SAA-POPC complexes , plasma HDL , and plasma LDL . This enhancement reflects the possibilities that: i ) SAA not only binds to phospholipid bilayers but also remodels them into smaller highly curved HDL-size particles ( ~8 nm ) that are readily hydrolyzed by sPLA2 ( Figure 1D , E ) , and ii ) SAA augments the action of sPLA2 in a manner that does not involve SAA binding to the substrate , as evident from the SAA-induced enhancement of LDL lipolysis ( Figure 2C , D ) . As we and others have shown that SAA binds FFA and lysoPC in vitro ( Tanaka et al . , 2017; Jayaraman et al . , 2018 ) , the latter effect could stem from interactions of SAA with the products of sPLA2 . This idea was tested as described below . The ~8 nm SAA-POPC complexes formed upon spontaneous solubilization of MLV using 1:10 to 1:100 protein:lipid molar ratioswere nearly invariant in size ( Figure 1C ) . By contrast , SAA incubated with small uninlamellar vesicles ( SUV ) of POPC formed particles that increased in size as protein:lipid ratio decreased ( Frame et al . , 2017 ) . Therefore , in the current study , we used POPC SUV to test the effect of particle size on lipolysis by sPLA2 . To test whether lipoprotein hydrolysis by sPLA2 involved changes in particle size in the absence and in the presence of SAA , we used non-denaturing PAGE to analyze model and plasma HDL before and after the lipolysis . First , SAA-POPC complexes that varied in size from about 8 nm to 22 nm were prepared by incubating SAA with POPC SUV at protein to lipid molar ratios ranging from 1:1 to 1:100 . These incubation mixtures were hydrolyzed by sPLA2 for 3 hr at 37°C as described in 'Materials and methods' . Non-denaturing PAGE showed that all parent particles were remodeled by sPLA2 into species that migrated at 7–7 . 5 nm ( Figure 3—figure supplement 1A ) . Next , we performed similar studies using plasma HDLs that were either native or enriched with exogenous SAA ( SAA-HDL ) as described in 'Materials and methods'; these HDLs were used as substrates for sPLA2 . Before hydrolysis , both native HDL and SAA-HDL ranged in size from about 8 . 5 to 12 nm . Hydrolysis of native HDL by sPLA2 caused little change in the particle size distribution and no significant protein release ( Figure 3—figure supplement 1B ) . By contrast , hydrolysis of SAA-HDL led to lipoprotein remodeling into two major protein-containing species , of 10–12 nm and 7–7 . 5 nm in size ( Figure 3—figure supplement 1B ) . Together , the results in Figure 3—figure supplement 1 suggest that sPLA2 hydrolysis of SAA-POPC particles and SAA-HDL , but not of native HDL , leads to a release of protein-containing species that have a hydrodynamic size of 7–7 . 5 nm . We tested whether similar species were formed upon direct interaction of SAA with hydrolyzed phospholipids . First , POPC SUV were incubated with sPLA2 for 3 hr at 37°C as described in 'Materials and methods' , leading to the hydrolysis of 40–50% of the POPC . Next , the hydrolyzed samples were incubated at 25°C for 6 hr with free SAA using protein to PC molar ratios ranging from 1:1 to 1:100 . Non-denaturing PAGE showed that the particle size distribution varied depending upon the initial protein to lipid ratio , yet at all ratios , the major protein-containing species were observed at 7–7 . 5 nm ( Figure 3A , right panel ) . A strikingly similar migration pattern was observed for the SAA-POPC complexes that were formed using 1:1 to 1:100 protein to PC ratios and then hydrolyzed ( Figure 3—figure supplement 1A ) . Therefore , regardless of the order of the events ( binding to SAA and hydrolysis by sPLA2 ) , SAA formed 7–7 . 5 nm complexes with the hydrolytic products of POPC , suggesting that these complexes represented stable , kinetically accessible species . Incubation of free SAA with sPLA2-treated HDL ( Figure 3B ) or LDL ( Figure 3C ) also led to the formation of distinct 7–7 . 5 nm species . Such species were released from parent lipoproteins only in the presence of both SAA and hydrolyzed phospholipids , and were detected in model systems and in plasma lipoproteins ( Figure 3A–C ) . We reasoned that all of these 7–7 . 5 nm species could represent stable complexes of SAA with the products of phospholipid hydrolysis , and tested this idea in the following experiments . To determine the properties of the 7–7 . 5 nm species formed in the presence of SAA and hydrolyzed phospholipids , these species were isolated by density gradient centrifugation . SAA-POPC particles of 8 nm in size , as well as SAA-containing samples of HDL or LDL — containing a 4:1 protein weight ratio of SAA to either apoA-I ( which is the major HDL protein ) or apoB ( which is the major LDL protein ) — were hydrolyzed with sPLA2 as described in Figure 3—figure supplement 1 . The 7–7 . 5 nm particles formed upon hydrolysis were isolated in the density range 1 . 16–1 . 20 g/ml . In control experiments , three density fractions were taken after centrifugation of SAA-POPC complexes: before hydrolysis ( 1 . 16–1 . 18 g/ml ) , hydrolyzed SAA-POPC ( 1 . 17–1 . 20 g/ml ) , and lipid-free SAA ( >1 . 22 g/ml ) . Non-denaturing PAGE detected no 7–7 . 5 nm species at 1 . 16–1 . 18 g/ml in the control experiments . By contrast , samples of hydrolyzed SAA-POPC clearly showed 7–7 . 5 nm species in the 1 . 17–1 . 20 g/ml density fraction ( Figure 4—figure supplement 1 ) . SAA-containing samples of hydrolyzed HDL and LDL also showed species in this range of size and density ( Figure 4—figure supplement 1B , C ) . After isolation by centrifugation , the migration pattern changed slightly and smaller particles became predominant; henceforth these are collectively termed ~7 nm species . To assess the number of SAA molecules per particle , SAA-POPC particles were cross-linked with glutaraldehyde . SDS PAGE of intact SAA-POPC showed sharp bands corresponding to protein monomers , dimers and trimers , whereas hydrolyzed SAA-POPC showed a prominent hexamer band ( Figure 4—figure supplement 1D ) , suggesting that each particle contained at least six protein molecules . We used SDS PAGE and mass spectrometry to determine the protein composition of this isolated ~7 nm species . The results showed both SAA and apoA-I in the species released from the hydrolyzed SAA-HDL ( Figure 4A , B ) . Only SAA was detected in the species released from hydrolyzed LDL in the presence of SAA ( Figure 4C , D ) . Lipid composition in this species was assessed by thin-layer chromatography and enzymatic assays . Both PC and lysoPC were observed in the ~7 nm species isolated from all hydrolyzed lipoproteins , including SAA-POPC , SAA-HDL , and SAA-containing LDL samples ( Figure 4E ) . FFA and phospholipid assays showed 30–45% FFA and 12–22% PC as a weight fraction of total lipids in these ~7 nm species . We conclude that SAA sequesters the FFA and lysoPC that are produced upon the lipolysis of diverse lipoproteins by sPLA2 , and removes these hydrolytic products from the parent particle in the form of ~7 nm protein-lipid complexes . These complexes are heterogeneous and their exact size and biochemical composition vary depending on the parent lipoproteins , yet they all contain SAA and the products of lipolysis . Previous studies showed that the binding of SAA to POPC and other phospholipids induces α-helical folding in this intrinsically disordered protein at ambient temperatures , greatly increasing the thermal stability of SAA and protecting it from proteolysis ( Takase et al . , 2014; Jayaraman et al . , 2015; Frame et al . , 2017; Lu et al . , 2014 ) . To probe whether the SAA-containing complexes that are released upon hydrolysis of these precursors also formed stable structures , these ~7 nm complexes were isolated by density in the 1 . 17–1 . 20 g/ml range , and their secondary structure and stability were assessed by circular dichroism ( CD ) spectroscopy . Far-UV CD spectra at 25°C showed a major conformational change , from a largely unfolded secondary structure in free SAA to ~40% α-helix in complexes with lipids . Notably , the lipid-bound secondary structure was very similar in the parent SAA-POPC particles and in the ~7 nm products that are released upon lipolysis ( Figure 5A ) . The structural stability of the ~7 nm products was assessed by measuring the CD signal at 222 nm as a function of temperature to monitor helical unfolding and refolding during heating and cooling from 5°C to 95°C . The heating data for the precursor particles and the ~7 nm product species partially overlapped and showed similar melting temperatures , Tm = 50 ± 2°C and 48 ± 2°C , respectively ( Figure 5B ) , much higher than that of free SAA ( Tm = 17 ± 2°C ) ( Jayaraman et al . , 2015; Frame et al . , 2017 ) . Therefore , SAA complexes with either intact or hydrolyzed POPC showed comparable thermal stability that was much higher than that of free SAA . Unlike the heating data , the cooling data for the two complexes significantly differed ( Figure 5B , gray and black solid lines ) . For the precursor SAA-POPC particles , the refolding upon cooling was observed at much lower temperatures than the unfolding upon heating . Such a hysteresis is a hallmark of thermodynamically irreversible transitions; in lipoproteins , it reflects irreversible structural remodeling such as fusion ( Jayaraman et al . , 2015 and references therein ) . By contrast , free SAA shows a reversible unfolding without a hysteresis ( Figure 5B , open circles ) . Notably , in the ~7 nm SAA complexes formed upon hydrolysis , the heating and cooling transitions were much closer than those in the precursor particles , and the hysteresis was nearly abolished ( Figure 5B , gray solid line ) . This observation is consistent with the relatively high protein to lipid ratio in the ~7 nm complexes , which is evident from their higher density and smaller size as compared to the precursor particles . The conformational stability of the ~7 nm complexes was further probed by limited proteolysis as described in 'Materials and methods' . In contrast to free SAA , which was largely fragmented within 5 min of incubation with trypsin at 22°C ( Figure 5C ) , the ~7 nm complexes and their SAA-POPC precursor particles resisted proteolysis , and they underwent no major fragmentation even after 12 hr of incubation ( Figure 5C ) . These results agree with the CD data showing more helical structure and increased stability in lipid-bound SAA when compared with free SAA ( Figure 5A , B ) . Together , the results showed that SAA forms ~7 nm complexes with the hydrolytic products of sPLA2 ( Figure 3—figure supplement 1 , Figure 4—figure supplement 1 , Figure 5 ) . Although these complexes migrate in the same size range as free SAA on the non-denaturing gel , their structure is distinct from that of either free SAA or the SAA-POPC precursor particles . Unlike free SAA , which shows unfolded secondary structure when examined by CD and is rapidly degraded by trypsin at ambient temperatures , SAA in the ~7 nm complexes is ~40% helical and resists proteolysis ( Figure 5A–C ) , thereby resembling the SAA-POPC precursors . In contrast to precursors , these ~7 nm complexes: i ) contain large amounts of FFA and lysoPC ( Figure 4 ) ; ii ) have smaller size ( 6 . 5–7 . 5 nm ) and higher density ( 1 . 17–1 . 20 g/mL ) indicative of a higher protein to lipid ratio; and iii ) undergo a more thermodynamically reversible thermal unfolding ( Figure 5B ) . The latter is consistent with the kinetically accessible character of these product complexes suggested by their similarity , regardless of the order of SAA binding and sPLA2 hydrolysis ( Figure 3A , Figure 3—figure supplement 1A ) . To our knowledge , such small stable protein-rich proteolysis-resistant complexes comprised of SAA and the products of phospholipid hydrolysis have not been reported previously . To ascertain that the high structural stability of the SAA complexes with the products of phospholipid hydrolysis is not limited to model systems , we analyzed the conformation and stability of the ~7 nm complexes isolated from a mixture of SAA and sPLA2-hydrolyzed plasma lipoproteins . The complexes released from the hydrolyzed SAA-HDL contained both SAA and apoA-I ( Figure 4A , B ) . Owing to difficulties in dissecting the contributions from individual proteins , these complexes were not explored in detail . Instead , we focused on the ~7 nm complexes released from hydrolyzed LDL in the presence of SAA . Such complexes contained SAA as their sole protein ( Figure 4C , D ) , facilitating a direct comparison with a model SAA-POPC system . Parent LDL was hydrolyzed with sPLA2 and incubated with SAA , leading to the formation of 7–7 . 5 nm SAA-only complexes ( Figure 3C ) that were isolated by density at 1 . 17–1 . 20 g/ml . Far-UV CD spectra of these complexes revealed a ~ 40% α-helical conformation ( Figure 5—figure supplement 1A ) , similar to that seen in the ~7 nm complexes released upon lipolysis of SAA-POPC ( Figure 5A ) . Moreover , similar to these model complexes , thermal unfolding of the LDL-derived SAA-containing complexes was observed with Tm = 45 ± 2°C and was largely thermodynamically reversible ( Figure 5—figure supplement 1B ) . Finally , limited proteolysis ascertained the high conformational stability of SAA in these complexes , which resisted fragmentation upon incubation with trypsin at 22°C for up to 24 hr ( Figure 5—figure supplement 1C ) . We conclude that SAA sequesters the hydrolytic products of sPLA2 from model and plasma lipoproteins and forms stable ~7 nm complexes with these products . These complexes migrate in the same size range on the non-denaturing gel as free self-associated mSAA1 . 1 . However , unlike free SAA that is structurally labile , the protein in these complexes acquires a stable highly α-helical proteolysis-resistant conformation at ambient temperatures . Hydrolysis of POPC by sPLA2 generates equimolar amounts of oleic acid ( OA ) and lysoPC . Upon incubation with OA , SAA was previously shown to form spontaneously binary SAA-OA complexes that migrate at ~7 . 5 nm; the protein in these complexes was ~40% α-helical at 25°C and resisted tryptic digestion ( Jayaraman et al . , 2018 ) . Here , we explored the formation and properties of binary complexes of SAA with lysoPC . The complexes were formed as described in the 'Materials and methods' by incubating SAA with lysoPC at 1:10 protein to lipid molar ratio; similar experiments with POPC were used as a control . Non-denaturing PAGE showed SAA-lysoPC particles of ~8 nm in size , similar to those of SAA-POPC formed at 1:10 molar ratio of protein to lipid ( Figure 5—figure supplement 2A ) . Far-UV CD spectra showed substantial α-helical content in SAA-lysoPC particles , slightly lower than that in SAA-POPC particles ( 35% versus 40% ) ( Figure 5—figure supplement 2B ) . Thermal unfolding of SAA-lysoPC particles was observed with Tm = 45°C , slightly lower than 50°C seen in SAA-POPC ( Figure 5—figure supplement 2C ) . Unlike SAA-POPC particles , SAA-lysoPC particles showed little hysteresis during thermal unfolding and refolding . Moreover , similar to other SAA-lipid complexes , SAA-lysoPC complexes resisted tryptic digestion ( Figure 5—figure supplement 2D ) . Together , our results showed that SAA forms binary complexes with OA and with lysoPC . The protein in these complexes is 35–40% α-helical and resists proteolysis at ambient temperatures; the thermal unfolding is centered at Tm = 45–48°C and is largely thermodynamically reversible . In this regard , these binary complexes of SAA-OA and SAA-lysoPC resemble the quaternary complexes containing SAA , OA , lysoPC and POPC ( Figure 5 ) . Clearly , SAA can sequester POPC and its hydrolytic products , either separately or together , to form highly α-helical proteolysis-resistant complexes that are thermodynamically stable at ambient temperatures . Next , we tested whether SAA can sequester the naturally occurring hydrolytic products from human lipoproteins . Our focus was on the FFA that are elevated in diabetes , inflammation and other diseases . LDL was isolated from the pooled plasma of subjects who were either normolipidemic or had diabetes mellitus , as previously described ( Jayaraman et al . , 2017b ) . The content of endogenous FFA in these diabetic LDL was 20% higher than that in normolipidemic LDL ( Jayaraman et al . , 2017b ) . SAA was incubated with LDL using 1:1 weight ratio of SAA to apoB . Non-denaturing PAGE showed that the SAA-containing ~7 . 5 nm complexes were released from diabetic as well as normolipidemic LDL ( Figure 6—figure supplement 1A ) , similar to those released from LDL upon hydrolysis by sPLA2 in the presence of SAA ( Figure 3C , Figure 2—figure supplement 1D ) . This result suggests that SAA sequesters exogenous and endogenous FAA from lipoproteins to form similar-size complexes . Under normal in vivo conditions , FFA and lysoPC are transported in plasma mainly by serum albumin . In acute inflammation , plasma levels of albumin and its ability to sequester FFA and lysoPC decrease , while the levels of sPLA2 and SAA increase ( Gabay and Kushner , 1999; Pruzanski and Vadas , 1991; Fichtlscherer et al . , 2004 ) . Is it possible for SAA to assume albumin’s function under these conditions ? To explore this , we compared the ability of SAA and albumin to remove endogenous FFA from plasma LDL . LDL from diabetic and healthy normolipidemic subjects ( 0 . 2 mg/ml apoB ) was incubated at pH 7 . 5 , 37°C , for 6 hr with 0 . 2 mg/ml SAA ( 2% w/v ) . A similar LDL incubation was carried out with 2% w/v human serum albumin . As a control , LDL was incubated under similar conditions without any added proteins . LDL was isolated by SEC and the FFA content was determined . The results showed that both albumin and SAA removed a large portion of endogenous FFA from diabetic as well as normolipidemic LDL , and that albumin was slightly more efficient than SAA at pH 7 . 5 ( Figure 6—figure supplement 1B ) . This finding suggests that SAA can potentially contribute to FFA removal at plasma pH , albeit less efficiently than albumin . Next , we tested whether SAA can remove FFA at acidic pH , which severely impairs albumin’s function ( Lähdesmäki et al . , 2009 ) . Previously , we showed that the SAA structure , stability and ability to remodel POPC MLV remain invariant at pH 7 . 5–5 . 5 but are altered at near-lysosomal pH ( Jayaraman et al . , 2017a ) . Here , we explored how pH influences the ability of SAA to remove FFA from human lipoproteins . Single-donor LDLs were hydrolyzed with sPLA2 and incubated either with SAA , with albumin , or alone ( control ) for 6 hr at pH ranging from 4 . 5 to 7 . 5 . Thereafter , LDL was isolated by density gradient centrifugation and the FFA content was measured . As expected , albumin was less efficient at removing FFA at acidic pH than at pH 7 . 5; this was evident from the higher residual content of FFA detected in LDL at pH 4 . 5–6 . 5 versus that at pH 7 . 5 ( Figure 6 , LDL + albumin ) . By contrast , SAA’s ability to remove FFA remained invariant at pH 4 . 5–7 . 5 ( Figure 6 , LDL + SAA ) . Importantly , at acidic pH , when the activity of albumin but not SAA was impaired , SAA removed more FFA than albumin per gram of protein ( Figure 6 , LDL + SAA and LDL + albumin compared ) . This finding strongly supports the role of SAA in FFA removal in vivo and suggests that this role becomes particularly important at the acidic pH that are typical ofinflammation sites . This role is expected to be important in inflammation and in other conditions in which SAA level increases while albumin level and activity declines . To ascertain the relevance of these findings to humans , selected experiments were performed using hSAA1 . 1 and either model or plasma lipoproteins that were hydrolyzed by sPLA2 . Human and murine SAA1 . 1 show striking sequence , structural and functional similarities ( Rennegarbe et al . , 2017 ) . Compared to mSAA1 . 1 , hSAA1 . 1 is less water-soluble in the lipid-free state and forms larger ~9 nm oligomers on the non-denaturing PAGE ( hSAA , Figure 6—figure supplement 2A ) relative to the ~7 . 5 nm oligomers formed by free mSAA1 . 1 ( Figure 1B , Figure 4—figure supplement 1A , Figure 6—figure supplement 1A ) . Importantly , non-denaturing PAGE showed that , similar to mSAA1 . 1 , hSAA1 . 1 spontaneously formed 8 . 5–9 nm particles upon incubation with POPC SUV at a 1:10 protein to lipid molar ratio ( hSAA-POPC , Figure 6—figure supplement 2A ) . Upon hydrolysis with sPLA2-IIa or sPLA2-III , these 8 . 5–9 nm particles were remodeled into ~7 . 5 nm complexes ( sPLA2-hydrolyzed , Figure 6—figure supplement 2A ) , similar to those formed by mSAA1 . 1 ( sPLA2-hydrolyzed 1:10 , Figure 3—figure supplement 1A ) . Further , when hSAA1 . 1 was incubated with LDL that have been hydrolyzed with sPLA2 , it also formed complexes that migrated at ~7 . 5 nm ( hSAA + LDL , Figure 6—figure supplement 2B ) . Complexes of similar size were formed by mSAA1 . 1 upon sequestration of lipolytic products from LDL ( SAA + hydLDL , Figure 4—figure supplement 1C ) . These comparisons suggest that , similar to mSAA1 . 1 , hSAA1 . 1 sequesters the products of phospholipid hydrolysis from model and plasma lipoproteins to form ~7 . 5 nm particles . Finally , FFA analysis of LDL from the plasma of normolipidemic and diabetic patients that have been incubated with SAA and then re-isolated by density showed that , similar to mSAA1 , hSAA1 . 1 sequestered a major fraction of FFA from these LDL ( Figure 6—figure supplement 2C ) . Together , these results suggest that , similar to mSAA1 . 1 , hSAA1 . 1 can sequester phospholipids and their lipolytic products into HDL-size particles .
This in vitro study demonstrates that two ancient acute-phase plasma proteins , sPLA2 and SAA , act in synergy to break down and remove phospholipids . The increased lipolytic activity of sPLA2 on HDL upon the addition of SAA was previously reported but the mechanism was unknown ( Pruzanski et al . , 1995 ) . The current study shows that SAA enhances the lipolysis by sPLA2 of diverse lipid assemblies , including model phospholipid vesicles as well as human HDL and LDL ( Figures 1 and 2 ) , and that the mechanism involves direct SAA-mediated enhancement of sPLA2 activity . We show that SAA augments the sPLA2 reaction both by generating highly curved substrates for sPLA2 ( Figure 1 ) and by removing its reaction products , FFA and lysophospholipids ( Figure 4 ) . The latter may well be particularly important because product removal determines the reaction rate of many lipases including sPLA2 ( Carman et al . , 1995 ) . Together , these findings expand the previously postulated housekeeping roles of SAA in solubilizing and clearing cellular membrane phospholipids from injured sites ( Frame et al . , 2017 ) and in providing an anti-oxidant for lipoproteins by sequestering lipid peroxides ( Jayaraman et al . , 2016 ) . Our results suggest that sequestration of membrane lipids and their degradation products by SAA and the safe removal of these bioactive products from the injured sites , which is a prerequisite for tissue healing , represents a vital role of this Cambrian protein . Unlike SAA , several other acute–phase reactants either have no effect on sPLA2 activity or inhibit it , which perhaps helps to control the resolution of inflammation ( Pruzanski et al . , 1996 ) . To our knowledge , SAA is the only acute-phase protein that enhances the activity of sPLA2 . The in vivo synergy between SAA and sPLA2 is probably facilitated by the simultaneous secretion of these proteins systemically as well as locally at the inflammation sites . Furthermore , both proteins show preference for highly curved lipid surfaces that can be generated by SAA ( Figure 1; Frame et al . , 2017; Jayaraman et al . , 2018 ) and are required for the activation of sPLA2 . This is expected to lead to spatiotemporal overlap between SAA and sPLA2 , facilitating their synergy at the inflammation sites . We propose that the ability of SAA to sequester diverse phospholipids and their degradation products , which was demonstrated in the current and previous studies by us and others ( Takase et al . , 2014; Tanaka et al . , 2017; Frame et al . , 2017; Jayaraman et al . , 2018; Jayaraman et al . , 2016 ) , underlies several beneficial effects . First , SAA can spontaneously solubilize diverse phospholipid bilayers in vitro ( Figure 1A , B; Jayaraman et al . , 2018 ) , and perhaps in vivo , which is particularly relevant with respect to dead cells whose normal lipid efflux to HDL via ATP-driven transporters is impaired ( Frame et al . , 2017 ) . By contrast , lipid sequestration by SAA is energy-independent ( Figure 1A ) . The resulting stable nanoparticles are hydrolyzed by sPLA2 ( Figure 1C; also see Jayaraman et al . , 2018 ) and/or perhaps are internalized by macrophages through scavenger receptors such as CD36 or SR-BI that bind SAA ( Eklund et al . , 2012; Frame et al . , 2017 and references therein ) . Second , SAA solubilizes lysoPC and FFA and sequesters them into small particles ( Figure 3 ) that are stable , substantially α-helical , and resist proteolysis at 37°C ( Figure 5 ) . This sequestration augments the sPLA2 reaction; it is also expected to facilitate the safe removal of toxic lipolytic products by SAA while protecting free SAA from unfolding and rapid degradation . Third , consistent with these findings , previous studies showed that SAA protects human lipoproteins from oxidation in vitro ( Jayaraman et al . , 2016 ) and in vivo ( Sato et al . , 2016 ) . This anti-oxidant effect is mediated mainly by free ( rather than HDL-bound ) SAA that probably sequesters lipid hydroperoxides and safely removes them from lipoproteins ( Jayaraman et al . , 2016 ) . in a manner similar to the SAA-mediated removal of FFA and lysoPC described in the current study . Together , these findings suggest that during inflammation , when oxidative stress and lipolysis are increased , SAA augments the action of serum albumin by removing FFA , lysoPC , lipid peroxides and , potentially , other products of lipid degradation . We speculate that free SAA , like albumin , serves as a lipid scavenger that sequesters diverse lipophilic molecules . Unlike albumin , SAA forms oligomers to sequester lipids . Although the structure of these oligomers is unknown , lipids are expected to bind in a hydrophobic cavity formed by concave apolar faces of amphipathic helices from several protein molecules ( Frame and Gursky , 2016; Frame et al . , 2017 ) . The high stability of SAA complexes with lipids and their hydrolytic products contrasts with the marginal in vitro stability of lipid-free SAA oligomers at pH above pH 5 ( Jayaraman et al . , 2017a and references therein ) . We speculate that these marginally stabile protein oligomers are primed for interaction with lipids to form stable complexes like those reported in the current study . Normally , serum albumin removes a major fraction of sPLA2 products . However , albumin is a negative acute-phase reactant whose plasma levels drop sharply in inflammation , with a concomitant steep increase in positive reactants , such as SAA and sPLA2 ( Gabay and Kushner , 1999 ) . In addition , albumin’s ability to remove FFA is compromised upon the protein’s post-translational modifications in inflammation ( Lee and Wu , 2015 ) . Moreover , albumin’s capacity to remove FFA decreases under acidic conditions ( Figure 6 ) , which are characteristic of inflammation sites ( Lähdesmäki et al . , 2009 ) . All together , these effects are expected to cause an imbalance between the massive generation of FFA and lysoPC at inflammation sites and the impaired removal of these molecules by albumin . Our finding that , per gram of protein , SAA is more efficient than albumin in removing FFA from plasma lipoproteins at acidic pH ( Figure 6 ) suggests that SAA acts locally at the sites of inflammation to compensate for the impaired albumin activity , and helps to remove excess lipids and the products of their hydrolysis and oxidation . We propose that the ability to sequester diverse lipids and their degradation products , which is rooted in the unique structure of SAA with its concave hydrophobic surface that has been highly conserved throughout evolution ( Frame and Gursky , 2016 ) , constitutes a previously unknown vital role of SAA in the immune response . This role remains to be tested in future cell-based and in vivo studies .
Recombinant murine SAA isoform 1 . 1 ( 103 amino acids , 11 . 6 kDa ) was used throughout this study; it is termed mSAA1 . 1 , or SAA for brevity . SAA was expressed in Escherichia coli and purified to 95% purity as described previously ( Kollmer et al . , 2016 ) . In selected experiments , we used recombinant human SAA isoform 1 . 1 ( hSAA1 . 1 , 104 amino acids , cat . # SRP4324 ) from Sigma . Essentially fatty acid-free human serum albumin ( cat . # A1887 ) was from Sigma . Lipids 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine ( POPC; C16:0 , C18:1 ) and lysoPC ( 16:0 ) were 97% + pure from Avanti Polar Lipids . Trypsin from bovine pancreas ( cat . # T1426 ) , group-III sPLA2 ( sPLA2-III , cat . # P9279 ) , lipoprotein lipase ( cat . # L2254 ) , sphingomyelinase from Bacillus cerus ( cat . # S9396 ) , and sodium oleate ( cat # O7501 ) were from Sigma . Human recombinant group-IIa sPLA2 ( sPLA2-IIa , cat . # RD172054100 ) was from Biovendor . Enzychrom free fatty acid assay kit ( cat . # EFFA-100 ) and Enzychrom phospholipid assay kit ( cat . # EPLP-100 ) were from Fisher Scientific . Ultrapure sodium phosphate buffer at pH 7 . 5 ( BB-148 ) was from Boston Bioproducts . All other chemicals were of the highest purity analytical grade . SAA stock solutions were prepared daily by dissolving lyophilized protein at 1 mg/ml in water and dialyzing it overnight against the standard buffer ( 50 mM sodium phosphate , 150 mM NaCl , pH 7 . 5 ) . SAA stock solutions were centrifuged at 10 , 000 g for 10 min prior to each experiment to remove protein aggregates . Recombinant hSAA1 was reconstituted according to the manufacturer's recommendations and was immediately diluted in standard buffer , followed by overnight dialysis in standard buffer prior to use . Protein concentrations were determined by a bicinchoninic acid assay . Unless otherwise stated , single-donor human lipoproteins from three healthy volunteers were used throughout this study . Plasma from anonymous healthy donors was obtained commercially from the local blood bank according to the rules of the institutional review board . Single-donor lipoproteins were isolated following published protocols ( Schumaker and Puppione , 1986 ) from fresh EDTA-treated plasma by KBr density gradient ultracentrifugation in the density range 0 . 94–1 . 006 g/mL for VLDL , 1 . 019–1 . 063 g/mL for LDL , and 1 . 063–1 . 21 g/mL for HDL . Lipoproteins from each class migrated as a single band on the agarose and non-denaturing gels . Lipoprotein stock solutions were prepared by extensive dialysis against 50 mM sodium phosphate buffer , 150 mM NaCl , 0 . 25 mM EDTA , 0 . 02% NaN3 , pH 7 . 5 , degassed , and stored in the dark at 4°C . Each stock solution was used within two weeks during which no protein degradation was detected by SDS PAGE and no changes in the lipoprotein electrophoretic mobility were observed by agarose PAGE . To obtain HDL enriched with exogenous SAA ( termed SAA-HDL ) , single-donor HDL were isolated from healthy subjects and were incubated with SAA at 37°C for 6 hr in standard buffer . The molar ratio of SAA to apoA-I varied from 0:1 ( SAA-free control ) to 4:1 . Free ( uncomplexed ) proteins , which contained excess SAA as well as apoA-I that was displaced from HDL by SAA , were removed by size exclusion chromatography ( SEC ) as previously described ( Jayaraman et al . , 2015 ) . The total incubation mixture before purification is termed SAA-HDL ( total ) and the SEC-purified lipoprotein fraction is termed SEC Fr ( Figure 2—figure supplement 1A , B ) . As a control , HDL were incubated without SAA under otherwise identical conditions; no changes in the particle size , composition or stability were detected upon such incubation . Total protein concentration was determined by bicinchoninic acid assay , and individual apolipoprotein content was determined as a weight fraction of the total protein by quantifying SDS gel bands using image J software , as described previously ( Jayaraman et al . , 2015 ) . Intact non-modified HDL contained 75% apoA-I and 25% apoA-II; SAA-HDL ( 1:1 mol:mol SAA:apoA-I , SEC fr ) contained 65% apoA-I , 23% apoA-II and 12% SAA; and SAA-HDL ( 4:1 , SEC fr ) contained 56% apoA-I , 17% apoA-II and 27% SAA ( Figure 2—figure supplement 1C ) . To explore the effects of SAA on LDL , SAA ( 1 mg/ml ) was incubated with single-donor nomolipidemic plasma LDL ( 1 mg/ml apoB ) in standard buffer at 37°C for 3 hr . The mixture was analyzed by SEC and non-denaturing PAGE to ascertain that SAA did not bind LDL ( Figure 2—figure supplement 1D ) . To compare human LDLs that vary in endogenous FFA levels , plasma was pooled from 25 patients diagnosed with type-2 diabetes mellitus and from 25 healthy normolipidemic subjects , as previously described ( Jayaraman et al . , 2017b ) . Plasma was obtained at the Lipid Laboratory of Hospital de Saint Pau and LDL were isolated by density gradient centrifugation in the laboratory of Dr . Jose Luís Sanchez-Quesada at the Hospital de Saint Pau ( Barcelona , Spain ) ; these studies were done with the written informed consent of the patients and upon approval by the institutional ethics committee , as previously described ( Jayaraman et al . , 2017b ) . The FFA content , which was quantified by an enzymatic assay described below , was 20% higher in diabetic than in normolipidemic LDL . Lipoproteins were reconstituted using a thin film evaporation method . POPC was dissolved in chloroform:methanol ( 2:1 v/v ) , the organic solvent was evaporated under nitrogen stream , and the samples were dried under vacuum overnight at 4°C . MLVs were prepared by dispersing the lipid film in standard buffer followed by vortexing . SUVs ( diameter ~22 nm ) were prepared by sonication of MLVs and were used within one day . SAA-containing lipoproteins of controled size were prepared by incubating SAA with POPC SUV at 25°C for 3 hr using an SAA to POPC molar ratio of either 1:100 ( to prepare SAA-containing SUVs of ~22 nm in size ) or 1:10 ( to prepare SAA-POPC complexes of 8 nm in size ) . Excess lipid was removed by centrifugation and excess protein was removed by SEC . The protein to lipid weight ratio in the final preparations was 1:42 ( for 1:100 mol:mol initial ratio ) or 1:28 ( for 1:10 mol:mol initial ratio ) . SAA complexes with oleic acid were reconstituted as described by Jayaraman et al . ( 2018 ) . To reconstitute SAA complexes with lysoPC , SAA was incubated with freshly prepared lysoPC at 37°C for 1 hr at protein:lipid molar ratios varying from 1:1 to 1:100 . This range encompassed lysoPC concentrations below and above its critical micelle concentration ( 4–8 . 3 mM ) . MLV clearance by SAA at room temperature was monitored by turbidity at 325 nm using a Varian Cary-300 UV-vis spectrophotometer . SAA ( 20 µg/ml ) was rapidly mixed with MLV suspension in standard buffer ( 40 µg/ml lipid ) , and the time course of turbidity changes was recorded as micron-size MLVs were converted into smaller lipoprotein nanoparticles . MLVs alone were used as controls . SAA-POPC complexes ( 10 µM lipid ) , as well as human plasma HDL and LDL ( 0 . 5–1 . 0 mg/ml total protein ) , were used as substrates for sPLA2 . First , the lipoproteins were dialyzed against 10 mM Tris buffer saline at pH 7 . 5 . Phospholipid lipolysis was performed using 50 nm human recombinant sPLA2-IIa or bee-venom sPLA2-III in TBS at pH 7 . 4 in the presence or absence of 2% w/v of fatty-acid-free bovine serum albumin and 2 mM CaCl2 . After a 3 hr incubation at 37°C , the reaction was terminated by adding EDTA ( final concentration 20 mM ) . To assess the spontaneous lipolysis of phospholipids , control experiments were carried out under identical conditions without sPLA2 . Hydrolysis of POPC SUV by sPLA2 without SAA was also quantified as a control . The extent of hydrolysis was assessed by measuring the FFA products using an enzymatic assay kit ( Wako Chemicals ) . FFA and total phospholipids were quantified using kits from Bioassay Systems ( EnzycrhomTM free fatty assay kit EFFA-100 , and EnzychromTM phospholipid assay kit EPLP-100 ) , according to the manufacturer’s instructions . Lipoprotein lipids were assayed by thin layer chromatography using samples containing 0 . 5 mg/ml protein . The lipids were extracted following published protocols ( Folch et al . , 1957 ) , the organic solvent was dried , and the lipids were spotted on the plate ( glass backed , plain silica gel ) . The tank was first saturated with a chloroform:methanol:water ( 32 . 5:12 . 5:2 v/v/v ) solvent system . The plate was developed for 1 hr . The spots were identified by charring with sulfuric acid spray . CD data were recorded using an AVIV 62DS spectropolarimeter to monitor protein secondary structure and thermal stability . Far-UV CD spectra were recorded at 190–250 nm from solutions containing 0 . 1 mg/ml SAA in standard buffer . Melting data were recorded at 222 nm to monitor changes in the α-helical structure during sample heating and cooling at a constant rate of either 70 °C/h or 10 °C/h . The data at these scan rates closely overlapped , consistent with previous studies ( Jayaraman et al . , 2015 ) . Buffer baselines were subtracted from the data; the results were normalized to the protein concentration and reported as molar residue ellipticity , [Θ] . Helical content was estimated on the basis of the value of [Θ]222 , as previously described ( Mao and Wallace , 1984 ) . For non-denaturing PAGE , Novex 4–20% Tris-glycine gels ( Invitrogen ) were loaded with 6 μg protein per lane and run to termination at 1500 V·h under non-denaturing conditions in Tris-glycine buffer . For SDS PAGE , Novex 16% or 18% Tris-glycine gels were loaded with 5 μg protein per lane and run at 200 V for 1 hr under denaturing conditions in SDS-Tris-glycine buffer . The gels were stained with Denville Blue protein stain ( Denville Scientific ) . SEC was performed with a Superose 6 10/300 GL column controlled by an ÄKTA UPC 10 FPLC system ( GE Healthcare ) . Elution by 10 mM PBS at pH 7 . 5 was carried out at a flow rate of 0 . 5 ml/min . SAA oligomerization in SAA-POPC complexes before and after their hydrolysis with sPLA2 was assessed using glutaraldehyde , which can cross-link free NH2 groups that are separated by up to 12 Å . SAA-POPC ( 0 . 5 mg/ml protein ) was incubated for 30 min at 24°C with glutaraldehyde ( 0 . 01–0 . 08% ) , the reaction was quenched by adding 100 mM Tris , and the samples were analyzed by SDS PAGE . The results were obtained using standard buffer ( 50 mM sodium phosphate buffer , 150 mM NaCl , pH 7 . 5 ) , which yielded less non-specific aggregation . The exact cross-linking pattern varied depending on the salt concentration , and more non-specific higher order oligomers were observed in a low-salt buffer . SAA , either in lipid-free form or in complexes with lipids , was incubated with trypsin at 1:1500 mg:mg enzyme:substrate ratio in standard buffer at room temperature . Tryptic digestion was quenched using 2 mM of a serine protease inhibitor , phenylmethylsulfonyl fluoride . The reaction was quenched after 5 min for lipid-free SAA , which was rapidly digested , and after 1 hr for lipid-bound SAA , which was digested much more slowly . The reaction products were analyzed by SDS PAGE and matrix-assisted laser desorption ionization – time of flight mass spectrometry . For mass spectrometry , the spectra were recorded on a Reflex-IV spectrometer ( Bruker Daltonics , Billerica , MA ) equipped with a 337 nm nitrogen laser . The instrument was operated in the positive-ion reflection mode at 20 kV accelerating voltage with time-lag focusing enabled . Calibration was performed in linear mode using a standard calibration mixture containing the oxidized B-chain of bovine insulin , equine cytochrome C , equine apomyoglobin , and bovine serum albumin . The matrix , cyano-4-hydroxycinnamic acid ( alpha cyano , Mw = 189 g/mol ) , was prepared as a saturated solution in 70% acetonitrile and 0 . 1% trifluoroacetic acid in water . Mass spectrometry results were reported as an average of three independent experiments . To compare the ability of SAA and albumin to remove FFA from LDL in a broad range of physiologically relevant pH , samples of single-donor human LDL containing 1 mg/ml apoB were hydrolyzed using 50 nM sPLA2-III at 37°C for 3 hr at pH 7 . 5 , 6 . 5 , 5 . 5 or 4 . 5 . The hydrolysis was performed either sPLA2-III alone , sPLA2-III with 2% ( w/v ) FFA-free human serum albumin , or sPLA2-III with 2% ( w/v ) SAA . After incubation , LDL were re-isolated by density gradient centrifugation , and their FFA content was quantified by an enzymatic assay . The FFA content in intact LDL from the same batch was also quantified and the results are shown in Figure 6 . To compare the ability of human and murine SAA1 . 1 to remove endogenous FFA from LDL , pooled plasma LDL from healthy normolipidemic subjects and from diabetic patients were used . LDL were incubated at 37°C for 3 hr at pH 7 . 5 with either hSAA1 . 1 or mSAA1 . 1 using a 1:1 SAA:apoB weight ratio . Thereafter , LDL were re-isolated by density centrifugation and their FFA content was measured . The FFA content of intact LDL from the same batch was also measured ( Figure 6—figure supplement 2C ) . To ensure reproducibility , all experiments in this study were performed three or more times , unless otherwise stated . The enzymatic assays were performed in technical triplicates of biological duplicates and are reported as an average of a technical triplicate with the corresponding standard errors of means . Statistical analysis was performed using the ANOVA t-test . | Cell boundaries are made up of fatty substances known as lipids . When cells get severely damaged , their lipid membranes break apart . These broken fragments of membrane become highly toxic , and must be removed as soon as possible to allow the tissue to heal . A small protein called serum amyloid A , SAA for short , was recently proposed to play a pivotal role in this process . In humans , SAA levels in the blood rapidly spike to over a thousand times their normal level following inflammation , injury or infection . Combined with the fact SAA has been conserved for over 500 million years , this suggests that SAA must be important for survival . But , it is not entirely clear how this protein works . One clue for how SAA works is its relationship to another ancient protein called secretory phospholipase A2 . This protein , also known as sPLA2 , is part of a big family of enzymes that break down lipids in the cell membrane . Notably , sPLA2 levels rise at the same time and place as SAA during inflammation . This led Jayaraman et al . to ask whether SAA and sPLA2 might be working together to clean up the cell membrane debris . To find out , Jayaraman et al . mixed mouse SAA with vesicles of membrane lipids , and then added sPLA2 . This revealed that SAA reshapes the lipid membrane into smaller ‘nanoparticles’ with tightly curved surfaces that are easier for sPLA2 to break down . As the sPLA2 breaks up these particles , SAA then gathers up and gets rid of the leftover toxic fragments . This suggests that SAA has two roles: helping sPLA2 break down the membrane , and removing any toxic debris . Clearing debris after injury is essential for proper healing . So , understanding how it works is crucial to find new ways to treat inflammation . Further work to understand SAA and sPLA2 could improve our understanding of how to treat acute and chronic inflammation and its life-threatening complications . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"immunology",
"and",
"inflammation"
] | 2019 | Synergy between serum amyloid A and secretory phospholipase A2 |
Experience governs neurogenesis from radial-glial neural stem cells ( RGLs ) in the adult hippocampus to support memory . Transcription factors ( TFs ) in RGLs integrate physiological signals to dictate self-renewal division mode . Whereas asymmetric RGL divisions drive neurogenesis during favorable conditions , symmetric divisions prevent premature neurogenesis while amplifying RGLs to anticipate future neurogenic demands . The identities of TFs regulating RGL symmetric self-renewal , unlike those that regulate RGL asymmetric self-renewal , are not known . Here , we show in mice that the TF Kruppel-like factor 9 ( Klf9 ) is elevated in quiescent RGLs and inducible , deletion of Klf9 promotes RGL activation state . Clonal analysis and longitudinal intravital two-photon imaging directly demonstrate that Klf9 functions as a brake on RGL symmetric self-renewal . In vivo translational profiling of RGLs lacking Klf9 generated a molecular blueprint for RGL symmetric self-renewal that was characterized by upregulation of genetic programs underlying Notch and mitogen signaling , cell cycle , fatty acid oxidation , and lipogenesis . Together , these observations identify Klf9 as a transcriptional regulator of neural stem cell expansion in the adult hippocampus .
In the adult mammalian brain , radial-glial neural stem cells ( RGLs ) in the dentate gyrus subregion of the hippocampus give rise to dentate granule cells and astrocytes ( Seri et al . , 2001; Garcia et al . , 2004; Ahn and Joyner , 2005; Lagace et al . , 2007; Bonaguidi et al . , 2011; Encinas et al . , 2011; Gonçalves et al . , 2016b; Moss et al . , 2016; Pilz et al . , 2018 ) , a process referred to as adult hippocampal neurogenesis ( Altman and Das , 1965; Eriksson et al . , 1998; Spalding et al . , 2013; Boldrini et al . , 2018; Sorrells et al . , 2018; Moreno-Jiménez et al . , 2019; Tobin et al . , 2019; Gage , 2019; Knoth et al . , 2010 ) . Adult-born dentate granule cells integrate into hippocampal circuitry by remodeling the network and ultimately contribute to hippocampal-dependent learning and memory and regulation of emotion ( Gonçalves et al . , 2016b; Anacker and Hen , 2017; Miller and Sahay , 2019 ) . Levels of adult hippocampal neurogenesis are highly sensitive to experience ( Cope and Gould , 2019; Vicidomini et al . , 2020 ) suggesting that neurogenesis may represent an adaptive mechanism by which hippocampal-dependent memory functions are optimized in response to environmental demands . Essential to this adaptive flexibility is the capacity of RGLs to balance long-term maintenance with current or future demands for neurogenesis ( ‘anticipatory neurogenesis’ ) in response to distinct physiological signals ( Bonaguidi et al . , 2011; Cope and Gould , 2019; Vicidomini et al . , 2020; Dranovsky et al . , 2011; Schouten et al . , 2020 ) . Depending on environmental conditions , RGLs make decisions to stay quiescent or self-renew asymmetrically or symmetrically . Whereas enriching experiences ( e . g . , complex environments , exploration , and socialization ) bias RGLs toward asymmetric divisions to generate astrocytes and neurons ( Dranovsky et al . , 2011; Song et al . , 2012 ) , unfavorable conditions promote RGL quiescence ( e . g . , chronic stress and aging ) or symmetric self-renewal to support neural stem cell ( NSC ) expansion at the expense of neurogenesis ( e . g . , social isolation , seizures , and aging ) ( Dranovsky et al . , 2011; Sierra et al . , 2015; Ibrayeva et al . , 2021 ) . Asymmetric self-renewal of RGLs predominates over symmetric self-renewal division mode in the adult hippocampus and it ensures maintenance of RGL numbers while supporting current neurogenic demands ( Pilz et al . , 2018; Vicidomini et al . , 2020 ) . Conversely , symmetric self-renewal decouples RGL divisions from differentiation and is thought to serve distinct functions . First , symmetric divisions prevent premature differentiation of RGLs in a nonpermissive or unhealthy niche , and consequently , avert aberrant integration of adult-born dentate granule cells detrimental to hippocampal functions ( Ibrayeva et al . , 2021; Cho et al . , 2015 ) . As such , RGL amplification anticipates future demands for neurogenesis upon return to favorable conditions . Second , RGL expansion may represent an efficient mechanism to replenish the adult RGL pool after injury . Third , symmetric stem cell divisions maybe more efficient than asymmetric divisions for long-term maintenance since fewer divisions are required to maintain RGL numbers . Furthermore , symmetric divisions may be associated with a lower rate of mutations and reduced replicative aging ( Shahriyari and Komarova , 2013 ) . Extracellular physiological signals recruit transcription factors ( TFs ) within adult hippocampal RGLs to execute quiescence-activation decisions and symmetric or asymmetric self-renewal divisions ( Vicidomini et al . , 2020; Andersen et al . , 2014; Urbán et al . , 2019 ) . A growing number of transcriptional regulators of quiescence and asymmetric ( neurogenic or astrogenic ) stem cell renewal have been identified ( Mukherjee et al . , 2016; Jones et al . , 2015; Zhang et al . , 2019; Ehm et al . , 2010; Imayoshi et al . , 2010 ) . Deletion of such factors results in loss of RGL quiescence , increased neurogenesis and ultimately , differentiation-coupled depletion of the RGL pool . In sharp contrast , the identities of TFs that regulate RGL expansion have remained elusive . Here , we report that expression of the ubiquitously expressed TF , Kruppel-like factor 9 ( Klf9 ) , a regulator of dendritic and axonal plasticity in postmitotic neurons ( Moore et al . , 2009; McAvoy et al . , 2016 ) , is elevated in nondividing RGLs compared to dividing RGLs . Inducible genetic upregulation of Klf9 in RGLs and progenitors decreased activation , whereas conditional cell-autonomous deletion of Klf9 in RGLs promoted an activated state . Clonal lineage tracing and longitudinal two-photon imaging of adult hippocampal RGLs in vivo directly demonstrated a role for Klf9 as a brake on symmetric self-renewal . In vivo translational profiling of RGLs generated a molecular blueprint for RGL expansion in the adult hippocampus: we found that loss of Klf9 in RGLs results in downregulation of a program of quiescence-associated factors and upregulation of genetic ( mitogen , notch ) and metabolic ( fatty acid oxidation and lipid signaling ) programs underlying RGL symmetric self-renewal . Together , these data identify Klf9 as a transcriptional regulator of NSC expansion in the adult hippocampus . Our study contributes to an emerging framework for how experiential signals may toggle a balance of transcriptional regulators of symmetric and asymmetric self-renewal of RGLs to amplify NSCs or asymmetrically divide and generate neurons and astrocytes .
To characterize Klf9 expression in RGLs in the adult dentate gyrus , we bred Klf9-LacZ knockin reporter mice ( Scobie et al . , 2009 ) with a Nestin GFP transgenic mouse line in which Nestin+ RGLs are genetically labeled with GFP ( Mignone et al . , 2004 ) . Quantification of Klf9 expression based on LacZ intensities in Klf9 LacZ/+ mice revealed enrichment in quiescent RGLs relative to activated RGLs ( MCM2+ ) ( one-way analysis of variance [ANOVA] , F = 17 . 07 , p = 0 . 003 ) ( Figure 1A , B ) . MCM2 expression captures activated cells that have exited quiescence . To refine this estimation that is based on a surrogate ( LacZ ) of Klf9 expression within the RGL compartment , we performed fluorescence in situ hybridization ( FISH ) using a Klf9-specific riboprobe and immunohistochemistry for GFP and BrdU on adult hippocampal sections obtained from Klf9+/+ or LacZ/LacZ;Nestin GFP transgenic mice perfused 2 hr following a BrdU pulse ( one-way ANOVA , F = 5 . 6 , p = 0 . 04 ) ( Figure 1C–E ) . No signal was detected with FISH using the Klf9 riboprobe on brain sections from Klf9LacZ/LacZ mice thus conveying specificity of the riboprobe ( Figure 1D ) . Quantification of Klf9 transcripts using Image J revealed significantly enriched expression in quiescent vs . activated ( Brdu+ Nestin GFP+ ) RGLs ( Figure 1C , E ) . We next asked what happens when we delete Klf9 in adult hippocampal RGLs . To address this question , we engineered Klf9 conditional mutant mice ( Klf9f/f ) to cell autonomously delete Klf9 in RGLs . We first validated our Klf9f/f mouse line by crossing it with the POMC-Cre mouse line that drives recombination in the dentate gyrus . In situ hybridization ( ISH ) on hippocampal sections from POMC-Cre:Klf9f/f revealed salt and pepper expression of Klf9 in the dentate gyrus consistent with the established pattern of POMC-Cre-dependent recombination in the dentate gyrus ( McHugh et al . , 2007 ) . No signal was detected by ISH using the Klf9 riboprobe on brain sections from Klf9LacZ/LacZ mice thus conveying specificity of the riboprobe ( Figure 1—figure supplement 1 ) . We bred Klf9f/f mice with ( GLI-Kruppel family member 1 ) Gli1CreERT2 to recombine Klf9 ( deletion of exon 1 ) in hippocampal RGLs ( Figure 1F , G ) . We chose the Gli1CreERT2 driver line because population-based lineage tracing and chronic in vivo imaging suggests that Gli1CreERT2-labeled RGLs contribute to long-term maintenance and self-renewal ( Ahn and Joyner , 2005; Bottes et al . , 2021 ) . We next generated Klf9f/f or +/+ mice harboring a Gli1CreERT2 allele and a Cre-reporter allele ( Ai14 , B6;129S6-Gt ( ROSA ) 26Sortm14 ( CAG-tdTomato ) Hze/J ) ( Madisen et al . , 2010 ) to indelibly label Gli1-positive RGLs and their progeny ( Figure 1F ) . By analysis of Klf9 transcript-associated fluorescence intensity in Gli1-positive tdTomato-labeled RGLs we estimated the recombination frequency of Klf9 ( i . e . , reduction in Klf9-associated signal ) to be approximately 32% in Gli1-positive tdTomato-labeled RGLs ( Figure 1—figure supplement 2 ) . We induced Klf9 recombination and tdTomato expression in RGLs of adult ( 2 months old ) Gli1CreERT2;Klf9f/f or +/+; Ai14 mice and processed brain sections for Nestin , tdTomato , and MCM2 immunohistochemistry 7 days postinjection ( 7 dpi ) to quantify activated RGLs ( Figure 1F , G ) . We found that conditional deletion of Klf9 in Gli1CreERT2-targeted adult hippocampal RGLs significantly increased the fraction of activated RGLs ( % of MCM2+ tdTomato + RGLs ) ( Figure 1G; unpaired t-tests , Figure 1G , p < 0 . 0001 ) . Complementing these results , we demonstrated that genetic overexpression of Klf9 in activated hippocampal NSCs and progenitors of adult Sox1 tTA;tet0 Klf9 mice ( McAvoy et al . , 2016; Venere et al . , 2012 ) significantly decreased the fraction of activated and dividing cells ( Figure 1—figure supplement 3 ) . Together , these data demonstrate that Klf9 expression is enriched in quiescent RGLs and that loss of Klf9 expression in RGLs either promotes or maintains an activated state of RGLs in the adult hippocampus . We next asked how Klf9 loss-of-function in RGLs affects self-renewal division mode . Population-level lineage tracing experiments at short-term chase time points suggested that Klf9 loss in Gli1+ RGLs increased RGL numbers ( data not shown ) . However , analysis of NSC dynamics at the population level is encumbered by changes in numbers of labeled progeny overtime ( Bonaguidi et al . , 2011; Bottes et al . , 2021 ) . The challenges of interpreting population-level analysis are exacerbated because Klf9 is also expressed in immature adult-born neurons and mature dentate granule cells . As such , changes in numbers of labeled descendants following loss of Klf9 in RGLs make population-level lineage tracing difficult to interpret . Therefore , to directly investigate whether loss of Klf9 in RGLs results in NSC expansion at a single clone level , we performed in vivo clonal analysis in adult Gli1CreERT2;Klf9f/f or +/+; Ai14 mice shortly after low-dose tamoxifen adminsitration . Single dose of TAM at 50 mg/kg body weight permitted sparse labeling of single tdTomato+ RGLs and visualization of labeled single RGL clones and their individual constituents . Analysis of clonal composition at 7 dpi revealed a significantly greater fraction of multi-RGL containing clones and a smaller fraction of single RGL containing clones ( Figure 2A–D , Figure 2—figure supplement 1 , Figure 2—videos 1–8; Figure 2B , two-way ANOVA , genotype × cell type p < 0 . 0001 , Bonferroni post hoc Klf9+/+ vs . Klf9f/f . 1 RGL n . s . , 2+ RGLs p < 0 . 0001 , 1 RGL+ p < 0 . 0001 , no RGL n . s . Figure 2D , left panel , two-way ANOVA , genotype × cell type p < 0 . 0001 , Bonferroni post hoc Klf9+/+ vs . Klf9f/f . 2+ RGLs p = 0 . 09 , 2 RGLs+ P + A p < 0 . 0001 , 2 RGLs+ p n . s . Figure 2D , right panel , two-way ANOVA , genotype × cell type p = 0 . 09 , Bonferroni post hoc Klf9+/+ vs . Klf9f/f . 1 RGL n . s . , 1 RGL+ P + A p = 0 . 01 , 1 RGL+ P p = 0 . 01 ) . Many of the multi-RGL containing clones also comprised of neural progenitors and astrocytes suggesting that loss of Klf9 biases RGL expansion but does not prevent RGL differentiation into progeny ( Figure 2D ) . To corroborate these findings and address any potential confound introduced by bias in the Ai14 genetic lineage tracer , we performed clonal analysis at 7 dpi using a different lineage tracing reporter , mTmG ( t ( ROSA ) 26Sortm4 ( ACTB-tdTomato , -EGFP ) Luo/J ) ( Muzumdar et al . , 2007 ) . Our analysis demonstrated a significant increase in multi-RGL clones and decrease in single RGL clones derived from RGLs lacking Klf9 in Gli1CreERT2;Klf9 f/fmTmG mice ( Figure 2E , F; Figure 2F , two-way ANOVA , genotype × cell type p < 0 . 0001 , Bonferroni post hoc Klf9+/+ vs . Klf9f/f . 1 RGL n . s . , 2 RGLs+ p = 0 . 0001 , 1 RGL+ p = 0 . 03 , no RGLs n . s ) . The lack of a difference in single RGL clones suggests that loss of Klf9 maintains the activated state associated with increased symmetric divisions . Alternatively , it may reflect a floor effect ( estimation of recombination frequency suggests that not all tdTomato RGLs undergo recombination for Klf9 ) in the assay that occludes detection of a decrease in number . These findings provide evidence for Klf9 in cell-autonomous regulation of RGL expansion and are suggestive of a role for Klf9 in inhibition of symmetric self-renewal of RGLs . To unequivocally establish clonal origin of labeled progeny and directly test the hypothesis that Klf9 inhibits symmetric self-renewal of RGLs in vivo , we performed longitudinal two-photon imaging ( Gonçalves et al . , 2016a ) of RGLs for up to 2 months and tracked symmetric and asymmetric division patterns ( Figure 3A , B; Figure 3—figure supplement 1 , Figure 3—videos 1–4 ) . We implanted Gli1CreERT2:Klf9f/f or +/+;Ai14 mice with a hippocampal window over CA1 for long-term imaging . After allowing 2 weeks for recovery from surgery , we injected mice with a single dose of tamoxifen ( 150 mg/kg ) to induce Cre recombination and tdTomato expression in Gli1+ RGLs ( as shown in Figure 1—figure supplement 2 ) . This resulted in sparse labeling that allowed us to image and track individual cells and their processes . Imaging sessions started 2 days post-tamoxifen injection ( dpi ) and were repeated daily up to 6 dpi in order to locate isolated labeled RGLs which were clearly identifiable by their tufted radial process . Astrocytes were occasionally labeled but were readily distinguishable from RGLs due to their lack of polar morphology and were disregarded . Post hoc histology analysis of morphological features and immunoreactivity for GFAP in brain sections was performed to corroborate our initial in vivo identification of a subset of RGLs ( Figure 3 , Figure 3—figure supplement 1 ) . After 6 dpi we track individual RGLs to quantify their first division event: we revisited each previously identified , RGL containing field-of-view every 3 days and compared it with previous time points in order to quantify the first cell division and classify it as symmetric or asymmetric . As previously described ( Pilz et al . , 2018 ) , asymmetric divisions resulted in motile daughter cells that migrated away from their progenitors within one to two imaging sessions ( 3–6 days ) and exhibited shorter and less stable processes , often undergoing further divisions and differentiation ( Figure 3B; Figure 3—video 1 ) . Conversely , and as shown previously ( Pilz et al . , 2018 ) , symmetric divisions resulted in the appearance of a faint radial process of a single static daughter cell that remained adjacent to its mother cell ( Figure 3B; Figure 3—videos 3; 4 ) . Over time the cell body of the daughter RGL emerged . For our analysis of cell division , we only considered the first division event from an identified RGL , disregarding subsequent divisions of the daughter cells and analysis of RGL-derived lineage trees . Deletion of Klf9 in RGLs resulted in a significantly greater number of symmetric cell divisions ( 39 symmetric , 38 asymmetric , 10 mice ) compared to Klf9+/+ RGLs ( 18 symmetric , 47 asymmetric , 8 mice ) ( Figure 3C ) . We made sure to have a similar number of division events across both genotypes so that we were confident that the differences in the mode of division are not due to under/over sampling each experimental group ( N = 8 control mice , 65 divisions , mean 8 . 125 divisions per mouse; 10 experimental mice , 77 divisions , mean 7 . 7 divisions per mouse ) ( Figure 3D ) . These data provide definitive evidence for Klf9 functioning as a brake on symmetric self-renewal of RGLs in the adult hippocampus . To understand how Klf9 regulates RGL division mode , we performed in vivo molecular profiling of RGLs lacking Klf9 . We generated Gli1CreERT2:Rpl22HAf/+:Klf9f/f or +/+ ( B6N . 129-Rpl22tm1 . 1Psam/J mice Ribotag ) mice ( Sanz et al . , 2009 ) to genetically restrict expression of a hemagglutinin ( HA ) epitope-tagged ribosomal subunit exclusively in Gli1+ RGLs ( Figure 4A ) . Four days following TAM injections to induce HA expression and Klf9 recombination in a sufficient number of Gli1+ RGLs and progeny arising from first division , we dissected the dentate gyrus subregion , biochemically isolated actively translated transcripts , generated cDNA libraries and performed Illumina sequencing ( Figure 4A–C ) . Analysis of the resulting data and gene ontology annotation ( gGOSt , https://biit . cs . ut . ee/gprofiler/gost ) of differentially expressed genes ( DEGs ) ( Supplementary file 1 ) broadly categorized signaling pathways and molecular programs associated with NSC activation and quiescence ( Mukherjee et al . , 2016; Zhang et al . , 2019; Mira et al . , 2010; Codega et al . , 2014; Shin et al . , 2015; Hochgerner et al . , 2018; Knobloch et al . , 2013; Figure 4C; Figure 4—figure supplement 1 , Supplementary files 2 and 3 ) . Functional categories enriched among upregulated DEGs ( 276 ) included phospholipase activity ( Pla2g7 , Pla2g4e , and Gpld1 ) , mitogen growth factor signaling ( Egfr , Fgfr3 , Ntrk2 , and Lfng ) , and ligand-gated ion channels ( Gabra1 , Chrna7 , Grin2C , and P2r × 7 ) . Additionally , our analysis revealed elevation of metabolic programs sustaining energy production and lipogenesis through generation of Acetyl-CoA: CoA- and fatty acid-ligase activity ( Acsl3 , Ascl6 , Acss1 , and Acsbg1 ) and oxidoreductase and aldehyde dehydrogenase activity ( Acad12 , Acox1 , Ak1b10 , Aldh3b1 , and Aldh4a1 ) ( Knobloch et al . , 2013; Namba et al . , 2021; Knobloch et al . , 2017; Xie et al . , 2016; Supplementary file 2 ) . A complementary set of modules overrepresented in the downregulated gene set ( 462 DEGs ) were quiescence growth factor signaling ( Bmp2 and Bmp4 ) , extracellular matrix binding ( Itga3 , Itga10 , and Igfbp3 , 6 , 7 ) , cell adhesion ( e . g . , Emb , Itga3 , and Itga5 ) , actin binding ( Iqgap1 ) , TFs ( NeuroD4 and Zic3 ) , and voltage-gate potassium channel activity ( Kcnj8 and Kcnq1 ) ( Supplementary file 3 ) . For validation of DEGs previously linked with NSC quiescence and activation ( Mukherjee et al . , 2016; Zhang et al . , 2019; Codega et al . , 2014; Shin et al . , 2015; Knobloch et al . , 2013; Baser et al . , 2019 ) , we performed qRT-PCR on an independent replicate of biochemically isolated mRNAs from this population of Gli1+ RGLs in vivo . We first confirmed downregulation of Klf9 in RGLs . Next , we validated downregulation of canonical quiescence signaling factors ( Bmp4 ) and upregulation of genes involved in lipid metabolism ( Pla2g7 ) , cell cycle ( Ccn1a ) , mitogen signaling ( epidermal growth factor receptor , Egfr ) , and Notch signaling ( Lunatic fringe , Lfng ) ( Figure 4D ) . Consistent with Lfng-mediated potentiation of Notch1 signaling through cleavage of the Notch1 intracellular domain ( NICD ) , we observed significantly elevated levels of NICD in Gli1+ RGLs lacking Klf9 ( Figure 4E; Hochgerner et al . , 2018; Zhao and Wu , 2018 ) . We infer from our loss-of-function data that high levels of Klf9 in RGLs induce BMP4 expression and repress gene modules specifying mitogen signaling , fatty acid oxidation , RGL differentiation , and cell-cycle exit to inhibit RGL expansion .
Central to experience-dependent regulation of neurogenesis is the ability of RGLs to constantly balance demands for neurogenesis and astrogenesis or RGL expansion with self-preservation through regulation of quiescence . Since interpretation of the external world is dependent on integration and convergence of physiological extracellular signals upon TFs in RGLs , enriching and adverse experiences are likely to modulate the balance between transcriptional programs that regulate RGL division modes supporting amplification or asymmetric self-renewal ( Vicidomini et al . , 2020 ) . However , in contrast to our knowledge of TFs that regulate asymmetrical self-renewal of RGLs in the adult hippocampus ( Mukherjee et al . , 2016; Jones et al . , 2015; Zhang et al . , 2019; Ehm et al . , 2010; Imayoshi et al . , 2010 ) , the identities of transcriptional regulators of symmetric self-renewal of RGLs have remained elusive . By combining conditional mouse genetics with in vivo clonal analysis and longitudinal two-photon imaging of RGLs , we demonstrated that Klf9 acts as a transcriptional brake on RGL activation state and expansion through inhibition of symmetric self-renewal ( Figure 5 ) . That Klf9 expression is higher in nondividing RGLs than in activated RGLs is consistent with gene expression profiling of quiescent adult hippocampal RGLs ( Bottes et al . , 2021; Knobloch et al . , 2013; Jaeger and Jessberger , personal communication ) and other quiescent somatic stem cells such as satellite cells ( Pallafacchina et al . , 2010 ) and NSCs in the subventricular zone ( Codega et al . , 2014; Morizur et al . , 2018; Renault et al . , 2009 ) . Loss of Klf9 in Gli1+ RGLs resulted in increased RGL activation . Based on our clonal analysis of RGL output and in vivo translational profiling , we think that this increased RGL activation reflects maintenance of an activated or cycling state ( also discussed later ) to support increased symmetric self-renewal ( Encinas et al . , 2011 ) . Our current knowledge of TFs that regulate symmetric self-renewal in the adult hippocampus can only be extrapolated from studies on hippocampal development ( Noguchi et al . , 2019 ) . Clonal analysis of Gli1-targeted RGLs revealed multi-RGL containing clones with progeny . This potentially reflects competition between TFs that dictate balance between symmetric and asymmetric divisions , compensation by downstream effectors of Klf9 or constraints on RGL expansion imposed by availability of niche factors . Such compensatory mechanisms may also explain why constitutive deletion of Klf9 does not overtly affect size of the dentate gyrus ( Scobie et al . , 2009 ) . Studies on adult hippocampal neural stem and progenitor cells have relied on assays that induce quiescence and activation in vitro ( Knobloch et al . , 2013 ) , unbiased single cell profiling of neurogenesis ( Shin et al . , 2015; Hochgerner et al . , 2018 ) or FACS sorting of neural stem and progenitor cells in vivo ( Zhang et al . , 2019 ) . Because asymmetric self-renewal is the predominant mode of division , it is most certainly the case that the RGL activation profile inferred from these studies is biased toward asymmetric , rather than symmetric , self-renewal . In contrast , our in vivo translational profiling of long-term self-renewing Gli1+ RGL population following cell-autonomous deletion of Klf9 allowed us to infer how changes in gene expression relate to RGL symmetric division mode and create an exploratory resource for the NSC research community . While ribosomal profiling does not allow us to isolate transcripts from single RGLs , it offers other advantages such as minimizing stress response associated with cell dissociation ( Machado et al . , 2021 ) . Since Gli1CreERT2 specifically targets RGLs and astrocytes ( but not progenitors ) and we performed biochemical profiling at 4 days postrecombination when we first observe RGL derived progeny , our analysis largely reflects changes in the RGL population , progeny arising from first division and astrocytes . That we observe an enrichment of genes expressed exclusively in RGLs permits us to link gene expression with changes in RGL numbers driven by division mode . Analysis of Klf9 levels by qRT-PCR suggests greater than 50% recombination efficiency of Klf9 in targeted cell populations . Our genome-wide expression analysis suggests that Klf9 functions as an activator or repressor depending on cellular context , although repression appears to be the dominant mode of gene regulation ( Knoedler et al . , 2017; Ying et al . , 2014 ) . Validation of specific DEGs in biochemically isolated transcripts from RGLs suggests that Klf9 may activate BMP4 expression in RGLs to suppress activation in vivo ( Mira et al . , 2010 ) . Additionally , Klf9 suppresses RGL proliferation through repression of mitogen signaling receptor tyrosine kinases ( EGFR ) , lipidogenesis ( Pla2g7 ) , and cell cycle ( CyclinA1 ) . Pla2g7 , interestingly , is expressed only in RGLs and astrocytes in the DG ( Shin et al . , 2015; Hochgerner et al . , 2018 ) and as such may represent a novel marker of activated RGLs . Given the dual roles of Notch signaling in regulation of active and quiescent RGLs ( Sueda and Kageyama , 2020 ) , we validated that Lfng is significantly upregulated in RGLs lacking Klf9 . Lfng is exclusively expressed in RGLs , promotes Notch1 signaling through glycosylation of Notch1 and generation of NICD following ligand binding , and dictates RGL activation in a ligand-dependent manner ( Semerci et al . , 2017 ) . Genetic overexpression of Lfng in T-cell progenitors sustained Notch1-mediated self-renewal and clonal expansion at expense of differentiation ( Yuan et al . , 2011 ) . Consistent with Lfng upregulation in RGLs , we observed elevated levels of NICD in Gli1+ RGLs lacking Klf9 indicative of enhanced Notch1 signaling . Bioinformatics analysis of our data identified enhanced fatty acid β-oxidation ( FAO ) , a substrate for energy production and lipogenesis as a metabolic program recruited to sustain RGL expansion ( Figure 5 ) . In fact , lineage tracing studies on embryonic neocortical NSCs have demonstrated a role for FAO in maintenance of NSC identity and proliferation ( Namba et al . , 2021 ) . Specifically , inhibition of Tmlhe ( a carnitine biosynthesis enzyme ) and carnitine-dependent long-chain FAO ( carnitine palmitoyltransferase I , CPT1 , which catalyzes the rate-limiting reaction in this process ) resulted in a marked increase in symmetric differentiating divisions at expense of both symmetric and asymmetric self-renewal of NSCs ( Xie et al . , 2016 ) . Inhibition of FAO prevented hematopoietic stem cell maintenance and promoted symmetric differentiating divisions of hematopoietic stem cells ( Ito et al . , 2012 ) . High levels of FAO are directly linked to intestinal stemness ( Mana et al . , 2021 ) and persistence of proliferative capacity across cancers ( Oren et al . , 2021 ) . In sharp contrast to these findings , it has been suggested that high levels of FAO are important for maintaining RGL quiescence . Specifically , deletion of Cpt1a ( and inhibition of FAO ) in adult hippocampal NSC and progenitors impaired expansion and reduced numbers of RGLs . However , it could not be determined if this was due to death and/or inhibition of symmetric self-renewal of RGLs ( Knobloch et al . , 2017 ) . Based on our data , we propose that NSCs , like other somatic stem cells and progenitors , require high levels of FAO for symmetric self-renewal or expansion . How does Klf9 function as a brake on RGL symmetric self-renewal ? We propose that Klf9 corepresses a suite of genes associated with maintenance of RGLs in symmetric division mode . Pioneering studies have implicated Notch signaling in sustaining symmetric divisions of neuroepithelial cells ( Egger et al . , 2010 ) , expansion of putative NSCs and progenitors ( Androutsellis-Theotokis et al . , 2006 ) and maintenance of radial glial cell like identity through inhibition of differentiation and cell-cycle exit ( Gaiano et al . , 2000; Yoon et al . , 2008 ) . Importantly , genetic gain-of-function of Notch1 signaling in RGLs in the adult DG maintains RGLs at the expense of hippocampal neurogenesis ( Breunig et al . , 2007 ) . Klf9 may also directly suppress a proneurogenic program in RGLs ( e . g . , NeuroD4 , downregulated DEG , Supplementary file 3; Masserdotti et al . , 2015 ) or indirectly via competitive interactions with TFs that regulate RGL asymmetric self-renewal . Taken together , loss of Klf9 in RGLs drives expansion through enhanced mitogen and cell-cycle signaling ( Berdugo-Vega et al . , 2020 ) , prevention of RGL differentiation , and elevation of lipogenic and FAO metabolic programs ( Figure 5 ) . Our findings stimulate discussion on how experiential signals regulate RGL activation and expansion . To date , GABA ( A ) R signaling and PTEN signaling ( by inhibiting PI3K–Akt pathway ) have been shown to promote quiescence and suppress RGL amplifying divisions ( Bonaguidi et al . , 2011; Song et al . , 2012 ) . It is plausible that Klf9 participates in these signaling pathways as a downstream actuator . Klf9 expression is reduced in NSCs lacking FoxO3 ( Renault et al . , 2009 ) . Thus , Akt-dependent regulation of NSC activation through inactivation of FoxO3 ( Urbán et al . , 2019 ) may require Klf9 downregulation . Since some of the identified Klf9 target genes are also regulated by other TFs ( e . g . , inhibition of EGFR and cyclinA1 by Notch2 [Zhang et al . , 2019] , activation of Pla2g7 by FoxO3 [Renault et al . , 2009] ) , we infer that these factors do not compensate each other , but instead , confer flexible integration of diverse physiological signals in RGLs to regulate activation . Inhibition of pulsatile glucocorticoid receptor signaling has also been shown to promote RGL quiescence ( Schouten et al . , 2020 ) . Because Klf9 expression is regulated by steroid hormone signaling and neural activity ( Scobie et al . , 2009; Datson et al . , 2011; Besnard et al . , 2018 ) and Klf9 represses gene expression through recruitment of a mSin3A corepressor complex ( Zhang et al . , 2001 ) , Klf9 may support an epigenetic mechanism for reversible , experiential regulation of NSC decision making . Our genome-wide dataset serves as a general exploratory community resource in several ways . First , it catalyzes further enquiry into mechanisms underlying NSC quiescence and expansion . By way of example , candidate genes such as the cell adhesion molecule Embigin ( downregulated DEG ) regulates quiescence of hematopoietic stem/progenitor cells ( Silberstein et al . , 2016 ) whereas the alpha7 nicotinic receptor ( upregulated DEG ) , ChrnA7 , has been shown to be required for maintaining RGL numbers ( Otto and Yakel , 2019 ) . Second , numerous genes identified in our blueprint are implicated in driving tumorigenesis and as such may guide differentiation-based strategies to block tumor proliferation ( Carracedo et al . , 2013 ) . Third , our work motivates assessment of how Klf9 may link extracellular , physiological signals with genetic and metabolic programs in RGLs . Fourth , our findings may guide investigation of functional significance of Klf9 enrichment in other quiescent neural ( SVZ ) ( Codega et al . , 2014; Morizur et al . , 2018; Renault et al . , 2009 ) and somatic stem cell populations ( Pallafacchina et al . , 2010 ) . Our study enables a more holistic assessment of how competing transcriptional programs in RGLs mediate decision making by including regulators of symmetric and asymmetric self-renewal . A deeper understanding of Klf9-dependent regulation of RGL homeostasis may guide genetic and metabolic strategies to replenish the RGL reservoir and restore neurogenesis following injury or expand the NSC pool in anticipation of future neurogenic demands to support hippocampal-dependent memory processing and emotional regulation ( Anacker and Hen , 2017; Miller and Sahay , 2019; McAvoy et al . , 2016 ) .
The following mouse lines were obtained from Jackson Labs: Klf9-lacZ knock-in ( Stock No . 012909 ) , Gli1CreERT2 ( Stock No . 007913 ) , Ai14 ( Stock No . 007908 ) , mT/mG ( Stock No . 007676 ) , B6N . 129-Rpl22tm1 . 1Psam/J ( RiboTag ) ( Stock No . 011029 ) , and POMC-Cre ( Stock No . 010714 ) . Sox1tTA transgenic mice ( Venere et al . , 2012 ) were obtained from Dr . Robert Blelloch ( University of California , San Fransisco ) . Klf9LacZ/LacZ mice were obtained from Dr . Yoshiaki Fujii-Kuriyama ( University of Tsukuba and is also available from Jackson Labs , Stock No . 012909 ) . tetO Kf9/Klf9 knockin mice were generated by us previously ( McAvoy et al . , 2016 ) . Nestin GFP mice ( Mignone et al . , 2004 ) were obtained from Dr . David Scadden at MGH . Klf9 conditional knockout mice were generated through homologous gene targeting using C57BL/6 ES cells by Cyagen . F0s were bred with C57BL/6J mice to generate F1s with germline transmission and mice were backcrossed with C57BL/6J mice for 5+ generations . A set of primers ( forward: GGTAGTCAAATGGCGCAGCTTTT; reverse: CCATCCATTCCTTCATCAGTCTCC ) was used to genotype Klf9+/+ or f/f mice to amplify 363 bp mutant band and 240 bp wildtype band . Gli1CreERT2:Klf9+/+ or f/f Ai14 and Gli1CreERT2:Klf9+/+ or f/f:mT/mG+/− , were generated by crossing Gli1CreERT2 mice with mT/mG or Ai14 and Klf9+/+ or f/f mice in a C57BL/6J background . For analysis of cell proliferation in dentate gyrus , mice were injected with BrdU ( 200 mg/kg body weight , i . p . ) and sampled 2 hr later . For analysis of long-term retaining cells in dentate gyrus , mice were given daily injection of BrdU ( 25 mg/kg body weight , i . p . ) for 14 days and sampled 24 hr after the last injection . Tamoxifen ( 20 mg/ml , Sigma , T5648 ) was freshly prepared in a 10% ethanol of corn oil ( Sigma C8267 ) . For population analysis , a dose of 150 or 250 mg/kg was intraperitoneally injected into 8-week-old male and female mice ( Figure 1F ) . For clonal analysis , a dose of 50 and 100 mg/kg were used in reporter lines of Ai14 and mT/mG , respectively ( Figure 2A , E ) . Mice were sampled 7 or 28 days post-tamoxifen injection . For two-photon imaging ( Figure 3A ) , one dose of 150 mg/kg tamoxifen was given 2 days prior to in vivo imaging . For ribosomal profiling , a dose of 250 mg/kg body weight was intraperitoneally injected into 2–3 months mice every 12 hr for three times . Mice were sampled 4 days after the last injection ( Figure 4A ) . 35 μm cryosections obtained from perfused tissue were stored in phosphate-buffered saline ( PBS ) with 0 . 01% sodium azide at 4°C . For immunostaining , floating sections were washed in PBS , blocked in PBS containing 0 . 3% Triton X-100% and 10% normal donkey serum and incubated with primary antibody overnight at 4°C overnight ( Rockland , rabbit anti RFP , 1:500; Millipore , chicken anti-GFAP , 1:2000; goat anti-GFP , Novus , 1:500; Santa Cruz , sc-8066 , Goat anti-DCX , 1:500 ) . The Mcm2 ( BD Biosciences , mouse anti-Mcm2; 1:500 ) , GFP ( Abcam , Chicken anti-GFP , 1:2000 ) , LacZ ( Promega , Mouse anti-beta Galactosidase , 1:2000 ) , and Nestin ( Aves lab , chicken anti-Nestin , 1:400 ) antigens were retrieved by incubating brain sections in citric buffer in pressure cooker ( Aprum , 2100 retriever ) for 20 min , followed by 60 min cooling to room temperature . BrdU antigen was retrieved by incubating brain sections in 2 N HCl for 30 min at 37°C following 15 min fixation in 4% paraformaldehyde ( PFA on previously processed fluorescent signal ) . On the next day , sections were rinsed three times for 10 min in PBS and incubated for 90 min with fluorescent-label-coupled secondary antibody ( Jackson ImmunoResearch , 1:500 ) . Sections were rinsed three times for 10 min each in PBS before mounting onto glass slides ( if applicable ) and coverslipped with mounting media containing DAPI ( Fluoromount with DAPI , Southern Biotech ) . NICD ( rabbit anti-cleaved Notch1 , Assay Biotech Cat# L0119 RRID:AB_10687460 at 1:100 ) immunostaining was performed as described ( Semerci et al . , 2017 ) . We used a transgenic mouse line that expresses GFP under the control of the Nestin promoter to label the cell bodies ( Mignone et al . , 2004 ) . Mice were sacrificed 2 hr after a single BrdU injection ( 200 mg/kg ) . Klf9 expression was detected by florescent in situ hybridization ( FISH ) using a Klf9 antisense probe complementary to exon 1 ( 530–1035 bp ) of Klf9 mRNA . Briefly , ISH was performed using dioxygenin-labeled riboprobes on 35 μm cryosections generated from perfused tissue as described ( McAvoy et al . , 2016 ) . Premixed RNA labeling nucleotide mixes containing digoxigenin-labeled UTP ( Roche Molecular Biochemicals ) were used to generate RNA riboprobes . Klf9 null mice were used as a negative control and to validate riboprobe specificity . Riboprobes were purified on G-50 Microspin columns ( GE Healthcare ) . Probe concentration was confirmed by Nanodrop prior to the addition of formamide . Sections were mounted on charged glass ( Superfrost Plus ) slides and postfixed for in 4% PFA . Sections were then washed in DEPC-treated PBS , treated with proteinase K ( 40 μg/ml final ) , washed again in DEPC-treated PBS , and then acetylated . Following prehybridization , sections were incubated with riboprobe overnight at 60°C , washed in decreasing concentrations of SSC buffer , and immunological detection was carried out with anti-DIG peroxidase antibody ( Roche ) at 4°C overnight and were visualized using Cy3-conjugated Tyramide Signal Amplification system ( Perkin-Elmer ) at room temperature . ISH was followed by immunostaining for GFP ( Goat anti-GFP , Novus , 1:500 ) and BrdU ( Rat anti-BrdU , Biorad , 1:500 ) incubated at 4°C overnight and followed by incubation of 488- and Cy5-conjugated secondary antibodies ( Jackson ImmunoResearch , 1:500 ) for 2 hr at room temperature . Klf9 ISH was performed on POMC-Cre:Klf9+/+ and f/f mice using Klf9 exon1 probe to validate the Klf9 conditional knockout mice . Immunological detection was carried out with anti-DIG antibody conjugated with alkaline phosphatase ( Roche ) . Color reaction was conducted with NBT/BCIP . Klf9 null mice were used as a negative control . Estimation of Klf9 recombination frequency in Gli1-positive tdTomato-labeled RGLs . Gli1CreERT2:Klf9+/+:Ai14 and Gli1CreERT2:Klf9f/f:Ai14 mice were given one dose of tamoxifen ( 150 mg/kg ) IP and were then perfused with DPEC-treated PBS and fixed with 4% PFA . 35 μm cryosections sections were mounted on the same slides . After hybridization with Klf9 riboprobe , slides were washed and blocked with NEN buffer for 1 hr at RT . The following antibodies were used for immunostaining: anti-DIG peroxidase antibody ( mouse , 1/8000 , Roche ) ; anti-RFP ( rabbit , 1/500 , Rockland ) . Slides were washed and coverslipped with mounting medium ( Southern biotech ) . Klf9 fluorescence intensity within the cell body was recorded . Images were obtained from one set of brain sections ( six sets generated for each brain ) for each immunostaining experiment ( set of antigens ) . Stained sections were imaged at ×20 or ×40 on a Nikon A1R Si confocal laser , a TiE inverted research microscope or a Leica SP8 confocal microscope . All of analysis were performed by an experimenter blind to group identity . LacZ intensity quantification . We used mice carrying a LacZ allele knocked into the endogenous Klf9 allele ( Klf9LacZ/LacZ or LacZ/+ mice ) ( Scobie et al . , 2009 ) . Klf9Lac/LacZ or LacZ/+ mice were crossed with Nestin GFP mice to generate Klf9LacZ//+;Nestin GFP mice . These mice were used to quantify LacZ expression levels in quiescent RGLs ( GFP+ MCM2 with radial process ) , activated RGLs ( GFP+ MCM2+ with radial process ) , and neural progenitor cells ( NPCs; ( GFP+ MCM2+ without radial processes ) ) . The distinction between RGLs and NPCs was determined through morphological analysis . Images ( 1024 resolution ) were acquired as 7 Z-stacks with a step size of 1 μm . Two to four stacks of images from each mouse were selected for further quantification . Since the LacZ gene had been knocked into the endogenous Klf9 locus , mean intensity of LacZ expression , assessed by fluorescent signal with LacZ immunostaining using ImageJ software in each GFP+ cell body , was used as a surrogate for Klf9 expression in Klf9LacZ/+ mice . Mean background intensity was obtained from LacZ negative regions being divided from the calculations in the same section . Klf9 FISH signal quantification . Images ( 2048 resolution ) were acquired by a Leica SP8 confocal microscope as 30 Z-stacks with a step size of 0 . 5 μm . Representative images were generated by exporting stacked confocal images at full resolution for three-dimensional visualization using Imaris . The distinction between NPCs and NSCs was determined through morphological analysis with GFP staining . Activated RGLs were differentiated from quiescent RGLs through BrdU antibody staining ( cell proliferation markers ) . Analysis and quantification of Klf9 signal intensity in each GFP+ cell body were conducted using automatic counting with ImageJ software . Images were converted into 1-bit images . Then Klf9 puncta were counted within GFP+ cell body boundaries through particle analysis allowing for number and average size of puncta to be recorded . Klf9 null mice crossed with Nestin GFP mice were used as a negative control . Clonal analysis was conducted with sparse labeling after optimizing dose of tamoxifen as previously described ( Bonaguidi et al . , 2011 ) . Ai14 and mTmG reporter mice were used to visualize the recombined cells . Serial coronal sections were generated and immunostained for GFAP , RFP , or GFP antigens . Images acquisition and analysis were restricted to entire dentate gyri ~2000 μm along the dorsal–ventral axis . RGLs were classified as cells that were located in the subgranular zone , had radial projections that extended into the granule cell layer , and were colabeled with GFAP and RFP or GFP . Cells with GFAP labeling without radial processes but exhibiting a bushy morphology were identified as astrocytes . Recombined GFP+ or RFP+ cells without GFAP labeling in close spatial proximity to other cells were identified as neuronal progeny cells . A ring with a radius of 50 μm from the center of the RGL was used to determine the clone composition . A single cell ( astrocyte or neuron ) was not counted as a clone . Images ( 1024 resolution ) were acquired using a Leica SP8 confocal microscope as 20–25 Z-stacks with a step size of 1 . 5 μm . Mice with less than two clones per hemisection on average were determined as standard for sparse labeling and were selected for clonal analysis . Except for the single RGL clone category , all the labeled cells within one clone were in close spatial proximity to each other . Clones were categorized according to the presence or absence of an RGL and the type of progeny . For imaris image analysis , Z-series confocal images were processed for all the channels . The intensity of each channel was adjusted and representative images were used to generate a TIFF file by taking a ‘screen snapshot’ . Twelve- to sixteen-week-old Gli1CreERT2:Klf9+/+ or f/f:Ai14 mice were used for intravital 2P imaging of RGLs . Window implantation: We followed an established protocol to implant a cranial window over the right hemisphere of the dorsal hippocampus ( Pilz et al . , 2018 ) . Briefly , we drilled a ~3-mm wide craniotomy , removed the underlying dura mater and aspirated the cortex and corpus callosum . A 3-mm diameter , 1 . 3-mm deep titanium implant , with a glass sealed to the bottom was then placed above the hippocampus . The implant and a titanium bar ( 29 × 3 . 8 × 1 . 3 mm ) were held in place with dental cement . A titanium bar was used in order to secure the animal to the microscope stage . Mice were given a single dose of dexamethasone ( 1 mg/kg , i . p . ) before surgery to reduce brain swelling , and carprofen ( 5 mg/kg , i . p . ) for inflammation and analgesic relief after surgery completion . Implanted animals were given 2 weeks to recover from surgery and allow any inflammation to subside . Two-photon imaging of aRGL divisions: In vivo imaging was done on a custom two-photon laser scanning microscope ( based on Thorlabs Bergamo ) using a femtosecond-pulsed laser ( Coherent Fidelity 2 , 1075 nm ) and a ×16 water immersion objective ( 0 . 8 NA , Nikon ) . We imaged mice under isoflurane anesthesia ( ~1% isoflurane in O2 , vol/vol ) and head-fixed to the microscope stage via a titanium bar implant while resting on a 37°C electrical heating pad ( RWD ThermoStar ) . Expression of the tdTomato fluorescent label in Gli1+ RGLs was induced with a single injection of Tamoxifen ( 150 µl/mg ) 2 weeks after window implantation . Imaging began 2 days after tamoxifen injection ( 2 dpi ) and continued every day until 6 dpi in order to locate sparse labeled RGLs . Afterwards , mice were imaged every 3 days , whenever possible and were imaged up to 60 days . Using a coordinate system , we marked locations of RGLs for recurrent imaging of the same cell . At each time point , we acquired a three-dimensional image stack of each field of view containing tdTomato-expressing cells and annotated their location so that the same cell could be imaged again in the following session . Cell division classification: Cell divisions were analyzed by two different experimenters blinded to genotype . We first compiled all Z-stacks into a single sum-projected image for each time point , and then we used FIJI-ImageJ to analyze the images . Only the first recorded cell division for a given clone was included in the analysis . We defined RGL symmetric division as a new RGL generated from the mother RGL , characterized by the development of a stable radial process and static behavior of cell bodies for at least 6 days after birth . We defined asymmetric division as new NPCs generated from the mother RGL that exhibited shorter and less stable processes . These NPCs often began to migrate away within one to two imaging sessions ( 3–6 days ) . We used Gli1CreERT2:Rpl22HAf/+:Klf9f/f or +/+ mice which enables expression of HA-tagged ribosomal protein RPL22 ( RPL22–HA ) following Cre recombination in Gli1+ Klf9/f or +/+ RGLs . RiboTag immunoprecipitation and RNA extraction were performed 4 days after last TAM injection following the original protocol with minor modifications ( Sanz et al . , 2009 ) . Six dentate gyri from three mice were pooled per sample and homogenized with a dounce homogenizer in 900 µl cycloheximide-supplemented homogenization buffer . Homogenates were centrifuged and the supernatant incubated on a rotator at 4°C for 4 hr with 9 µl anti-HA antibody ( CST Rb anti-HA #3724 , 1:100 ) to bind the HA-tagged ribosomes . Magnetic IgG beads ( Thermo Scientific Pierce #88847 ) were conjugated to the antibody–ribosome complex via overnight incubation on a rotator at 4°C . RNA was isolated by RNeasy Plus Micro kit ( Qiagen 74034 ) following the manufacturer’s protocol . Eluted RNA was stored at −80°C . For enrichment analysis , 45 µl of homogenate ( pre- anti-HA immunoprecipitation ) was set aside after centrifugation , kept at −20°C overnight , and purified via RNeasy Micro kit as an ‘input’ sample , and used to determine NSC enrichment . RNA quantity and quality were measured with a Tape Station ( Agilent ) and Qubit fluorimeter ( Thermo Fisher Scientific ) . Sequencing libraries were prepared using Ultra Low Input RNA Kit ( Clontech ) . NGS libraries were constructed from total RNA using Clontech SMARTer v4 kit ( Takara ) , followed by sequencing on an Illumina HiSeq 2500 instrument , resulting in 20–30 million 50 bp reads per sample . The STAR aligner ( Dobin et al . , 2013 ) was used to map sequencing reads to transcriptome in the mouse mm9 reference genome . Read counts for individual genes were produced using the unstranded count function in HTSeq v . 0 . 6 . 0 ( Anders et al . , 2015 ) , followed by the estimation of expression values and detection of differentially expressed transcripts using EdgeR ( Robinson et al . , 2010 ) and including only the genes with count per million reads >1 for one or more samples ( Anders et al . , 2013 ) . DEGs were defined by at least 1 . 2-fold change with p < 0 . 05 . NCBI GEO accession number GSE164889 . qRT-PCR mRNA was biochemically pooled and isolated as described above for ribosomal profiling . The first-stranded complementary DNA was generated by reverse transcription with SuperScript IV first-strand synthesis system ( Thermo Fisher Scientific ) . For quantification of mRNA levels , aliquoted cDNA was amplified with specific primers and PowerUp SYBR Master Mix ( BioRad ) by CFX384 Touch Real-Time PCR detection system ( BioRad ) . Primers were optimized and designed to hybridize with different exons . Primers are listed here ( name and sequence 5′ → 3′ are indicated ) . Statistical analysis was carried out using GraphPad Prism software . Both data collection and quantification were performed in a blinded manner . Data in figure panels reflect several independent experiments performed on different days . An estimate of variation within each group of data is indicated using standard error of the mean . Comparison of two groups was performed using two-tailed Student’s unpaired t-test unless otherwise specified . Comparison of one group across time was performed using a one-way ANOVA with repeated measure . Comparison of two groups across treatment condition or time was performed using a two-way repeated measure ANOVA and main effects or interactions were followed by Bonferroni post hoc analysis . In the text and figure legends , ‘n’ indicates number of mice per group . Detailed statistical analyses can be found in Supplementary file 4 . For statistical analysis of DEGs , please see RNA-seq analysis section for details . Two-photon imaging: In order to compare differences in the modes of RGL division between the two genotypes , we used the R statistical analysis software to fit a generalized linear mixed effects model to the division numbers across different mice , using genotype as a fixed effect , and including animal identity as a random effect in order to account for differences between individual animals [DivisionType ~ Genotype + ( 1|MouseIdentity ) ] . p values were calculated with a likelihood-ratio test comparing our model to a null model with no genotype information and identical random effects [DivisionType ~ 1 + ( 1|MouseIdentity ) ] . | In humans and other mammals , a region of the brain known as the hippocampus plays important roles in memory . New experiences guide cells in the hippocampus known as radial-glial neural stem cells ( RGLs ) to divide to make new neurons and other types of cells involved in forming memories . Each time an RGL divides , it can choose to divide asymmetrically to maintain a copy of itself and make a new cell of another type , or divide symmetrically ( a process known as symmetric self-renewal ) to produce two RGLs . Symmetric self-renewal helps to restore and replenish the pool of stem cells in the hippocampus that are lost due to injury or age , allowing us to continue making new neurons . Proteins known as transcription factors are believed to control how RGLs divide . Previous studies have identified several transcription factors that regulate the RGLs splitting asymmetrically to make neurons and other cells . But the identities of the transcription factors that regulate symmetric self-renewal in the adult hippocampus have remained elusive . Here , Guo et al . searched for transcription factors that regulate symmetric self-renewal of RGLs in mice . The experiments found that RGLs that are resting and not dividing ( referred to as ‘quiescent’ ) have higher levels of a transcription factor called Klf9 than RGLs that are actively dividing . Loss of the gene encoding Klf9 triggered quiescent RGLs to start dividing , and further experiments showed that Klf9 directly inhibited symmetric self-renewal . Guo et al . then used an approach called in vivo translational profiling to generate a blueprint that revealed new insights into the molecular processes involved in this symmetric division . These findings pave the way for researchers to develop strategies that may expand the numbers of stem cells in the hippocampus . This could eventually be used to help replenish brain circuits with neurons and improve the memory of individuals with Alzheimer’s disease or other conditions that cause memory loss . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"neuroscience"
] | 2022 | Transcriptional regulation of neural stem cell expansion in the adult hippocampus |
Many small endotherms use torpor to reduce metabolic rate and manage daily energy balance . However , the physiological ‘rules’ that govern torpor use are unclear . We tracked torpor use and body composition in ruby-throated hummingbirds ( Archilochus colubris ) , a long-distance migrant , throughout the summer using respirometry and quantitative magnetic resonance . During the mid-summer , birds entered torpor at consistently low fat stores ( ~5% of body mass ) , and torpor duration was negatively related to evening fat load . Remarkably , this energy emergency strategy was abandoned in the late summer when birds accumulated fat for migration . During the migration period , birds were more likely to enter torpor on nights when they had higher fat stores , and fat gain was positively correlated with the amount of torpor used . These findings demonstrate the versatility of torpor throughout the annual cycle and suggest a fundamental change in physiological feedback between adiposity and torpor during migration . Moreover , this study highlights the underappreciated importance of facultative heterothermy in migratory ecology .
Facultative hypothermia is an energy conservation strategy that allows many mammalian and some avian species to survive periods of resource unavailability or to optimize their energy budgets in certain environments or life stages ( McKechnie and Lovegrove , 2002; Ruf and Geiser , 2015 ) . During facultative hypothermia , metabolic rates and body temperatures are reduced to varying extents across species and environmental conditions ( Ruf and Geiser , 2015 ) . As some of the smallest avian species , hummingbirds ( Trochilidae ) are known for their ability to use daily torpor , a deep , short-term form of facultative hypothermia , to cope with energetic challenges they face daily and throughout their annual cycle ( Carpenter , 1974; Hainsworth et al . , 1977 ) . Studies that have investigated the use of torpor in hummingbirds in relation to food availability and body mass suggest that torpor initiation is controlled by an endogenous mechanism sensitive to an energy-store threshold ( Hainsworth et al . , 1977; Hiebert , 1992; Powers et al . , 2003 ) . This model predicts that a bird will initiate torpor if its energy stores reach critically low levels with enough time remaining in the night to achieve net energy savings . Hummingbirds typically rewarm 1–2 hr before sunrise , so they rarely enter torpor after approximately 75% of the night has elapsed; even if critically low energy levels are reached late in the night , birds may avoid entering torpor at this point if the energetic benefits are outweighed by the potential costs ( e . g . predation , moult delay ) ( Bouma et al . , 2010; Carr and Lima , 2013; Hainsworth et al . , 1977; Hiebert , 1992; Hiebert , 1990 ) . Previous studies also suggest that the function of torpor shifts seasonally , from an energy emergency survival mechanism to an energy-storage maximization strategy during migration ( Carpenter and Hixon , 1988; Hiebert , 1993 ) . However , this threshold has not been repeatedly and accurately measured in individual birds spanning life history stages , and the relationship between torpor use and the components of body composition ( fat and lean mass ) remains unclear . We explored torpor use in ruby-throated hummingbirds ( Archilochus colubris ) , which breed in eastern North America in the early and mid-summer , and migrate to wintering grounds in Mexico and Central America in the late summer . In the breeding period , birds maintain relatively lean body compositions to optimize aerial agility , important for successful courtship displays and competitive interactions that allow them to maintain secure access to food resources ( Altshuler et al . , 2010; Hou and Welch , 2016 ) . But like most long-distance migrants , ruby-throated hummingbirds substantially increase their body mass prior to migratory departure to fuel their journey ( Hou and Welch , 2016 ) . We repeatedly and non-invasively quantified the relationship between torpor and endogenous energy stores ( fat ) in ruby-throated hummingbirds to investigate the underlying rules of torpor use during the breeding and migration periods . We predicted that in the breeding period , when birds reached a low energy-store threshold before 75% of the night had elapsed , they would enter torpor to avoid complete energy depletion , and that birds that remained normothermic until this point would not enter torpor if it would not achieve net energy savings . We also predicted that this threshold would be abandoned in the late summer to facilitate premigratory fattening . Furthermore , we predicted that in the breeding period , torpor use would be primarily driven by evening fat content and the rate at which those energy reserves were used , and that in the migration period , the amount of torpor used would be driven primarily by night length , allowing the birds to spare a maximum amount of fat stores , irrespective of longer late-summer nights . We measured the torpor use and body composition of captive adult ( and one juvenile ) male ruby-throated hummingbirds ( n = 16; capture mass: 2 . 5–3 . 2 g ) that experienced semi-natural photoperiods , on 158 focal bird-nights throughout the summer . On all days and nights , the birds experienced air temperatures of approximately 20°C ( 19 . 7°C ± 0 . 0°C ) to control for the potential effect of air temperature as a proximate cue for torpor use . We used respirometry to calculate rates of energy expenditure and the rate of oxidation of stored fat , and quantitative magnetic resonance ( QMR ) to measure body composition ( Guglielmo et al . , 2011; Lighton , 2008 ) . We identified the start of torpor entry and arousal by evaluating the slope of each smoothed V´O2 trace , and we calculated the fat content at the time of torpor entry as the amount of evening fat ( g ) minus the estimated amount of fat expenditure prior to torpor entry , divided by the morning body mass . We aimed to determine the rules governing the use of torpor and whether these differed throughout the summer . We evaluated changes in the relationships between evening body composition and torpor occurrence , torpor duration , time of torpor entry , fat content at torpor entry , and amount of energy expended before torpor entry , within and among periods of consistent change in body mass . We also evaluated the influence of night length on these variables . To specifically investigate the role of torpor in driving premigratory increases in body mass , we evaluated the effects of mean torpor duration and mean daily food consumption on the amount of mass gained and duration of the fattening period .
Throughout the study period , 13 of 16 birds exhibited relatively low morning body masses , indicative of breeding condition , until late August or early September when they substantially increased their body mass; the birds subsequently maintained high body masses . We analyzed daily changes in morning body mass and defined ‘breeding’ , ‘fattening’ , and ‘migration’ periods , respectively ( Materials and methods , Supplementary file 3 ) . Additionally , three ‘non-fattener’ birds maintained relatively constant body masses throughout the summer ( Figure 1; Supplementary file 1 ) . QMR scans indicated that changes in body mass were driven primarily by increases in fat ( r ( 109 ) =0 . 94 , 95% CI [0 . 92 , 0 . 96] p < 0 . 001; Appendix 1—figure 1A ) , and that body mass and lean mass were slightly negatively correlated ( r ( 107 ) =−0 . 26 , 95% CI [−0 . 43 , –0 . 08] , p = 0 . 006; Appendix 1—figure 1B ) . Individual variation in daily food consumption and presumably activity were the primary factors leading to nightly variation in evening body composition , as environmental factors such as air temperature , humidity , and food availability were consistent throughout the study period . During the mid-summer breeding period , birds maintained consistently low morning body masses ( 2 . 77 ± 0 . 05 g; slope: 0 . 00 ± 0 . 00 g⋅day–1; p = 0 . 338; Figure 1; Table 1; Supplementary file 1 ) . On average , birds used torpor on 61 . 6% ± 11 . 1% of focal bird-nights ( Supplementary file 1 ) . When birds started the night with less fat , they were more likely to enter torpor ( slope: –0 . 70 ± 0 . 22; p = 0 . 001; Appendix 1—figure 2 ) , entered torpor earlier in the night ( slope: 0 . 55 ± 0 . 09 hentry ⋅%fat–1; p < 0 . 001 ) , and remained torpid longer ( slope: –0 . 63 ± 0 . 10 hr⋅%fat–1 , p < 0 . 001; Figure 2A ) . Neither torpor propensity ( slope: –0 . 31 ± 0 . 47 , p = 0 . 452 ) , the time of torpor entry relative to the start of the night ( slope: 0 . 52 ± 0 . 30 hentry⋅hnight , p = 0 . 054 ) nor torpor duration ( slope: 0 . 52 ± 0 . 32 htorpor⋅hnight , p = 0 . 077 ) were significantly related to night length ( Supplementary file 1 ) . Furthermore , when birds started the night with greater fat content , they expended more energy before initiating torpor ( slope: 0 . 39 ± 0 . 06 kJ⋅%fat–1 , p < 0 . 001 , Appendix 1—figure 3A ) , and lost more fat mass overnight ( slope: 1 . 00 ± 0 . 17 mgfat⋅%fat–1 , p < 0 . 001; Figure 2A ) . Additionally , on longer nights , birds spent significantly more energy before entering torpor ( slope: 0 . 40 ± 0 . 22 kJ⋅hnight–1 , p = 0 . 041 ) , and lost more fat ( slope: 1 . 35 ± 0 . 59 mgfat⋅hnight–1 , p < 0 . 001 ) . Throughout the breeding period , birds consistently entered torpor at a time when they had a relatively low remaining fat level ( 5 . 56% ± 0 . 79%; 14 . 27 ± 3 . 77 mgfat; Figure 3B; Supplementary file 1 ) . The fat content at the time of torpor entry did not vary with night length ( slope: 0 . 84 ± 0 . 78 %fat⋅hnight–1 , p = 0 . 253 ) , or the time of torpor entry ( slope: 0 . 38% ± 0 . 31% fat⋅hentry–1; p = 0 . 191; Supplementary file 1 ) . Consistent with published observations , birds never entered torpor after approximately 75% of the night had elapsed , except on one night in which the bird entered torpor at 80% of the night and aroused within 1 hr ( Hiebert , 1992; Figure 3 ) . On 47 of 55 ( 85% ) bird-nights , the birds either entered torpor when their fat contents reached approximately 5% of body mass , or did not enter torpor if their fat content passed this threshold after 75% of the night had elapsed ( Supplementary file 2 ) . Substantial changes in torpor use accompanied changes in body composition in the late summer when the birds fattened prior to migration . In late August and September , 13 of 16 birds increased body mass ( slope: 0 . 02 ± 0 . 00 g⋅day–1 , p < 0 . 001; Supplementary file 1 ) over the span of 10 ± 1 days ( range: 6–18 days; Figure 1; Supplementary file 1 ) . In this short period , birds increased body mass by an average of 0 . 58 ± 0 . 05 g ( range: 0 . 34–0 . 90 g ) . Relative to the start of fattening , birds increased their body mass by an average of 19 . 6% ± 1 . 6% ( range: 11 . 6–29 . 0%; Supplementary file 1 ) . In every period , there was a negative relationship between torpor duration and overnight fat mass loss ( slopes: breeding: –1 . 72 ± 0 . 05 mgfat⋅hr–1 , fattening: –1 . 67 ± 0 . 10 mgfat⋅hr–1 , migration: –1 . 53 ± 0 . 07 mgfat⋅hr–1 , p < 0 . 001; Appendix 1—figure 4; Supplementary file 1 ) . Because longer torpor durations invariably spared more fat , we predicted that more frequent and longer torpor use , in addition to higher food consumption during the fattening period , would enhance the rate of premigratory fattening . Most interestingly , birds that used torpor for longer on average achieved greater mass gains during the fattening period ( slope: 0 . 07 ± 0 . 01 g⋅hr–1; p = 0 . 004; n = 11; Figure 4A; Supplementary file 1 ) . Contrary to our predictions , food consumption did not significantly affect the amount of fattening ( slope: 0 . 00 ± 0 . 00 g⋅kJ–1 , p = 0 . 996 ) , and neither mean torpor duration ( slope: –0 . 60 ± 0 . 75 days⋅hr -1 , p = 0 . 468 ) nor mean food consumption ( slope: –0 . 27 ± 0 . 25 days⋅kJ–1 , p = 0 . 343 ) was related to the length of the fattening period ( Figure 4; Supplementary file 1 ) . In the migration period , birds maintained greater morning body masses compared to the summer ( 3 . 73 ± 0 . 05 g , p < 0 . 001 ) , and the migration period body masses remained stable ( slope: 0 . 00 ± 0 . 00 g⋅day–1 , p = 0 . 079; Figure 1; Table 1; Supplementary file 1 ) . Despite beginning nights with three to five times more fat than they would need to remain normothermic for the entire night at 20°C ( mean overnight fat loss on normothermic nights was 19 . 42 ± 0 . 43 mgfat in the breeding period; 19 . 60 ± 0 . 42 mgfat during fattening; and 19 . 92 ± 0 . 47 mgfat during the migration period [pairwise comparisons: p > 0 . 745] , when accounting for the effect of night length [p < 0 . 001] ) and not approaching the critical threshold apparent in the breeding period , birds used torpor in the migration period at similar frequencies to those of the breeding period ( breeding: 61 . 6% ± 11 . 1%; migration: 65 . 1% ± 11 . 1% , p = 0 . 946; Supplementary file 1 ) . However , in stark contrast to the breeding period , birds were more likely to enter torpor on nights when they started with greater fat stores ( slope: 0 . 26 ± 0 . 11; p = 0 . 017 ) and on longer nights ( slope: 2 . 06 ± 0 . 94 , p = 0 . 028; Appendix 1—figure 2 ) . Additionally , the time of torpor entry ( slope: –0 . 08 ± 0 . 07 hentry⋅%fat–1; p = 0 . 230 ) , energy expenditure before torpor entry ( slope: –0 . 06 ± 0 . 05 kJ⋅%fat–1; p = 0 . 204; Figure 3A ) , and overnight fat mass loss ( slope: –0 . 15 ± 0 . 12 mgfat⋅%fat–1; p = 0 . 145; Figure 2B ) were not significantly related to evening fat content . Time of torpor entry relative to the start of the night ( slope: –0 . 51 ± 0 . 56 hentry⋅hnight–1 , p = 0 . 333 ) , energy expenditure before torpor entry ( slope: –0 . 41 ± 0 . 41 kJ⋅hnight–1 , p = 0 . 276 ) , and overnight fat mass loss ( slope: –0 . 02 ± 1 . 07 mgfat⋅hnight–1 , p = 0 . 982 ) were consistent across the migration period and not significantly related to night length ( Supplementary file 1 ) . Conversely , torpor duration was not significantly related to evening fat stores ( slope: 0 . 10 ± 0 . 07 hr⋅%fat–1; p = 0 . 124; Figure 2A ) but was positively correlated with night length ( slope 1 . 18 ± 0 . 58 htorpor⋅hnight–1 , p = 0 . 031 ) . In contrast to the breeding period , average fat content at the time of torpor entry was substantially greater in the migration period ( 32 . 94% ± 0 . 77%; 125 . 15 ± 3 . 85 mgfat; p < 0 . 001; Figure 3B; Supplementary file 1 ) . Additionally , while there was a significant negative relationship between time of torpor entry and the fat content at that time ( slope: –1 . 20 ± 0 . 35 % fat⋅hentry–1 , p < 0 . 001; Figure 3A; Supplementary file 1 ) , fat content at torpor entry did not vary with night length ( slope: 0 . 99 ± 1 . 41%fat⋅hnight–1 , p = 0 . 450 ) .
Ruby-throated hummingbirds exhibit substantial shifts in body composition and use of torpor between the summer breeding period and the migration period as a means to achieve premigratory fat gains ( Table 1 ) . In the summer , torpor is sensitive to critically low endogenous energy reserves; however , when the birds fatten for migration , this rule is abandoned , and the birds enter torpor with high levels of fat . While it does not appear that torpor initiation in the migration period is simply sensitive to a high energy threshold , the abandonment of the emergency threshold indicates that the rules governing torpor use are dependent on life history stage , and that hummingbirds may employ torpor as part of various energy management strategies throughout the annual cycle . In the summer breeding period , ruby-throated hummingbirds reserve torpor for times when they face critically low fat stores during the night . At air temperatures of 20°C , which free-living birds commonly experience , torpor initiation is primarily driven by instantaneous fat stores . On nights when they started with lower evening fat stores birds depleted energy stores to critically low levels earlier in the night , and thus remained torpid longer , irrespective of night length . These results support the hypotheses that torpor use is sensitive to a low , consistent threshold of fat during the breeding period , and that torpor is an energy emergency survival strategy mechanism that protects hummingbirds from depleting energy stores during the night or before they can reach their first meal in the morning ( Hainsworth et al . , 1977; Hiebert , 1992; Powers et al . , 2003 ) . The birds obeyed the average threshold on the clear majority of focal nights ( 47 out of 55 nights ) . On eight nights , 6 of the 13 fattener birds either crossed the threshold and did not enter torpor , or entered torpor without ever crossing the threshold , but we argue it is not surprising that there were a few exceptions to the threshold rule . The threshold could occasionally be overridden to remain normothermic despite reaching critical energy levels if extended bouts of torpor on subsequent nights impede regenerative processes associated with normothermy and sleep , or increase risk of predation ( Bouma et al . , 2010; Carr and Lima , 2013; Hiebert , 1990 ) . Similarly , we and other researchers have observed hummingbirds prematurely rewarming after inadvertent light or noise disturbances , which suggests that the threshold override and emergency arousal mechanisms could be related ( A . Shankar , E . Eberts , personal observations ) . Although we did not observe any obvious signs of greater stress in some individuals than others , it is also possible that some birds ( e . g . those that were more recently captured ) were more sensitive to handling and confinement in the respirometry chamber and therefore initiated torpor despite not having depleted fat levels to the threshold level ( Hiebert et al . , 2000 ) . Furthermore , although temperature , precipitation , and food availability were controlled in this study , these factors would likely affect free-living hummingbirds’ torpor use decisions by effecting the amount of fat the birds start each night with and the rate that they expend that fat ( Hainsworth et al . , 1977; McGuire et al . , 2021 ) . For example , the birds were fasted to ensure accurate body composition measurements , but if the birds were allowed to eat during the last 2 hr of the day , it is likely that they would have amassed larger evening fat stores and therefore would have entered torpor later in the night or not at all ( Eberts et al . , 2019 ) . Additionally , the level of the energy-store threshold could be modulated in response to anticipated energy demand . For instance , free-living birds at our study site during the breeding period experienced minimum nighttime air temperatures between 10°C and 25°C ( London International Airport , Ontario , Canada ) . At colder temperatures , normothermic hummingbirds would need to sustain a higher resting metabolic rate and may therefore anticipate an energy emergency earlier in the night ( Hiebert , 1990 ) . While these factors likely play a role in the complex decision-making that governs free-living hummingbird torpor use during the breeding season , our evidence strongly supports the existence of an ‘adipostat’ mechanism that initiates compensatory physiological changes depending on the instantaneous level of endogenous energy stores ( Boyer and Barnes , 1999; Powers et al . , 2003 ) . The patterns and role of torpor use drastically shifts as hummingbirds accumulate fat stores prior to migration . Contrary to the breeding period , torpor propensity and torpor duration were primarily driven by night length rather than evening fat content , and torpor entry , energy expenditure before torpor entry , and overnight fat loss did not vary with night length . Instead of initiating torpor to survive the night whenever they reach a critically low threshold , hummingbirds appear to enter torpor after a consistent amount of time , so that they expend a predictable amount of energy and achieve greater energy savings as they experience progressively longer nights in the late summer . This suggests that during the migration period , birds maximize their time in torpor , but must remain normothermic for a consistent period of time , perhaps accounting for the time it takes to process their final evening meal and their blood sugar declines to a level allowing for torpor initiation ( Eberts et al . , 2019 ) . This normothermic period could also allow the birds to achieve sufficient sleep before entering torpor when the regenerative and immunological benefits of sleep are unlikely to occur ( Bouma et al . , 2010 ) . Overall , the seasonal change in torpor use , from a survival strategy initiated by an ‘adipostat’ mechanism to a more routine use of torpor to maximize energy savings and build fat stores prior to migration , shows that torpor is a critical energy management strategy that allows migratory hummingbirds to balance fuel supply and demand during a particular season . Similar torpor use patterns have been well studied in mammals , though studies investigating avian endocrine mechanisms are needed to elucidate the proximate factors of torpor initiation across taxa . The similarities between the life histories of North American hummingbirds and bats provide an important lens through which to interrogate the ultimate drivers of torpor use throughout the annual cycle ( McGuire et al . , 2012 ) . Using torpor while roosting could allow both hummingbirds and bats animals to maintain high fuel stores when feeding opportunities are constrained by migratory priorities . Unlike most nocturnal migrant birds that can replenish fuel stores during the day , ruby-throated hummingbirds are diurnal migrants that do not forage at night . In an opposite but parallel manner , North American bats migrate at night when they would otherwise forage and do not forage during the day . In both of these types of animals , foraging , migrating , and sleeping can be mutually exclusive endeavors , but torpor can allow them to conserve energy and avoid making extended refueling stopovers ( McGuire et al . , 2014 ) . While we did not study free-living birds while they migrated , the drastic shift in the relationship between hummingbird body composition and torpor use , and of the link between torpor duration and mass gains in the late summer compellingly support the ‘torpor-assisted migration’ hypothesis ( Carpenter and Hixon , 1988; Hiebert , 1993; Hou and Welch , 2016; McGuire et al . , 2012 ) . The magnitude and timing of fattening in our captive birds resembled those documented in some free-living ruby-throated hummingbirds ( present study: 0 . 58 ± 0 . 05 g , 19 . 6% ± 1 . 6% over 10 ± 1 days; Hou and Welch , 2016: ~ 0 . 65 g , or 17% over 4 days ) . However , not all birds in this study showed such substantial body mass changes; there was continuous variation in magnitude and duration of fattening within the fattener birds , and three non-fattener birds maintained relatively lean body compositions throughout the study period . The ranges in magnitude and timing of fattening are not surprising because hummingbirds are asocial birds that do not migrate in flocks , and we would not expect them to have finely synchronized timing of migratory preparation , especially in the absence of natural ecological cues . Furthermore , our captive birds varied in wing feather wear and activity level , suggesting that the amount of fat they could deposit was limited by how much extra weight they could carry , determined by wing morphology , pectoral mass , and power output ( Chai , 1997; Dakin et al . , 2020 ) . These interindividual differences could also reflect disparate energy management strategies observed among free-living individuals . For example , a study examining migratory paths of juveniles suggested that adults take a more direct route than individuals ( Zenzal and Moore , 2016 ) . These disparate migratory strategies could be driven by differences in competitive abilities or experience from previous migratory journeys ( Carpenter et al . , 1993b; Carpenter et al . , 1993a; Hou and Welch , 2016; Kodric-Brown and Brown , 1978; Welch et al . , 2008 ) . This hypothesis is supported by the fact that one of the three non-fattener birds was the only juvenile in our study , and ongoing work is investigating potential age/sex class differences in an ecologically relevant context . This is the first study to non-terminally and repeatedly sample the body composition of individuals to accurately define a consistent rule governing torpor use in hummingbirds: birds will enter torpor when their fat stores reach a consistently low fat threshold ( 5% ) , during life history stages where a relatively lean body composition is advantageous . This rule may explain the typically low degree of torpor use in dominant individuals and large species , and the more frequent use of torpor in subordinate individuals and smaller species ( Powers et al . , 2003 ) . Moreover , our study is consistent with the long standing , but heretofore unproven , assumptions of the key role of torpor in premigratory fattening and in refueling at migratory stopover sites . Hummingbirds dramatically shift their rules for torpor use and enter torpor at high fat levels during times when it is advantageous to accumulate excess energy stores . These findings demonstrate the versatility of torpor as an energy management mechanism throughout the annual cycle and have important implications for understanding the physiological basis of torpor initiation .
Adult ( and one juvenile ) male ruby-throated hummingbirds ( A . colubris; n = 16; capture mass: 2 . 54–3 . 2 g ) were captured with a modified box trap ( drop door trap ) in London , ON , Canada , at the University of Western Ontario . Captive hummingbirds were housed individually in EuroCage enclosures ( Corners Ltd , Kalamazoo , MI ) , measuring 91 . 5 cm W × 53 . 7 cm H × 50 . 8 cm D , at the University of Western Ontario’s Advanced Facility for Avian Research . Once captive , birds were fed ad libitum on a 20% ( w/v ) solution of a Nektar-Plus ( Guenter Enderle , Tarpon Springs , FL ) , and were housed at 20°C and approximately 50% relative humidity . Birds experienced semi-natural photoperiods that were changed approximately weekly , ranging from 15 hr light/9 hr dark to 12 hr light/12 hr dark . These photoperiods are reflective of the birds’ natural summer photoperiod , as the data were collected in the summers 2018 and 2019 . Lights were abruptly turned on in the morning and shut off in the evening . The birds transitioned from a breeding condition in the beginning of the study period to a migratory condition in end of the study period . Details of animal husbandry and all experiments were approved by the University of Toronto ( protocol # 20011649 ) and the University of Western Ontario Animal Care Committees ( protocol #2018–092 ) . Hummingbirds were captured under Ontario Collecting Permit SC-00041 . This study uses QMR to measure body composition . QMR is a technology developed in the last 15 years that allows for non-invasive measurement of the masses of fat , lean tissue , and total body water ( Guglielmo et al . , 2011 ) . QMR allows for short scan times ( 2–3 min ) , high precision and accuracy , and the ability to measure resting , non-anesthetized animals ( Guglielmo et al . , 2011 ) . We used an Echo-MRI ( Echo Medical Systems , Houston , TX ) with an A10 antenna for measuring birds < 10 g . We calibrated the QMR machine with 1 . 5 g canola oil and 10 g water standards and scanned these standards daily to check the calibration; scans were set at three accumulations . On focal nights , birds were scanned three to five times in the evening and the morning; the means of the values from these scans were calculated . QMR , paired with respirometry allows us to non-invasively and accurately estimate the level of endogenous energy stores throughout the night , and specifically at the time of torpor initiation . This study uses respirometry to calculate rates of energy expenditure . Oxygen consumption and carbon dioxide production rates overnight were obtained via push-flow respirometry using an FC-1B oxygen analyzer , a CA-2A carbon dioxide analyzer ( Sable Systems International , Las Vegas , NV ) . Air was first passed through a dew point generator set at 10–15°C and then was flowed into the chambers through Bev-a-line tubing at a rate of 150 mL/min . The excurrent airstream was subsampled at 50 mL/min . Subsampled air first passed through a water vapor meter , which measured water vapor pressure ( kPa ) ( RH-300; Sable Systems International ) . The air then passed through a column containing indicating Drierite ( W . A . Hammond DRIERITE , Xenia , OH ) , the carbon dioxide gas analyzer , and the oxygen analyzer . Analogue voltage outputs from the thermoresistor , oxygen , and carbon dioxide analyzers , flow meter , water vapor pressure , and in-line barometric pressure sensors were recorded at 1 s intervals over the duration of the trial ( 9–12 hr ) using EXPEDATA software ( v . 1 . 9 . 27; Sable Systems International ) and were recorded on a laptop computer . Raw data were corrected to standard temperature and pressure , and rates of oxygen consumption ( V´o2 ) and carbon dioxide production ( V´co2 ) were calculated in Expedata using standard equations ( 12; equations 10 . 6 , 10 . 7 , respectively ) . The rate of oxygen consumption ( V´o2 ) , the respiratory exchange ratios ( RER = V´o2 / V´co2 , indicates primary metabolic fuel type ) , and the oxyjoule equivalent were used to calculate the rate of energy expenditure ( kJ⋅min–1 ) ( Erate= ( 16 + 5 . 164*RER ) * V´o2 ) ( Lighton , 2008 ) . Where RER was extraneously below 0 . 71 or above 1 . 0 it was bound at these limits to satisfy the assumptions of this equation . Total nighttime and metabolic state-specific energy expenditures were calculated by integrating the metabolic rate ( Erate ) over time . In each overnight experiment , birds were food-deprived for approximately 2 hr ( 1 . 66 ± 0 . 44 hr ) prior to lights-off to ensure that crop stores were emptied ( for accurate body composition measurements ) and that the only available sources of energy were endogenous fat and lean mass . Body composition was measured using QMR before the birds were placed in respirometry chambers ( 10 cm W × 10 cm H × 20 cm D ) at 20°C and the lights were turned off . Torpid birds could not be scanned because disturbing them would cause them to unnaturally arouse and because their cooler body temperatures could decrease scan precision and consistency ( Guglielmo et al . , 2011 ) . Immediately following lights-on in the morning , body composition was measured . Air temperature was measured directly outside the chamber via a thermoresistor; although air temperature inside the chamber was not recorded , experiments show the inside and outside air temperatures were within 1°C . Between June and September 2018 and 2019 , we recorded repeated overnight measurements on 15 adult male birds and 1 male juvenile , for a total of 158 bird-nights . The birds began exhibiting increased body masses ( indicative of migratory condition ) at different times in the late summer ( mean: August 28; range: August 12–September 8 ) . In order to identify periods of distinct rates of change in body mass , we analyzed the rate of change in morning body mass across the study period ( Supplementary file 3 ) . We first smoothed the morning body mass trace with a smoothing parameter ( 0 . 35 ) that we identified through an iterative process . We calculated the first derivative of this smoothed body mass trace , and defined periods based on bird-specific ‘cut-off’ slopes that we calculated as 75% of the maximum slope ( g⋅day–1 ) . We defined ‘breeding’ as periods where the rate of change was less than the cut-off slope , ‘fattening’ as periods where the rate of change was greater than the cut-off slope , and ‘migration’ as periods where the rate of change was less than the cut-off slope and after the start of the fattening period . When this analysis yielded periods that clearly disagreed with visual inspection of the curve , we slightly adjusted the bounds of the fattening period to fit a more realistic pattern . Three birds that did not fatten were categorized as ‘non-fatteners’ and were excluded from the statistical models that regard seasonal changes in torpor use . We calculated torpor propensity as the percentage of nights the birds entered torpor of the total number of observation nights for each bird within each period . In order to calculate the energetic characteristics of torpor , such as the fat content at the time of torpor entry and duration , the temporal characteristics of torpor must be clearly defined . In much of the avian and mammalian torpor literature , torpor entry is defined by phrases such as when the metabolic rate ‘abruptly declines’ , or by criteria such as a threshold value of a set proportion of the average normothermic resting values of body temperature or metabolic rate ( Ruf and Geiser , 2015; Shankar et al . , 2020; Wolf et al . , 2020 ) . However , these various and often vague definitions are problematic for repeatability and our understanding of energy metabolism at specific stages of torpor . In order to identify accurate and repeatable periods of consistent metabolic states , we analyzed the rate of change in V´o2 across the night ( Appendix 1—figure 6 ) . We first linearly interpolated V´o2 and smoothed this trace using a smoothing parameter ( 0 . 6 ) that we identified through an iterative process . We calculated separate smoothed traces for the time before ( containing entry ) and after ( containing arousal ) the end of torpor/start of arousal . To determine this intermediate point we calculated the derivative of a smoothed trace of the entire night ( using night-specific smoothing factor ) , and preliminarily identified the approximate end of torpor/start of arousal as the minute the rate of change was greater than an arousal cut-off slope of 0 . 005 V´o2 ⋅min–1 . We then calculated the first derivative of each of the entry and arousal smoothed V´o2 traces , and defined states based on entry and arousal cut-off slopes . We defined ‘entry’ as periods where the rate of change was greater than two standard deviations of the rate of change during the normothermic period before torpor ( which was identified by analyzing the preliminary smoothed curve where the rate of change was less than –0 . 003 V´o2 ⋅min–1 ) . We defined ‘arousal’ as periods where the rate of change of was greater than four standard deviations higher than the mean rate of change during the normothermic period before torpor; we defined the end of the arousal period as the point when the bird exhibited peak V´o2 . We defined ‘torpor’ as periods where the V´o2 was stable and between entry and arousal . We also annotated points before the start of entry and after the end of arousal as ‘normo-pre’ and ‘normo-post’ , respectively . This process yielded accurate and repeatable metabolic state annotations of each minute . The initial evening and the final morning percent fat content were calculated as fat mass ( g ) /body mass ( g ) . We used the rate of energy expenditure , cumulative energy expenditure , and initial fat mass measurements to estimate instantaneous percent fat content throughout each night . These data allowed for the estimation of the amount of energy reserves , relative to body mass , of each bird at the time of torpor initiation . We calculated this value by subtracting the fat mass equivalent ( 1 gfat/37 kJ ) of cumulative energy expenditure at torpor entry from evening fat mass and dividing the result by morning body mass . We calculated torpor duration ( hr ) as the time between the start of torpor entry until the beginning of arousal ( excluding arousal ) . We calculated overnight mass losses as the fat mass equivalent of the amount of overnight energy expenditure . Lastly , we calculated body mass increases as the change in body mass from the beginning to the end of the fattening period , relative to the smoothed morning body mass trace used to determine the periods . We used mixed effects analyses to evaluate the relationships between various response and predictor variables , while accounting for intra-individual differences ( Bates et al . , 2015 ) . First , we used linear mixed effects analyses to determine how body mass , fat content , lean mass , and food consumption changed within and among ( with respect to date ) each period of distinct mass change ( with day length as a covariate for the food consumption model ) . We evaluated the overall relationships between body mass and fat mass , and body mass and lean mass , using a repeated measures correlation test ( ‘rmcorr’ R package; Bakdash and Marusich , 2017 ) . We used a linear mixed effects model to compare mean torpor propensity among periods , and a logistic mixed effects model to compare the influence of evening fat content on probability of occurrence of torpor within each period . We also used linear mixed effects models to determine how torpor duration , energy expenditure before torpor entry , and overnight fat mass loss changed with respect to evening fat content within and among periods , and with respect to night length within and among periods . We used linear mixed effects models to determine how % fat at the time of torpor entry changed with respect to date and the time of night , with night length as a covariate , within and among periods . We also used linear mixed effects models to determine how overnight mass loss changed with respect to torpor duration within and among periods . Lastly , we used linear models to evaluate the effect of mean torpor duration and mean daily food consumption within the fattening period on the magnitude and duration of the fattening period . We performed all statistical analyses using R Development Core Team , 2020 . To generate linear and logistic mixed effects models , we used the ‘lme4’ package , with Bird ID as a random effect ( Bates et al . , 2015 ) . For linear models we used the ‘lm’ function . For each response variable , we iteratively compared several combinations of relevant fixed effects , and used the AICs to determine the most parsimonious model . The model with the lowest AIC by at least two points was considered the best , and we verified that residuals of this model showed homoscedasticity and normality . We determined p-values using the ‘anova’ function on each model generated using the ‘lmerTest’ package , and made pair-wise comparisons within and among periods using the ‘emmeans’ package . All values are given as estimated marginal means ± standard error , unless otherwise indicated , and significance was taken at α < 0 . 05 . | Torpor is an energy-saving strategy used by warm-blooded animals , including birds and small mammals . Similar to hibernation , although shorter in duration , torpor is a state of minimal activity , low body temperatures and reduced metabolism that helps animals conserve energy in unfavorable conditions . Some animals use torpor to survive times when food is not readily available . Hummingbirds , for example , eat nectar all day long to meet their high energy needs , but must build fat reserves to see them through their overnight fast . If they go to sleep with too little fat , they can descend into torpor to stretch out that limited energy supply and survive until morning . Many hummingbirds migrate to areas with warmer weather , where food remains available , for the winter months . The ruby-throated hummingbird ( Archilochus colubris ) , for example , travels over 5 , 000 kilometers in its fall migration . Like most long-distance migrants , ruby-throated hummingbirds increase their fat stores before departing , using these stores to fuel their journey . It is thought that this bird may use torpor as a way to accelerate fat build up before its annual migration . However , it remained unclear whether hummingbirds switched from using torpor strictly in energy emergencies , to using it as strategy to prepare for migration . To shed light on this question , Eberts , Guglielmo and Welch investigated when , why and how hummingbirds save energy using torpor during the summer , and whether there are seasonal shifts in their use of torpor coinciding with migration . Eberts , Guglielmo and Welch hypothesized that a bird would initiate daily torpor if its energy stores fall below a critical level during the night , but that they may abandon this threshold ( triggering torpor at higher fat levels ) in late summer as a way to spare energy and gain fat before their annual migration . To test their hypotheses , Eberts , Guglielmo and Welch tracked body composition , food intake , energy expenditure and torpor use throughout summer in a group of captive ruby-throated hummingbirds . In the middle of the summer , the birds entered torpor and remained torpid for longer when they went to sleep with low fat stores . In late summer , however , the same birds were more likely to enter torpor at consistent times and when they had higher fat stores . Eberts , Guglielmo and Welch also observed that the more time birds spent in torpor , the more fat they gained . This suggests that in late summer , hummingbirds switch from using torpor as a survival strategy to using it to maximize energy savings before migration . These results clearly define the physiological rules governing torpor use in hummingbirds . They also support the long-standing assumption that torpor helps migratory species save energy and accumulate fat stores before long-haul flights . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology",
"short",
"report",
"evolutionary",
"biology"
] | 2021 | Reversal of the adipostat control of torpor during migration in hummingbirds |
Hox genes are essential regulators of embryonic development . Their step-wise transcriptional activation follows their genomic topology and the various states of activation are subsequently memorized into domains of progressively overlapping gene products . We have analyzed the 3D chromatin organization of Hox clusters during their early activation in vivo , using high-resolution circular chromosome conformation capture . Initially , Hox clusters are organized as single chromatin compartments containing all genes and bivalent chromatin marks . Transcriptional activation is associated with a dynamic bi-modal 3D organization , whereby the genes switch autonomously from an inactive to an active compartment . These local 3D dynamics occur within a framework of constitutive interactions within the surrounding Topological Associated Domains , indicating that this regulation process is mostly cluster intrinsic . The step-wise progression in time is fixed at various body levels and thus can account for the chromatin architectures previously described at a later stage for different anterior to posterior levels .
Mammalian Hox genes encode proteins that are essential for patterning along the rostral-to-caudal body axis of the developing embryo ( Duboule and Morata , 1994; Krumlauf , 1994 ) . Mouse and human Hox genes are organized in four genomic clusters ( HoxA to HoxD ) , where the relative position of the genes strongly impacts upon their patterns of expression . This structure-function relationship was initially described in Drosophila ( Lewis , 1978 ) and further extended to vertebrates ( Gaunt et al . , 1988; Duboule and Dolle , 1989; Graham et al . , 1989 ) , where an additional correspondence exists between gene position and the timing of transcriptional activation ( ‘temporal colinearity’ , Izpisua-Belmonte et al . , 1991; Deschamps and van Nes , 2005 ) . In murine embryos , transcription of Hox genes can be divided in several phases and is first detected at around embryonic day 7 ( E7 ) at the most posterior aspect of the primitive streak region ( Deschamps and Wijgerde , 1993; Forlani et al . , 2003 ) . Over time , Hox genes are sequentially activated following their chromosomal order and transcripts encoded by the last Hox group 13 genes can be detected at around E9 , that is two days after the onset of activation ( Deschamps et al . , 1999; Kmita and Duboule , 2003; Deschamps and van Nes , 2005 ) . This transcriptional progression ( the ‘Hox clock’ , Duboule , 1994 ) thus extends over several days . In the pre-somitic mesoderm ( PSM ) , this sequential activation needs to be coordinated with the time-sequenced production of body segments ( the ‘segmentation clock’ , Pourquie , 2003 ) , such that newly produced somites acquire distinct genetic identifiers ( Dubrulle et al . , 2001; Zakany et al . , 2001 ) . Next , the various states of Hox gene activity are fine-tuned and memorized , ultimately leading to domains along the rostral to caudal axis where partially overlapping sets of HOX products can be observed ( ‘spatial colinearity’ ) . As a result , genes located at 3′positions ( e . g . , groups 3 , 4 ) are transcribed almost along the entire embryonic axis , including the lateral plate mesoderm , paraxial mesoderm and neural tube , whereas the 5′-located group 10 or 11 are active in the posterior trunk and group 13 in the tip of the tail bud only ( Deschamps et al . , 1999; Kmita and Duboule , 2003; Deschamps and van Nes , 2005 ) . While both temporal and spatial colinear processes likely reflect one and the same organizational principle , they are nevertheless implemented with distinctive features . Spatial colinearity could be recapitulated by several single-gene transgenes ( e . g . , [Puschel et al . , 1991; Whiting et al . , 1991] ) , yet not in all instances ( Tschopp et al . , 2012 ) . Indeed , a systematic analysis of modified HoxD clusters in vivo revealed that , at a late stage , the sustained transcription of these genes at the correct body level primarily relies upon local regulatory elements ( Tschopp et al . , 2009 ) , which are present in transgenic constructs . In contrast , the precise timing of Hoxd gene activation depends on the integrity of the full cluster , a genomic situation observed thus far in all animals developing following a temporal rostral to caudal progressive strategy during their early development ( Duboule , 1994 ) . The genomic clustering of Hox genes is thus considered as an essential feature for temporal colinearity to properly process , whereas it may not be as important for the correct distribution of HOX products along the AP-axis , at least in the late phase of spatial colinearity ( Duboule , 2007; Tschopp et al . , 2009; Noordermeer and Duboule , 2013 ) . Even though the mechanisms underlying temporal and spatial colinearities are becoming increasingly understood , many aspects of how genomic topology is translated into sequential transcriptional activation remain to be clarified . In vertebrates , two conceptual frameworks have been proposed to account for this process , the first relying on bio-molecular mechanisms ( e . g . , Duboule , 1994 ) and the second involving biophysical forces ( Papageorgiou , 2001 ) . In embryonic stem ( ES ) cells , that is cells that reflect best the state of Hox genes before their activation , Hox clusters are decorated by both repressive ( H3K27me3 ) and activating ( H3K4me3 ) marks ( Bernstein et al . , 2006; Schuettengruber et al . , 2007; Soshnikova and Duboule , 2009; Noordermeer and Duboule , 2013 ) . Subsequently , cells that activate these genes in a time sequence resolve this bivalent chromatin state and show two opposing distributions of histone marks over the HoxD cluster: transcribed genes carry large domains of H3K4me3 marks , whereas inactive genes are covered by H3K27me3 marks only ( Soshnikova and Duboule , 2009 ) . The same dichotomy in chromatin marks over Hox gene clusters was observed in various parts of the E10 . 5 embryonic trunk , in parallel with the spatial colinear distribution of transcripts ( Noordermeer et al . , 2011 ) . The analysis of the 3D chromatin organization at this stage revealed a bi-modal compartmentalization , whereby active genes labeled by H3K4me3 are clustered together and physically separated from the inactive genes , labeled by H3K27me3 that are also found in a defined spatial structure ( Noordermeer et al . , 2011 ) . These 3D compartments , whose sizes correlate with the number of active vs inactive genes , may reinforce the proper maintenance of long-term transcriptional states at various AP levels by isolating Hox clusters from their surrounding chromatin and reducing interference between the active and inactive chromatin domains . Such distinct bimodal 3D organizations , associated with transcriptional regulation at Hox clusters , have been observed in other instances , either in the embryo ( Montavon et al . , 2011; Andrey et al . , 2013 ) or in mouse and human cultured cells ( Fraser et al . , 2009; Ferraiuolo et al . , 2010; Wang et al . , 2011; Rousseau et al . , 2014 ) . However , in-embryo conformation studies were reported so far only in the context of spatial colinearity , that is by comparing samples from different body levels at the same developmental stage . Consequently , a potential association between these bimodal chromatin structures and the progressive activation of transcription along the Hox gene clusters , rather than its maintenance , remained to be assessed . In this study , we describe the 3D organization of Hox gene clusters at high resolution during the implementation of temporal colinearity in the PSM and show that their stepwise activation occurs in parallel with their physical transition from a negative to a positive domain . We also show that this process is accompanied by series of long-range contacts with the flanking gene deserts , even though these contacts remain largely invariable throughout temporal colinearity , unlike what was observed during limb development ( Andrey et al . , 2013 ) . We discuss whether this stepwise transition of genes from one domain to the other may guide temporal colinearity or , in contrast , is a mere consequence of a sequential transcriptional activation .
In order to monitor the 3D organization of Hox clusters during their sequential activation , we considered ES cells as a starting point of our time curve . These cells indeed represent early embryonic cells related to blastocyst inner cell mass cells , that is when Hox genes are all supposedly silent . We hypothesized that these cells reflect the ground state 3D architecture of the Hox clusters , which we assessed by using high-resolution 4C-seq ( Circular Chromosome Conformation Capture; Noordermeer et al . , 2011; van de Werken et al . , 2012 ) and a variety of viewpoints within all four Hox clusters . These various baits generated similar interaction profiles with the majority of sequence reads covering the gene clusters and extending within several kilobases ( kb ) on either sides , as illustrated by Hoxd13 , Hoxd9 and Hoxd4 ( Figure 1A , Figure 1—figure supplements 1–4 ) . Additional contacts were scored in the flanking gene deserts , though with significantly lower frequencies ( see section ‘Temporal colinearity within a constitutive framework of long-range interactions’ ) . The overall size of the strong interaction profiles exactly matched the distribution of bivalent chromatin marks in these cells , with a moderate level of H3K27me3 covering the cluster and rather weak H3K4me3 peaks labeling promoters ( Figure 1A , Figure 1—figure supplements 1–4; Bernstein et al . , 2006; Soshnikova and Duboule , 2009 ) . Therefore , prior to their activation , Hox clusters are already organized into 3D chromatin compartments that physically separate the chromatin decorated by bivalent marks from the genomic surroundings , even though some contacts are established at a larger scale , outside the gene cluster itself ( see section ‘Temporal colinearity within a constitutive framework of long-range interactions’ ) . 10 . 7554/eLife . 02557 . 003Figure 1 . Hox clusters in ES cells are organized as 3D compartments . ( A ) Quantitative local 4C-seq signal for the Hoxd13 ( top ) , Hoxd9 ( middle ) and Hoxd4 ( bottom ) viewpoints in ES cells . Below , the H3K27me3 and H3K4me3 ChIP-seq signals are aligned . The boundaries of the inactive Hox gene compartments are indicated by dashed lines . The locations of Hox genes ( red ) and of other transcripts ( black ) are shown below . ( B ) Quantitative local 4C-seq signal for the Hoxd13 ( left ) and Hoxb9 ( right ) viewpoints , either in ES ( orange ) or in E10 . 5 forebrain ( green ) cells . Below , the H3K27me3 and H3K4me3 ChIP-seq signals are aligned . Ratios between the 4C-seq signals in ES cells and E10 . 5 forebrain are indicated between the profiles , with signal in one color indicating that the viewpoint interacts more with this fragment in the sample represented by this color . Regions of increased interactions outside the 3D Hox gene compartments in ES cells are highlighted in orange . ( C ) Distribution of ratios inside and outside the inactive 3D Hox gene compartments in both ES and E10 . 5 forebrain cells . Fragments are classified either as positive in ES cells ( orange ) , or positive in E10 . 5 forebrain cells ( green ) . The number of fragments is indicated below . Significance between distribution inside and outside 3D compartments was calculated using a G-test of independence . ( D ) Model of 3D compartmentalization of the inactive HoxD and HoxB clusters in both ES cells and E10 . 5 forebrain cells . The increased contacts with the surrounding chromatin in ES cells are illustrated by invading grey lines . DOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 00310 . 7554/eLife . 02557 . 004Figure 1—figure supplement 1 . 3D compartments in the HoxD cluster are less discrete in ES cells than in embryonic brain cells . Comparison of quantitative local 4C-seq signals for replicate samples with the indicated viewpoints , either in ES ( orange ) or E10 . 5 forebrain ( green ) cells . All six comparisons between two replicates in each condition are given . Viewpoints are indicated with arrowheads and regions excluded around the viewpoints are indicated with light grey boxes . Below , the H3K27me3 and H3K4me3 ChIP-seq signals are aligned . The ratios between 4C-seq signals are indicated between the corresponding profiles . The locations of Hoxd genes ( red ) and other transcripts ( black ) are shown below . Only regions covered by the random 4C-seq libraries are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 00410 . 7554/eLife . 02557 . 005Figure 1—figure supplement 2 . 3D compartments in the HoxD and HoxB cluster are less discrete in ES cells than in embryonic brain cells . Comparison of quantitative local 4C-seq signals for replicate samples with the indicated viewpoints , either in ES ( orange ) or E10 . 5 forebrain ( green ) cells . All six comparisons between two replicates in each condition are given . Viewpoints are indicated with arrowheads and regions excluded around the viewpoints are indicated with light grey boxes . Below , the H3K27me3 and H3K4me3 ChIP-seq signals are aligned . The ratios between 4C-seq signals are indicated between the corresponding profiles . The locations of Hoxd and Hoxb genes ( red ) and other transcripts ( black ) are shown below . Only regions covered by the random 4C-seq libraries are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 00510 . 7554/eLife . 02557 . 006Figure 1—figure supplement 3 . 3D compartments in the HoxB cluster are less discrete in ES cells than in embryonic brain cells . Comparison of quantitative local 4C-seq signals for replicate samples with the indicated viewpoints , either in ES ( orange ) or E10 . 5 forebrain ( green ) cells . All six comparisons between two replicates in each condition are given . Viewpoints are indicated with arrowheads and regions excluded around the viewpoints are indicated with light grey boxes . Below , the H3K27me3 and H3K4me3 ChIP-seq signals are aligned . The ratios between 4C-seq signals are indicated between the corresponding profiles . The locations of Hoxb genes ( red ) and other transcripts ( black ) are shown below . Only regions covered by the random 4C-seq libraries are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 00610 . 7554/eLife . 02557 . 007Figure 1—figure supplement 4 . 3D compartments in the HoxC and HoxA cluster are less discrete in ES cells than in embryonic brain cells . Comparison of quantitative local 4C-seq signals with the indicated viewpoints in ES ( orange ) or E10 . 5 forebrain ( green ) cells . Viewpoints are indicated with arrowheads and regions excluded around the viewpoints are indicated with light grey boxes . Below , the H3K27me3 and H3K4me3 ChIP-seq signals are aligned . The ratios between 4C-seq signals are indicated between the corresponding profiles . The locations of Hoxc and Hoxa genes ( red ) and other transcripts ( black ) are shown below . Only regions covered by the random 4C-seq libraries are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 00710 . 7554/eLife . 02557 . 008Figure 1—figure supplement 5 . Distribution of ratios inside and outside the inactive 3D Hox gene compartments in both ES and E10 . 5 forebrain cells . The comparison between replicate samples one ( as used in the main text ) is indicated on the left , the comparison between combined replicate samples is indicated at the center left , the comparison between ES cell replicates is indicated at the center right and the comparison between E10 . 5 forebrain replicates is indicated on the right . Fragments are classified as positive either in ES ( orange ) or in E10 . 5 forebrain ( green ) cells within the region covered by the random 4C-seq libraries . The number of fragments is indicated below . Significance between distribution inside and outside the 3D compartments was calculated using a G-test of independence . DOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 00810 . 7554/eLife . 02557 . 009Figure 1—figure supplement 6 . Different discretion of 3D compartments is not due to overall increased background signal . ( A ) Distribution of 4C-seq signal on chromosome 2 from viewpoints in the HoxD cluster . On the right , a schematic representation of chromosome 2 is given , with color codes for the three categories that have been quantified in ES cells and E10 . 5 forebrain indicated below . Comparison of distributions between ES cells and E10 . 5 forebrain show that TAD signal in ES cells are considerably increased , but that more distal signal is reduced . Elevated signal in the TADs in ES cells is therefore not a representation of generally increased background signal . ( B ) Distribution of 4C-seq signal on chromosome 11 from viewpoints in the HoxB cluster . Similar effects are observed as for viewpoints in the HoxD cluster . DOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 00910 . 7554/eLife . 02557 . 010Figure 1—figure supplement 7 . Increased Hox background transcription in ES cells . ( A ) Expression levels of Hox genes and of four housekeeping genes in both ES and E10 . 5 forebrain cells , as determined by RNA-seq . The large majority of Hox genes show low level activity in ES cells , whereas only few , very low , transcribed Hox genes are identified in E10 . 5 forebrain . In contrast , the expression levels of selected housekeeping genes are within a similar range ( maximum threefold difference ) . ( B ) Overall gene expression patterns in ES and E10 . 5 forebrain cells are not significantly different . Box plots showing the overall distribution of RNA-seq signals per gene ( RPKM ) , with colored boxes indicating the 25 to 75% range and whiskers indicating the 10 to 90% range . Differences between distributions were scored using a two-sided Welch two samples t test . ( C ) Quantitation of selected spliced Hox gene transcripts in ES cell and E10 . 5 forebrain samples as determined by RT-qPCR , with amounts in each sample relative to the Tubb2c gene . Below each sample , the specific product of a representative qPCR reaction is displayed . Color-coded dots are used to classify the different outcomes ( see legend ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 010 These global 3D domains including the Hox clusters and their immediate flanking DNAs resemble the chromatin architecture found in embryonic forebrain cells , where the silent Hox clusters are covered by high levels of H3K27me3 only ( Noordermeer et al . , 2011 ) . A more quantitative comparison in 3D architectures between ES cells and E10 . 5 forebrain cells nevertheless indicated that in ES cells , Hox genes interacted more with the outside chromatin , relative to their interactions within the cluster , as compared to forebrain cells ( Figure 1B , C , Figure 1—figure supplements 1–5 ) . Therefore , despite the fact that the clusters are presumably inactive in both situations , the presence of bivalent marks in ES cells coincided with a 3D domain that has elevated relative levels of interactions with the directly surrounding regions , when compared to its counterpart in brain cells ( Figure 1D , left and Figure 1—figure supplement 6A ) . This difference is more pronounced at the HoxB cluster ( Figure 1B , right ) . In embryonic forebrain cells , HoxB forms a single 3D compartment , excluding the 80 kb large repeat-rich intergenic region located between Hoxb13 and Hoxb9 , which loops out ( Noordermeer et al . , 2011 ) . In contrast , both the Hoxb13 and Hoxb9 viewpoints revealed local 3D compartments in ES cells , matching again the extent of bivalent histone marks , yet these two compartments remained separated and did not fuse . Rather , they displayed increased interactions with the nearby chromatin , as if decreased internal interactions would increase contacts outside the cluster ( Figure 1B , C , Figure 1—figure supplements 2 , 3 and 5 ) . In ES cells , the HoxB cluster is thus organized in two 3D compartments , which have more interactions with their genomic surroundings than in forebrain cells ( Figure 1D , right and Figure 1—figure supplement 6B ) . To have a possibly more unbiased view on how 3D compartments and the presence of H3K27me3 and H3K4me3 modifications relate to each other , in both ES and brain cells , we devised an approach to correlate 4C-seq signals with either H3K27me3 or H3K4me3 ChIP-seq signal ( Table 1; ‘Materials and methods’ ) . In both cell types , H3K27me3 marks strongly correlated with the 3D organization , suggesting a direct link between these two readouts . A considerably lower correlation was scored for HoxB , perhaps related to the absence of clustering of the two H3K27me3 marked sub-domains . In contrast , no particular correlation was observed between the 3D organization and the presence of H3K4me3 marks , in the bivalent state ( Table 1 ) , suggesting that H3K4me3 marks and/or the associated factors do not noticeably contribute to the formation of 3D compartments in ES cells . 10 . 7554/eLife . 02557 . 011Table 1 . Spearman's rank correlation coefficient between pairs of 4C-seq and ChIP-seq samplesDOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 011ChIP-seq4C-seqInputH3K27me3H3K4me3Hoxd13 ES cells 1−0 . 140 . 520 . 24Hoxd13 ES cells 2−0 . 070 . 400 . 22Hoxd13 E8 . 5 PSM−0 . 030 . 580 . 13Hoxd13 E10 . 5 Forebrain 1−0 . 120 . 670 . 26Hoxd13 E10 . 5 Forebrain 2−0 . 090 . 690 . 25Hoxd13 E10 . 5 Anterior trunk−0 . 070 . 800 . 30Hoxd9 ES cells 1−0 . 080 . 630 . 28Hoxd9 ES cells 2−0 . 130 . 590 . 26Hoxd9 E8 . 5 PSM−0 . 050 . 310 . 29Hoxd9 E10 . 5 Forebrain 1−0 . 080 . 660 . 26Hoxd9 E10 . 5 Forebrain 2−0 . 120 . 610 . 28Hoxd9 E10 . 5 Anterior trunk−0 . 150 . 670 . 47Hoxd4 ES cells 10 . 010 . 480 . 11Hoxd4 ES cells 2−0 . 070 . 500 . 29Hoxd4 E8 . 5 PSM−0 . 040 . 040 . 38Hoxd4 E10 . 5 Forebrain 1−0 . 050 . 590 . 24Hoxd4 E10 . 5 Forebrain 2−0 . 040 . 580 . 27Hoxd4 E10 . 5 Anterior trunk−0 . 070 . 160 . 59Hoxc13 ES cells 1−0 . 030 . 390 . 20Hoxc13 E8 . 5 PSM−0 . 030 . 55−0 . 03Hoxc13 E10 . 5 Forebrain 1−0 . 070 . 570 . 18Hoxc13 E10 . 5 Anterior trunk−0 . 050 . 820 . 00Hoxb13 ES cells 1−0 . 050 . 120 . 02Hoxb13 ES cells 2−0 . 08−0 . 010 . 15Hoxb13 E8 . 5 PSM0 . 100 . 29−0 . 17Hoxb13 E10 . 5 Forebrain 10 . 020 . 480 . 09Hoxb13 E10 . 5 Forebrain 20 . 080 . 440 . 10Hoxb13 E10 . 5 Anterior trunk−0 . 030 . 490 . 26Hoxb9 ES cells 10 . 010 . 470 . 09Hoxb9 ES cells 20 . 030 . 340 . 04Hoxb9 E8 . 5 PSM−0 . 04−0 . 300 . 57Hoxb9 E10 . 5 Forebrain 10 . 020 . 630 . 19Hoxb9 E10 . 5 Forebrain 20 . 030 . 590 . 16Hoxb9 E10 . 5 Anterior trunk0 . 06−0 . 010 . 69Hoxa13 ES cells 10 . 100 . 520 . 14Hoxa13 E8 . 5 PSM0 . 100 . 580 . 12Hoxa13 E10 . 5 Forebrain 10 . 070 . 600 . 22Hoxa13 E10 . 5 Anterior trunk0 . 060 . 730 . 20Spearman's rank correlation coefficient between pairs of 4C-seq and ChIP-seq samples in different samples ( see section ‘Material and methods’ for methodology ) . For each 4C-seq sample , the highest correlating ChIP-seq sample is highlighted in bold . It was previously reported that genes covered by bivalent marks in ES cells can be transcribed at low levels , resulting in detectable spliced transcripts ( Stock et al . , 2007 ) . We assessed whether the observed difference in the strength and homogeneity of the interaction profiles between ES cells and embryonic brain cells was associated with distinct levels of background transcription . In ES cells , RNA-seq detected transcription for most Hox genes , though generally at very low level ( Figure 1—figure supplement 7A , B ) . RT-qPCR of a subset of transcripts confirmed that some of these low-level transcripts ( particularly the Hoxd13 and Hoxb13 transcripts ) constitute genuine processed transcripts ( Figure 1—figure supplement 7C ) . In contrast , transcription of Hox genes in E10 . 5 forebrain cells was rarely detected , and no reliable spliced transcripts were detected ( Figure 1—figure supplement 7 ) . Therefore , when Hox genes are decorated by bivalent chromatin marks , they appear more permissive for background transcription as compared to other cell types where they are covered by H3K27me3 marks only , likely illustrating the increased resistance to transcription of the latter condition . In this context , posterior Hox genes seems to be more prone to background transcriptional activation in ES cells than more anterior Hox genes , in contrast to their subsequent dynamics of activation in future embryonic tissues where anterior genes come first . This may reflect the presence of strong enhancers in their vicinity ( Montavon et al . , 2011 ) . Next , we assessed whether this large 3D domain observed in ES cells is modified when Hox genes become activated in the pre-somitic mesoderm ( PSM ) or instead , whether the previously observed positive and negative compartments are only established at a later stage to fix and memorize particular combinations of Hox gene activities determined at earlier stages and at various body levels . For this purpose , we compared the 4C-seq profiles from ES cells with those obtained from early embryonic E8 . 5 PSM cells dissected out at Theiler stage 13 , posterior from the approximate level of the 12th to 14th forming somite ( Figure 2A , scheme; Figure 2—figure supplements 1 and 2 ) . In the most caudal aspect of this latter cellular territory , transcriptional activation had progressed up to the Hoxd9 gene , whereas the Hoxd10 to Hoxd13 loci remained silent ( Soshnikova and Duboule , 2009 ) . This cellular population was thus composed of a mixture of cells positive and negative for Hoxd9 expression , whereas all cells were negative for Hoxd13 . Conversely , the majority of cells expressed Hoxd4 . The inactive Hoxd13 viewpoint interacted mostly with the domain labeled by H3K27me3 , at the centromeric side of the cluster ( Figure 2A , bottom left ) . In contrast , the active Hoxd4 gene essentially interacted with the other transcribed genes on the telomeric side of the cluster , labeled by H3K4me3 marks ( Figure 2A , bottom right ) . The same bi-modal 3D organization was observed for all Hox gene clusters ( Figure 2—figure supplements 1 and 2 ) . 10 . 7554/eLife . 02557 . 012Figure 2 . Bi-modal 3D organization of Hox clusters upon sequential activation . ( A ) Quantitative local 4C-seq signal for the Hoxd13 ( left , centromeric side of HoxD cluster ) and Hoxd4 ( right , telomeric side of HoxD cluster ) viewpoints , either in ES ( orange ) , or E8 . 5 pre-somitic mesoderm ( cyan ) cells . Below , the H3K27me3 and H3K4me3 ChIP-seq profiles are aligned . The colinear expression status of Hoxd genes in each sample is schematized below the ChIP-seq profiles , with active genes in blue and inactive genes in red . Ratios between the 4C-seq signals in different samples are indicated between the profiles . The boundaries separating active from inactive Hox gene compartments are indicated by dashed lines . The locations of Hoxd genes ( red ) and other transcripts ( black ) are shown below . The samples are shown on the left and cartoons summarizing the genome organizations are indicated on the right . ( B ) Spearman's rank correlation coefficient between pairs of 4C-seq and ChIP-seq samples , in early and late embryonic material . DOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 01210 . 7554/eLife . 02557 . 013Figure 2—figure supplement 1 . Upon sequential activation , the HoxD cluster adopts a bi-modal 3D organization . Quantitative local 4C-seq signals for the indicated Hoxd gene viewpoints . Profiles are displayed for ES ( orange ) and E8 . 5 pre-somitic mesoderm ( cyan ) cells . The viewpoints are indicated with arrowheads and excluded regions around the viewpoints are indicated with light grey boxes . Below , the H3K27me3 and H3K4me3 ChIP-seq signals are aligned . The ratios between 4C-seq signals are indicated between the respective profiles , with the signal in one particular color indicating that the viewpoint interacts more with fragments in the sample of the same color . The locations of Hoxd genes ( red ) and other transcripts ( black ) are shown below . Only the region covered by the random 4C-seq library is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 01310 . 7554/eLife . 02557 . 014Figure 2—figure supplement 2 . Upon sequential activation , other Hox clusters adopt a bi-modal 3D organization as well . Quantitative local 4C-seq signals for the indicated Hox gene viewpoints in other Hox clusters . Profiles are displayed for ES ( orange ) and E8 . 5 pre-somitic mesoderm ( cyan ) cells . The viewpoints are indicated with arrowheads and excluded regions around the viewpoints are indicated with light grey boxes . Below , the H3K27me3 and H3K4me3 ChIP-seq signals are aligned . The ratios between 4C-seq signals are indicated between the respective profiles , with the signal in one particular color indicating that the viewpoint interacts more with fragments in the sample of the same color . The locations of both Hox genes ( red ) and other transcripts ( black ) are shown below . Only regions covered by the random 4C-seq libraries are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 014 We correlated 4C-seq signals with ChIP-seq data in both early and late embryonic samples ( Figure 2B; Table 1 ) and Hoxd13 always strongly correlated with H3K27me3 histone marks . In contrast , in both the E8 . 5 tail bud and the E10 . 5 anterior trunk , the interactions of the active Hoxd4 gene correlated primarily with H3K4me3 marks . The contacts established by Hoxd9 correlated both with H3K27me3 and H3K4me3 , either in E8 . 5 tailbuds , or in E10 . 5 anterior trunk , likely due to the presence of both expressing and non-expressing cells . At both stages where Hox clusters are partially active , the patterns of 3D compartmentalization and histone marks thus strongly correlated . Therefore , step wise Hox gene transcriptional activation , at least for the Hoxd9 to Hoxd13 genes , is accompanied by a conformational separation between active and inactive domains , which pre-figures their 3D organization at later developmental stages along the AP-axis ( Figure 2; Noordermeer et al . , 2011 ) . Temporal colinearity was initially defined as the sequential activation of Hox genes according to their positions in the clusters ( Izpisua-Belmonte et al . , 1991; Duboule , 1994 ) . However , studies on the global transcriptional organization of the HoxD cluster , at least in the developing spinal cord , revealed two large and regulatory-independent modules , which separate ‘posterior’ genes ( the AbdB-related Hoxd9 to Hoxd13 genes ) from the rest of the gene cluster ( Tschopp et al . , 2012 ) . Also , in different developmental contexts such as the limbs and the cecum , groups of neighboring Hoxd genes are activated as single regulatory blocks ( Montavon et al . , 2011; Andrey et al . , 2013; Delpretti et al . , 2013 ) . We thus assessed whether the transition in chromatin domains also occurred stepwise or , alternatively , if large domains consisting of multiple genes were initially organized in space , followed by sequential gene transcription within these domains . We first compared the 3D cluster architecture over the course of embryonic development , between the E8 . 5 PSM and dissected E9 . 5 tail buds ( Figure 3 ) . This latter sample was obtained after cutting off the most caudal part of E9 . 5 embryos ( Theiler stage 15 ) right after the incipient hind limb bud , that is at ca . somite 26–27 level . Accordingly , this sample contained the tail bud proper as well as some tissue localized slightly more rostral . During this 24 hr time interval , the Hoxd10 and Hoxd11 genes become robustly activated in these cells , which are derived from a sub-population of the sample dissected at E8 . 5 . In the E8 . 5 PSM , Hoxd10 and Hoxd11 are still silenced . 10 . 7554/eLife . 02557 . 015Figure 3 . Activated Hoxd genes switch compartments . Quantitative local 4C-seq signals for the Hoxd13 , Hoxd11 Hoxd9 and Hoxd4 viewpoints in either E8 . 5 pre-somitic mesoderm ( cyan ) , E9 . 5 tail bud ( brown ) or E10 . 5 tail bud ( purple ) cells . The colinear expression status of Hoxd genes is schematized below each profile and , on the left , below each cartoon . Ratios between 4C-seq signals in different samples are indicated between the corresponding profiles . The boundaries between active and inactive Hox gene compartments are indicated by dashed lines and regions displaying important changes in interactions , as discussed in the text , are highlighted . Black arrows point towards opposing interacting behaviors due to the heterogeneous activity state of the viewpoint in the sample . The locations of Hoxd genes ( red ) and other transcripts ( black ) are shown below . DOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 015 The re-organization of compartmentalization occurring along with gene activation over this 24 hr period was clearly revealed by comparing the profile obtained when using Hoxd11 as a viewpoint with those obtained with either Hoxd4 or Hoxd13 ( Figure 3 , top ratio between E8 . 5 PSM and E9 . 5 Tail bud ) . When switching from an inactive to an active state , Hoxd11 re-deployed its interactions from the inactive , centromeric compartment ( Figure 3 , top ratio: blue shaded area ) to the active telomeric compartment ( Figure 3 , top ratio: brown shaded area ) . These negative and positive compartments can be identified by the interaction profile of either Hoxd13 ( Figure 3 , top left ) or Hoxd4 ( Figure 3 , top right ) , respectively . Accordingly , Hoxd4 shifted its interactions towards the centromeric ( active ) part of the cluster in both E9 . 5 and E10 . 5 samples ( Figure 3; right ) , to contact Hoxd10 , Hoxd11 and , to some extent , Hoxd12 ( Figure 3 , top ratio: brown shaded area ) . Of note , the dissected E8 . 5 PSM contained a mixture of cells either positive or negative for Hoxd9 transcription , which coincided with this gene showing conspicuous contacts with both extremities of the gene cluster , depending whether it was active ( right ) or inactive ( left ) ( Figure 3 , top , black arrows ) . In contrast , the E9 . 5 dissection contains a more homogenous cell population , strongly expressing this gene . As a consequence , in this latter sample , Hoxd9 contacts more strongly the now expressed Hoxd10 and Hoxd11 genes , whereas the interactions with Hoxd13 or Evx2 are strongly diminished ( Figure 3; compare top with middle panels ) . In the E10 . 5 tail bud , the terminal part of the cluster containing Hoxd12 and Hoxd13 has been activated , as shown by the increased contacts established by Hoxd13 with the telomeric part of the cluster , indicating that the full HoxD array had been processed and that all genes were now ( at least in part ) included into a ‘positive’ compartment ( Figure 3 , bottom ratio: purple shading ) . While considerably increased , contacts of Hoxd13 with the telomeric part of the cluster were however weak ( Figure 3 , bottom , black arrows ) , most likely reflecting the restricted expression of Hoxd13 at this stage , in a small subset of the dissected cells . In the same context , contacts established both by Hoxd4 and Hoxd9 with the centromeric part of the cluster extended towards the end of the cluster along with the developmental stage , such that both reached Hoxd12 in E10 . 5 samples , whereas Hoxd13 was only weakly contacted , corresponding to the results obtained when using Hoxd13 as a bait ( Figure 3; bottom left ) . Therefore , it appears that the colinear time sequence in Hox genes activation is paralleled by a progressive transition in the chromatin structure , with a positive domain gaining in size along with time , at the expense of the negative domain , as best seen by the extension of Hoxd4 contacts . At E8 . 5 , these interactions extended up to Hoxd8-Hoxd9 . In E9 . 5 samples Hoxd10 was clearly contacted , and in E10 . 5 Hoxd11 and Hoxd12 were also involved ( Figure 3 , right column ) . These dynamic topologies suggest a stepwise transition of the genes from the negative to the positive compartment , rather than the switch of large groups of multiple transcription units , following a discrete and global chromatin re-organization . During axial extension , Hox genes are activated in the most posterior aspect of the elongating embryo ( Deschamps and van Nes , 2005 ) . It is thus possible that cells implementing this stepwise transition in chromatin domains can fix and memorize their bimodal distribution once they exit the posterior zone of activation , leading to the colinear Hox conformations observed along the AP-axis ( Noordermeer et al . , 2011 ) . Accordingly , one would expect cellular territories along the developing body axis to maintain the same bimodal combinations as those established at the time of their origin , during early axial extension . We looked at the similarities in bimodal profiles between posterior samples dissected at different times on the one hand , and various samples micro-dissected at different body levels , from E10 . 5 embryos , on the other hand ( Figure 4 ) . 10 . 7554/eLife . 02557 . 016Figure 4 . The bimodal 3D organization of Hox cluster may help memorize states of colinear expression . Quantitative local 4C-seq signals for the Hoxd13 , Hoxd11 Hoxd9 and Hoxd4 viewpoints , in samples taken at various anterior to posterior positions along the developing body axis from E10 . 5 embryos . Anterior trunk ( red ) , lumbo-sacral trunk ( blue ) and tail bud ( purple ) tissues were used and the approximate expression status of Hoxd genes in every sample is schematized below each profile ( as for Figure 3 ) . Ratios between 4C-seq signals in the different samples are indicated between the corresponding profiles . The boundaries between active and inactive Hox gene compartments are indicated by dashed lines and regions displaying important changes in interactions , as discussed in the text , are highlighted . The locations of Hoxd genes ( red ) and other transcripts ( black ) are shown below . On the right , cartoons summarizing the 3D genome organization of the HoxD cluster are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 016 The profile obtained from E8 . 5 PSM ( Figure 3 , top ) , right at the onset of Hoxd9 activation globaly aligned with that observed in the ‘anterior trunk’ sample at E10 . 5 ( Figure 4 , top ) , that is a cellular domain with a posterior boundary positioned approximately at the Hoxd9 anterior limit of expression . In both cases , Hoxd9 clearly contacted the negative domain , as defined by the Hoxd13 contacts ( Figure 4 , top ) , whereas some weak contacts were also scored with the positive domain , as determined by Hoxd4 contacts , indicating that the posterior limit of the dissection was slightly below the Hoxd9 boundary . These contacts were somehow stronger , in proportion , in the E8 . 5 than in the E10 . 5 dissection . At E9 . 5 ( Theiler stage 14 ) , the ‘tail bud’ ( i . e . , from the start of the non-segmented mesoderm ) was dissected from ca . somite 22 to 25 and caudally . At this stage , Hoxd9 is robustly transcribed , whereas Hoxd11 has just started transcriptional activation . The interaction profiles obtained in this tissue were most similar to those obtained when a fragment of E10 . 5 trunk was dissected out that grossly corresponded to the future lumbo-sacral region , at levels 22 to 28 , that is the AP levels supposedly produced in the E9 . 5 tail bud ( Figure 4 , middle ) . At this AP level , Hoxd9 is fully activated and this was reflected by the quasi absence of contact with Hoxd13 whereas , conversely , strong interactions appeared with the active part of the cluster ( Figure 4 , top ratio: red and blue shading ) . This was controlled by using Hoxd4 as bait , since contacts were now clearly scored with Hoxd9 and Hoxd10 and , to a lesser extent , with Hoxd11 ( Figure 4 , top ratio: blue shading ) . In this lumbo-sacral sample , neither Hoxd12 nor Hoxd13 are as yet transcribed , which coincided with the absence of contact between Hoxd13 and the active part of the gene cluster ( Figure 4 , middle , left ) . On the other hand , Hoxd11 expectedly displayed a mixed interaction profile , contacting both the negative and positive domains , likely reflecting the presence of both expressing and non-expressing cells ( Figure 4 , middle ) . In the most caudal piece of the E10 . 5 mouse embryo , interactions between Hoxd12 , Hoxd13 and the positive domain were finally detected , suggesting that the entire cluster falls into a single spatial domain ( Figure 4 , bottom ratio: purple shading ) . Here again , however , though the interactions were significant , they were not particularly strong , suggesting the presence of a mixed cell population . Based on these data , we propose that the bimodal distributions are frozen in those cells leaving the zone of proliferation , at the caudal aspect of the embryo where temporal colinearity is potentially processed . These 3D structures , and hence the Hox transcription programs , will thus be maintained and memorize the various AP levels from which they originate . In different developmental contexts , the transcriptional activity of Hoxd genes coincides with an overall remodeling of long-range chromatin interactions with the flanking gene deserts , which harbor essential enhancer elements active in these developing tissues ( Montavon et al . , 2011; Andrey et al . , 2013; Berlivet et al . , 2013; Delpretti et al . , 2013 ) . Colinear activation of Hoxd genes along the developing trunk is thought to primarily rely on regulatory influences intrinsic to the gene cluster itself ( Spitz et al . , 2001 ) . However , and even though their importance remains unclear , contributions of the flanking regulatory landscapes in this process have been proposed ( Tschopp et al . , 2009; Tschopp and Duboule , 2011b ) . Therefore , we assessed whether or not the reported changes in local interactions are associated with variations in long-range contacts during temporal colinearity , as was observed during limb and intestinal development ( Montavon et al . , 2011; Andrey et al . , 2013; Delpretti et al . , 2013 ) . By using a recently developed analytical methodology ( Woltering et al . , 2014 ) , we found that all interrogated Hoxd genes displayed substantial interactions with the flanking gene deserts ( Figure 5A , Figure 5—figure supplement 1A ) . The quantification of interactions over both the centromeric and telomeric gene deserts revealed a gene-specific interaction preference towards either one or the other desert ( Figure 5B ) , similar to what was previously described in limb bud cells ( Montavon et al . , 2011; Andrey et al . , 2013 ) . However , in marked contrast , the dynamics of these long-range chromatin interactions were moderate , if any , and no clear modification in the contact profiles were detected between the inactive state in ES cells , and the subsequent transcriptional activation ( Figure 5B , C ) . Hierarchical clustering of global patterns of long-range interactions revealed that the Hoxd4 , Hoxd9 and Hoxd11 viewpoints systematically cluster together , whereas the Hoxd13 viewpoint always behaves as outlier ( Figure 5—figure supplement 1B ) . 10 . 7554/eLife . 02557 . 017Figure 5 . Sequential Hoxd gene activation occurs without drastic remodeling of long-range interactions . ( A ) Distribution of long-range contacts in both the centromeric and telomeric gene deserts surrounding the HoxD cluster . Smoothed 4C-seq signals ( 11 fragment window size ) are shown for the Hoxd13 and Hoxd4 gene viewpoints in ES and E9 . 5 tail bud cells . The analyzed genomic interval is the same as in Woltering et al . ( 2014 ) . The location of topological domains ( TADs ) in ES cells are obtained from Dixon et al . ( 2012 ) and indicated on the top with the HoxD cluster and both the centromeric and telomeric gene deserts indicated by arrows . The dashed lines demarcate the domain of high signal over the HoxD cluster , which is excluded from the analysis . ( B ) Summaries of the distributions in long-range signals within the centromeric and telomeric gene deserts surrounding the HoxD cluster , for all Hoxd genes assayed at various stages of their sequential activation . Each Hoxd gene specifically interacts with either the centromeric or the telomeric gene desert and these privileged contacts remain largely invariant during transcriptional activation . ( C ) Cumulative signals over the centromeric and telomeric gene deserts and the HoxD cluster for all Hoxd genes assayed at various stages of their sequential activation . DOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 01710 . 7554/eLife . 02557 . 018Figure 5—figure supplement 1 . Temporal colinearity occurs without dynamic long-range interactions . ( A ) Distribution of long-range contacts in the centromeric and telomeric gene deserts surrounding the HoxD cluster . Smoothed 4C-seq signals ( 11 fragment window size ) for the indicated HoxD viewpoints either in ES ( orange ) , E8 . 5 pre-somitic mesoderm ( cyan ) , E9 . 5 tail bud ( brown ) or E10 . 5 tail bud ( purple ) cells over the same genomic interval as analyzed in Woltering et al . ( 2014 ) . Genomic location of the HoxD cluster and surrounding genes is indicated below . TADs observed in ES cells ( from Dixon et al . 2012 ) are indicated on the top . The positions of both the HoxD cluster and the centromeric and telomeric gene deserts are indicated by arrows . The dashed lines demarcate the domain of high signals over the HoxD cluster , which is excluded from the analysis . ( B ) Hierarchical clustering of global patterns of long-range interactions in the surrounding gene deserts , for Hoxd viewpoints in ES cells and at different stages of sequential Hox gene activation . The Hoxd4 , Hoxd11 and Hoxd13 viewpoints are consistently clustered together , with the Hoxd13 behaving as an outlier . The correlations between samples ( indicated by heatmaps ) were calculated using Spearman's ranking of smoothed 4C-seq signals ( 11 fragment window size ) over the combined genomic intervals as used in Woltering et al . ( 2014 ) , with the HoxD cluster itself excluded . The samples were subsequently clustered ( top ) according to standard hierarchical clustering . ( C ) Hierarchical clustering of global patterns of long-range interactions in the surrounding gene deserts for Hoxd viewpoints in autopod ( digits ) and zeugopod ( limbs ) cells . Data are from Woltering et al . ( 2014 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 01810 . 7554/eLife . 02557 . 019Figure 5—figure supplement 2 . Comparison between HiC and 4C-seq datasets obtained in ES cells . ( A ) Virtual 4C carried out from HiC datasets , using bins covering the indicated Hoxd genes as viewpoints . Bins used as viewpoints are indicated in red . The interactions with bins covering the surrounding centromeric and telomeric TADs are given in light orange . TADs in ES cells ( obtained from Dixon et al . 2012 ) are indicated on the top and the location of the HoxD cluster is indicated below . The dashed lines demarcate the assigned TAD boundaries in ES cells . ( B ) Comparison of the distribution of long-range signals in both the centromeric and telomeric gene deserts , as obtained either by virtual 4C ( light orange , data from Dixon et al . 2012 ) or by 4C-seq ( bright orange , this study ) . Despite large differences in both the size of the viewpoints and the resolution , the distribution is largely similar . The distribution of the HiC bin covering the promoters of the Hoxd13 and Hoxd11 genes behaves as a mix of the two individual 4C-seq viewpoints . ( C ) Coordinates of the centromeric and telomeric TADs surrounding the HoxD cluster ( from Dixon et al . 2012 ) . ( D ) Detailed location of the HiC bins covering the HoxD cluster . ( E ) 4C-seq and virtual 4C patterns obtained when using a viewpoint covering the regulatory region CNS39 ( Andrey et al . , 2013 ) , within the telomeric gene desert . In contrast to Hoxd gene viewpoints , the interactions observed with the centromeric gene desert are near background . DOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 019 This clustering of interactions matches with the position of a previously mapped boundary between ‘topological associated domains’ ( TADs; Dixon et al . , 2012; Nora et al . , 2012 ) . In ES cells indeed , two TADs cover approximately the gene deserts on either side and have their border at the level of the Hoxd12-Hoxd11 genes ( Dixon et al . , 2012; Figure 5A , Figure 5—figure supplement 1A ) . Virtual 4C from HiC data with bins that cover the Hoxd genes show highly similar priming of interactions with the surrounding TADs when compared to our 4C-seq analysis , confirming that both approaches score similar chromatin behavior ( Figure 5—figure supplement 2; Sexton et al . , 2012 for analysis strategy ) . During limb development , genes located near this boundary ( Hoxd9 to Hoxd11 ) change their tropism and switch their contacts from one TAD to the other , such as to interact sequentially with the appropriate enhancers ( Andrey et al . , 2013 ) . This structural re-organization is clearly illustrated in our hierarchical clustering with Hoxd11 changing its association from Hoxd4 to Hoxd13 ( Figure 5—figure supplement 1C ) . During temporal colinearity , however , such structural re-organization is not observed and hence the stepwise transcriptional activation of Hoxd genes appears to occur within a largely constitutive framework of long-range interactions; genes up to Hoxd11 interact mostly with the telomeric domain , either before or after their activation , and Hoxd13 always interacts with the centromeric domain . This opposed tropism for Hoxd genes in ES cells , as revealed by HiC and the consequent TAD structures , is somewhat at odds with the local clustering of Hoxd genes when in a negative state , which we report here by using 4C . This paradox is discussed below .
We also observe that the sequential transcriptional activation of Hox genes in the PSM coincides with a gene-by-gene transfer or positioning from the inactive H3K27me3-decorated compartment to a newly formed 3D compartment containing active genes only . This indicates that the presence of Hox genes in 3D compartments of various extents , along the developed body axis ( Noordermeer et al . , 2011 ) is not an a posteriori mechanism used to fix and secure the long-term maintenance of various states of activity , fixed earlier by ‘classical’ transcriptional regulations acting in trans . Instead , it suggests that such spatial structures are instrumental in the precise regulation of their transcriptional timing . This is observed at least for the Hoxd9 to Hoxd13 genes and we infer that the same process occurs in the part of the cluster containing from Hoxd1 to Hoxd8 . It is however not possible to assess this experimentally due to technical limitations associated with the size of the embryonic material at the corresponding developmental stages . Temporal colinearity was originally proposed as a mechanism to translate time into spatial coordinates , in different ontogenic contexts ( Duboule , 1994; Gerard et al . , 1997; Durston et al . , 2012 ) . While our results support this idea , the transcriptional timing associated with gene clustering may not be an absolute prerequisite to achieve the proper spatial distributions of Hox genes products , as suggested by the multiple cases where single mammalian Hox transgenes could largely recapitulate the major expression specificities along the AP-axis ( Krumlauf , 1994; Duboule , 2007; Tschopp et al . , 2009; Tschopp and Duboule , 2011a ) . In this context , it is possible that the progressive transition of Hox genes from an inactive to an active 3D compartment reflects the existence of a mechanism whose major aim would not be to precisely regulate a time sequence but instead , to protect the most ‘posterior’ Hox genes from a premature exposure to activating factors , a situation shown to block posterior elongation and hence to be detrimental to the embryo ( Young et al . , 2009; Mallo et al . , 2010 ) . The necessity to actively prevent the most posterior genes from premature activation is supported by their basal transcriptional activity in ES cells , where Hox clusters are less discrete than in subsequent negative tissues such as fetal brain cells . Such a basal activity , which was not scored in these latter cells , may reflect the rather generic nature of the activating signals , the general mechanism underlying temporal colinearity thus relying on de-repression . Studies using internal Hox cluster deletions and duplications indeed showed that the relative position of Hox genes , rather than their promoters , determines their responses to activating signals ( Tschopp et al . , 2009 ) . In this view , graded signals emanating from the posterior aspect of the developing embryo would lead to a progressive de-repression of Hox clusters , implying that these clusters would display some directional sensitivity . While the nature of the activating factors is elusive , a link with the segmentation clock was proposed ( Dubrulle et al . , 2001; Zakany et al . , 2001 ) . Concerning the directional sensitivity , Polycomb group ( Pc-G ) gene products may play an important role in this process , as the distribution of H3K27me3 marks correlate with the size of the inactive 3D compartments . Recently , a somewhat graded distribution of both EZH2 and RING1B , two proteins members of the PRC2 and PRC1 complexes , respectively , was described over the HoxD cluster in ES cells , with the highest signals covering the most ‘posterior’ genes ( Li et al . , 2011 ) . Directionality may therefore derive from a weaker ‘anterior’ repression exerted by the Pc system . In this context , progressive alterations of the repressive system should sensitize the transcriptional threshold , while keeping on with directionality . This effect was observed in Cbx2−/− mutant embryos ( a component of the PRC1 complex formerly known as M33 ) , where the efficiency of the PRC1 complex was moderately decreased: RA treatment resulted in premature yet colinear activation of Hoxd genes ( Bel-Vialar et al . , 2000 ) . Alternatively , collinear activation may rely upon a different kind of model involving for example biophysical forces ( Almirantis et al . , 2013 ) . Future experiments where the process will be witnessed at the cellular level in real time may be informative in this context . Hox genes are originally activated in the most posterior aspect of the gastrulating embryo . This initial wave of activation seems to involve first a poised transcriptional status ( Forlani et al . , 2003 ) , followed by an apparent anterior forward spreading ( Deschamps and Wijgerde , 1993; Gaunt and Strachan , 1994; Gaunt , 2001 ) , which will ultimately lead to the positioning and initiation of the expression domains in the pre-somitic mesoderm ( PSM ) . The colinear processing of this early phase may involve preparatory modifications in the chromatin status , making the system poised for activation by factors emanating from posterior cells ( Forlani et al . , 2003 ) . In this view , the observed anterior forward spreading in expressing cells ( Deschamps and Wijgerde , 1993; Gaunt and Strachan , 1994; Gaunt , 2001 ) may reflect a prolonged exposure to low levels of signals diffusing from the posterior end of the primitive streak ( Forlani et al . , 2003 ) . A second ( non-exclusive ) possibility is that it illustrates the initial difficulty to maintain a robust boundary in Pc repression in a gene cluster where some anterior genes are fully active , with a tendency for the nearby-located genes to be de-repressed and activated . However , our results suggest that once the expression is finally established within the PSM , the boundary between the active and inactive compartments remain rather stable for the next couple of days , until the axial skeleton is fully determined . In this view , these chromatin domains may represent part of the machinery used to fix a given state of activation and thus translate a temporal parameter into spatial coordinates . As such , early heterochronies in Hox gene activation within the PSM will lead to subsequent re-positioning of the expression boundary , as previously observed ( Gerard et al . , 1997 ) . By using genetic approaches , it was previously argued that the time-sequenced activation of Hoxd genes primarily uses regulatory influences located within the gene cluster itself ( Spitz et al . , 2001 ) , with some contributions coming from more distant flanking regions ( Tschopp et al . , 2009; Tschopp and Duboule , 2011b ) . We now report that such a transcriptional activation is implemented with little-if any-differences in the interaction profiles between the target genes and their neighboring gene deserts , unlike the situation observed during limb development where new contacts appear upon gene activation ( Montavon et al . , 2011; Andrey et al . , 2013 ) . However , temporal colinearity does occur within a framework of constitutive long-range interactions , which may provide a scaffold helping the bimodal separation of active and inactive genes to take place . Further experiments with mice carrying large re-arrangements of these two gene deserts will be necessary to clearly weight the importance of flanking regions in the implementation of the Hox clock . Finally , while the comparison between published HiC data ( Dixon et al . , 2012 ) and our 4C datasets are generally highly consistent ( e . g . , Figure 5 , Figure 5—figure supplement 1; Andrey et al . , 2013 ) , the data reported here using ES cells raise an apparent paradox . HiC analysis in ES cells identified a boundary between topological domains positioned around the Hoxd12 to Hoxd11 gene ( Dixon et al . , 2012; Figure 5—figure supplement 2D ) and such boundaries are thought to impose or reflect a physical separation between the two interaction landscapes ( e . g . , Nora et al . , 2013 ) . As a consequence , Hoxd13 should display more interactions with its flanking gene desert than with the other part of the HoxD gene cluster . Yet , by using several viewpoints in a 4C set-up , the HoxD cluster in ES cells appears to form a single negative compartment , despite the interspersed presence of this TAD boundary ( Figure 1 ) . In fact , a detailed analysis of the HiC dataset reveals that the HoxD cluster itself forms a ‘micro-TAD’ , displaying strong internal interactions , in agreement with the 4C results reported here . As such , we consider it likely that the TAD boundary identified by HiC in ES cells ( Dixon et al . , 2012 ) represents an average description of two distinct configurations ( Figure 6 ) . For each allele , either the most posterior Hoxd13 gene forms stable interactions within the TAD on the centromeric side or , alternatively , the Hoxd11 to Hoxd1 genes interact with the TAD on the telomeric side ( Figure 6 ) . As a consequence , for each allele the entire HoxD 3D compartment becomes located towards a single TAD , on one side of the cluster with a physical separation from the other side . Molecule ( s ) causing these interactions are elusive and may include proteins that mediate constitutive loops between Hoxd gene promoters and their regulatory elements . The CTCF protein , which may play a role in scaffolding TADs , binds multiple sites around the Hoxd13 to Hoxd8 region ( Ferraiuolo et al . , 2010; Phillips-Cremins et al . , 2013 ) . Because formaldehyde crosslinking has a very short range of action ( Orlando et al . , 1997 ) , the system generates a graded pattern of 4C and HiC interactions from the location of the actual binding sites . Therefore , while in ES cells and for each allele , TAD borders are likely located at either side of the HoxD 3D compartment ( Figure 6A , black lines ) , our analysis of a large population of cells reflects the equilibrium that exists between these two situations ( Figure 6E ) . At later stages , when the HoxD cluster adopts a bimodal 3D organization , the tethering of interactions , as illustrated by the existence of TADs on either side , may help implement the separation between activated and repressed Hox genes , thereby potentially reducing deleterious regulatory interferences and premature activation of the most posterior Hox genes . 10 . 7554/eLife . 02557 . 020Figure 6 . Model of dynamic bi-modal 3D compartmentalization during temporal colinearity . ( A ) Schematic organization of topological domains in ES cells ( from Dixon et al . 2012 ) matching the centromeric and telomeric gene deserts , with an apparent boundary assigned near the Hoxd11 gene ( grey diagonal lines ) . All Hoxd genes in ES cells have considerable interactions on either side of the cluster , suggesting that this border is more diffuse and hence the entire HoxD cluster can be integrated in either TAD ( diagonal black lines ) . ( B ) Various states of activity for Hoxd genes in different samples , analyzed during sequential activation . The assigned TAD boundary in ES cells is indicated by the dashed line . ( C ) Conceptual 2D representation of chromatin organization within the HoxD cluster chromatin compartment and surrounding centromeric and telomeric TADs in ES cells . ( D ) Schemes illustrating the dynamics of local 3D compartmentalization for the HoxD cluster ( red and blue compartments ) vs the constitutive nature of interactions in the context of the surrounding TADs during sequential activation . ( E ) A dynamic equilibrium to explain the paradox in the observed local vs long-range interactions . Genes located at the centromeric or telomeric extremities of the HoxD cluster form stable interactions with DNA sequences located with the flanking gene deserts , thereby dragging the HoxD 3D chromatin compartments into either one of the TADs . Within a cellular population , this process is in equilibrium , resulting in a read-out where Hoxd genes have a graded preference to interact with either the centromeric or the telomeric deserts , despite being organized into a single 3D chromatin compartment . DOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 020
All experiments were performed in agreement with the Swiss law on animal protection ( LPA ) . Tissue samples were isolated at the indicated time points ( with maximum 6 hr delay ) , with day E0 . 5 being noon on the day of the vaginal plug . Tissue pieces for 4C-sequencing , ChIP-sequencing , RNA-sequencing and Reverse Transcriptase-qPCR were isolated in PBS and subsequently transferred to PBS supplemented with 10% Fetal Calf Serum . 4C-seq and ChIP-seq material was incubated for 45 min with 1 mg/ml collagenase ( Sigma-Aldrich , St . Louis , MO ) , and 4C-seq material was further made single cell using a cell strainer ( BD Falcon ) . Mouse ES cells were grown under feeder-free conditions on gelatinized plates in Dulbecco's modified Eagle's medium ( DMEM , Life Technologies , Carlsbad , CA ) supplemented with 17% fetal calf serum , 1 × non-essential amino acids ( Life Technologies ) , 1 × Pen–Strep ( Life Technologies ) , Sodium Pyruvate ( Life Technologies ) , 0 . 1 mM β-mercaptoethanol , and 1000 U/ml LIF . Embryonic 4C-seq samples consisted of pooled material from multiple embryos: 129 embryos for E8 . 5 pre-somatic mesoderm samples , 196 embryos for E9 . 5 tail bud , 143 embryos for E10 . 5 tail bud or E10 . 5 lumbo-sacral trunks and around 20 embryos for each E10 . 5 forebrain sample . For embryonic ChIP samples , 50 μg of chromatin was cross-linked at a time , of which 10 μg was used per ChIP . To obtain 50 μg of chromatin , 750 E8 . 5 pre-somatic mesoderm samples or 10 E10 . 5 forebrains were pooled . Total RNA from E10 . 5 forebrain was isolated from single embryos . ES cell 4C-seq and ChIP-seq samples were prepared from samples consisting of 20 million cells . 10 μg of ES cell chromatin was used per ChIP . Total RNA was isolated from 1 million cells . 4C–seq libraries were constructed as previously described ( Noordermeer et al . , 2011 ) . NlaIII ( New England Biolabs , Ipswich , MA ) was used as the primary restriction enzyme and DpnII ( New England Biolabs ) was used as the secondary restriction enzyme . For each viewpoint , a total of 1 μg ( E9 . 5 tail bud , E10 . 5 tail bud , E10 . 5 lumbo-sacral trunks , E10 . 5 forebrain and ES cells ) or 50 ng ( E8 . 5 pre-somatic mesoderm ) of each 4C-seq library was amplified using 16 individual PCR reactions with inverse primers including Illumina Solexa adapter sequences ( primer sequences in Table 2 ) . Illumina sequencing was done on multiplexed samples , containing PCR amplified material of up to 7 viewpoints , using 100 bp Single end reads on the Illumina HiSeq system according to the manufacturer's specifications . 10 . 7554/eLife . 02557 . 021Table 2 . 4C-seq Inverse primer sequencesDOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 021ViewpointInverse primerSequenceHoxd13iHoxd13 forward*AATGATACGGCGACCACCGAACACTCTTTCCCTACACGACGCTCTTCCGATCTAAAAATCCTAGACCTGGTCATGchr2:74504328-74504348iHoxd13 reverse*CAAGCAGAAGACGGCATACGAGGCCGATGGTGCTGTATAGGchr2:74505579-74505598Hoxd11iHoxd11 forward*AATGATACGGCGACCACCGAACACTCTTTCCCTACACGACGCTCTTCCGATCTAAGCATACTTCCTCAGAAGAGGCAchr2:74523621-74523643iHoxd11 reverse*CAAGCAGAAGACGGCATACGACTAGGAAAATTCCTAATTTCAGGchr2:74523881-74523903Hoxd9iHoxd9 forward*AATGATACGGCGACCACCGAACACTCTTTCCCTACACGACGCTCTTCCGATCTACGAACACCTCGTCGCCCTchr2:74536168-74536185iHoxd9 reverse*CAAGCAGAAGACGGCATACGACCCTCAGCTTGCAGCGATchr2:74536797-74536814Hoxd4iHoxd4 forward*AATGATACGGCGACCACCGAACACTCTTTCCCTACACGACGCTCTTCCGATCTAAGGACAATAAAGCATCCATAGGCGchr2:74561330-74561353iHoxd4 reverse*CAAGCAGAAGACGGCATACGATCCAGTGGAATTGGGTGGGATchr2:74562171-74562191Hoxc13iHoxc13 forward*AATGATACGGCGACCACCGAACACTCTTTCCCTACACGACGCTCTTCCGATCTAGATAATTTTCCTGAGACATTGTAACchr15:102756108-102756132iHoxc13 reverse*CAAGCAGAAGACGGCATACGAGCTCAATGTTCCCTTCCCTAACGchr15:102755251-102755273Hoxb13iHoxb13 forward*AATGATACGGCGACCACCGAACACTCTTTCCCTACACGACGCTCTTCCGATCTAGGACTGTTCCTCGGGGCTATchr11:96057673-96057692iHoxb13 reverse*CAAGCAGAAGACGGCATACGAATCTGGCGTTCAGAGAGGCTchr11:96057448-96057467Hoxb9iHoxb9 forward*AATGATACGGCGACCACCGAACACTCTTTCCCTACACGACGCTCTTCCGATCTAAGATTGAGGAGTCTGGCCACTTchr11:96136070-96136091iHoxb9 reverse*CAAGCAGAAGACGGCATACGATCATCAAACCAAGCAGGGCAchr11:96136671-96136690Hoxa13iHoxa13 forward*AATGATACGGCGACCACCGAACACTCTTTCCCTACACGACGCTCTTCCGATCTAACACTTGCACAACCAGAAATGCchr6:52212211-52212232iHoxa13 reverse*CAAGCAGAAGACGGCATACGAGGCGAGGCTCAGGCTTTTATchr6:52212476-52212495CNS ( 39 ) iCNS ( 39 ) forward†AATGATACGGCGACCACCGAACACTCTTTCCCTACACGACGCTCTTCCGATCTATCCAAGGAGAAAGGTGTTGGTCchr2:74975258-74975279iCNS ( 39 ) reverse†CAAGCAGAAGACGGCATACGACAGGGCGTTGGGTCACTCTchr2:74975670-74975687Location of primers according to NCBI37 ( mm9 ) . *Primers from Noordermeer D , Leleu M , Splinter E , Rougemont J , De Laat W , Duboule D . 2011 . The dynamic architecture of Hox gene clusters . Science 334:222–225 . †Primers from Andrey G , Montavon T , Mascrez B , Gonzalez F , Noordermeer D , Leleu M , Trono D , Spitz F , Duboule D . 2013 . A switch between topological domains underlies HoxD genes collinearity in mouse limbs . Science 340:1234167 . 4C-seq reads were sorted , aligned , and translated to restriction fragments using the 4C-seq pipeline of the BBCF HTSstation ( available at http://htsstation . epfl . ch; Noordermeer et al . , 2011; David et al . , 2014 ) according to ENSEMBL Mouse assembly NCBIM37 ( mm9 ) . 4C-seq patterns were corrected vs previously generated random 4C–seq libraries ( Noordermeer et al . , 2011 ) , consisting of BACs covering the mouse Hox clusters ( HoxD: RP23-331E7; HoxC: RP23-430C12; HoxB: RP23-381I12 and RP23-196F5; HoxA: RP24-298M24 ) . After random correction , three restriction fragments were removed that returned aberrant values ( HoxD: chr2:74’597’000-74’597’732; chr2:74’608’796-74’609’312 , HoxB: chr11:95’999’958-96’000’916 ) due to sequence abnormalities in the BAC template ( confirmed by Sanger sequencing; not shown ) . Normalization and further data processing was done as previously described ( Noordermeer et al . , 2011 ) . Quantitative log2 ratios were calculated by dividing the quantitative fragment count between tissue samples . Unprocessed 4C-seq data is available from the Gene Expression Omnibus ( GEO ) repository under accession number GSE55344 . Random corrected tracks are available from http://duboule-lab . epfl . ch/data . The directionality of long-range interactions was calculated as previously described ( Woltering et al . , 2014 ) . In Figure 5A , the smoothed 4C-seq patterns ( running mean , window size 11 ) were obtained using the 4C-seq pipeline of the BBCF HTSstation ( available at http://htsstation . epfl . ch; David et al . , 2014 ) . HiC data on topological associated domains ( TADs ) from ES cells were obtained from ( http://chromosome . sdsc . edu/mouse/hi-c/database . php; Dixon et al . , 2012 ) . Two TADs located centromeric and telomeric of the clusters were selected , covering genomic coordinates chr2:73400000-75960000 ( discussed in Woltering et al . , 2014 ) . Spearman correlation of long-range patterns was done over the region covering these TADs , with signal on the HoxD cluster itself removed ( excluded region: chr2:74484971-74607492 ) . Conventional hierarchical clustering was done to score for relationships between viewpoints . ChIP was performed as previously described ( Noordermeer et al . , 2011 ) . Cells were fixed for 5 min in a 2% formaldehyde solution at room temperature . ChIP-seq samples were fragmented to a range of 200–500 bp using tip sonication ( Misonix S4000 , Misonix , Farmingdale , NY ) , For all ChIP assays , 10 μg of cross-linked chromatin was used . Antibodies used: anti Histone H3K27me3 ( #17-622; Millipore , Billerica , MA ) and anti H3K4me3 ( #17-614; Millipore ) . ChIP-seq libraries were constructed from 6 to 10 nanograms of immune-precipitated DNA according to the manufacturers instructions ( Illumina , San Diego , CA ) . Sequencing was done using 50 or 100 bp Single end reads on the Illumina HiSeq system according to the manufacturer's specifications . ChIP-seq reads were mapped to ENSEMBL Mouse assembly NCBIM37 ( mm9 ) , and extended to 100 bp if read lengths smaller than 100 bp were used , using the ChIP-seq pipeline of the BBCF HTSstation ( available at http://htsstation . epfl . ch; David et al . , 2014 ) . ChIP-seq data is available from the Gene Expression Omnibus ( GEO ) repository under accession numbers GSE55344 and GSE31570 . Random corrected 4C-seq and ChIP-seq samples were correlated by ranking experimental values within restriction fragments ( Table 1 ) . First , to each NlaIII restriction fragment covered by the random 4C tracks within the regions visualized in Figures 1 and 2 ( HoxD cluster: chr2:74454783-74622413; HoxC cluster: chr15:102715179-102909417; HoxB cluster: chr11:95992344-96244915; HoxA cluster: chr6:52058584-52234371 ) , the average ChIP-seq signal was assigned for each condition . Restriction fragment within individual samples were ranked based on their 4C-seq or ChIP-seq value and subsequently the Spearman's rank correlation coefficient was calculated between pairs of samples . Total RNA from tissue samples was isolated using Trizol LS reagent ( Life Technologies ) . Total RNA from ES cell samples was isolated using Trizol reagent ( Life Technologies ) . For RNA-seq , the RNA was depleted from rRNAs and , subsequently , strand-specific total RNA-seq libraries were constructed according to the manufacturers instructions ( Illumina ) . Sequencing was done using 50 bp Single end reads on the Illumina HiSeq system according to the manufacturer's specifications . RNA-seq reads were mapped to ENSEMBL Mouse assembly NCBIM37 ( mm9 ) and translated into reads per gene ( RPKM ) using the RNA-seq pipeline of the BBCF HTSstation ( available at http://htsstation . epfl . ch; David et al . , 2014 ) . RNA-seq data is available from the Gene Expression Omnibus ( GEO ) repository under accession numbers GSE55344 . For RT-qPCR , cDNA was synthesized after DNAseI treatment ( Life Technologies ) using SuperScript III ( Life Technologies ) and oligo-dT primers ( Life Technologies ) , using the manufacturer's instructions . For ES cells and E10 . 5 forebrain , 2 μg of RNA was used as input for the cDNA synthesis , for E10 . 5 posterior trunk 1 μg of RNA was used . Products were quantified by qPCR using EXPRESS SYBR GreenER mixes ( Life Technologies ) on a CFX96 PCR Detection System ( BioRad , Hercules , CA ) . Sequences of intron-spanning primers are provided in Table 3 . 10 . 7554/eLife . 02557 . 022Table 3 . RT-qPCR primer sequencesDOI: http://dx . doi . org/10 . 7554/eLife . 02557 . 022FragmentPrimerSequencemRNAmRNA Tubb2c forward*GCAGTGCGGCAACCAGAT chr2:25080064-25080081Tubb2cmRNA Tubb2c reverse*AGTGGGATCAATGCCATGCT chr2:25079711-25079730mRNAmRNA Tbp forward*TTGACCTAAAGACCATTGCACTTC chr17:15644342-15644365TbpmRNA Tbp reverse*TTCTCATGATGACTGCAGCAAA chr17:15650497-15650518mRNAmRNA Hoxd13 forward*GGTGTACTGTGCCAAGGATCAG chr2:74507077-74507098Hoxd13mRNA Hoxd13 reverse*TTAAAGCCACATCCTGGAAAGG over intron boundrymRNAmRNA Hoxd9 forward*GCAGCAACTTGACCCAAACA over intron boundryHoxd9mRNA Hoxd9 reverse*GGTGTAGGGACAGCGCTTTTT chr2:74537278-74537298mRNAmRNA Hoxd4 forwardTCAAGCAGCCCGCTGTGGTC chr2:74565709-74565728Hoxd4mRNA Hoxd4 reverseTCTGGTGTAGGCCGTCCGGG chr2:74566355-74566374mRNAmRNA Hoxb13 forwardGTCCATTCTGGAAAGCAG chr11:96056334-96056351Hoxb13mRNA Hoxb13 reverseAAACTTGTTGGCTGCATACT chr11:96057389-96057408mRNAmRNA Hoxb9 forwardGGCAGGGAGGCTGTCCTGTCT chr11:96133282-96133302Hoxb9mRNA Hoxb9 reverseGCCAGTTGGCAGAGGGGTTGG chr11:96135938-96135958Location of primers according to NCBI37 ( mm9 ) . *Primers from Montavon T , Le Garrec JF , Kerszberg M , Duboule D . 2008 . Modeling Hox gene regulation in digits: reverse collinearity and the molecular origin of thumbness . Genes Dev 22:346–359 . | Most animals are symmetrical about an imaginary line that runs from the head to the tail . A family of genes called the Hox family ensures that the cells in an animal embryo develop into the correct body parts along this head-to-tail axis . Hox genes—which are found in animals as diverse as flies and humans—are often clustered on the chromosomes , and their order within a cluster affects when and where each Hox gene is ‘switched on’ . In mammals , Hox genes at one end of a cluster are switched on first and along almost the entire length of the embryo . Hox genes near the other end of the cluster are expressed later and only towards the hind end of the animal . And Hox genes at the furthest end of the cluster are expressed last and in the very tip of the developing tail . The time when a Hox gene is expressed depends largely on its relative position within the gene cluster . However , it is not clear how the ordering of the genes within a cluster is translated into a schedule whereby the genes are sequentially switched on during development . Much of the DNA in a chromosome is wrapped around proteins to form a structure called chromatin; chromatin is normally tightly packed , but ‘unpacking’ it allows the genes to be accessed and switched on . Now , Noordermeer et al . have used a technique called ‘circular chromosome conformation capture’ to follow how the packing of the chromosomes that carry the Hox gene clusters changes during embryonic development . Harvesting cells from mouse embryos of different ages , and cross-linking the DNA to the proteins , allowed those genes that are packed in the chromatin to be distinguished from those that have been unpacked and activated . When the embryo is still just a ball of almost identical cells , all the Hox genes are switched off and packed into inactive chromatin . However , Noordermeer et al . found that , as the embryo develops and when each Hox gene is switched on in turn , the relevant region of DNA is also unpacked and moved into more active chromatin . This mechanism likely prevents Hox genes that direct the development of the hind end of the mouse from being switched on too early , and hence it avoids body parts being misidentified and developing incorrectly . Further , the patterns of active chromatin vs inactive chromatin can be fixed at each section along head-to-tail axis , such that it will be memorized in all daughter cells produced subsequently from each particular body section . Future challenges will be to uncover the trigger behind the step-wise transition of every Hox gene from inactive chromatin to active chromatin , and to crack the underlying ‘clock’ that controls the timing of this process . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2014 | Temporal dynamics and developmental memory of 3D chromatin architecture at Hox gene loci |
Recent studies suggest de novo mutations may involve the pathogenesis of autism and attention-deficit/hyperactivity disorder ( ADHD ) . Based on the evidence that excessive alcohol consumption may be associated with an increased rate of de novo mutations in germ cells ( sperms or eggs ) , we examine here whether the risks of autism and ADHD are increased among individuals with a family history of alcohol use disorders ( AUDs ) . The standardized incidence ratios ( SIRs ) of autism and ADHD among individuals with a biological parental history of AUDs were 1 . 39 ( 95% CI 1 . 34–1 . 44 ) and 2 . 19 ( 95% CI 2 . 15–2 . 23 ) , respectively , compared to individuals without an affected parent . Among offspring whose parents were diagnosed with AUDs before their birth , the corresponding risks were 1 . 46 ( 95% CI 1 . 36–1 . 58 ) and 2 . 70 ( 95% CI 2 . 59–2 . 81 ) , respectively . Our study calls for extra surveillance for children with a family history of AUDs , and further studies examining the underlying mechanisms are needed .
Autism is a neural development disorder characterized by impaired social interaction and verbal and non-verbal communication ( Biederman and Faraone , 2005; Levy et al . , 2009 ) . Attention-deficit/hyperactivity disorder ( ADHD ) is characterized by inappropriate levels of inattention and hyperactivity , resulting in functional impairment ( Biederman and Faraone , 2005 ) . Autism and ADHD typically begin in childhood and often persist into adulthood ( Biederman and Faraone , 2005; Levy et al . , 2009 ) . A recent study suggests that autism and ADHD share common genetic mutations ( Cross-Disorder Group of the Psychiatric Genomics Consortium , 2013 ) , which may provide the genetic basis for the associations of the two disorders . The etiology of autism and ADHD is , however , largely unknown , although a few genetic and environmental factors have been suggested to be associated with their developments ( Szpir , 2006; Walsh et al . , 2008 ) . Epidemiological studies suggest that the prevalence of both disorders is increasing ( Fombonne , 2005; Parner et al . , 2008; Boyle et al . , 2011 ) , and that advanced parental age at the time of conception is associated with increased risk of autism ( Kong et al . , 2012; Parner et al . , 2012 ) . A subsequent study showed that the risk of autism was also higher in grandchildren with an older maternal grandmother compared to those with a younger one ( Golding et al . , 2010 ) . In addition , twin studies showed that the concordance rate for autism is much higher in monozygotic twins than in dizygotic twins ( Nordenbaek et al . , 2013 ) , suggesting de novo mutations may partly explain the development of autism ( Callaway , 2012; Kong et al . , 2012 ) . Previous studies found that a number of de novo mutations and epigenetic modifications in germ cells could lead to increased rates of autism , schizophrenia , and other mental disorders ( Malaspina , 2001; Fombonne , 2005; Boyle et al . , 2011; Callaway , 2012; Kong et al . , 2012; Parner et al . , 2012; Gratten et al . , 2013; Zaidi et al . , 2013 ) . It is thus important to identify potential environmental factors that contribute to the increased rate of de novo mutations and/or epigenetic modifications in the parents ( Kinney et al . , 2010 ) . We assume that excessive alcohol consumption may be associated with an increased rate of de novo mutations in germ cells ( sperms or eggs ) , because previous studies showed that exposures to alcohol could lead to increased rates of mutations in germ cells ( Narod et al . , 1988; Yamauchi et al . , 2012 ) . It is thus necessary to examine whether the risks of autism and ADHD are increased among individuals with a family history of alcohol use disorders ( AUDs ) . In addition , we examined the offspring risks of autism and ADHD when their adoptive parents were diagnosed with AUDs to disentangle the potential contribution of high levels of psychological tension and stress in AUD families .
In Table 1 , we present the basic characteristics of individuals with autism and ADHD . A total of 24157 and 49348 individuals were identified with autism and ADHD in Sweden between 1987 and 2010 . Men were more often diagnosed with these two disorders than women , and the median age at diagnosis was for both disorders 16 years . The incidences of the two disorders increased greatly during the last decade . 10 . 7554/eLife . 02917 . 003Table 1 . Basic characteristics of patients with autism and ADHDDOI: http://dx . doi . org/10 . 7554/eLife . 02917 . 003CharacteristicAutismADHDNo . %No . %SexMale1680869 . 63349167 . 9Female734930 . 41585732 . 1Age ( years ) <10599524 . 8812916 . 510–19967140 . 02285446 . 320–29421517 . 4772715 . 730+427617 . 71063821 . 6Time period1987–19904251 . 82070 . 41991–19956562 . 73810 . 81996–200011995 . 013072 . 62001–2005645726 . 7995320 . 22006–20101542063 . 83750076 . 0All24157100 . 049348100 . 0 The risks of autism and ADHD among individuals with a family history of AUDs are presented in Table 2 . During more than 10 million person-years of follow-up , a total of 3136 individuals were diagnosed with autism and 10 , 047 individuals were diagnosed with ADHD . The overall risks of autism and ADHD were 1 . 39 ( 95% CI 1 . 34–1 . 44 ) and 2 . 19 ( 95% CI 2 . 15–2 . 23 ) , respectively , among individuals with an affected parent with AUDs compared to those without an affected parent . The risks were similar in affected sons and daughters . The PAF was 3 . 6 and 11 . 0% , respectively , for autism and ADHD . 10 . 7554/eLife . 02917 . 004Table 2 . Risk of autism and ADHD in offspring when their parents were diagnosed with alcohol use disorder ( AUD ) DOI: http://dx . doi . org/10 . 7554/eLife . 02917 . 004AUD in parentNo . of offspring at riskPerson-years of follow-upAutismADHDOSIR95% CIOSIR95% CIRisk in sonsFather235696448570717931 . 391 . 331 . 4555482 . 172 . 112 . 22Mother6921413447665671 . 551 . 431 . 6919522 . 732 . 612 . 85Parents289763554036222031 . 411 . 351 . 4768002 . 202 . 152 . 25Risk in daughtersFather22531743061947531 . 311 . 221 . 426092 . 122 . 042 . 21Mother65 , 47912755072451 . 501 . 321 . 79702 . 772 . 602 . 95Parents27643953052859331 . 341 . 251 . 4332472 . 182 . 102 . 25Risk in offspringFather461013879190125461 . 361 . 311 . 4281572 . 152 . 112 . 20Mother13469326202738121 . 541 . 431 . 6529222 . 742 . 642 . 84Parents5662021084564731361 . 391 . 341 . 4410 , 0472 . 192 . 152 . 23PAF3 . 6%11%O , observed number of cases; SIR , standardized incidence ratio; CI , confidence interval . Bold type , 95% CI does not include 1 . 00 . PAF , population attributable fraction . To examine whether the observed association in Table 2 is possibly due to putative germ cell mutation in parents with alcohol consumption , we studied the risk of autism and ADHD among offspring whose parents were diagnosed with AUDs before their birth ( Table 3 ) . Only 95 , 003 offspring had a parent diagnosed with AUDs before their birth , which accounted for 16% of all offspring with a family history of AUDs . After 922 , 618 person-year of follow-up , the overall risks of autism and ADHD were 1 . 46 ( 95% CI 1 . 36–1 . 58 ) and 2 . 70 ( 95% CI 2 . 59–2 . 81 ) , respectively , as compared to those without an affected parent . The increase risks were similar in affected sons and daughters . 10 . 7554/eLife . 02917 . 005Table 3 . Risk of autism and ADHD in offspring when their parents were diagnosed with alcohol use disorder ( AUD ) before the birth of the offspringDOI: http://dx . doi . org/10 . 7554/eLife . 02917 . 005AUD in parentNo . of offspring at riskPerson-years of follow-upAutismADHDOSIR95% CIOSIR95% CIRisk in sonsFather418324103274421 . 441 . 311 . 5915552 . 602 . 482 . 74Mother877680 , 689991 . 671 . 352 . 033473 . 142 . 823 . 49Parents487044725965141 . 461 . 341 . 5917782 . 612 . 492 . 73Risk in daughtersFather396843904291451 . 401 . 181 . 645653 . 052 . 803 . 31Mother845777259361 . 791 . 252 . 481193 . 442 . 854 . 12Parents462994500221751 . 471 . 261 . 706343 . 002 . 773 . 24Risk in offspringFather815168007565871 . 431 . 321 . 5521202 . 712 . 592 . 83Mother172331579481351 . 701 . 422 . 014663 . 222 . 933 . 52Parents950039226186891 . 461 . 361 . 5824122 . 702 . 592 . 81O , observed number of cases; SIR , standardized incidence ratio; CI , confidence interval . Bold type , 95% CI does not include 1 . 00 . In Table 4 , we present the risk of autism and ADHD among adoptees when either their biological or adoptive parents were diagnosed with AUDs . The risks of autism and ADHD were significantly increased when their biological parents were diagnosed with AUDs with a SIR of 1 . 75 and 1 . 91 , respectively . However , the risks were not significant when their adoptive parents were diagnosed with AUDs , possibly because of limited numbers of cases . 10 . 7554/eLife . 02917 . 006Table 4 . Risk of autism and ADHD in adoptees when their biological or adoptive parents were diagnosed with alcohol use disorder ( AUD ) DOI: http://dx . doi . org/10 . 7554/eLife . 02917 . 006No . of offspring at riskPerson-years of follow-upAutismADHDOSIR95% CIOSIR95% CIBiological parents with AUD Father217551746162 . 151 . 233 . 50532 . 631 . 973 . 44 Mother138132745101 . 810 . 863 . 35181 . 210 . 721 . 92 Parents325177310191 . 751 . 052 . 74581 . 911 . 452 . 47Adoptive parents with AUD Father45710 , 83910 . 750 . 004 . 2871 . 780 . 703 . 68 Mother2295443041 . 930 . 504 . 98 Parents6541554610 . 510 . 002 . 9391 . 590 . 723 . 04O , observed number of cases; SIR , standardized incidence ratio; CI , confidence interval . Bold type , 95% CI does not include 1 . 00 .
In this population-based cohort study from Sweden of 24 , 157 and 49 , 348 individuals with autism and ADHD , respectively , we found that the risks were significantly increased among individuals with a family history of AUDs , after adjusting for potential confounding factors . The incidence rates of autism and ADHD were somewhat more pronounced when the diagnosis of AUDs in parents occurred before the birth of their offspring . The time-relationship supports that the observed association between parental AUDs and autism and ADHD in offspring may be causal . In addition , we studied the incidence among adoptees to disentangle the contribution by environmental factors . However , we could not draw definite conclusions from these results , as the cases were very few among adoptees with an affected adoptive parent . The PAF was 3 . 6 and 11 . 0% for autism and ADHD , respectively , if excess alcohol consumption could be avoided in their parents . There are a few possible biological explanations for the high observed risks of autism and ADHD among individuals with a family history of AUDs . Firstly , they could be explained by an increased rate of de novo mutations among individuals with AUDs , confirming our hypothesis . It should be noted that although a few de novo mutations have been found to be related to autism ( Callaway , 2012; Kong et al . , 2012; Zaidi et al . , 2013 ) , associations with ADHD are still lacking , calling for more research to examine possible mutations among patients with ADHD . However , a recent study found that autism and ADHD share common genetic mutations ( Cross-Disorder Group of the Psychiatric Genomics Consortium , 2013 ) , which may partly explain their associations with parental AUD . In addition , the increased risks of autism and ADHD may be related to epigenetic modifications of specific genes because exposure to alcohol can alter the methylation status in sperm cells ( Govorko et al . , 2012 ) . How epigenetic modifications may affect the development of these disorders should , however , be examined in further studies . It should be noted that there might be common genetic predispositions linked to AUD , autism and ADHD . An additional explanation for the observed associations is that children of individuals with alcohol problems may face high levels of tension and stress at home , and that a stressful home environment may lead to communication problems , bad school performances , and , ultimately , to the development of autism and ADHD ( Berger , 1993 ) . However , our study found that the risk of autism and ADHD was increased only when the biological parents were diagnosed with AUDs , suggesting that genes as well as a stressful home environment have an impact on the increased risk of autism and ADHD . One possible reason for the high risks of autism and ADHD among offspring with mothers with AUDs is that mothers with AUDs may continue drinking alcohol during pregnancy , causing long-term structural , behavioral , and cognitive damage to their children . It is known that alcohol can be carried to the placenta and that the concentration of alcohol in the unborn baby's bloodstream is the same as that in the mother . Mothers who consume alcohol during pregnancy may give birth to a baby with fetal alcohol syndrome ( FAS ) . In general , the more severe the mother's drinking problem is during pregnancy , the more severe are the symptoms of FAS in infants ( Landgren et al . , 2010 ) . Children with FAS may be diagnosed with autism or ADHD because these conditions are all characterized by an impaired quality of social interaction ( Bishop et al . , 2007 ) . Around 4% of autism and 11% of ADHD in offspring can be avoided if parents avoid excessive alcohol consumption . It should be noted that even moderate alcohol consumption during pregnancy may affect the child's cognitive score ( Lewis et al . , 2012 ) , suggesting that the proportion of the two diseases related to parental alcohol consumption would be even higher if individuals with moderate alcohol drinking were also included in the calculation of PAF . An important strength of this study is that all the data were retrieved from Swedish registers with high quality and coverage . In addition , the number of patients included is large enough to guarantee reliable risk estimates . The prospective study design and the completeness of the follow-up of patients are other major advantages of the present study . Moreover , we adjusted for a number of confounding factors and thereby minimized confounding . One limitation of this study is that only individuals with autism and ADHD who visited either primary health care or a hospital were included . Another limitation is the lack of information on some individual-level risk factors , such as parental drinking habits of moderate nature , psychosocial factors , and sociocultural behaviors . In addition , it is likely that the identification of AUD in parents could be underreported . In summary , individuals with a family history of AUDs had increased risks of autism and ADHD , which calls for further study to examine the underlying mechanisms and for clinical attention and extra surveillance for children with a family history of AUDs .
This study was approved by the Ethics Committee at Lund University , Sweden . Individuals with autism and ADHD were identified from four Swedish Registers: the primary health care registers covering the counties of Skåne ( 1987–2010 ) and Stockholm ( 2001–2007 ) and the National Swedish Hospital Discharge Register ( 1987–2010 ) and Outpatient Register ( 2001–2010 ) . The primary health care register in Skåne ( the southernmost province of Sweden ) , PASIS , contains data on individuals ( around 1 . 2 million ) who visited primary health care in Skåne between 1987 and 2010 . The primary health care register in Stockholm covers 75 primary health care centers for the period 2001–2007 . The Swedish Hospital Discharge Register was founded in 1964–65 by the National Board of Health and Welfare , and has had complete nationwide coverage since 1987 . The Outpatient Register contains data on all visits to outpatient clinics in Sweden since 2001 . A recent external review suggested that the overall accuracy of the Swedish Hospital Registry is approximately 85–95% ( Ludvigsson et al . , 2011 ) . Individuals with autism were identified according to the International Classification of Diseases ( ICD-9 code 299B and ICD-10 code F840 ) . Individuals with ADHD were identified according to ICD-9 code 314 and ICD-10 code F90 . We further linked individuals with autism and ADHD to the Swedish Multi-Generation Register to identify their parents ( Ekbom , 2011 ) . The Swedish Multi-Generation Register contains data on the biological parents of index persons born in or since 1932 and registered in Sweden since 1961 . The Multi-Generation Register contains data on more than 3 . 2 million families and 12 million individuals . In addition , we identified 22 , 996 individuals who were born after 1950 and had been adopted , with information available on both adoptive parents and at least one biological parent . Adoptees adopted by biological relatives or by an adoptive parent living with a biological parent were excluded . Diagnoses of AUDs in parents ( biological or adoptive ) were identified from three national registries , including the Hospital Discharge and Outpatient Registers using the ICD-9 ( 291A–291F , 291W , 291X , 303 , 305A ) and ICD-10 codes ( F10 ) , the Crime Register ( which contains data on individuals who committed AUD-related crimes in 1973–2007 ) , and the Swedish Pharmacy Register ( 2005–2010 ) . Data from the latter were based on prescriptions of disulfiram , naltrexone , and acamprosate . Using these three sources , we identified 420 , 489 unique individuals with AUDs ( lifetime prevalence in Sweden of ∼3 . 8% ) . Additional linkages were made to the Statistics Sweden's Total Population Register ( ≥1990 ) and the Population a Housing Census ( <1990 ) to obtain information on individual-level characteristics , such as year of birth , sex , and country of birth; to the Cause of Death Register to identify date of death; and to the Emigration Registry to identify date of emigration . All linkages were performed using individual national identification numbers , which were replaced with serial numbers in order to preserve anonymity . We calculated person-years at risk of autism and ADHD among individuals with a family history of AUDs ( either biological or adoptive parents with AUDs ) from the date of birth , immigration , or 1 January 1987 , whichever came last , until the date of diagnosis of autism or ADHD , death , emigration , or the end of the study period ( 31 December 2010 ) , whichever came first . Standardized incidence ratios ( SIRs ) were calculated as the ratio of the observed and expected numbers of cases ( Breslow and Day , 1987; Rothman and Greenland , 1998 ) . The expected number of cases was calculated according to the incidence rate for all individuals without a family history of AUDs . SIRs were standardized by 5-year age group , sex ( male and female ) , 5-year time period , family history of autism or ADHD , and individual disposable income ( Esteve et al . , 1994 ) . 95% confidence intervals for the SIRs were calculated assuming a Poisson distribution , and were rounded to the nearest two decimals ( Esteve et al . , 1994 ) . The population attributable fraction ( PAF ) is the proportion of the disease burden in the population that could be prevented if exposure to a particular risk factor could be avoided ( dos Santos Silva , 1999 ) . The calculation of PAF is based on a relative risk estimate . In this study , SIRs were used to calculate PAF as follows: PAF = proportion of autism and ADHD patients with a family history of AUD × ( SIR − 1 ) /SIR . All analyses were performed using SAS version 9 . 1 ( SAS Institute , Cary , NC ) . | Children learn to talk , manage their emotions , and control their behavior in a period when the brain is developing rapidly . The first signs of several developmental disorders , such as autism and attention-deficit/hyperactivity disorder ( ADHD ) , may also emerge during this period . Children with autism may have difficulties with social interactions and communication , while those with attention-deficit/hyperactivity disorder may struggle to pay attention to a task and may be more active than other children . Autism or ADHD are diagnosed based on the child's behavior because the underlying causes of the disorders are not well understood . Both genes and the environment have been linked to the conditions; and it was recently suggested that certain common genetic mutations are more common in children with ADHD or autism . However , as some of the mutations linked to autism are not found in the genes of the affected children's parents , it is likely that they occurred in either of the sperm or the egg cell from the parents . Exposure to harmful substances in the environment can cause mutations in egg or sperm cells , or alter the expression of genes without changing the gene sequence . Excessive alcohol consumption is one environmental factor that can mutate genes or alter gene expression . Here , Sundquist et al . have looked to see if there is a relationship between a child having a parent with an alcohol use problem and the child's risk of developing autism or ADHD . Examining national medical registries identified 24 , 157 people with autism and 49 , 348 with ADHD in Sweden between 1987 and 2010 . Sundquist et al . discovered that autism and ADHD were more common in individuals who had a parent with a history of an alcohol use disorder than in those whose parents had no history of an alcohol use disorder . There was also an even greater risk of either condition if the parent had been diagnosed with an alcohol use problem before the birth of the child . Adopted children who had a biological parent with an alcohol use disorder were at a greater risk of autism and ADHD than those whose adoptive parent had an alcohol use disorder . However , as very few adopted parents were diagnosed with an alcohol use problem , it is important to be cautious about drawing firm conclusions from this observation . Sundquist et al . estimate that around 4% of autism cases and 11% of ADHD cases could be avoided if parents abstained from heavy alcohol consumption . Though these findings are consistent with parents with an alcohol use disorder being more likely to pass on mutations to their children , there are also other possible explanations . As such , further research examining the underlying cause is still needed . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"epidemiology",
"and",
"global",
"health"
] | 2014 | Autism and attention-deficit/hyperactivity disorder among individuals with a family history of alcohol use disorders |
Defensive behaviors reflect underlying emotion states , such as fear . The hypothalamus plays a role in such behaviors , but prevailing textbook views depict it as an effector of upstream emotion centers , such as the amygdala , rather than as an emotion center itself . We used optogenetic manipulations to probe the function of a specific hypothalamic cell type that mediates innate defensive responses . These neurons are sufficient to drive multiple defensive actions , and required for defensive behaviors in diverse contexts . The behavioral consequences of activating these neurons , moreover , exhibit properties characteristic of emotion states in general , including scalability , ( negative ) valence , generalization and persistence . Importantly , these neurons can also condition learned defensive behavior , further refuting long-standing claims that the hypothalamus is unable to support emotional learning and therefore is not an emotion center . These data indicate that the hypothalamus plays an integral role to instantiate emotion states , and is not simply a passive effector of upstream emotion centers .
Across the animal kingdom , appropriate defensive behavior is key to survival . Accordingly , it is not surprising that the brain has evolved multiple circuits to control such behaviors . For example , studies in rodents have led to the conclusion that there are multiple , anatomically distinct pathways controlling learned ( conditioned ) and innate defensive responses ( reviewed in [Rosen , 2004; LeDoux , 2012] but see [Cezario et al . , 2008; Martinez et al . , 2011; Gross and Canteras , 2012] ) . In these pathways , sensory inputs converge on the amygdala , whose output is relayed by the hypothalamus to downstream structures that control the defensive response . However , different subdivisions of the amygdala and hypothalamus are thought to control learned vs innate responses , in a parallel manner ( Figure 1A; reviewed in [Davis , 1992; Fanselow , 1994; LeDoux , 1995 , 2000; Gallagher and Chiba , 1996; Fanselow and LeDoux , 1999; Canteras , 2002; Rosen , 2004; Swanson , 2005; Gross and Canteras , 2012; LeDoux , 2012; Saper and Lowell , 2014] ) . 10 . 7554/eLife . 06633 . 003Figure 1 . Characterization of SF1+ neurons and their optogenetic activation . ( A ) Schematic illustrating brain circuits involved in defensive behaviors . ( B ) Coronal section of the mouse brain showing the location of VMHdm ( Allen Brain Atlas ) . VMH is indicated by the blue outline . ( C–E ) Representative images of the VMH in a wild type mouse showing double-label immunostaining for SF1 ( green ) and progesterone receptor ( PR ) , a marker of VMHvl neurons involved in social behaviors ( red ) . ( F ) Percentage of cells in VMHdm/c ( white bars ) and VMHvl ( black bars ) that are SF1+ or PR+ . n = 3 animals for each condition . ( G–I ) Representative images of VMH from an SF1-Cre transgenic mouse injected in VMH with a Cre-dependent AAV encoding mCherry ( red ) and immunolabeled with anti-SF1 antibody ( green ) . ( J ) Percentage of overlap between VMHdm/c SF1+ cells and mCherry . n = 3 animals for each condition . ( K–M ) Representative images of VMH as in ( G–I ) , double labeled for mCherry+ ( red ) and PR+ cells ( green ) . ( N ) Percentage of total neurons that are mCherry+ in VMHdm/c and VMHvl ( defined by domain of PR expression ) . n = 3 animals for each condition . ( O ) Schematic illustrating preparation for whole-cell patch clamp recordings of SF1+ neurons . ( P ) Representative photomicrograph of ChR2-eYFP-expressing ( SF1+ ) cells ( red arrow ) patched for recording; DIC , differential interference contrast . ( Q ) Photostimulation-evoked spiking in neurons recorded as in ( P ) . ( R ) Percent spike fidelity in ( Q ) . n = 7 cells . ( S ) Schematic for in vivo electrophysiological response recordings from VMHdm/c in mice expressing ChR2 in SF1+ neurons . ( T ) Time-course of mean firing rate change in vivo in response to photostimulation . n = 6 units . ( U ) Average number of Fos+ neurons per section of VMHdm/c from photostimulated mice expressing ChR2 in SF1+ neurons . Control non-stimulated contralateral side within each animal . n = 4 animals for each condition . Values are represented as mean ± SEM . *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 00310 . 7554/eLife . 06633 . 004Figure 1—figure supplement 1 . Projection profile of SF1+ ( Nr5a1+ ) VMHdm/c neurons . Yellow fibers are a computational representation of the projections of SF1+ neurons from VMHdm , revealed by the expression of a Cre-dependent GFP AAV reporter virus ( Oh et al . , 2014 ) sterotaxically injected into the VMHdm of NR5a1-Cre transgenic mice . Data were collected using serial two-photon tomography ( Ragan et al . , 2012 ) and deposited in the Allen Mouse Brain Connectivity Atlas ( http://connectivity . brain-map . org ) . Source experiment: Nr5a1-Cre:VMH-LHA , Experiment 114290225 ( http://connectivity . brain-map . org/ ? searchMode=source&sourceDomain=693&primaryStructureOnly=true&transgenicLines=177839331&initImage=TWO_PHOTON&experimentCoordinates=7100 , 6800 , 6200&experiment=114290225 ) . Data were imported into Brain Explorer 2 ( http://mouse . brain-map . org/static/brainexplorer ) , for computational reconstruction in 3D . Data were filtered to show the highest-density/highest intensity projections . Projection targets ( green ) and associated coronal sections from the Allen Brain Atlas ( http://connectivity . brain-map . org/ ? searchMode=source&sourceDomain=693&primaryStructureOnly=true&transgenicLines=177839331&initImage=TWO_PHOTON&experimentCoordinates=7100 , 6800 , 6200&experiment=114290225 ) were identified by sampling terminals in high-density target sites , following which a volumetric rendering of relevant target structure was computationally superimposed on the projection map . A subset of the highest-density projection targets is illustrated . Individual panels are shown in slightly different 3D orientations to more clearly illustrate the pattern of projections to the indicated structure . The relative orientation of the entire brain in each panel can be viewed in the 3D volumetric model illustrated in the top right corner of each panel . Upper left panel depicts the VMH ( injection site ) . Remaining panels depict the following projection sites: AHN , BNST , LHA , MeA , PAG . These illustrations are presented in accordance with AIBS policy and citation guidelines ( see http://www . alleninstitute . org/terms-of-use/ and http://www . alleninstitute . org/citation-policy/ ) . These projection profiles can be viewed in 3D here in these two supplemental Videos ( Video 9 , 10 ) . Projections from Nr5a1+ neurons in VMH to the following structures are labeled: AHN ( orange ) , CeA ( blue-grey ) , CoA ( turquoise ) , LHA ( red ) , MeA ( pale blue ) , PAG ( violet ) , LA ( green ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 004 In mammals , at least , defensive behaviors reflect internal emotion states ( Darwin , 1872 ) , which are subjectively perceived by humans as ‘fear’ or ‘anxiety’ ( Adolphs , 2010 , 2013; LeDoux , 2012 ) . A large body of evidence has established the amygdala , principally the lateral ( LA ) , basolateral ( BLA ) and central ( CEA ) subdivisions ( Figure 1A ) , as a brain region that plays a central role in the implementation of emotion states , based on its involvement in conditioned fear ( reviewed in [Davis , 1992; Fanselow , 1994; LeDoux , 1995 , 1996 , 2000] ) ; for convenience we will henceforth use the shorthand term ‘emotion center’ to refer to such regions . In contrast , the hypothalamus is viewed primarily as a relay between the output of the amygdala , and downstream structures that generate observable behavioral , autonomic and endocrine components of a conditioned defensive response ( Davis , 1992; LeDoux , 1995 , 2000 , 2012; LeDoux and Damasio , 2013 ) . Because the identification of emotion centers has been rooted in their ability to mediate emotional learning ( LeDoux , 1996; Panksepp , 1998 , 2011b ) , it has been challenging to ascertain whether innate defensive behaviors also reflect underlying emotion states , and therefore to investigate whether structures that mediate these behaviors , such as the medial hypothalamus ( Figure 1A; reviewed in [Canteras , 2002; Rosen , 2004; Swanson , 2005; Sternson , 2013; Saper and Lowell , 2014] ) serve as emotion centers . Indeed , classical studies reporting that electrical stimulation of the hypothalamus is unable to condition learned defensive responses ( Masserman , 1941; Wada and Matsuda , 1970 ) have been used as evidence that the hypothalamus is not itself an emotion center , a view reflected in contemporary textbooks ( LeDoux and Damasio , 2013 ) . Instead , it has been assumed , by analogy to circuits mediating conditioned defensive responses , that emotion centers for innate defensive behaviors would be located in the medial amygdala ( MEA ) , and that downstream hypothalamic targets would similarly serve as passive relays for amygdala output ( Gross and Canteras , 2012; LeDoux , 2012; LeDoux and Damasio , 2013 ) . Experimental testing of this assumption , however , has been hindered by the lack of more general criteria to identify and study emotion states in systems mediating unlearned defensive responses . We have recently proposed that emotion states have several key properties that generalize across emotions and species . These properties include scalability ( the magnitude and/or nature of the behavioral response varies with the level of arousal or intensity of the associated internal state ) , valence ( positive or negative ) , generalization ( a given state can be induced by multiple stimuli , and can control multiple behavioral responses ) and persistence: they endure long after a threat is no longer present ( Russell , 2003; Posner et al . , 2005; Anderson and Adolphs , 2014 ) . Furthermore we argue , in line with Darwin ( Darwin , 1872 ) , Cannon ( Cannon , 1927 ) and others ( Panksepp , 1998 , 2011b ) , that these emotion states play a causative role in controlling behavior ( Anderson and Adolphs , 2014 ) . If one accepts this premise , then behaviors that exhibit the general properties described above can be taken as evidence of an underlying emotion state with similar properties . This broader and more general view of emotion states ( Anderson and Adolphs , 2014 ) , together with the availability of genetically based tools for cell type-specific manipulation of neuronal function ( Luo et al . , 2008; Yizhar et al . , 2011; Tye and Deisseroth , 2012 ) , provides an opportunity to revisit the role of hypothalamic neurons in controlling emotion states . The application of such tools in turn requires the identification of molecular markers for the cell types of interest . Recently , Silva et al . ( 2013 ) reported that pharmacogenetic silencing of neurons in the ventromedial hypothalamus , dorsomedial/central region ( VMHdm/c; Figure 1B ) , which express the nuclear co-receptor Nr5a1 ( also called SF1 ) ( Dhillon et al . , 2006 ) , caused a reduction in defensive responses to a predator , but not to other types of threats , such as an aggressive conspecific or a footshock ( Silva et al . , 2013 ) . Here we have used time-resolved optogenetic gain-of-function manipulations of SF1+ neurons , as well as cell-specific ablation , to investigate their role in defensive behaviors and associated emotion states . We demonstrate that direct activation of these neurons , a manipulation that anatomically bypasses amygdala input , is sufficient to evoke multiple defensive behaviors , whose collective properties are consistent with the induction of an underlying defensive emotion state ( Anderson and Adolphs , 2014 ) . Ablation of SF1+ neurons , moreover , attenuates defensive behaviors in a variety of contexts . Finally , we show that SF1+ neurons can condition learned defensive responses to initially neutral contextual cues , further refuting earlier claims to the contrary ( Masserman , 1941; Wada and Matsuda , 1970 ) . Together these findings suggest that SF1+ neurons contribute directly and causally to a defensive internal emotion state .
Neurons that express the gene Nuclear receptor subfamily 5 , group a ( Nr5a1 ) , which is also referred to as Steroidogenic factor 1 ( SF1 ) ( Dhillon et al . , 2006; Silva et al . , 2013 ) constitute about 60% of cells in VMHdm/c , and are not found in other hypothalamic or amygdalar regions ( www . brain-map . org , Nr5a1 in situ hybridization data ) . Double-labeling indicated that these neurons are essentially non-overlapping with subjacent VMHvl neurons mediating social behaviors such as aggression ( Yang et al . , 2013; Lee et al . , 2014 ) ( Figure 1C–F ) . To express different effectors in these neurons , we obtained a BAC transgenic mouse line that expresses Cre-recombinase under SF1 regulatory elements ( SF1-Cre ) ( Dhillon et al . , 2006 ) . We validated Cre-specific recombination in SF1+ neurons by injecting stereotaxically , into the VMHdm/c of SF1-Cre mice , an adeno-associated virus ( AAV ) expressing a Cre-dependent mCherry reporter , and double labeling with an anti-SF1 antibody ( Figure 1G–J ) . This analysis indicated that 80% of mCherry-expressing neurons were SF1+ , while little expression was observed in VMHvl ( Figure 1K–N ) . In order to optogenetically manipulate SF1+ cells in VMHdm/c , SF1-Cre transgenic mice were infected with a Cre-dependent adeno-associated virus 2 ( AAV2 ) containing an EF1α promoter-driven channelrhodopsin-2 ( ChR2 H134R ) fused to eYFP ( AAV-DIO-ChR2-eYFP ) ( Boyden et al . , 2005; Aravanis et al . , 2007 ) . We characterized the physiological response of SF1+ neurons to optogenetic stimulation using patch clamp recordings in acute VMH slices ( Figure 1O–R ) , as well as by in vivo extracellular recordings ( Figure 1S–T ) . In both cases , time-locked spiking was evoked by photostimulation , with 100% spike fidelity maintained up to a stimulation frequency of 20 Hz ( Figure 1Q , R ) . Additionally , we observed strong Fos induction in the VMHdm/c region following photostimulation , providing further evidence of activation in vivo ( Figure 1U ) . When threatened , animals will display species-specific defensive reactions ( Bolles , 1970; Blanchard et al . , 1990 , 1998 , 2005; Fanselow , 1994 ) , such as freezing . Therefore , we initially asked whether freezing could be triggered by optogenetic stimulation of SF1+ neurons . We found that when a 10-s blue light stimulation ( 20 Hz , 20 ms pulse width ) was administered in the animals' home cage ( Figure 2A ) , ChR2 virus-injected mice exhibited a short-latency ( 0 . 45 ± 0 . 09 s ) freezing response ( Figure 2B–E , J; Video 1 ) . Control eYFP mice did not show any changes in behavior ( Figure 2E ) . 10 . 7554/eLife . 06633 . 005Figure 2 . Optogenetic stimulation of SF1+ neurons induces freezing and/or activity bursts . ( A ) Representative tracking traces ( red ) of SF1-ChR2 expressing mice before ( ‘Prelight’ ) or during ( ‘Light’ ) optogenetic stimulation . Red dot in lower image reflects immobility of animal . ( B ) Representative velocity trace displaying light-elicited freezing ( arrow ) in a ChR2 mouse . Blue shading represents period of photostimulation . Inset , expanded view of region in dashed box . ( C ) Percentage of photostimulation trials evoking freezing . ( D ) Percentage of time spent freezing during photostimulation averaged across trials . ( E ) Percentage of ChR2-expressing or control eYFP-expressing animals showing photostimulation-evoked freezing behavior . n = 17–18 animals for each group . ( F ) Representative tracking traces of a SF1-ChR2 expressing mouse ( ‘Prelight’ ) or during ( ‘Light’ ) optogenetic stimulation . Wider spacing between points in ‘Light’ indicates higher velocity . ( G ) Representative velocity trace displaying light-induced activity burst behavior in an SF1-ChR2 mouse . Note period of freezing prior to activity burst . Inset , expanded view of region in dashed box . ( H ) Percentage of stimulation trials evoking activity bursts following freezing . ( I ) Average velocity during activity burst period . ( J ) Average onset latency for freezing vs activity burst . n = 9 animals for each condition . Values are displayed as mean ± SEM . ****p < 0 . 0001; ***p < 0 . 001; **p < 0 . 01; *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 00510 . 7554/eLife . 06633 . 006Figure 2—figure supplement 1 . ChR2-eYFP quantification in the VMHdm/c of freezing only and freezing + activity burst groups . ( A ) Representative images of ChR2-eYFP expression in the VMHdm/c for mice that showed freezing only ( top ) or freezing followed by an activity burst ( bottom ) . ( B ) Average pixel value of native ChR2-eYFP expression in the VMHdm/c in the freezing only group and freezing + activity burst group . n = 5–6 animals for each condition . Values are displayed as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 00610 . 7554/eLife . 06633 . 007Video 1 . ChR2-induced freezing . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 007 In about 50% of animals , we observed that towards the end of the photostimulation period freezing behavior was followed by an activity burst , defined as a dramatic and sharply delimited episode of undirected , high velocity ( >30 cm/s ) movement often including vigorous jumping ( Figure 2F–J; Video 2 ) . This activity burst was similar to activity bursts observed in rodents exposed to a footshock ( Fanselow , 1982; Kiernan et al . , 1995 ) and is considered part of the repertoire of rodent defensive behaviors , presumably evoked in nature by a high intensity and/or proximate threat ( Anisman and Waller , 1973; Bolles and Riley , 1973; Fanselow and Lester , 1988; Fanselow , 1994 ) . Importantly , such activity bursts , when they occurred , were exhibited following a sustained ( 8 . 07 ± 0 . 39 s ) period of freezing ( Figure 2G , J ) . The activity burst terminated upon photostimulation offset ( offset latency: 0 . 31 ± 0 . 04 s ) . Activity bursts were not observed in photostimulated control mice expressing eYFP in SF1+ neurons ( data not shown ) , indicating that this behavior is not a consequence of heating the brain during the photostimulation trial . 10 . 7554/eLife . 06633 . 008Video 2 . ChR2-induced freezing and activity burst . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 008 We investigated the reason for this variability in activity burst behavior . Histological quantification of ChR2 expression revealed that mice showing activity burst responses exhibited a trend towards slightly higher levels of ChR2 expression in VMHdm/c , compared to animals showing freezing only ( Figure 2—figure supplement 1 ) . While this trend did not reach significance , it prompted us to investigate whether the level or intensity of SF1+ neuronal activation might influence the type of optogenetically evoked behavior . To directly test whether activity bursts required a higher level of SF1+ neuron activation , we utilized SF1-Cre animals that were bilaterally injected with Cre-dependent AAV ChR2 and implanted with bilateral ferrule optic fibers , which enabled us to take advantage of the ability to independently stimulate one , the other , or both fibers within the same animal ( Figure 3A ) . Assuming similar levels of ChR2 expression in each hemisphere , bilateral photostimulation should activate approximately twice as many SF1+ neurons as unilateral photostimulation ( on either side ) in the same animal . Indeed , episodes of freezing followed by activity bursts were only observed using bilateral stimulation ( Figure 3B ) . Unilateral stimulation under these conditions evoked freezing but no ensuing activity bursts . These data suggest that activity burst behavior requires activation of more SF1+ neurons than does freezing . 10 . 7554/eLife . 06633 . 009Figure 3 . Optogenetic stimulation of SF1+ neurons induces freezing and/or activity bursts depending on strength and duration of photostimulation . ( A ) Schema illustrating unilateral vs bilateral optogenetic stimulation . Each mouse was implanted with bilateral optic fibers , and stimulation was delivered either to one side ( ‘unilateral #1’ ) , the contralateral side ( ‘unilateral #2’ ) or to both sides ( ‘bilateral’ ) . ( B ) Percent trials evoking freezing only ( white ) or freezing followed by an activity burst ( black ) in unilateral vs bilaterally stimulated ChR2 mice . **p < 0 . 01; Two-Way ANOVA , Bonferroni correction . n = 2 animals for each condition , each animal was stimulated either through one or the other of the two optic fibers ( ‘Unilateral #1 , Unilateral #2’ ) , or through both ( ‘Bilateral’ ) . ( C ) Threshold stimulation intensity required to generate freezing alone or freezing followed by an activity burst , during the photostimulation period . n = 9 per condition . ( D ) Percentage of trials evoking freezing alone , or freezing followed by an activity burst , at respective stimulation intensities . **p < 0 . 0001; Two-Way ANOVA , Bonferroni correction . ( E ) Representative images of Fos+ ( red ) and SF1+ ( green ) neurons from animals exhibiting optogenetically induced freezing alone ( upper ) , or freezing followed by an activity burst . Last column is higher magnification view of boxed area in adjacent ‘Overlay’ column . Arrow indicates cells double labeled for Fos+ ( red ) and SF1+ ( green ) ( F ) Percentage of SF1+ neurons that are Fos+ in eYFP ( white ) or ChR2 ( black ) mice following photostimulation trials eliciting freezing alone ( ‘Freezing’ ) or freezing followed by an activity burst ( ‘Activity Burst’ ) . n = 3–5 mice per condition . ( G ) Percentage of trials evoking freezing alone , or freezing followed by an activity burst , in response to different photostimulation frequencies . n = 5 animals per condition . ( H ) Onset latency for activity burst as a function of increasing light intensity ( x-axis ) and frequency ( colored bars ) . Arrow indicates condition that elicited activity burst without prior freezing . ( I ) Representative velocity trace displaying light-induced activity burst without preceding freezing . Arrow indicates expanded trace from boxed region . n = 2 animals for each condition . Values are displayed as mean ± SEM . ****p < 0 . 0001; ***p < 0 . 001; **p < 0 . 01; *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 009 To investigate further whether freezing required a lower level of SF1 neuron activation than did activity bursts , we systematically varied the intensity or frequency of optogenetic activation using bilateral photostimulation . The threshold light intensity required to evoke an activity burst was ∼fivefold higher than that required for freezing ( Figure 3C–D ) . Our previous studies of optogenetically evoked social behavior in VMHvl Esr1+ neurons indicated that more active neurons are recruited with increasing light intensity ( Lee et al . , 2014 ) . Consistent with that conclusion , Fos was expressed in a significantly higher proportion of SF1+ neurons in animals that exhibited activity bursts ( light intensity 5 . 25 mW/mm2 ) , compared to those exhibiting only freezing ( 0 . 65 mW/mm2; Figure 3E–F ) . Activity bursts following freezing could also be elicited using a higher frequency of photostimulation at a fixed light intensity ( Figure 3G ) , suggesting that increasing the average spiking rate among these neurons can also shift the behavioral output from freezing towards the activity burst . The observation that activity bursts are typically observed after several seconds of freezing ( Figure 3H–I ) raised the question of whether freezing behavior per se was a prerequisite for an activity burst . Therefore , we investigated whether certain stimulation conditions could elicit an activity burst without observable prior freezing . To do so , we used a more penetrant and highly expressing viral serotype , AAV5 , for delivery of ChR2 to SF1+ neurons ( Aschauer et al . , 2013 ) . Systematic manipulation of stimulation parameters yielded a high-intensity , high-frequency condition ( 10 . 5 mW/mm2 and 100 Hz ) that evoked a short-latency activity burst following stimulation offset , without prior freezing ( Figure 3H–I , arrow ) . During the ramp-up to such a condition , the latency to the onset of the activity burst during freezing gradually decreased . Taken together , these data suggest that SF1+ neurons can trigger either freezing and/or activity burst behavior , depending on the intensity and duration of optogenetic activation . The undirected nature of the defensive responses evoked by photostimulation of SF1+ neurons in the animals' home cage left open the question of whether activation of these cells can promote avoidance or withdrawal . To investigate this question , we tested whether photostimulation of SF1+ neurons was sufficient to generate real-time place aversion ( RTPA ) ( Stamatakis and Stuber , 2012; Kim et al . , 2013 ) . Mice expressing ChR2 in SF1+ neurons were randomly placed on one side of a contextually identical two-chamber place preference box ( Stamatakis and Stuber , 2012 ) ( Figure 4A ) . Photostimulation was delivered using a manual closed-loop protocol: the laser was switched on by the observer as soon as the mouse spontaneously entered the side opposite the one in which he had initially been placed; stimulation was continued until the animal moved to the non-stimulated side , at which point the laser was switched off . This stimulation regime was carried out over 20 min . Light pulses were delivered at low intensity ( 0 . 01 mW/mm2 ) , below the threshold required to elicit robust freezing or activity bursts . 10 . 7554/eLife . 06633 . 010Figure 4 . Optogenetic stimulation of SF1+ neurons induces aversion and interrupts ongoing consummatory behaviors . ( A ) Representative tracking traces of ChR2 mouse ( top ) and eYFP control mouse ( bottom ) in a real-time place avoidance assay ( RTPA ) . Photostimulation ( blue bar ) was delivered in a manual closed-loop manner depending on the animal's behavior ( see text ) . ( B ) Percentage of total time ( 20 min ) spent in stimulated side during 20-min trial . ( C ) Average latency to withdraw from the stimulated side . ( D ) Average velocity to enter or exit the stimulated side for the first ( left ) vs last trial ( right ) . ( E ) Latency to withdraw from the stimulated side for the first vs last trials . n = 6–7 animals for each condition in D and E . ( F ) Sample video still frames taken from consummatory behavioral assays . ( G ) Percentage of indicated behavior episodes terminated by light stimulation during the behavior within 6 s of photostimulation onset . ( H ) Latency to terminate respective consummatory behavior during photostimulation . n = 4–6 mice per condition . Values are displayed as mean ± SEM . ****p < 0 . 001; ***p < 0 . 001; **p < 0 . 01; *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 010 During the stimulation period , ChR2-expressing mice spent significantly less time on the stimulated side of the chamber ( Figure 4B ) , and withdrew from that side with shorter latency ( Figure 4C ) and higher velocity ( Figure 4D ) than did photostimulated eYFP-expressing control mice . Interestingly , the latency to withdraw from the stimulated side decreased significantly by the last trial for ChR2-injected mice compared to controls ( Figure 4E; Video 3 , 4 ) , suggesting that sensitization of the withdrawal response occurred with repeated exposures to the stimulus . However , mice did not exhibit progressively longer latencies to re-enter the stimulated side across trials , suggesting that under these conditions no lasting association between photostimulation and the chamber was formed ( but see below ) . 10 . 7554/eLife . 06633 . 011Video 3 . eYFP control in RTPA . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 01110 . 7554/eLife . 06633 . 012Video 4 . ChR2–induced withdrawal in RTPA . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 012 These results demonstrate that weak activation of SF1+ neurons can produce withdrawal from a context distinct from the home cage . This suggests that SF1+ neuron activation has a negative valence , in the behavioral sense of promoting avoidance rather than approach . In addition to simple avoidance , defensive states often actively inhibit positively valenced appetitive behaviors such as feeding or mating , which otherwise increase the animal's vulnerability to predation ( Petrovich et al . , 2009; Sukikara et al . , 2010 ) . Indeed , the sudden interruption of consummatory activity is often the most sensitive index of a perceived threat or anxiety state ( Estes and Skinner , 1941 ) . We thus investigated whether activation of SF1+ neurons would terminate ongoing social and consummatory behaviors , including mating , aggression and feeding ( Figure 4F–H ) . Indeed , all three behaviors were rapidly interrupted by modest photostimulation ( 1 . 0 mW/mm2 ) , in comparison to eYFP-expressing controls ( Figure 4G–H ) . Although continued photostimulation under these conditions eventually resulted in freezing , appetitive behaviors could also be interrupted using shorter periods of photostimulation , during which freezing had not yet occurred ( data not shown ) . These data indicate that SF1+ neuronal activation not only promotes adaptive defensive behaviors , but also can abrogate consummatory and social behaviors . One hallmark of an emotional reaction is that it often persists beyond the stimulus that evoked it ( Blanchard and Blanchard , 1989a , 1989b; Blanchard et al . , 1990; Adamec et al . , 2004; Anderson and Adolphs , 2014 ) . The preceding experiments indicate that SF1+ neuron activation can evoke avoidance , freezing and activity bursts in a stimulus-bound and time-locked manner . We next investigated whether such activation could also produce persistent effects that endured beyond the photostimulation period . Initial evidence of such persistence was observed during experiments in the home cage . We noticed that when optogenetic stimulation was terminated during an ongoing activity burst , the animal did not simply return to normal activity , but rather exhibited a period of freezing ( Figure 5A–C; Video 5 ) . Freezing typically occupied 40% of the time during a 10-s post-stimulation period ( Figure 5C ) , and lasted from ∼2–8 s depending on the animal ( Figure 5—figure supplement 1 ) . This observation suggests that photostimulation caused a period of residual defensive arousal , which gradually decayed over time . 10 . 7554/eLife . 06633 . 013Figure 5 . Stimulation of SF1+ neurons produces persistent defensive responses in multiple behavioral assays . ( A ) Representative velocity trace for ChR2 mouse displaying light-induced activity burst and post-light freezing behavior ( arrow ) . Note that freezing was observed during the photostimulation period prior to the activity burst ( inset , boxed region ) . See also Figure 2G . ( B ) Representative tracking traces ( red ) for pre-light vs post-light behavior in ChR2 mice . ( C ) Average percentage of time spent freezing during the 10-s pre-light and post-light bins averaged across trials and all mice . ( D ) Representative velocity trace for ChR2 mouse displaying freezing during photostimulation and elevated locomotion/jumping during the post-stimulation period . ( E ) Representative tracking traces for pre-light and post-light behavior in a ChR2 mouse . Vertical trace on left side of cage indicates jump ( ‘Postlight’ ) . ( F ) Average velocity during pre-light and post-light bins ( 10 s each ) averaged across trials and mice . n = 9 animals for each condition . ( G ) Representative tracking traces of ChR2 mouse ( top ) and eYFP control mouse ( bottom ) in the RTPA assay . The white dashed box marks the center area of the non-stimulated side used to index thigmotaxic behavior . Modified from image in Figure 4A , to illustrate thigmotaxic behavior . ( H ) Percentage of total assay period ( 20 min ) spent in the center of the non-stimulated side . ( I ) Average time spent in the center of the non-stimulated side , per individual entry . n = 6–7 animals for each condition . ( J ) Protocol to measure the resumption of mating behavior immediately following photostimulation of SF1+ neurons . ( K ) Representative raster plot illustrating mating episodes with ChR2 activation ( blue ) and control light ( yellow ) activation . The yellow wavelength does not activate ChR2 and is used as an internal control . ( L ) Latency to terminate mating following photostimulation with blue vs yellow light . ( M ) Latency to re-initiate mating after mating termination following blue vs yellow light . n = 7 animals for each condition . Values are displayed as mean ± SEM . **p < 0 . 01; *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 01310 . 7554/eLife . 06633 . 014Figure 5—figure supplement 1 . Persistent responses following light induced activity burst and freezing . ( A ) Average duration of freezing exhibited by individual ChR2 animals during post-stimulation period following stimulation-induced activity burst . n = 9 animals . ( B ) Trial-averaged velocity plot for ChR2 mice displaying pre-light , light-induced freezing , and post-light elevated locomotion behavior . In black , the trial-averaged velocity following stimulus offset . The solid pink line indicates the time-averaged velocity computed over the 10-s window prior to stimulus onset; the dashed pink line is + one standard deviation of the trial-averaged velocity in the same window . In red , we fit an exponential of the form a + b*exp ( −t/tau ) to the trial-averaged velocity starting at t = 0 ( time of stimulus offset ) . The fit exponential had a time constant of 25 . 8 s; the trial-averaged velocity took 47 . 03 s to decay to within one standard deviation of the baseline velocity prior to stimulation ( marked by red dot . ) ( n = 28 trials ) . Values are displayed as mean ± SEM . ( C ) Average number of jumps exhibited in the 20-s time bins pre-light and post-light for ChR2 mice . n = 9 animals for each condition . Values are displayed as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 01410 . 7554/eLife . 06633 . 015Figure 5—figure supplement 2 . Activation of SF1+ neurons produces an increase in neuroendocrine responding . ( A ) Illustration of testing for stimulation induced changes in corticosterone . ( B ) Optogenetic stimulation of SF1+ neurons increases serum corticosterone levels in ChR2 mice compared to eYFP controls . n = 11–12 animals for each condition . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 01510 . 7554/eLife . 06633 . 016Video 5 . ChR2-induced activity burst ands post-stimulation freezing . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 016 Alternatively , if photostimulation was terminated while the animals were freezing , it was often followed by increased locomotion that was directed around the perimeter of the cage ( average velocity ∼6 cm/s ) . This directed locomotion was qualitatively and quantitatively distinct from the undirected , high velocity ( ∼30 cm/s ) activity burst behavior described earlier . Moreover , some animals exhibiting such post-stimulation increases in locomotor activity also attempted transiently to jump out of the cage ( Figure 5D–F , Video 6; Figure 5—figure supplement 1 ) . This increase in locomotor velocity decayed slowly , over a period of 40–60 s ( Figure 5—figure supplement 1 ) , while the period of jumping lasted only 15–20 s ( Figure 5D , Figure 5—figure supplement 1 ) . This latter observation suggests that post-stimulation induced jumping may require a higher level of residual defensive arousal than does increased locomotion . 10 . 7554/eLife . 06633 . 017Video 6 . ChR2-induced freezing and post-stimulation elevated locomotion and jumping . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 017 We also investigated whether persistent effects of photostimulation could be observed in the real-time place avoidance assay . Notably , we observed that after ChR2 mice withdrew from the photostimulated side into the non-stimulated chamber , in contrast to control eYFP expressing animals , they avoided the center of the latter side , showing a dramatic increase in thigmotaxis ( preferential occupation of the perimeter of an open space ) ( Figure 5G–I ) . Thigmotaxis is considered a key measure of anxiety ( Simon et al . , 1994 ) . Thus , in both the home cage and in a different context , optogenetic stimulation of SF1+ neurons leads to persistent defensive or anxiety-like behaviors . We also investigated whether transient activation of SF1+ neurons caused a persistent inhibitory influence on appetitive behavior , by testing whether it increased the latency to re-initiate mating , following its interruption ( Figure 5J ) . We observed significantly longer latencies for ChR2 mice to re-initiate mating after SF1+ neurons were activated , compared to interleaved internal controls in which yellow rather than blue light was delivered ( Figure 5K–M ) . Thus , activation of SF1+ neurons causes persistent defensive responses in a variety of behavioral assays , suggesting that it engenders an associated internal defensive state . Consistent with this idea , optogenetic stimulation of these neurons produced nearly a twofold elevation in serum corticosterone ( Figure 5—figure supplement 2 ) . There has been conflicting evidence as to whether hypothalamic stimulation can support conditioning , considered by some to be a sine qua non property of an emotion state ( Masserman , 1941; Cohen et al . , 1957; Roberts , 1958; Wada and Matsuda , 1970; Panksepp , 2011a , 2011b ) . Therefore , we asked whether activation of SF1+ neurons could serve as a US for the formation of associative fear memories . To do this , we used a modified two-chamber , real-time conditioned place aversion assay ( RTPA ) ( Figure 4A ) , which we refer to simply as conditioned place aversion ( CPA ) . We introduced several modifications for the CPA assay . First , we distinguished the two chambers by lacing them with different odors , and providing distinct mesh flooring and different colored plastic wall inserts . Second , for training we used bilateral stimulation with high light intensity ( 5 . 25 mW/mm2 ) , which evoked freezing and/or activity bursts , rather than the low-level stimulation employed for the RTPA assay , which produced withdrawal , but no associative memory ( Figure 4 ) . The experimental design is illustrated in Figure 6A . Before conditioning , we performed a 5 min pre-training test to determine each animal's initial chamber preference . Photostimulation during training was then carried out on each animal's initially preferred side ( IPS ) , to determine whether training would overcome this initial preference . Training was performed over a 20-min period , using the manual closed-loop procedure described for the RTPA assays ( see above ) . It consisted of a series of photostimulation trials lasting 10 s , or until the animal withdrew from the stimulation chamber , whichever occurred first . Trials were repeated at 10-s intervals . If the animal withdrew from the training chamber during a photostimulation trial , the next trial was administered after the animal spontaneously re-entered the stimulation chamber . Over the 20-min training period , animals that initially exhibited freezing or activity bursts in the photostimulation chamber during the first few training trials eventually responded to photostimulation during later trials by rapid withdrawal from the IPS/training chamber . 10 . 7554/eLife . 06633 . 018Figure 6 . Activation of SF1+ neurons produces conditioned place avoidance learning and memory . ( A ) Protocol for conditioned place avoidance assay . Pre-training phase used to determine each animal's initially preferred side . ( B ) Representative tracking traces during pre-training , training , short-term memory ( STM ) and long-term memory ( LTM ) test phases as indicated in overlying schematic . Blue bar represents light delivery on the initially preferred side exhibited by each individual animal . ( C ) Percentage of total time for each phase ( see panel A ) spent in initially preferred side during pre-training , training ( blue shading ) , STM test , and LTM test . Animals spontaneously spent ∼75% of their time in one of the two chambers ( defined as the initially preferred side ) , during pre-training period . ( D ) Preference score ( percent total time spent in the initially preferred minus the initially non-preferred side ) , for each experimental phase . Negative value indicates that the animal spends more time in the initially non-preferred side , than in the initially preferred side . Preference score measures distribution of animals between the two chambers during a given testing phase . ( E ) Preference scores from ( D ) replotted for comparison of pre-training vs LTM scores for ChR2-expressing ( black bars ) vs eYFP control mice ( green bars ) . ( F ) Difference scores ( percent time spent in the initially preferred side during each respective phase minus the percent time spent in the same side during the pre-training phase ) for training , STM and LTM tests . Difference score measures change in time spent in initially preferred side during pre-training phase vs a given testing phase . Difference score during LTM test is ∼38% of that measured during STM test , indicating some retention of avoidance conditioning . ( G ) Mean velocity during each experimental phase revealed no differences . n = 7–10 animals for each condition . All values are displayed as mean ± SEM . ****p < 0 . 001; ***p < 0 . 001; **p < 0 . 01; *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 018 Following the 20-min training period , animals were observed for an additional 10-min period and the percentage of time they spent in each of the two chambers during this period was measured . For convenience , we refer to this post-stimulation test period operationally as a ‘short term memory’ ( STM ) test . Following this STM test , animals were returned to their home cage for 24 hr and then tested for their chamber preference once again; we refer to this latter test as the ‘long term memory’ ( LTM ) test . As expected , ChR2-expressing mice exhibited a dramatic avoidance of the photostimulation chamber during the training period , as compared to eYFP controls ( Figure 6B–C ) . Importantly , robust avoidance of the stimulated ( initially preferred ) chamber persisted during the 10-min STM test ( Figure 6B–C ) . This reversal of preference was reflected in a large negative value of the preference score , calculated as the percentage of time spent in the initially preferred chamber minus the percentage of time spent in the initially non-preferred chamber ( Figure 6D ) during the STM test ( or other phase of the experiment ) . It was also reflected in a significantly more negative ( in comparison to eYFP controls ) difference score , calculated as the percent of time spent in the IPS during the STM test minus the percentage of time spent in the IPS before training ( Figure 6F ) . ( For both metrics , a negative value is indicative of avoidance of the initially preferred chamber ) . Other features of behavior , such as average velocity , were unaffected by conditioning ( Figure 6G ) . Re-testing the animals 24 hr later ( LTM test ) revealed a modest but statistically significant retention of conditioning in ChR2-expressing animals , compared to controls ( Figure 6B–C ) as indicated by both the preference score and the difference score ( Figure 6D–F ) . Importantly , a comparison between pre-training and LTM preference scores ( Figure 6E , LTM ) revealed that the long-term reduction in preference for the IPS was robust ( 95% confidence interval: −72 . 30 to −18 . 14; Effect size: 1 . 2173; Power: 0 . 927 ) . In contrast , eYFP control mice showed no significant change in preference score between the pre-training phase and the LTM test ( Figure 6E ) . The retention of this conditioned response was confirmed by the significantly more negative LTM difference score , in comparison to eYFP controls ( Figure 6F ) . Based on a comparison of the difference scores between the STM and LTM tests ( Figure 6F ) , ChR2 mice showed ∼38% retention of their aversion memory . Thus , activation of SF1+ neurons can serve as a US for the formation of a conditioned place avoidance memory . The foregoing gain-of-function experiments raised the question of the context ( s ) in which the function of SF1+ neurons is normally required , and the precise role they play in such contexts . Previous studies have shown that the VMHdm/c is activated by predator cues ( Dielenberg et al . , 2001; Martinez et al . , 2008; Silva et al . , 2013; Stowers et al . , 2013 ) , and that SF1+ neurons are necessary for predator defensive responses , but not for other types of threat responses ( Silva et al . , 2013 ) . These data have led to the view that VMHdm/c primarily mediates innate defensive responses to predators ( Gross and Canteras , 2012; LeDoux , 2012 ) . Alternatively , SF1+ neurons may control a defensive state that is employed in a broader variety of contexts . To address this issue , we performed loss-of-function experiments to determine whether SF1+ neurons are required for a diverse set of threat-evoked defensive behaviors . Using a caspase-mediated cell ablation method ( Yang et al . , 2013 ) , SF1+ neurons in the VMHdm/c were selectively killed . Animals were bilaterally injected with a Cre-dependent AAV encoding activated caspase3 ( Figure 7A ) . We observed more than 90% elimination of SF1+ neurons in the VMH as compared to SF1-Cre negative littermates under similar conditions ( Figure 7B , D ) . Adjacent cells in the VMHvl and ARH neurons were unaffected by this ablation , as shown by double labeling with the VMHvl and ARH-specific marker Estrogen receptor 1a ( Esr1a ) ( Lee et al . , 2014 ) , confirming the restriction of caspase-mediated viral selective ablation to the SF1+ population in VMHdm/c ( Figure 7C , E , F ) . 10 . 7554/eLife . 06633 . 019Figure 7 . SF1+ neurons are necessary for predator aversion and conditional fear . ( A ) Schematic for bilateral injection of Cre-dependent apoptotic effector virus into the VMHdm/c of SF1-Cre mice . ( B ) Representative images of SF1+ neurons ( red ) in control and SF1+ ablated mice . ( C ) Representative images of Esr1a+ ( VMHvl ) neurons ( red ) in control and SF1+ ablated mice . ( D ) Number of total SF1+ neurons in VMHdm/c of control vs SF1+ ablated mice . ( E ) Number of total Esr1a+ neurons in VMHvl of control vs SF1+ ablated mice . ( F ) Number of total of Esr1a+ neurons in the arcuate nucleus ( ARH ) of control and SF1+ ablated mice . n = 4–6 animals , 3–4 sections per injection site . ( G ) Representative tracking traces of an SF1+ ablated mouse ( top ) and a control mouse ( bottom ) in a predator avoidance task . The rat predator ( constrained within a mesh cage ) is outlined in blue and the mouse in yellow . ( H ) Percentage of total test time ( 3 min ) spent by mice in each zone , with Z1 representing the closest zone and Z3 the furthest . ( I ) Average velocity across entire predator avoidance test . n = 5–7 animals for each condition . ( J ) Still video frame from the cued fear conditioning assay . Mice were given 5 tone-footshock pairings . ( K ) Average motion index ( au ) units during the two-minute , pre-conditioning baseline ( ‘BL’ ) period and during the activity burst elicited by each footshock-US . ( L ) Average percent time spent freezing during baseline ( BL period ) and during each 30 s tone-CS presentation preceding delivery of footshock . n = 9–10 animals for each condition . Values are displayed as mean ± SEM . ***p < 0 . 001; **p < 0 . 01; *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 019 Following successful ablation , we tested mice across a variety of behavioral paradigms involving defensive behavior . Initially , we tested whether this manipulation could replicate the effect of pharmacogenetic silencing of SF1+ neurons to reduce predator responses ( Silva et al . , 2013 ) , using a live rat as a stimulus ( Yang et al . , 2004; Blanchard et al . , 2005 ) . We found that males with ablated SF1+ neurons showed a striking deficit in predator avoidance behavior compared to controls , as demonstrated by a significant increase in the time spent in the zone closest to the rat ( Figure 7G , H , Z1 ) . In fact , some experimental animals even appeared to actively investigate the rat , hanging onto its mesh enclosure and attempting to poke their nose through the holes ( Video 7 , 8 ) . 10 . 7554/eLife . 06633 . 020Video 7 . Control ablated mouse in predator avoidance test . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 02010 . 7554/eLife . 06633 . 021Video 8 . SF1+ ablated mouse in predator avoidance test . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 021 Next , we examined the behavioral effects of SF1+ neuronal ablation on auditory cued fear conditioning ( Figure 7J–L ) ( Wehner and Radcliffe , 2004 ) . Males with ablated SF1+ neurons were fear conditioned using five tone-conditional stimulus ( CS ) presentations co-terminating with a footshock-US . These mice showed a significant reduction in the magnitude of their activity bursts during the footshock , for some ( but not all ) of the training trials ( Figure 7K ) . In addition , they exhibited a significant retardation in their ability to acquire conditional freezing to the tone-CS during training , as measured during the 30-s tone CS presentation prior to delivery of the footshock ( Figure 7L ) . However , tone CS-induced freezing eventually reached the same asymptotic level as that observed in controls . Lastly , we tested whether SF1+ neurons play an essential role in anxiety ( Gordon and Hen , 2004; Davis et al . , 2010 ) . To do this , we compared the behavior of males with ablated SF1+ neurons vs controls in three different anxiety assays: the elevated plus maze , novel object test , and the light–dark box ( Gordon and Hen , 2004 ) ( Figure 8A–C ) . To increase baseline anxiety and avoid ‘floor effects’ , animals tested in this assay were exposed to inescapable footshock 3 days prior to testing ( Maier and Watkins , 2005 ) . Ablation of SF1+ neurons significantly reduced measures of anxiety in each of the three different assays ( Figure 8D–F ) . Neither entries nor velocity was significantly affected ( Figure 8G–L ) , suggesting that this effect was not caused by differences in locomotor activity . Thus , SF1+ neurons are required for either the induction or expression of anxiety ( or both ) . 10 . 7554/eLife . 06633 . 022Figure 8 . SF1+ neurons are necessary for anxiety . ( A ) Representative tracking traces in the novel object test for control ( left ) and ablated ( right ) mice . Mice are outlined in yellow . The dashed white box marks the center of the chamber . ( B ) Representative tracking traces in the light–dark box test . Note higher density of traces in the light side for the Ablation condition , in comparison to the control . ( C ) Representative tracking traces in the elevated plus maze . ( D ) Percentage of time spent in the center of the novel object test . ( E ) Percentage of time spent in the light side of the light–dark box . ( F ) Percentage of time spent in the open arms of elevated plus maze . n = 5–7 animals for each condition . ( G ) Center entries in the novel object test for control ( black bars ) and SF1-ablated ( brown bars ) mice . ( H ) Stimulated-side entries in the light–dark box assay . ( I ) Total open arm entries in the elevated plus maze . ( J ) Average velocity in the novel object test . ( K ) Average velocity in the light–dark box . ( L ) Average velocity in the elevated plus maze . n = 5–7 animals for each condition . Values are displayed as mean ± SEM . ***p < 0 . 001; **p < 0 . 01; *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 02210 . 7554/eLife . 06633 . 023Figure 8—figure supplement 1 . SF1+ neurons are not necessary for defensive responses elicited by looming visual stimuli . ( A ) Still video frame image of looming assay box . The experimental mouse is labeled in yellow , a nest occupies the corner and a computer monitor for displaying the looming visual stimulus ( i . e . , a disk ) is mounted directly above the apparatus . Mice are placed in the looming assay apparatus and given a 10-min acclimation period followed by a series of 10 looming stimulus presentations in 10 s . ( B ) Percentage of animals showing escape and freezing behaviors in response to the looming stimulus . ( C ) Duration of freezing following escape into the hide . n = 9–10 animals for each condition . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 023 Finally , we investigated the requirement of SF1+ neurons for defensive responses triggered by a visual threat . When presented for the first time with an overhead looming shadow , mice exhibit rapid and robust escape or freezing responses ( Yilmaz and Meister , 2013 ) . Interestingly ablation of SF1+ neurons did not significantly diminish these behavioral responses to the shadow ( Figure 8—figure supplement 1 ) . Thus these neurons are dispensable for rapid responses to a visual threat .
The role of VMHdm/c neurons in the control of innate defensive responses has been investigated for over a quarter century ( reviewed in [Gross and Canteras , 2012; LeDoux , 2012] ) . Early lesion or other inhibitory manipulations of VMHdm produced conflicting results regarding the valence of its role in defensive behaviors ( Turner et al . , 1967; Grossman , 1970 , 1972; Weisman and Hamilton , 1972; Colpaert , 1975 ) . More recently , pharmacogenetic inhibition of SF1+ neurons mice was reported to cause a decrease in rat-evoked ‘defensive behavior’ , a metric combining stretch postures and freezing or immobility ( Silva et al . , 2013 ) . Here we used targeted genetic ablation ( Yang et al . , 2013 ) to address the necessity of these neurons for defensive behaviors . These loss-of-function experiments confirm a role for these neurons in predator avoidance , but also provide evidence of a broader role in defensive behaviors and associated emotion states . In contrast to our findings , Silva et al . ( 2013 ) reported that chemogenetic inhibition of SF1+ neurons using DREADDs selectively impaired defensive responses to a predator ( rat ) . The broader role of SF1+ neurons revealed by our loss-of-function experiments may reflect the fact that genetically targeted ablation ( Yang et al . , 2013 ) causes a more profound elimination of SF1+ neuronal function than does chemogenetic inhibition , whose extent and efficacy in vivo is difficult to assess . A caveat however , is that our genetically targeted ablation may have caused damage to neighboring populations of neurons , in a non cell-autonomous manner , and that such ‘collateral damage’ , if it occurred , could contribute to some of the phenotypes reported here . However , our demonstration that the number of Esr1+ cells in the adjacent VMHvl and neighboring ARH is quantitatively unaffected following ablation suggests that this possibility is less likely . Furthermore , the observed reduction in predator defense behavior caused by ablation of SF1+ neurons ( Figure 7 ) is similar to the effect of chemogenetic inhibition of these neurons reported earlier ( although the behavioral metrics were different ) ( Silva et al . , 2013 ) . Together , these considerations suggest that the ablation of SF1+ neurons is responsible for the behavioral phenotypes we observed . The role of VMHdm in defensive behavior has also been investigated previously using gain-of-function manipulations such as electrical or pharmacologic activation . In rats or non-human primates , such stimulation induced freezing and escape ( Lipp and Hunsperger , 1978; Lammers et al . , 1988; Silveira and Graeff , 1992; Freitas et al . , 2009 ) . However , the pharmacological methods ( Silveira and Graeff , 1992; Freitas et al . , 2009 ) lacked cellular specificity and high spatial resolution , while electrical stimulation ( Lipp and Hunsperger , 1978; Lammers et al . , 1988 ) could not exclude activation of fibers of passage . Moreover , loss- ( Silva et al . , 2013 ) and gain-of-function manipulations ( Lipp and Hunsperger , 1978; Lammers et al . , 1988; Silveira and Graeff , 1992; Freitas et al . , 2009 ) were reported in different studies from different laboratories , using different species and different assays , making direct comparisons difficult . Here we have performed both optogenetic activation , and targeted ablation , of a genetically defined subset of VMHdm/c neurons in mice , using a battery of behavioral assays including those where a predator was not present ( e . g . , anxiety assays ) . Our data argue that the function of SF1+ neurons is not restricted to predator defense per se , but rather that this population controls features of an internal defensive emotion state , which generalize across different contexts and different types of threats . Our results also argue against the more trivial interpretation that VMHdm/c serves exclusively as a permissive , sensory relay for predator-derived cues ( Papes et al . , 2010 ) —in essence , an ‘internal nose’—a possibility that could not be excluded from earlier loss-of-function studies ( Silva et al . , 2013 ) . However , our results do not exclude a role for SF1+ neurons in the transformation of sensory representations into an internal emotion or motivational state . Interestingly , we observed that ablation of VMHdm/c SF1+ neurons did not impair innate freezing or flight evoked by an overhead expanding shadow ( Yilmaz and Meister , 2013 ) . This negative result could reflect redundancy in the circuits that mediate the shadow response . Alternatively , SF1+ neurons may play no role in this paradigm , suggesting that they either mediate defensive responses to terrestrial but not to aerial threats , or that the visual response uses a specialized pathway , similar to the shadow-induced jump response in Drosophila ( Allen et al . , 2006 ) . Whatever the explanation , these neurons are not essential for all forms of predator defense . Changes in defensive behaviors during an encounter with a predator are often associated with a graded increase in the level of the underlying internal emotion or arousal state . These changes can be observed either as a quantitative increase in the amplitude or frequency of a given behavior ( e . g . , increased locomotor velocity ) , or as a qualitative shift in behavior ( e . g . , from freezing to flight ) ( Fanselow and Lester , 1988; Blanchard et al . , 1998 , 2001 , 2003a; Anderson and Adolphs , 2014 ) . Interestingly , the nature of the defensive behaviors evoked in this study depended on the intensity of the optogenetic stimulation: avoidance was evoked by low intensity stimulation , while freezing and the interruption of ongoing appetitive behavior required a higher intensity , and activity bursts yet more intense stimulation . By comparing unilateral vs bilateral stimulation of SF1+ neurons in the same animal , we demonstrated directly that activation of a larger number of SF1+ neurons was required to evoke an activity burst than to evoke freezing . Similarly , in the two-chamber assay , a higher level of photostimulation was required for associative memory formation , than simply to cause avoidance . Similar threshold-dependent changes in behavior have been observed during optogenetic stimulation of medial amygdala and hypothalamic cell types that mediate social interactions ( Hong et al . , 2014; Lee et al . , 2014 ) , suggesting that it may be a general property of some behavior control circuits . Whether this scalable control is achieved through an increase in ensemble size and activity within a homogenous population of SF1+ neurons , or reflects different subpopulations with different thresholds for activation ( Lee et al . , 2014 ) , will be an interesting topic for future study . Whatever the answer , the observation that a common circuit node can control multiple defensive behaviors , according to its level of activity , argues against alternative views invoking parallel processing models , in which anatomically distinct pathways control different types of behavioral responses depending on cues or contexts ( Fanselow , 1994; Mobbs et al . , 2007 ) . Interestingly , we observed that photostimulation conditions that initially evoked freezing were often followed , after a delay of several seconds , by activity bursts during the stimulation period . This observation suggests that the brain may be able to integrate the cumulative effects of SF1+ neuron activation over time , in a manner that changes the type of defensive behavioral output as different thresholds are reached . Such an integrative function is consistent with our observation that activation of these cells produces persistent behavioral effects , as persistent activity is a hallmark of neural integrators ( Major and Tank , 2004; Goldman et al . , 2007 ) . Alternatively , the transition from freezing to activity burst in our experiments might reflect a time-dependent inactivation or habituation of freezing neurons during photostimulation , which in turn releases from inhibition a second population that controls the activity burst in an antagonistic manner . Whatever the explanation , the ability of SF1+ neurons to integrate signals that change in their quality or intensity over time could allow an animal to express an appropriate behavioral response ( freezing , escape ) as a predator threat escalates , as encapsulated by ‘Predatory Imminence’ theories ( Fanselow and Lester , 1988; Blanchard and Blanchard , 1989b; Blanchard et al . , 1998 , 2003a; McNaughton and Corr , 2004 ) . The neural mechanisms underlying such integration and persistent activity , and whether they are instantiated in VMHdm/c or in a downstream target , remain to be investigated . The prevailing , textbook view that the amygdala is the central orchestrator of emotion states ( Kandel et al . , 2013 ) is rooted deeply in its capacity to mediate forms of emotional learning , such as fear conditioning ( LeDoux , 1995 , 2000; Gallagher and Chiba , 1996; Maren and Fanselow , 1996; Fanselow and LeDoux , 1999; LeDoux , 2003; Phelps and LeDoux , 2005; Pessoa and Adolphs , 2010 ) . However this criterion is more difficult to apply to circuits that mediate unlearned ( innate ) defensive behavior . Indeed , the failure of hypothalamic electrical stimulation to condition learned defensive responses has been used to argue that the hypothalamus is not itself an emotion center ( Masserman , 1941; Wada and Matsuda , 1970 ) , despite some evidence to the contrary ( Cohen et al . , 1957; Roberts , 1958 ) . Independent of learning , manipulations of VMHdm and other hypothalamic nuclei in rodents have been interpreted as evidence that these structures control innate ‘fear’ ( Gross and Canteras , 2012 ) , a conclusion consistent with the observation that electrical stimulation of this region in humans evoked anxiety and panic attacks ( Wilent et al . , 2010 , 2011 ) . However the attribution to animals of ‘fear’ , a subjective human experience , has recently been questioned ( LeDoux , 2014 ) , on the grounds that it can only be assessed by verbal report in humans ( Adolphs , 2013 ) . We have recently proposed objective criteria for identifying emotion states in animal models , based on general properties or features common to different emotions within a species , and to similar emotions across species ( Anderson and Adolphs , 2014 ) , and which are independent of anthropocentric attributions of human emotions such as ‘fear’ . These general properties include scalability , persistence , valence and generalization ( Russell , 2003; Posner et al . , 2005 ) . The ability to mediate emotional learning is but one facet of these general properties , and not necessarily an essential one . If one accepts this view , then structures or neurons whose activation can evoke behaviors exhibiting these collective properties are good candidates for implementing emotion states . The data presented here provide evidence that activation of SF1+ neurons in VMHdm/c is able to evoke defensive behaviors exhibiting the aforementioned general features of an underlying causal emotion state . To our knowledge , this study is the first to provide evidence of an emotion state in an animal model , using the set of objective and general criteria described above ( Anderson and Adolphs , 2014 ) . In addition , we find that optogenetic activation of SF1+ neurons in VMHdm/c can indeed serve as an unconditional stimulus ( US ) for associative learning , in a conditioned place avoidance assay . These data , together with earlier studies of conditioning in VMH ( Colpaert and Wiepkema , 1976; Santos et al . , 2008; Santos and Brandão , 2011 ) and associated hypothalamic nuclei ( Pavesi et al . , 2011 ) , provide definitive evidence against the view that the hypothalamus is not an emotion center ( Masserman , 1941; Wada and Matsuda , 1970 ) . Yet this perspective is still common in textbook views of emotion ( [LeDoux and Damasio , 2013] , in [Kandel et al . , 2013] ) , which place the amygdala as the central ‘orchestrator’ of emotion systems , and the hypothalamus as a motor effector or relay of amygdala output . The data presented here demonstrate that direct , optogenetic activation of a specific hypothalamic cell population , in a manner that anatomically bypasses the amygdala , can evoke a persistent , scalable and generalizable emotion state . These observations argue that the prevailing , ‘amygdalo-centric’ view of emotion systems should be expanded to include specific hypothalamic structures such as VMHdm , and its associated circuitry ( see below ) . While the VMHdm/c receives input from the anteriodorsal and posterioventral regions of the medial amygdala ( MEAad and MeApv ) ( Dong and Swanson , 2004 ) and the basomedial amygdala ( BMA ) ( Petrovich et al . , 2001 ) , recent data suggest that MeA functions primarily to encode sensory cues ( Bergan et al . , 2014 ) . If so , then the transformation of such sensory input into an internal emotion state may , arguably , be carried out primarily at the level of VMHdm , or other interconnected hypothalamic nuclei ( Risold et al . , 1994 ) , rather than in the amygdala itself . It should be noted the VMHdm also receives strong projections from the lateral parabrachial ( PB ) area , which transmits noxious stimuli ( Bester et al . , 1997 ) ; these projections may also provide sensory input to VMHdm important in the encoding of emotion states . That said , we cannot formally exclude the possibility that the effects of optogenetically stimulating SF1+ neurons are mediated by ascending ( feedback ) projections that activate the amygdala; in that case VMHdm/c would be ‘upstream’ , rather than ‘downstream’ , of the amygdala . However , high-resolution anatomical mapping of SF1+ neurons indicates that recurrent projections to amygdala nuclei are relatively weak ( Figure 1—figure supplement 1 and http//:connectivity . brain-map . org , VMH , Nr5a1-Cre experiments 114290225 and 182337561 , sections 64–82 ) . This issue could be addressed , in principle , by combining bilateral activation of SF1+ neurons with bilateral lesions of the amygdala . However such an experiment is challenging in mice because of the relatively small size of their brain , and the highly invasive nature of such an experiment . Thus , while descending input from the MEA is likely to contribute to VMHdm/c activation during defensive responses in an unmanipulated animal , our data show that one can experimentally bypass such amygdala input and evoke a persistent emotion state by direct activation of SF1+ neurons . VMHdm SF1+ neurons lie within a densely interconnected network of hypothalamic and midbrain nuclei ( Canteras , 2002; Gross and Canteras , 2012 ) . Therefore the ability of SF1+ neuronal activation to implement a persistent emotion state could be mediated by other nodes in this circuitry , rather than within VMHdm itself . VMHdm/c SF1+ neurons send projections to the BNST , AHN , lateral hypothalamus ( LHA ) , PMd , MeA and dorsal peri-aqueductal gray ( dPAG ) , as well as to other structures ( see Figure 1—figure supplement 1; Video 9 , 10 ) ( Canteras et al . , 1994 ) . Previous studies have shown that perturbations of some of these targets , including the dPAG or PMd , can influence some defensive behaviors ( Di Scala et al . , 1987; Di Scala and Sandner , 1989; Canteras et al . , 1997; Blanchard et al . , 2003b; Bittencourt et al . , 2005; Cezario et al . , 2008; Pagani and Rosen , 2009; Sukikara et al . , 2010; Pavesi et al . , 2011; Santos and Brandão , 2011; Kincheski et al . , 2012 ) . However , many of these earlier studies did not exclude a role for stimulation of fibers of passage , and lacked the cellular specificity and spatio-temporal resolution of the methods employed here . Furthermore , as in the case of VMHdm/c , the high degree of connectivity between these structures makes it difficult to ascribe specific functions to any individual node . 10 . 7554/eLife . 06633 . 024Video 9 . SF1+ projections . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 02410 . 7554/eLife . 06633 . 025Video 10 . SF1+ projections II . DOI: http://dx . doi . org/10 . 7554/eLife . 06633 . 025 Among potential downstream targets that may mediate the effects of SF1+ neuronal activation , the dPAG is a particularly noteworthy candidate . Activation of dPAG induces freezing and flight ( Brandao et al . , 1982 , 1986; Vianna et al . , 2001; Bittencourt et al . , 2005 ) , supports conditioning ( Di Scala et al . , 1987; Kincheski et al . , 2012 ) , and induces fear sensation in humans ( Amano et al . , 1982 ) . It may also play role in anxiety ( Gomes et al . , 2014 ) and interruption of other appetitive behaviors ( Sukikara et al . , 2010 ) . In preliminary experiments we have observed that direct optogenetic activation of vGlut2+ neurons in dPAG induces freezing and activity bursts ( data not shown ) . Consistent with this , while this manuscript was in its final production stages , Wang et al ( 2015 ) reported that activation of SF1+ projections to the dPAG evoked immobility . In contrast , activation of projections to the AHN evoked low-intensity escape responses . However , individual SF1+ neurons collateralized to both the dPAG and the AHN ( Wang et al . , 2015 ) , raising the question of how independent control of these different defensive behaviors can be achieved . Future work will clearly be required to resolve this issue . The results presented here characterize an important function for a specific , genetically defined hypothalamic cell type in promoting defensive behaviors , in a manner suggesting that these neurons induce or implement a persistent , scalable , generalizable , negatively valenced internal and causal emotion state . This state of apparent threat arousal or defensive motivation may share properties in common with the emotions that humans subjectively experience as ‘fear’ or ‘anxiety’ ( Phelps and LeDoux , 2005; LeDoux , 2012; Adolphs , 2013; Anderson and Adolphs , 2014 ) . While these results raise many important and unanswered mechanistic questions at the level of connectivity and neuronal activity dynamics , at the very least they should prompt a re-evaluation of the prevailing , ‘amygdalo-centric’ view of emotion control systems , by providing evidence that the hypothalamus is not simply a passive relay or effector of amygdala output , but can serve to implement a central emotion state itself .
These experiments were approved by the institutional animal care committee ( IACUC ) at the California Institute of Technology ( protocol number 1602 , 1600 and 1552 ) . SF1-Cre mice were provided by Dr Brad Lowell ( Dhillon et al . , 2006 ) and backcrossed to C57Bl/6N wildtype mice ( Charles River , Burlington , MA ) at the Caltech animal facility . Heterozygous male mice or their littermates aged 12–20 weeks were used for behavioral studies . Heterozygous females aged 8 weeks were used for slice electrophysiology experiments . Mice were maintained on a reversed , 14-hr light cycle and all experiments were conducted during the dark cycle . Long-Evans rats aged 12–16 weeks were ordered from Charles River for use in the predator exposure experiment . All procedures described here adhere to the NIH guidelines for animal research . AAV-EF1a-DIO-eYFP , AAV-EF1a-DIO-ChR2-eYFP , and AAV-EF1a-DIO-taCasp3-TEVp ( Yang et al . , 2013 ) were purchased from the University of North Carolina vector core facility . AAV-EF1a-DIO-mCherry ( Anthony et al . , 2014; Lee et al . , 2014 ) was constructed in house ( Dr Todd Anthony ) and packaged by the University of Pennsylvania vector core facility . Mice were stereotaxically injected with viruses as previously described ( Cai et al . , 2014 ) . Briefly , viruses were pressure injected ( Mico4Controller , World Precision Instruments; Nanojector II , Drummond Scientific ) unilaterally ( Figure 1O–U , Figure 4F–H and Figure 5J–M ) or bilaterally into the VMHdm/c using a pulled glass needle aimed at the VMHdm/c ( ML ± 0 . 5 , AP-4 . 65 , DV-5 . 5 ) following a high resolution atlas ( Aravanis et al . , 2007 ) . A total volume of 600 nl/site was injected at the rate of 100 μl/min . The needle was left in place for an additional 10 min to control for potential virus drag across the needle tract . A custom made bilateral ferrule fiber ( 200 μm in core diameter , Doric Lenses ) or a unilateral cannula ( 24 gauge , Plastics One ) was then placed 0 . 5 mm above the injection site . Fibers were cemented in place ( Metabond ) . Following surgery , mice were allowed to recover on a heat pad and thereafter closely monitored for an additional 5 days during which they received medicated water ( Septra and Motrin ) . Mice were single-housed for 4 weeks before commencing experiments to ensure surgical recovery and optimized Cre-mediated recombination . Animals were anaesthetized briefly using isoflurane to connect the fiberoptic cable to the unilateral cannula or bilateral ferrule . Mice were allowed to recover for 30 min in their home cage . They were then brought into an adjacent behavioral testing room for digital video capture of homecage behavior . The fiberoptic cable was then connected to a laser ( 473 nm for ChR2 stimulation and 593 nm for control stimulation , Shanghai laser ) using a bilateral commutator ( DoricLenses ) . A signal generator ( World Precision Instruments ) was used to control duration , frequency and pulse width of the light . A 20 Hz , 20 ms pulse width was used in all experiments except where mentioned otherwise . Laser intensity was calculated for a distance of 0 . 5 mm below the fiber tip . Testing for freezing and activity burst behavior in the homecage was comprised of a period of baseline behavior recorded in the homecage followed by a series of optogenetic stimulations . Each stimulation was 10 s in duration , except in the case of stimulation-induced activity bursting , where the laser was turned off immediately upon the production of an activity burst , regardless of whether this period comprised less than 10 s . Each animal was given six stimulation ‘trials’ per optogenetic condition , with at least an average inter-stimulation-interval of 90 s . Behavior during stimulations was averaged for data analysis . Freezing behavior was assessed by a complete lack of mobility except that required for respiration for 2 s or more using a custom designed behavioral scoring program in MATLAB ( Yang et al . , 2013 ) . An activity burst was defined as a sharp , random movement with high locomotion ( >20 cm/s of velocity , sustained for at least 1 s ) . Jumps were determined by assessing whether a mouse moved upwards with all four-legs off the ground . The total number of jumps occurring within the 20 s pre- , during , and post-light stimulation were calculated per mouse . Animals were rested for a week for subsequent tests . Fos induction in response to optogenetic stimulation was assessed in ChR2 SF1-Cre mice receiving blue light stimulation ( 473 nm , 20 Hz 20 ms , 10 s on and 10 s off for 20 min ) in their home cage . Following optogenetically-induced freezing and/or activity bursts , mice were kept in isolation and perfused 90 min later . Brains were extracted and harvested for subsequent sectioning and antibody staining . Interruption of aggression and mating behavior was tested using the resident intruder assay ( Hong et al . , 2014; Lee et al . , 2014 ) . Group housed , wildtype BALB/c males and females ( Charles River ) aged 12 weeks were used as intruders for aggression and mating testing , respectively . Females were selected for their receptivity beforehand to achieve robust baseline mating behavior . Resident males had at least 1 week of mating experience prior to surgery to increase their level of aggression ( Lee et al . , 2014 ) . Resident males that failed to exhibit aggression or mating behavior were excluded from analyses . To test for the ability to interrupt ongoing social behavior , mice were administered either blue ( 473 nm ) or control yellow ( 593 nm ) light ( 20 Hz , 20 ms ) once a behavior was underway . Light was delivered until the behavioral episode was terminated . The test was continued until at least seven blue activation and yellow control trials were recorded . The order of ChR2 and control activations were counter-balanced across animals . The order of aggression and mating tests were counter-balanced across animals . Behavior in the resident intruder assay was recorded with a video camera mounted in front of the homecage and manually scored by an observer blind to experimental conditions . Scoring was performed using a custom designed behavioral scoring program in MATLAB ( Yang et al . , 2013 ) . Following mating and aggression testing , mice were tested for interruption of feeding behavior as described previously ( Cai et al . , 2014 ) . Mice were food-deprived for 24-hr , placed into a novel cage , and presented with a food pellet . Feeding behavior was interrupted using a stimulation protocol identical to that used for mating and aggression interruption ( see above ) . Behavior was recorded using a video camera and subsequently scored by an observer blind to experimental conditions using a behavioral annotation software tool written in MatLab ( Yang et al . , 2013 ) . Real-time place avoidance ( RTPA ) was performed as described previously ( Stamatakis and Stuber , 2012 ) . The apparatus ( 100 × 50 × 25 cm; black pexiglass wall; white pexiglass floor ) was comprised of two identical sides that were connected by an opening ( 12 . 5 cm ) in the center . Animals were placed pseudo-randomly in one side of the chamber ( starting side was counterbalanced across mice ) and blue light ( 20 Hz , 20 ms , 0 . 01 mW/mm2 ) was delivered as soon as the mouse entered the alternate side of the apparatus by at least 50% of its body . Stimulation continued until the animal returned to the non-stimulated control side . The assay lasted a total of 20 min . Behavior during the session was recorded using a camera mounted above the apparatus and analyzed recording using Mediacruise recording software ( Canopus ) . Total time spent in each chamber , chamber entries , and latency to depart the chamber following stimulation using Ethovision . A custom-designed RTCPA apparatus was built for use in our conditioned place avoidance ( CPA ) assay . The apparatus measured 100 × 50 × 25 cm in dimensions . The two sides of the apparatus were made contextually distinct . One chamber side was covered with black plastic and fine mesh flooring while the other side was left white and had coarse mesh flooring . The different sides were also distinguished by odor ( 2 . 5% of Acetophenone or ethyl acetate ) . CPA testing was carried out over 2 days . Day 1 involved a 5-min pre-training session to habituate animals to the apparatus and to determine each mouse's place preference . Animals that showed more than 90% of preference during pre-training were excluded from the analysis . Each animal was pseudo-randomly placed in one side ( counterbalanced across mice ) . This was followed by a 20-min ‘conditioning’ session in which blue light stimulation ( 20 Hz 20 ms , 10 s on and 10 s off , 5 . 5 mW/mm2 ) was administered in the preferred side until the mouse returned to the non-preferred side . Following conditioning , animals were allowed to move freely for another 10 min without light stimulation to determine their post-stimulation preference . After 24 hr , mice were returned to the apparatus to test for long-term aversion memories . Mice were placed in the stimulated side in order to access the aversion memory associated with the context . As in the RTPA task , time spent in each side during the CPA assay was assessed using Ethovision . Mice with ablated SF1+ neurons in VMHdm/c were used to test for intact predator avoidance ( Blanchard et al . , 2005 ) . Predator rats weighting 300–500 gm were used to induce avoidance . A custom made testing apparatus measuring ( 36 × 18 × 40 cm ) was designed to be attached to a mouse's homecage . The test rat was confined to a mesh enclosure ( 16 × 11 × 15 cm ) and put on one side of the home cage . D-amphetamine ( 5 . 0 mg/kg , Sigma ) was injected ( i . p . ) 20 min prior testing to trigger uniform movements in the rat stimulus . Rats were lowered into the mesh enclosure and mouse behavior was assessed across a 3-min time period . In order to assess how much time a mouse spent close or far from the rat predator , the home cage was divided into three equal ‘zones’ , with Zone1 being closest to the rat and Zone 3 farthest . Time spent and frequency of entries into each zone was calculated using EthovisionXT software ( Noldus ) . SF1-ablated and control mice were placed in a conditioning chamber ( Med Associates ) and fear conditioned as previously described ( Haubensak et al . , 2010 ) . After 2 min of habituation ( baseline period , ‘BL’ ) , five training trials were delivered with an inter-trial interval of 1 min . Each trial consisted of a 85 dBA , 2k Hz tone for 30 s that co-terminated with a 2-s , 0 . 6 mA foot shock . Freezing and activity burst ( measured by motion index ) responses to the tone and shock , respectively , were analyzed using Video Freeze software ( Med Associates ) . Open field , novel object , light–dark box and elevated plus maze tests ( Anthony et al . , 2014; Cai et al . , 2014 ) were utilized to assess levels of anxiety in SF1-ablated mice . Mice were tested in the above-mentioned sequence of tests with at least a 4-hr rest period in between tests . The novel object test , which lasted 5 min , was done after the open field test . A stainless steel cup was placed at the center of the box . The center area for the novel object comprised 25% of total area . Time spent in the center as opposed to the borders of the apparatus was assessed . In the light–dark box , animals were initially placed on the light side of box and behavior was assessed across 10 min . In the elevated plus maze , animals were initially placed at the center of the maze and behavior was assessed across 10 min . Ethovision software was used to analyze time spent , entries , and velocities for each anxiety test . SF1-ablated mice were tested for behavior in a looming visual stimulus test , as described elsewhere ( Yilmaz and Meister , 2013 ) . Wild type littermate sibling mice were used as controls . Animals were placed in an open-top pexiglass box ( 48 × 48 × 30 cm ) . A triangular shaped nest ( 20 × 12 cm ) was placed in one corner . Recording using Nerovision software was done under illumination provided by Infrared LEDs ( Marubeni ) . After 10 min of habituation , a looming stimulus was provided from above when an animal was in the center . The stimulus of 0 . 5-s duration was repeated 10 times with an inter-stimulus interval of 0 . 5 s . Mice were given a post-stimulation period of 10 min . Behaviors were recorded using Nerovision software control or Mediacruise recording software ( Canopus ) . Annotation was carried out manually on a frame-by-frame basis by an experimenter blind to experimental conditions using a behavioral annotation software tool written in MatLab and/or using EthovisionXT . Corticosterone measurement was done as described previously ( Anthony et al . , 2014 ) . Mice were photostimulated with 473 nm light at 5 . 5 mW/mm2 , 20 Hz , 20 ms pulse with 10 s on and 10 s off for 30 min , immediately euthanized and decapitated for blood collection for corticosterone measurement using an immunoassay ( Enzo Life Sciences ) . Sectioning and immunostaining were done as described previously ( Haubensak et al . , 2010 ) . Mice were perfused using 4% PFA . Brains were cryoprotected ( 15% sucrose ) and frozen at −80°C until sectioning . Brains were sectioned on a cryostat ( Leica , Biosystems ) at either 30 μm thick ( for direct mounting onto slides ) or 60 μm thick ( for free-floating sections collected in a staining disc ) . The following antibodies were used: rabbit anti-SF1 antibody ( 1:500 , TransGenic ) , rabbit anti-SF1 antibody ( 1:500 , Upstate ) , goat anti-c-Fos ( Santa Cruz , 1: 500 ) , rabbit anti-Esr1a ( 1: 500 , Santa Cruz ) , mouse anti PR ( 1:500 , Thermoscientific ) , rabbit anti-GFP ( 1:500 , Invitrogen ) . The following secondary antibodies were used—donkey anti-goat IgG Alexa 546 ( 1:500 , Invitrogen ) , goat anti-rabbit IgG Alexa 488 ( 1:500 , Invitrogen ) , goat anti-rabbit IgG Alexa 546 ( 1:500 , Invitrogen ) , goat anti-rabbit IgG Alexa 633 ( 1:500 , Invitrogen ) . NeuroTrace fluorescent Nissl stains ( 1:200 , Invitrogen ) or DAPI ( 1:200 , Invitrogen ) was used to counterstain sections and label cell bodies . At least three representative coronal sections spaced equally along the AP axis were used for quantifications . Prism 5 ( GraphPad ) software was used for statistical analysis of behavioral and histological data . Data are presented as mean ± sem . The cutoff set for significance for all experiments was alpha <0 . 05 . Data were tested for uniform distribution using three different normality tests ( Kolmogorov–Smirnov test , D'Agostino and Pearson omnibus normality test and Shapiro–Wilk normality test ) . If data passed these normality tests , parametric tests were used . Otherwise , non-parametric tests were used . Unpaired t tests and Mann–Whitney tests were used for comparison between subjects , and paired t tests and Wilcoxon matched-pairs signed rank tests for within-subjects comparisons . For data involving two or more independent variables , two-way ANOVAs were used and Bonferroni posthoc tests , correcting for multiple comparisons , were used . Brain slices were prepared from 3-month-old mice approximately 4 weeks after virus injection ( Haubensak et al . , 2010; Cai et al . , 2014 ) . Coronal brain sections of 250 μm thickness were cut with a Leica VT1000S vibratome , using ice-cold glycerol-based ACSF containing ( in mM ) : 252 glycerol , 1 . 6 KCl , 1 . 2 NaH2PO4 , 1 . 2 MgCl2 , 2 . 4 CaCl2 , 18 NaHCO3 , 11 Glucose , oxygenated in carbogen ( 95% O2 balanced with CO2 ) for at least 15 min before use . Brain slices were recovered for ∼1 hr at 32°C and then kept at room temperature in regular ACSF containing ( in mM ) : 126 NaCl , 1 . 6 KCl , 1 . 2 NaH2PO4 , 1 . 2 MgCl2 , 2 . 4 CaCl2 , 18 NaHCO3 , 11 Glucose , oxygenated with carbogen . The fluorescence of the SF1+ neurons was detected by a fluorescence video microscopy camera ( Olympus BX51 ) . Whole-cell voltage or current clamp recordings were performed with a MultiClamp 700B amplifier and Digidata 1440A ( Molecular Devices ) . The patch pipette with a resistance of 5–8 MΩ was filled with an intracellular solution containing ( in mM ) : 135 potassium gluconate , 5 EGTA , 0 . 5 CaCl2 , 2 MgCl2 , 10 HEPES , 2 MgATP and 0 . 1 GTP , pH 7 . 2 , 290–300 mOsm . Data were sampled at 10 kHz , filtered at 3 kHz and analyzed with pCLAMP10 software . Optogenetic photostimulation was delivered by a 473 nm laser ( Shanghai Dream Laser , 473 nm ) controlled by an Accupulser Signal Generator ( World Precision Instruments ) . The estimated power at the specimen was set to 1 mW/mm2 , as measured with a photodiode ( Thorlabs ) . In vivo electrophysiological recordings were performed using custom-built electrode bundles or optrodes , as published before ( Lin et al . , 2012; Lee et al . , 2014 ) . The electrode bundle was affixed to a movable microdrive stage that could be lowered in steps of 55 µm when required . The electrode bundle was implanted to the stereotaxic coordinates that correspond to the dorsal extent of the VMHdm and lowered at least one step for every recording session . For single unit recordings during optogenetic photostimulation , we integrated a 62 . 5 µm core optical fiber into the 16-microwire electrode bundle ( Lee et al . , 2014 ) . Data was collected from neurons in the SF1-Cre mouse line with the expression of ChR2 using AAV2 . EF1α . FLEX . ChR2-eYFP , as identical to that used in the behavioral experiments . Photostimulation parameters for a given optrode were calibrated prior to implantation so that the transmitted light would irradiate the brain tissue at 1 . 0–1 . 5 mW/mm2 , measured under constant illumination . Hardware and software provisions for eliminating photoelectric artifact were used ( Kvitsiani et al . , 2013 ) . All spikes recorded at a single microwire electrode crossing a threshold two standard-deviations over baseline ( spike wavelength around 1 ms and interspike interval greater than 2 ms ) were sorted into clusters using PC analysis and were considered to represent individual units . Units were recorded under the same photostimulation parameters as those used in the behavioral experiments that is , at 20 Hz with 20 ms pulse-width . Neural activity was recorded over a baseline period of 40 s , followed by a photostimulation period of 40 s . | Animals have evolved a large number of ‘defensive behaviors’ to deal with the threat of predators . Examples include reptiles camouflaging themselves to avoid discovery , fish and birds swarming to confuse predators , insects releasing toxic chemicals , and humans readying themselves to fight or flee . In mammals , defensive behaviors are thought to be mediated by a region of the brain called the amygdala . This structure , which is known as the brain's ‘emotion center’ , receives and processes information from the senses about impending threats . It then sends instructions on how to deal with these threats to other regions of the brain including the hypothalamus , which pass them on to the brain regions that control the behavioral , endocrine and involuntary responses of the mammal . For many years it has been thought that the role of the hypothalamus is to serve simply as a relay for emotion states encoded in the amygdala , rather than as an emotion center itself . However , Kunwar et al . have now challenged this assumption with the aid of a technique called optogenetics , in which light is used to activate specific populations of genetically labeled neurons . When light was used to directly activate neurons within the ventromedial hypothalamus in awake mice , the animals instantly froze and/or fled , just as they would when faced with a predator . Given that the optical stimulation had completely bypassed the amygdala , this suggested that the hypothalamus must be capable of generating this defensive response without any input from the amygdala . The freezing and fleeing responses resembled the responses to a predator in a number of key ways . Mice chose to avoid areas of their cage in which they had received the stimulation , suggesting that—like a predator—these areas induced an unpleasant emotional state , perhaps akin to anxiety or fear . Freezing and fleeing persisted for several seconds after the stimulation had stopped , just as freezing and fleeing responses to predators do not immediately cease after the threat has gone . And finally , destroying the neurons targeted by the stimulation made mice less likely to avoid one of their main predators , the rat . It also made the animals less anxious . Overall the results suggest that the hypothalamus may be more than simply a relay for the amygdala , and that ‘amygdala-centric’ views of emotion processing may need to be re-visited . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Ventromedial hypothalamic neurons control a defensive emotion state |
Stu2p/XMAP215 proteins are essential microtubule polymerases that use multiple αβ-tubulin-interacting TOG domains to bind microtubule plus ends and catalyze fast microtubule growth . We report here the structure of the TOG2 domain from Stu2p bound to yeast αβ-tubulin . Like TOG1 , TOG2 binds selectively to a fully ‘curved’ conformation of αβ-tubulin , incompatible with a microtubule lattice . We also show that TOG1-TOG2 binds non-cooperatively to two αβ-tubulins . Preferential interactions between TOGs and fully curved αβ-tubulin that cannot exist elsewhere in the microtubule explain how these polymerases localize to the extreme microtubule end . We propose that these polymerases promote elongation because their linked TOG domains concentrate unpolymerized αβ-tubulin near curved subunits already bound at the microtubule end . This tethering model can explain catalyst-like behavior and also predicts that the polymerase action changes the configuration of the microtubule end .
Microtubules are dynamic polymers of αβ-tubulin that have critical roles in chromosome segregation and intracellular organization ( reviewed in Desai and Mitchison , 1997 ) . The polymerization dynamics of microtubules are regulated by multiple cellular factors . Evolutionarily conserved proteins in the Stu2p/XMAP215 family ( Gard and Kirschner , 1987; Ohkura et al . , 1988; Wang and Huffaker , 1997 ) regulate microtubule dynamics by promoting fast microtubule elongation . These essential proteins use multiple αβ-tubulin binding tumor overexpressed gene ( TOG ) domains to selectively recognize the growing microtubule end and promote its elongation ( Al-Bassam et al . , 2006; Brouhard et al . , 2008; Widlund et al . , 2011; Al-Bassam et al . , 2012 ) . A significant advance in understanding occurred with a landmark study of XMAP215 in which in vitro reconstitution assays demonstrated that the polymerase affected the rate of microtubule elongation without affecting the apparent equilibrium ( Brouhard et al . , 2008 ) . This and other observations ( e . g . , Shirasu-Hiza et al . , 2003; van Breugel et al . , 2003 ) formed the basis for describing XMAP215 as a catalyst for microtubule elongation . Catalytic action in turn led to the concept that TOG-containing polymerases might stabilize an otherwise rate-limiting intermediate along the polymerization pathway ( Brouhard et al . , 2008 ) . However , an understanding of mechanism has been limited because the nature of this intermediate , and how TOG domains might selectively promote it , remained unclear . A recent study from our group revealed that the TOG1 domain from Stu2p binds preferentially to a ‘curved’ conformation of αβ-tubulin that cannot be incorporated into the body of the microtubule ( Ayaz et al . , 2012 ) . Our study also showed that a TOG1-TOG2 construct could bind two αβ-tubulins ( Ayaz et al . , 2012 ) . This latter observation suggested that two TOG domains might cooperate to stabilize an αβ-tubulin:αβ-tubulin interface , and consequently that cooperative binding to αβ-tubulin might contribute to polymerase activity . Our study did not determine how the polymerase recognizes the extreme microtubule end , but we speculated based on apparent biochemical differences between the TOG1 and TOG2 domains ( Al-Bassam et al . , 2006; Ayaz et al . , 2012 ) that selective interactions between TOG2 and a different , end-specific conformation of αβ-tubulin might be important . In the present study we sought to gain insight into the mechanism of end recognition by determining the conformation of αβ-tubulin recognized by TOG2 and by testing whether cooperative TOG:αβ-tubulin interactions contributed to polymerase activity . We first determined the crystal structure of a TOG2:αβ-tubulin complex . This structure reveals that TOG2 binds to the same curved conformation of αβ-tubulin that TOG1 does . Our biochemical experiments underscore this structural similarity by demonstrating that the two TOG domains have comparable affinities for αβ-tubulin , with KD ∼100 nM . Next , we used analytical ultracentrifugation to show that in TOG1-TOG2 the two linked TOG domains bind non-cooperatively to two αβ-tubulins . Non-cooperative binding indicates that TOG1-TOG2 does not stabilize an αβ-tubulin:αβ-tubulin interface . Together with biochemical and genetic experiments , our results lead to a model that explains how the polymerase activity can emerge from the action of two tethered TOG domains that each bind independently to a conformation of αβ-tubulin that is incompatible with the microtubule lattice . We propose that the polymerase activity arises because linked TOG domains selectively increase the effective concentration of αβ-tubulin near weakly bound , curved αβ-tubulins already on the microtubule end . A computational realization of this model supports our proposal by recapitulating catalyst-like behavior . The model further suggests that the polymerase achieves its effect in part by transiently altering the configuration of the growing end .
We previously showed that the TOG1 domain from Stu2p binds preferentially to a curved conformation of αβ-tubulin ( Ayaz et al . , 2012 ) . However , TOG1 is dispensable for the plus-end binding of Stu2p ( Al-Bassam et al . , 2006 ) , and because of apparent differences in the biochemical behavior of TOG1 and TOG2 ( Al-Bassam et al . , 2006; Ayaz et al . , 2012 ) we speculated that TOG2 might bind to a different , lattice-induced conformation of αβ-tubulin ( Ayaz et al . , 2012 ) . We solved the crystal structure of a TOG2:αβ-tubulin complex ( Figure 1 ) to resolve this ambiguity . The structure was determined by molecular replacement from crystals that diffracted anisotropically to a minimum Bragg spacing of 2 . 8 Å ( Table 1 ) . The final model has good geometry ( Table 1; Molprobity [Chen et al . , 2010] clash score 1 . 79; 95 . 6% favored residues in Ramachandran plot ) and has been refined to an Rfree of 0 . 259 ( Rwork = 0 . 217 ) . 10 . 7554/eLife . 03069 . 003Figure 1 . TOG2 binds to curved αβ-tubulin analogously to TOG1 . ( A ) Structure of the TOG2:αβ-tubulin complex ( TOG2: slate , α-tubulin: pink , β-tubulin: green ) , with the important binding residues W341 and R519 represented as spheres . The semi-transparent gray cartoon shows the previously observed binding mode of TOG1 , with its binding residues W23 and R200 depicted as spheres . ( B ) Close-up of the TOG2:αβ-tubulin interface , colored as in A , and showing in spheres important interacting residues based on an earlier study . ( C ) Structural superposition of αβ-tubulin-bound ( slate ) and unbound ( gray ) TOG2 ( PDB 2QK1 ) . The two structures of TOG2 show only small local deviations , arguing against any significant conformational change associated with αβ-tubulin binding . ( D ) Structural superposition of the αβ-tubulin conformations in the TOG2 ( colored ) and TOG1 ( gray ) complexes . In both complexes αβ-tubulin adopts very similar curved conformations . ( E ) Structural superposition of TOG2-bound αβ-tubulin ( colored ) and the straight conformation of αβ-tubulin ( PDB 1JFF , gray ) , showing the ∼13° rotation of β-tubulin relative to α-tubulin that is characteristic of the curved conformation . DOI: http://dx . doi . org/10 . 7554/eLife . 03069 . 00310 . 7554/eLife . 03069 . 004Table 1 . Data collection and refinement statisticsDOI: http://dx . doi . org/10 . 7554/eLife . 03069 . 004Data collection Space groupC2 Cell dimensions a , b , c ( Å ) 111 . 91 , 89 . 57 , 135 . 51 β ( ° ) 112 . 31 Resolution ( Å ) 50 . 0–2 . 81 ( 2 . 92–2 . 81 ) * Rsym0 . 143 ( 0 . 924 ) <I>/<σI>9 . 6 ( 1 . 1 ) Wilson B-value ( Å ) 48 . 9 Anisotropy ( Å ) relative to best direction ( 001 ) ΔB in ( 100 ) direction , ΔB in ( 010 ) direction+29 . 95 , +8 . 38 CC1/2 in high resolution shell0 . 542 Completeness ( % ) 98 . 2 ( 91 . 3 ) Redundancy4 . 1 ( 3 . 2 ) Refinement Resolution ( Å ) 2 . 81 No . reflections26 , 235 Completeness ( % ) 86 . 5† ( 35 . 3 ) Rwork/ Rfree ( % ) 21 . 8/25 . 9 ( 33 . 0/41 . 3 ) Maximum likelihood estimated coordinate error ( Å ) 0 . 42 No . atoms8524 Protein ( non-hydrogen ) 8437 Ligand/ion66 Water21 B-factors Protein44 . 7 Ligand/ion52 . 0 Water27 . 3 Rms deviations Bond lengths ( Å ) 0 . 003 Bond angles ( ° ) 0 . 66 Ramanchandran plot Favored ( % ) 95 . 0 Allowed ( % ) 4 . 25 Disallowed ( % ) 0 . 75 Rotamer outliers ( % ) 3 . 2 Molprobity clash score1 . 5*Highest resolution shell is shown in parenthesis . †The data were corrected for anisotropy in HKL2000 . This treatment eliminated weak reflections and reduced the completeness of the data used for refinement compared to the completeness reported for data collection . The structure of the TOG2:αβ-tubulin complex is remarkably similar to that of the TOG1:αβ-tubulin complex ( Figure 1; Ayaz et al . , 2012 ) . Conserved residues like W341 and R519 on the tubulin-binding surface of TOG2 make very similar contacts with αβ-tubulin as their equivalents in TOG1 ( W23 and R200 , respectively ) ( Figure 1A , B ) . The structure of TOG2 bound to αβ-tubulin superimposes with the structure of ‘free’ TOG2 ( Slep and Vale , 2007 ) with 0 . 7 Å rmsd over 237 Cα atoms ( Figure 1C ) . Because the TOG2 domains pack differently in the ‘bound’ and ‘free’ crystals , the similarity in structure suggests that these TOG domains are rigid modules that do not change conformation when they bind αβ-tubulin . Even though the structure reported here was obtained from a new crystal form , the conformation of αβ-tubulin in complex with TOG2 is nearly identical to that seen for TOG1-bound αβ-tubulin ( Figure 1D ) ( 12 . 3° of curvature in the TOG2 complex vs 13 . 1° in the TOG1 complex ) , and is characteristically distinct from the straight conformation of αβ-tubulin ( 1° of curvature by our measure ) ( Löwe et al . , 2001; Figure 1E ) . Underscoring this similarity , individual tubulin chains between the TOG1 and TOG2 complexes superimpose on each other with 0 . 4 Å rmsd over Cα atoms . Thus , both TOG1 and TOG2 bind preferentially to the same , curved conformation of αβ-tubulin . By extension , other TOGs in this family probably also bind to curved αβ-tubulin . The shared preference of TOG1 and TOG2 for curved αβ-tubulin has implications for the mechanism of end recognition . Because TOG:αβ-tubulin interactions are required for plus-end localization of Stu2p/XMAP215 polymerases ( Al-Bassam et al . , 2006 ) , preferential binding of TOGs to curved αβ-tubulin suggests that curved αβ-tubulin itself is the distinctive end-specific feature the polymerase recognizes . In contrast , other plus-end tracking proteins like Eb1 recognize lattice-specific features and consequently show ‘comet-like’ localization that extends into the microtubule body ( Nakamura et al . , 2012; Maurer et al . , 2014 ) . By invoking binding to an epitope that cannot exist in the body of the microtubule , our model explains how Stu2p/XMAP215 polymerases localize to the extreme microtubule end ( Nakamura et al . , 2012; Maurer et al . , 2014 ) . The shared preference for fully curved αβ-tubulin also poses an apparent paradox , because it indicates that the polymerase is constructed from ‘parts’ that bind most strongly to a conformation of αβ-tubulin that cannot exist in the microtubule lattice . We explore and propose a resolution for this apparent contradiction in later sections . To determine if the structural similarity between the TOG1 and TOG2 complexes with αβ-tubulin extends to a biochemical similarity , we measured the affinity of TOG:αβ-tubulin interactions . Previously we had used polymerization-blocked αβ-tubulin mutants ( Johnson et al . , 2011 ) for TOG binding assays . To eliminate the possibility that blocking mutations and/or fluorescent labeling might mask or alter αβ-tubulin:αβ-tubulin interactions , we developed a label-free assay in which polymerization-competent αβ-tubulin could be used . Analytical ultracentrifugation monitored by absorbance at 230 nm ( A230 ) allowed us to work at concentrations of αβ-tubulin at which higher-order oligomers of αβ-tubulin were nearly undetectable ( Figure 2A ) . Indeed , at αβ-tubulin concentrations ranging from 0 . 08 to 1 μM there is a single dominant peak at 5 . 8 S with only 1 . 7 and 2 . 8% of the material sedimenting faster at the lowest and highest concentration tested . This uniform sedimentation behavior made it possible for us to use analytical ultracentrifugation as a quantitative binding assay . 10 . 7554/eLife . 03069 . 005Figure 2 . TOG1 and TOG2 bind αβ-tubulin with comparable affinity . ( A ) Sedimentation velocity analytical ultracentrifugation of polymerization competent yeast αβ-tubulin does not show signs of self-association between 80 nM and 1 μM concentration . The main plot shows c ( s ) distributions for a range of αβ-tubulin concentration . The inset shows that the s20 , w is not increasing with αβ-tubulin concentration . Data points are color coded to match the c ( s ) distribution for that concentration . c ( s ) , signal population as a function of s; s20 , w , sedimentation coefficient standardized to pure water and 20°C . ( B ) Analysis of TOG1:αβ-tubulin interactions by sedimentation velocity . The main plot shows c ( s ) distributions ( color coded by concentration ) for seven concentrations of TOG1 ( 44 nM–4 . 5 μM ) titrated into 0 . 35 μM αβ-tubulin . The inset shows the fit ( gray line ) of a 1:1 binding isotherm to the signal average sfast ( dots colored to match the c ( s ) distribution for that concentration of TOG1 ) resulting in a dissociation constant of 70 nM . ( C ) Analysis of TOG2:αβ-tubulin interactions by sedimentation velocity . Plots are as described in ( B ) . The fitted dissociation constant is 160 nM . DOI: http://dx . doi . org/10 . 7554/eLife . 03069 . 005 We measured the affinity of TOG1 and TOG2 for αβ-tubulin by separately titrating variable amounts of each TOG domain into a constant amount of αβ-tubulin and analyzing the resulting sedimentation behavior ( Figure 2B , C ) . We fit the concentration-dependent sedimentation profiles using single-site binding isotherms ( Dam and Schuck , 2005; Figure 2B , C ) . The fits indicate that both TOG domains form relatively tight , 1:1 complexes with αβ-tubulin: TOG1 binds with KD = 70 nM and TOG2 binds with KD = 160 nM . The KD for the TOG2:αβ-tubulin complex reported here is consistent with our prior measurement using fluorescence anisotropy and polymerization blocked αβ-tubulin ( Ayaz et al . , 2012 ) . Thus , in addition to a shared preference for the same curved conformation of αβ-tubulin , TOG1 and TOG2 also bind αβ-tubulin with comparable affinities . From both a biochemical and a structural perspective , the TOG1 and TOG2 domains are remarkably similar to each other . When present together in a TOG1-TOG2 construct , each of the TOG domains can engage its own αβ-tubulin ( Ayaz et al . , 2012 ) . We speculated that the two TOG-bound αβ-tubulins might also interact with each other , and that this cooperativity might make important contributions to polymerase function ( Figure 3A ) . 10 . 7554/eLife . 03069 . 006Figure 3 . In TOG1-TOG2 , the two TOG domains bind two αβ-tubulins without positive cooperativity . ( A ) Cartoons illustrating three different possible arrangements of a TOG1-TOG2: ( αβ ) 2 complex: independent ( left ) denotes that an αβ-tubulin:αβ-tubulin interface does not provide additional stability to the complex , cooperative ( right ) denotes that an αβ-tubulin:αβ-tubulin interface ( either longitudinal or lateral ) provides additional stability to the complex . Predicted sedimentation coefficients ( calculated using HYDROPRO [García De La Torre et al . , 2000] ) are indicated . αβ-tubulin is represented in pink and green , TOG1-TOG2 in shades of blue . ( B ) Sedimentation behavior of a TOG1-TOG2: ( αβ ) 1 complex by sedimentation velocity AUC using two different mutations ( W314A and R519A ) that impair TOG2:αβ-tubulin interactions . The ‘one tubulin’ complex sediments at 7 . 1 S . ( C ) Placing limits on the sedimentation behavior of a TOG1-TOG2: ( αβ ) 2 complex by sedimentation velocity AUC . At ∼5 molar equivalents of αβ-tubulin to TOG1-TOG2 , the resulting complex sediments at 9 . 1 S . The inset shows the predicted fraction of TOG1-TOG2 engaged in ‘two αβ-tubulin complex’ under different assumptions about cooperativity . ( D ) Concentration dependence of TOG1-TOG2:αβ-tubulin interactions . Seven concentrations of TOG1-TOG2 were mixed with 0 . 3 μM αβ-tubulin and analyzed by sedimentation velocity AUC . Red dots indicate the signal-weighted sw values for the seven runs . The blue and pink swaths show the predicted behavior for TOG1 and TOG2 binding αβ-tubulin with 10-fold positive or negative cooperativity , respectively , and assuming the sedimentation coefficient of TOG1-TOG2: ( αβ ) 2 falls in the range 9 . 1–10 . 9 S ( see text ) . The gold swath shows the predicted behavior for noncooperatively binding TOGs using the same range of sedimentation coefficient for TOG1-TOG2: ( αβ ) 2 . The data are not consistent with cooperative binding of TOG1-TOG2 to two αβ-tubulins . Instead , they are much better described by independently binding TOG domains . DOI: http://dx . doi . org/10 . 7554/eLife . 03069 . 006 We used analytical ultracentrifugation to determine if cooperative interactions stabilized the TOG1-TOG2: ( αβ-tubulin ) 2 complex . First , using mutations on TOG1-TOG2 that disrupt the slightly weaker TOG2:αβ-tubulin interactions , we determined that TOG1-TOG2 bound to a single αβ-tubulin sediments at 7 . 1 S ( Figure 3B ) . Next , using a mixture in which αβ-tubulin was superstoichiometric with respect to TOG1-TOG2 ( Figure 3C ) , we observed a larger species sedimenting at 9 . 1 S . This value places a lower limit on the sedimentation coefficient for the TOG1-TOG2: ( αβ-tubulin ) 2 complex ( Figure 3C , inset ) that is consistent with hydrodynamic calculations ( García De La Torre et al . , 2000 ) using extended models in which the two αβ-tubulins do not contact each other and the TOG domains are separated by varying distances . Hydrodynamic calculations predict that compact models containing longitudinally or laterally of αβ-tubulins should sediment around 10 . 5–10 . 9 S ( Figure 3A , C ) . Knowledge about the sedimentation behavior of the one and two tubulin complexes allowed us to analyze a titration of TOG1-TOG2 into αβ-tubulin under different assumptions about cooperativity ( Figure 3D ) . The resulting signal weighted average sedimentation coefficients are not consistent with models that assume even modest 10-fold positive or negative cooperativity ( Figure 3D ) . Instead , the data are much better described by a model in which each TOG domain in TOG1-TOG2 interacts independently with its αβ-tubulin ( the best fit was obtained from a model invoking less than twofold negative cooperativity , not shown ) . Noncooperative binding contradicts our initial expectation and indicates that αβ-tubulin:αβ-tubulin contacts do not provide additional stability to the TOG1-TOG2: ( αβ-tubulin ) 2 complex . The lack of positive cooperativity stabilizing the TOG1-TOG2: ( αβ-tubulin ) 2 complex is striking because in this complex the two bound αβ-tubulins are physically constrained to occupy a relatively small volume and therefore are effectively at quite high concentration relative to each other . Indeed , allowing 55 Å for the length of each TOG domain and assuming that the ∼75 amino acid linker is flexible and can maximally span 220 Å we estimate that the two TOG1-TOG2-bound αβ-tubulins are minimally at an effective concentration of roughly 200 μM , likely higher because the linker will rarely be fully extended . This effective concentration of TOG1-TOG2-bound αβ-tubulin is in the range of current estimates for the longitudinal KD for αβ-tubulins ( e . g . , Gardner et al . , 2011 ) . Observing non-cooperative binding therefore suggests that some property of the TOG1-TOG2 linker antagonizes or counterbalances longitudinal interactions between αβ-tubulins ( see below for experiments concerning the linker sequence ) . Our data do not rule out that linked TOG domains might stabilize lateral interactions between αβ-tubulins because these lateral interactions are thought to be much weaker ( only molar affinity [Gardner et al . , 2011] ) : even at effective αβ-tubulin concentrations approaching 1 mM a weak interface like this would not be populated and therefore would contribute little additional stability to the TOG1-TOG2: ( αβ-tubulin ) 2 complex . Whatever the underlying mechanism , the unexpected observation that the two linked TOG domains behave as independent αβ-tubulin binding modules places significant constraints on biochemical models for the mechanism of these polymerases . Stu2/XMAP215 family polymerases contain at least two different TOG domains . This property might indicate that the two domains have different functional specialization , but the structural and biochemical similarity of TOG1 and TOG2 ( described above ) does not seem consistent with separation of function . To investigate if different TOG domains are required for polymerase function we constructed a ‘TOG-swapped’ and other variants of Stu2p and assayed their ability to rescue the conditional depletion of endogenous Stu2p using a previously described assay ( Kosco et al . , 2001; Al-Bassam et al . , 2006; Ayaz et al . , 2012 ) . We first performed rescue assays using a ‘full-length’ Stu2p construct that retains the ability to dimerize and thus that contains two identical TOG1-TOG2 segments linked through the coiled-coil dimerization interface . A construct ( TOG2-TOG2 ) in which TOG1 was replaced by a second copy of TOG2 rescued as well as did the wild-type ( TOG1-TOG2 ) ( Figure 4A ) . We also tried to replace TOG2 with a second copy of TOG1 , but control experiments indicated that this variant was not stable ( data not shown ) and we have not yet pursued it further . Instead , we used site-directed mutagenesis as an alternate way to ablate the αβ-tubulin binding activity of TOG1 or TOG2 . Constructs in which either TOG was impaired for αβ-tubulin binding ( TOG1 ( R200A ) -TOG2 or TOG1-TOG2 ( R519A ) ) also gave full rescue ( Figure 4A ) . Ablating the function of individual TOGs is not without functional consequences under more stringent conditions , because these same constructs do show compromised rescue under conditions of microtubule stress ( Ayaz et al . , 2012 ) . These data indicate that under normal conditions , the polymerase can function with only TOG1 or only TOG2 domains . Thus , having different TOG domains does not appear to be essential for polymerase function . 10 . 7554/eLife . 03069 . 007Figure 4 . Two TOG domains are required for Stu2 function , but they do not have to be different . ( A ) Yeast carrying plasmid-based rescue constructs coding for dimerization-competent variants of Stu2p were plated at serial dilutions on media that was unmodified ( control ) or that contained 500 μM CuSO4 ( to deplete endogenous Stu2p; see text ) . All constructs , including those with debilitated TOG1 or TOG2 domains , showed full rescue . TOG domains are shown in blue , and the basic region in red . The coiled-coil is cartooned as a zipper . ( B ) As in A but using rescue constructs that are dimerization-impaired because the coiled-coil dimerization domain was deleted . In this more stringent background insults to either TOG domain abolished rescue activity . Replacing TOG1 with a second copy of TOG2 does not have adverse effects . DOI: http://dx . doi . org/10 . 7554/eLife . 03069 . 007 Using dimeric rescue constructs did not allow us to test Stu2p variants that contained only a single functional TOG domain of either kind . We therefore introduced the same mutations into a previously characterized dimerization-impaired variant of Stu2p in which the coiled-coil dimerization element had been deleted ( Stu2p-Δcc ) ( Al-Bassam et al . , 2006; Figure 4B ) . Consistent with prior observations ( Al-Bassam et al . , 2006 ) , dimerization impaired Stu2p only partially compensated for the depletion of endogenous Stu2p . Even in this sensitized background , however , we found that as long as there were two functional TOG domains they did not need to be different . Stu2p ( TOG2-TOG2 ) -Δcc showed rescue efficiency very similar to that of Stu2p ( TOG1-TOG2 ) -Δcc ( Figure 4B ) . Dimerization impaired variants of Stu2p in which TOG1 or TOG2 was defective for αβ-tubulin binding did not show any rescue activity ( Figure 4B ) , consistent with a prior in vitro study of XMAP215 that demonstrated a requirement for at least two TOG domains ( Widlund et al . , 2011 ) . More importantly , the ability of TOG2 to substitute for TOG1 in this more stringent , dimerization-impaired background strengthens the conclusion that the polymerase does not require different TOG domains for its function . The dimeric Stu2p ( TOG1-TOG2 ( R519A ) ) and Stu2p ( TOG1 ( R200A ) -TOG2 ) variants rescued the depletion of endogenous Stu2p in spite of the fact that the two functioning TOG domains were linked through the coiled-coil dimerization segment ( and in the case of TOG1-TOG2 ( R519A ) with a defective TOG domain in the linking sequence ) . This result suggested that how two TOG domains were linked was relatively unimportant , as long as they were linked . To explore this more systematically we determined how randomizing and/or shortening the TOG1-TOG2 linker in dimerization impaired Stu2p affected its rescue activity . To test if the primary sequence of the TOG1-TOG2 linker was important for function , we prepared two variants of Stu2pΔcc in which the order of the central 65 amino acids of the TOG1-TOG2 linker ( residues 252–316 ) was randomized ( Figure 5A ) . This ‘shuffling’ strategy preserves the overall amino acid composition but should disrupt any local features specific to the natural sequence . Both shuffled linkers gave rescue activity nearly indistinguishable from the wild-type linker ( Figure 5A ) , consistent with the robust rescue activity of alternatively linked functional TOGs in dimeric Stu2p ( TOG1-TOG2 ( R519A ) ) and Stu2p ( TOG1 ( R200A ) -TOG2 ) ( Figure 4A ) . Thus , the sequence linking the two functional TOG domains tolerates significant variation . 10 . 7554/eLife . 03069 . 008Figure 5 . Stu2p function tolerates variation in the primary sequence and in the length of the TOG1-TOG2 linker . Rescue assays were performed as in Figure 4 , using dimerization-impaired rescue constructs . ( A ) Stu2p variants with ‘shuffled’ ( randomized ) linker sequences rescue the depletion of endogenous Stu2p comparably to those with the natural linker . ( B ) Stu2p function is substantially abolished when the TOG1-TOG2 linker is truncated by 60 amino acids . Smaller truncations only show slightly compromised rescue activity . ( C ) Histogram illustrating the distribution of TOG1-TOG2 linker lengths in ∼300 orthologs . The distribution shows that the linker length can vary but has a minimum tolerable length of ∼40 amino acids . DOI: http://dx . doi . org/10 . 7554/eLife . 03069 . 008 We also made a series of internal deletions in the TOG1-TOG2 linker to determine if shortening the linker affected rescue activity . Deleting 14 amino acids had little effect on rescue activity ( Figure 5B ) . Rescue activity was mildly compromised by deletions of 22 , 32 , 40 , and 50 amino acids ( Figure 5B ) . Deleting 60 amino acids from the TOG1-TOG2 linker substantially abolished rescue activity ( Figure 5B ) . These data suggest that Stu2p function is compromised when the sequence linking two TOG domains becomes too short . To examine the question of linker length more generally , we analyzed the distribution of TOG1-TOG2 linker length in ∼300 Stu2p/XMAP215 orthologs . This analysis showed that linker lengths below ∼40 amino acids occur very rarely ( Figure 5C ) , supporting the notion that there is a minimal linker length below which polymerase function is compromised . Any model for the polymerase function must recapitulate the catalyst-like properties documented previously ( Brouhard et al . , 2008 ) while also accounting for the seemingly paradoxical observation that both TOG1 and TOG2 from Stu2p bind preferentially to a curved conformation of αβ-tubulin that cannot exist in the body of the microtubule . A good model must likewise be consistent with the lack of detectable positive cooperativity stabilizing the TOG1-TOG2: ( αβ-tubulin ) 2 complex , and should not require different TOG domains or a non-spacer-like role for the sequence linking the two TOG domains . In Figure 6 we cartoon a simple ‘tethering’ mechanism that incorporates these new constraints , using a minimal ‘two TOG’ polymerase ( Widlund et al . , 2011 ) to simplify the representation . The central elements of this model are: ( i ) plus-end localization of the polymerase occurs by a TOG domain binding to curved αβ-tubulin at the microtubule end; ( ii ) full curvature of αβ-tubulin at the microtubule end requires that it not have any lateral neighbors; ( iii ) αβ-tubulin bound to the other TOG domain is physically tethered to the microtubule end and therefore associates with it more frequently . When not engaged with αβ-tubulin at the microtubule end , we assume based on results from others ( Brouhard et al . , 2008 ) that the polymerase diffuses on the microtubule lattice by virtue of its basic domain , thereby rapidly ‘searching’ for another curved αβ-tubulin . 10 . 7554/eLife . 03069 . 009Figure 6 . A tethering model for the polymerase function that incorporates our structural and biochemical observations . The model posits that MT plus-end recognition occurs through TOG-mediated recognition of curved ( black outline , ‘kinked’ αβ-tubulin cartoon ) , not straight ( gray outlines ) , αβ-tubulin on the MT end , and that the linked TOG domains ( blue; the basic region is indicated in red ) serve to ‘tether’ an unpolymerized αβ-tubulin to the MT end . Polymerase activity is predicted to arise because increasing the effective concentration of αβ-tubulin near the MT end should also increase the rate of αβ-tubulin:MT encounters . DOI: http://dx . doi . org/10 . 7554/eLife . 03069 . 009 Can this tethered delivery model recapitulate the catalyst-like action of these polymerases ? To address this question , we took advantage of a kinetic model for microtubule assembly that we have been developing as part of other ongoing work . Our model for microtubule elongation is very similar to ( and was inspired by ) one previously developed by Odde and Cassimeris ( VanBuren et al . , 2002 ) . Alternative and more complicated models exist ( e . g . , VanBuren et al . , 2005; Margolin et al . , 2012 ) , but this relatively simple one captures the essence of the biochemistry—that the microtubule end presents a mix of high and low affinity binding sites , depending on neighbor context–in a reasonable way and using a small number of adjustable parameters ( ‘Materials and methods’ ) . Our model simulates microtubule polymerization one subunit association or dissociation at a time using kinetic Monte Carlo simulations on a two-dimensional lattice with staggered periodic boundary conditions to mimic the cylindrical structure of a 13 protofilament microtubule ( Figure 7A ) . Because XMAP215 has been shown to promote the elongation of microtubules in the presence of a hydrolysis resistant GTP analog ( Brouhard et al . , 2008 ) , we performed ‘GTP-only’ simulations ( no GTP hydrolysis ) to simplify the behavior by eliminating catastrophe . We performed a manual grid search of longitudinal and corner interactions ( in the model these interactions are the dominant contributors to growth rate ) to identify parameters that reproduced experimentally observed microtubule growth rates . To validate our program , we first fit the same experimental data ( Walker et al . , 1988 ) as did Odde and Cassimeris ( VanBuren et al . , 2002 ) , obtaining similar parameters ( Figure 7—figure supplement 1 ) . Other measurements of microtubule dynamics ( e . g . , Mitchison and Kirschner , 1984; Drechsel et al . , 1992; Hyman et al . , 1992; Brouhard et al . , 2008; Gardner et al . , 2011] ) have shown much lower critical concentrations than the Walker et al . data ( Walker et al . , 1988 ) do , so we also searched for parameters to approximate the growth rates observed in the presence of GMPCPP ( Brouhard et al . , 2008; Gardner et al . , 2011 ) . This procedure identified a different set of parameters ( Figure 7C; VanBuren et al . , 2002 ) . These models provide a starting point for exploring potential polymerase mechanisms . 10 . 7554/eLife . 03069 . 010Figure 7 . An implicit model can recapitulate the catalytic nature of polymerase activity . We developed a kinetic model for microtubule elongation and altered it to explore models for the polymerase . ( A ) Cartoon of the cylindrical microtubule ( left; pink and green spheres represent α- and β-tubulin , respectively ) alongside a two-dimensional representation of the MT lattice ( gray boxes , right; the darker boxes represent the microtubule ‘seed’ used to template elongation in our simulations ) . ( B ) The model parameterizes all six possible neighbor states for αβ-tubulin in the lattice , but the two that dominate the elongation behavior are the longitudinal ( top left ) and corner ( top right ) interactions . ( C ) A grid search identified parameters capable of recapitulating the concentration dependence of microtubule elongation rates in the presence of GMPCPP . The black line summarizes the trend from experimental observations ( Brouhard et al . , 2008; Gardner et al . , 2011 ) ; red dots represent the results from our simulations . Using a kon of 4 × 106 M−1s−1 , we obtained a good match to observed growth rates from KDlong = 8 mM and KDcorner = 33 nM . ( D ) ( left ) Cartoon illustrating the tethering model , with the polymerase ( TOG domains in blue , basic region in red ) localized to a curved αβ-tubulin bound at the MT end by pure longitudinal association , ( middle ) simulated growth rates obtained at increasing association rates for the tethered αβ-tubulin . Enhanced trapping of longitudinally-associated αβ-tubulin through the tethering effect shows catalyst-like activity: growth rates and apparent on-rate constant ( slope ) both change significantly but the apparent equilibrium for growth ( x-intercept ) does not . ( right ) Plot of fold increase in growth rate vs the fold increase in tethered αβ-tubulin association rate , using values at 0 . 5 μM αβ-tubulin concentration as a reference ( black dots; the black line shows the fit of a hyperbolic curve to the growth rates ) . The model for the polymerase gives a relatively modest change in growth rates compared to the fold-increase in tethered association rate . Simulations with progressively stronger ( dark green: 1 . 7 mM; light green , 0 . 35 mM ) longitudinal interactions show higher maximal polymerase activity . The polymerase activity is related to the population of longitudinally-associated αβ-tubulin at the MT end . ( E ) Examining an alternative tethering model in which the polymerase promotes incorporation at a ‘corner’ site . This model yields much greater stimulation of elongation ( middle ) because there is always at least one corner site at the microtubule end . The predicted response also does not appear to saturate with increased tethering effect ( right , linear fit ) . This alternative model does not describe the polymerase action because it fails to produce catalyst-like output . DOI: http://dx . doi . org/10 . 7554/eLife . 03069 . 01010 . 7554/eLife . 03069 . 011Figure 7—figure supplement 1 . An implicit model as in Figure 7 but trained against different measured growth rates . ( A ) A grid search identified parameters capable of recapitulating the concentration dependence of microtubule elongation rates measured by Walker et al . ( 1988 ) . The black line summarizes the trend from experimental observations; red dots represent the results from our simulations . Using a kon of 4 × 106 M−1s−1 , we obtained a good match to observed growth rates from KDlong = 3 mM and KDcorner = 4 μM . These parameters are similar to those obtained by Odde and Cassimeris using a similar model trained on the same data ( VanBuren et al . , 2005 ) . Using kon of 2 × 106 M−1s−1 or 6 × 106 M−1s−1 yielded longitudinal affinities of 0 . 4 and 18 mM , respectively . ( B ) ( left ) Cartoon illustrating the tethering model , with the polymerase ( TOG domains in blue , basic region in red ) localized to a curved αβ-tubulin bound at the MT end by pure longitudinal association , ( middle ) simulated growth rates obtained at increasing association rates for the tethered αβ-tubulin . Enhanced trapping of longitudinally-associated αβ-tubulin through the tethering effect shows catalyst-like activity: growth rates and apparent on-rate constant ( slope ) both change significantly but the apparent equilibrium for growth ( x-intercept ) does not . ( right ) Plot of fold increase in growth rate vs the fold increase in tethered αβ-tubulin association rate , using values at 10 μM αβ-tubulin concentration as a reference ( black dots; the black line shows the fit of a hyperbolic curve to the growth rates ) . The model for the polymerase gives a relatively modest change in growth rates compared to the fold-increase in tethered association rate . Simulations with stronger ( purple: 0 . 4 mM ) or weaker ( light green , 18 mM ) longitudinal interactions show higher or lower maximal polymerase activity . Thus , the polymerase activity is related to the population of longitudinally-associated αβ-tubulin at the MT end . ( C ) Examining an alternative tethering model in which the polymerase promotes incorporation at a ‘corner’ site . This model yields much greater stimulation of elongation ( middle ) because there is always at least one corner site at the microtubule end . The predicted response also does not appear to saturate with increased tethering effect ( right , linear fit ) . This alternative model does not describe the polymerase action because it fails to produce catalyst-like output . DOI: http://dx . doi . org/10 . 7554/eLife . 03069 . 011 An explicit model for the polymerase requires too many adjustable parameters to be feasible at this time . To address the question ‘can tethering give catalyst-like activity’ , we made a number of simplifying assumptions to create an implicit representation of polymerase action that isolates the ‘tethered delivery’ steps while ignoring other aspects of polymerase function and biochemistry . We assumed that every curved , longitudinally-bound ‘singleton’ αβ-tubulin at the MT end is engaged by one TOG domain of a polymerase ( Figure 7D , left panel ) . This assumption provides end-localized polymerase in the model without having to parameterize/describe its diffusive movement on the microtubule lattice ( this movement is mediated by the basic domain , which is required for polymerase function ) , or if it arrives at the microtubule end with its TOGs pre-loaded with αβ-tubulin ( XMAP215 has been observed to carry αβ-tubulin while diffusing on the lattice ) . To mimic the tethering-induced local increase in αβ-tubulin concentration , in the model we accelerated the rate of αβ-tubulin association to the site neighboring the polymerase-bound singleton ( Figure 7D , arrow ) . We do not assume that the linker is directing the tethered subunit to a particular site–separate simulations showed that accelerating the rate of αβ-tubulin association on top of the polymerase-bound singleton had no effect on growth rates ( data not shown ) . We also assumed that when the polymerase bound singleton acquired a lateral neighbor it became straight ( Rice et al . , 2008; Buey et al . , 2006 ) , and that this straightening was linked to TOG disengagement ( Ayaz et al . , 2012 ) . To explore if enhanced trapping of singleton αβ-tubulins could recapitulate catalyst-like activity , we simulated microtubule elongation rates using a range of accelerated tethered association rates . In reality , the tethered association rate is likely to be more or less constant , with maximal polymerase activity instead determined by the degree to which it saturates its binding sites on the plus end . Varying the tethered association rate provides a way to mimic this ‘saturation’ effect without having to model it explicitly . Strikingly , the simulated microtubule elongation rates increase with the tethered αβ-tubulin association rate while the apparent equilibrium constant for elongation ( x-intercept ) remains approximately constant ( Figure 7D , middle; Figure 7—figure supplement 1 ) . Affecting the rate but not the apparent equilibrium indicates that tethering-enhanced associative trapping of weakly-bound singletons can recapitulate the essence of catalyst-like activity . The increase in simulated growth rate ( less than 10-fold ) is significantly less than the maximal increase in tethered αβ-tubulin association rate we explored ( Figure 7D , right; Figure 7—figure supplement 1 ) . Thus , in addition to capturing catalyst-like action , our simple model also shows the modest overall stimulation that is characteristic of these polymerases . Why did the model only stimulate elongation modestly , as the polymerases do ? We speculated that because longitudinally associated αβ-tubulin is weakly/transiently bound at the microtubule end ( ∼mM affinity in our fitted parameters and those of comparable models ( VanBuren et al . , 2002; Gardner et al . , 2011 ) ; this compares to working concentrations of αβ-tubulin that are typically at most 20 μM ) , the low population and lifetime of this ‘substrate’ was limiting the activity obtainable in our model . We probed the linkage between ‘substrate scarcity’ and polymerase activity in two ways . We first created an alternative model in which the polymerase accelerated delivery of αβ-tubulin adjacent to higher affinity and much longer-lived ‘corner’ ( longitudinal + lateral interactions ) sites , at least one of which is always present at the microtubule end ( Figure 7E , left ) . In this alternative model the simulated elongation rates increase linearly with the increased effective concentration ( Figure 7E , right ) , yielding much stronger stimulation that also lacks catalyst-like characteristics ( Figure 7E , middle ) . This result is consistent with the idea that catalyst-like activity is related to the low affinity/lifetime of the end-bound singleton substrate . To explore this implication more directly we obtained parameter sets with higher affinity longitudinal interactions by using different assumed values of kon . As observed previously ( VanBuren et al . , 2002 ) , these alternative parameterizations can approximate the experimental rates of ( uncatalyzed ) microtubule elongation by compensating for faster/slower kon with weaker/stronger longitudinal affinity ( but retaining the same ‘corner’ [longitudinal + lateral] affinity ) . We observed in simulations that stronger longitudinal affinity gave higher maximal polymerase activity ( Figure 7D , right; see also Figure 7—figure supplement 1 ) . These results support the notion that the polymerase activity in our model is related to the population/lifetime of longitudinally-associated αβ-tubulin singleton at the microtubule end . The tethering model suggests a plausible transition state/collision complex that limits the rate of microtubule polymerization in the absence of a polymerase ( Figure 8 ) . The idea is that at the MT end , longitudinally-bound αβ-tubulins tend to dissociate much faster than the rate at which unpolymerized αβ-tubulins associate at neighboring sites . Accordingly , most of the ‘singleton’-type associations are unproductive for elongation , and the MT grows primarily by additions into ‘corner’ sites . In the model we propose , Stu2p/XMAP215 polymerases enhance elongation by selectively concentrating αβ-tubulin near singleton sites , thereby specifically promoting their associative trapping . We propose that the ‘trapping complex’—two side-by-side αβ-tubulins without additional neighbors to either side–may be the transition state . In our model the polymerase catalyzes formation of this ‘side-by-side’ state by selectively concentrating ( tethering ) the reactants that lead to it . Associative trapping of singletons also ‘roughens’ the microtubule end through the creation of additional , longer-lived ‘corner’ sites . By virtue of having more favorable sites of interaction , these rougher microtubule ends can transiently capture unpolymerized subunits more efficiently and independently of the polymerase , through ‘filling in’ ( a contribution from filling-in was anticipated previously [Brouhard et al . , 2008] ) . 10 . 7554/eLife . 03069 . 012Figure 8 . Schematic cartoons illustrating the origin of catalytic action . The microtubule end has multiple sites where αβ-tubulin can associate , but elongation is largely dominated by additions into the few , high-affinity ‘corner’ sites ( left panel ) because pure longitudinal associations are weak . By preferentially recognizing curved αβ-tubulin with one of its TOG domains , the polymerase ( TOG domains in blue , basic region in red ) can selectively localize to these ‘unproductive’ binding sites ( second from left ) . The tethering action greatly enhances the rate at which these weakly bound subunits are trapped by neighboring association of another αβ-tubulin ( middle two panels ) . Polymerization-induced straightening of αβ-tubulin releases the polymerase for another round of catalysis ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03069 . 012 Because our model does not explicitly represent the polymerase , it cannot explain the molecular mechanisms underlying processive action . Perhaps basic-domain mediated rapid diffusion on the microtubule lattice , as has been measured for XMAP215 ( Brouhard et al . , 2008 ) , is sufficient for the liberated polymerase to engage a newly added αβ-tubulin . In summary , we propose that Stu2p/XMAP215 polymerases localize to the plus end by virtue of preferential interactions between TOG domains and curved αβ-tubulin . Our proposal is consistent with the view that curved , GTP-bound αβ-tubulin is an on-pathway polymerization intermediate ( e . g . , Buey et al . , 2006; Rice et al . , 2008; Nawrotek et al . , 2011 ) . It also explains the need for at least two TOG domains ( Widlund et al . , 2011 ) because one TOG is required to bind curved αβ-tubulin on the microtubule and another linked TOG is required for tethering an unpolymerized αβ-tubulin . In this model , side-by-side trapping and the release of TOG domains will be linked because of the polymerization-associated conformational changes that occur in αβ-tubulin . The model provides a simple molecular explanation for the catalytic-like action of the polymerase , recapitulates the modest maximal stimulation , and rationalizes how a polymerase can be built from domains that bind preferentially to a microtubule-incompatible conformation of αβ-tubulin .
Plasmids to express TOG2 and wild-type or polymerization-blocked yeast αβ-tubulin have been described previously ( Johnson et al . , 2011; Ayaz et al . , 2012 ) and were used without further modification . Some plasmids ( to express ‘full-length’ Stu2p and its R200 and R519 mutants ) for the rescue assay were also previously described ( Ayaz et al . , 2012 ) . ‘Dimerization-impaired’ plasmids were constructed using mutagenesis primers to delete the region that codes for the coiled-coil dimerization element ( corresponding to amino acids 658–761 ) . These primers contained two regions that were complementary to the coding sequences just upstream and downstream of the deleted region . ‘Linker truncation’ plasmids were constructed using a similar strategy to remove increasingly large sections from the middle of the sequence linking TOG1 to TOG2 ( deleted regions are: Δ14:273-286; Δ22:277-298; Δ32:273-304; Δ40:257-296; Δ50:251-300; Δ60:251-310 ) . The ‘linker shuffled’ and ‘TOG swapped’ plasmids were made using GeneArt seamless cloning ( Invitrogen , Carslbad , CA ) . Shuffled linkers were inserted starting from synthetic , codon-optimized cDNA ( IDT , Coralville , IA ) , so we also prepared a codon optimized natural linker sequence to control for possible effects on expression levels . DNA coding for shuffled versions of the 65 central amino acids ( residues 252–316 ) of the TOG1-TOG2 linker was obtained from IDT and amplified using primers to append flanking nucleotides of wild-type upstream and downstream flanking sequence . The parent plasmid , excluding the region coding for the to-be-replaced linker sequence , was amplified in a separate reaction . The two products were mixed , incubated , and transformed according to the manufacturer's instructions . A similar strategy was used to replace TOG1 ( residues 11–245 ) with a second copy of TOG2 ( residues 318–560 ) . The entire coding region of all plasmids was verified by DNA sequencing . The TOG1 ( 1–317 ) , TOG2 ( 318–560 ) , and TOG1-TOG2 ( 1–560 ) were expressed in bacteria with C-terminal His6 tags and purified using Ni-affinity and ion exchange chromatography ( Ayaz et al . , 2012 ) . Wild-type or polymerization-blocked yeast αβ-tubulin ( β:T175R , V179R ) were overexpressed in Saccharomyces cerevisiae ( Johnson et al . , 2011 ) . All proteins were purified as previously described ( Johnson et al . , 2011; Ayaz et al . , 2012 ) . For crystallization , pure polymerization blocked yeast αβ-tubulin ( β:T175R , V179R ) and TOG2 domain were dialyzed into RB150 ( 25 mM Tris pH 7 . 5 , 150 mM NaCl , 1 mM MgCl2 , 1 mM EGTA ) and then mixed at equimolar stoichiometry . One tenth volume of 2 M L-proline was added before concentrating the protein using 30 kDa cutoff Amicon Ultra concentrators . When the desired concentration was reached ( ∼2–3 mg/ml ) , GTP was added from a 100 mM stock to achieve a final concentration of 1 mM . Sparse-matrix crystallization screening ( typically mixing protein with precipitants at 1:1 and 2:1 protein:reservoir ratios , using 200 nl of reservoir solution ) was performed using a Phoenix DT Drop Setter ( Rigaku , The Woodlands , TX ) . Needle- and blade-like crystals were obtained in multiple PEG-containing conditions , and these hits were optimized using finer screens . The crystal used for data collection was grown from 25% ( vol/vol ) PEG 3350 , 0 . 25 M ( NH4 ) 2SO4 , 0 . 1 M MES pH 6 . 1 ) . For harvesting , cryoprotectant ( reservoir solution supplemented with 100 mM NaCl and 40% [vol/vol] glycerol ) was added directly to the drop before looping crystals for freezing in liquid nitrogen in liquid nitrogen . Diffraction data were collected at Argonne National Laboratory using APS beamline 19ID using remote data collection . Diffraction data were processed using HKL2000 ( Otwinowski and Minor , 1997 ) . Crystals adopt space group C2 with one complex in the asymmetric unit . The diffraction was anisotropic ( Table 1 ) . We used 2 . 81 Å as the high-resolution cut-off to avoid excessive loss of completeness . All crystallographic calculations after diffraction data processing were performed using Phenix ( Adams et al . , 2010 ) . Phases were obtained by molecular replacement , using as search models the yeast α- and β-tubulin chains from PDB 4FFB ( Ayaz et al . , 2012 ) with GTP removed , and the structure of TOG2 from Stu2p ( PDB 2QK1 [Slep and Vale , 2007] ) . Model building was performed in Coot ( Emsley et al . , 2010 ) . Disordered regions were removed , and GTP and new sidechains were manually placed where the electron density indicated that was appropriate . The model was refined conservatively , and based on the behavior of Rfree and other indicators we decided that a combination of TLS with two grouped B-factors for each residue represented an optimal strategy . We also ran tests to optimize the relative weighting of the X-ray and chemical restraints , consistently observing that optimal results ( low Rfree and low Rfree-Rwork ) were obtained when the covalent geometry was tightly restrained to ideal values . Samples for analytical ultracentrifugation ( TOG1 , TOG2 , TOG1-TOG2 , and polymerization-competent yeast αβ-tubulin ) were dialyzed into RB100 ( 25 mM Tris pH 7 . 5 , 1 mM MgCl2 , 1 mM EGTA , 100 mM NaCl ) containing 20 μM GTP . Samples were mixed and incubated at 4°C for at least one hour prior to the experiment . All analytical ultracentrifugation experiments were carried out in an Optima XL-I centrifuge using an An50-Ti rotor ( Beckman–Coulter , Brea , CA ) . Approximately 390 μl of each sample were placed in charcoal-filled , dual-sector Epon centerpieces . Sedimentation ( rotor speed: 50 , 000 rpm ) was monitored using absorbance optics , and centrifugation was conducted at 20°C after the centrifugation rotor and cells had equilibrated at that temperature for at least 2 . 5 hr . Protein partial-specific volumes , buffer viscosities , and buffer densities were calculated using SEDNTERP ( Laue et al . , 1992 ) . The c ( s ) distributions were generated using SEDFIT ( Schuck , 2000; Schuck et al . , 2002 ) . Isotherms for TOG1 , TOG2 , and TOG1-TOG2 binding to αβ-tubulin were assembled by importing c ( s ) distributions into GUSSI ( available at http://biophysics . swmed . edu/MBR/software . html ) and integrating them to obtain the weighted-average sedimentation coefficients as a function of TOG concentration . These isotherms were exported to SEDPHAT ( Dam and Schuck , 2005 ) for evaluation . For the binding of TOG1 and TOG2 to αβ-tubulin , a simple bimolecular binding model was used . For the binding of TOG1-TOG2 to αβ-tubulin , a two-site binding model ( i . e . , two αβ-tubulins binding to one TOG1-TOG2 ) with two microscopic association constants was used . Quantities measured in separate experiments ( s-values of the individual components , s-values of TOG1-TOG2: ( αβ-tubulin ) 1 , and the association constants derived with the single-domain TOG constructs ) were treated as fixed parameters . We used a range of possible s-values for the TOG1-TOG2: ( αβ-tubulin ) 2 , informed by different assumptions about cooperativity as shown in Figure 5C and by hydrodynamic calculations ( García De La Torre et al . , 2000 ) . To calculate sedimentation coefficients , we first prepared compact ‘longitudinal’ and ‘lateral’ models by docking the TOG1 and TOG2 complexes onto the αβ-tubulins in the stathmin complex ( pdb 1SA0 [Ravelli et al . , 2004] ) and a section of microtubule ( coordinates provided by Ken Downing , Lawrence Berkeley lab ) , respectively . Extended models were generated by translating one TOG complex relative to the other . To calculate coordinates for the linker , we used the Modeller ( Sali and Blundell , 1993 ) plugin to UCSF Chimera ( Pettersen et al . , 2004 ) , calculating five different linker traces for each model in an attempt to capture some of the likely variability . Each resulting model was passed to HYDROPRO ( García De La Torre et al . , 2000 ) , and sedimentation coefficients were calculated using a solvent density of 0 . 99823 g/ml and otherwise default settings . The genetic rescue assays were adapted from Al-Bassam et al . ( 2006 ) and performed as previously described ( Ayaz et al . , 2012 ) . The parent strain was CUY1147ΔLEU2 ( CUY1147 ( Kosco et al . , 2001 ) in which the LEU2 gene was replaced with a marker coding for G418 resistance ) . This strain expresses a tagged version of Stu2p that is selectively degraded upon exposure to copper . The parent plasmid was pWP70 ( Wang and Huffaker , 1997 ) , a CEN plasmid that expresses ( non-degradable ) HA-tagged Stu2p from its endogenous promoter . For rescue assays CUY1147ΔLEU2 transformed with rescue plasmid or empty vector was grown overnight in CSM-Leu media , normalized to A600 = 1 , and plated at serial 10-fold dilutions onto CSM-Leu plates with or without 500 μM CuSO4 . Plates were incubated at 30°C or room temperature and imaged after several days . To collect a large number of TOG-domain containing sequences , we performed psiblast ( Altschul et al . , 1997 ) ( non-redundant [nr] database , with e-value = 0 . 005 over three iterations ) on amino acid sequences from TOG domains of known structure: the TOG1 and TOG2 domains from Stu2p ( S . cerevisiae ) , the TOG2 domain from XMAP215 ( Xenopus laevis ) ( Slep and Vale , 2007 ) , and the TOG3 domain from Zyg9 ( Caenorhabditis elegans ) ( Al-Bassam et al . , 2007 ) . The full protein sequences for all hits from each of the four runs were combined and identical hits were removed . They were then clustered using CLANS ( Frickey and Lupas , 2004 ) ( default parameters ) over 130 iterations . The Stu2 cluster was extracted and aligned using mafft ( Katoh and Standley , 2013 ) and passed to Jalview ( Waterhouse et al . , 2009 ) . Proteins that were missing residues in highly conserved regions of TOG1 and TOG2 , in particular in the boundaries adjacent to the start and end of the linker , were manually pruned . Based on the structures , we defined the boundaries for the end of TOG1 and beginning of TOG2 as K243 and L319 , respectively . A linker for a given protein was measured counting these boundary residues and all residues in between . To reduce ‘overcounting’ artifacts , before making the histogram we thinned the set of sequences using the ‘remove redundancy’ ( threshold = 0 . 99 ) feature of Jalview . We wrote a computer program to perform kinetic Monte Carlo simulations of microtubule elongation . We implemented an algorithm similar to one described previously ( VanBuren et al . , 2002 ) . In the program , the microtubule lattice is represented by a two dimensional lattice with a staggered periodic boundary condition to mimic the cylindrical microtubule structure . A 5 × 13 section of the lattice was designated as a ‘seed’ and considered to be permanently occupied . The program simulates microtubule elongation one biochemical reaction ( in this case subunit addition or dissociation , GTP hydrolysis is ignored ) at a time . A lattice site is considered available for subunit addition if it is empty and a neighboring site is occupied . The rate of subunit addition into any available site is given by kon*[αβ-tubulin] . Occupied sites ( excepting the seed ) are considered available for dissociation , with the rate of dissociation given by kon*KD where KD is the affinity of interaction , which is determined by the neighbor state ( number and type of lattice contacts ) and obtained from longitudinal and corner affinities through thermodynamic coupling ( Erickson and Pantaloni , 1981; VanBuren et al . , 2002 ) . Simulations begin with only 13 possible events , each on an association onto the end of a protofilament . Possible events are stored in an indexed priority queue , sorted by their ‘execution times’ ( Gibson and Bruck , 2000 ) . Execution times are determined by first sampling a random number x between 0 and 1 , and then calculating the time as − ( 1/rate ) *ln ( x ) , where rate gives the appropriate first- or pseudo-first order rate constant . At each step the event with the shortest execution time is implemented , the simulation time is advanced accordingly , and the list of possible events and their associated rates is updated to account for changes in subunit neighbor state . To obtain the length ( in μm ) of a simulated microtubule at a given time , we divide the number of subunits by 1625 , the number of αβ-tubulin subunits in 1 μm of microtubule . To model our proposed mechanism for polymerase action , we accelerated the rate at which unpolymerized subunits add adjacent to longitudinally associated ‘singletons’ at the microtubule end ( or next to more tightly bound ‘corner’ subunits ) ( Figure 7 ) . This acceleration mimics a ‘tethering’ effect . Control experiments showed that accelerating the rate for adding on top of a longitudinally-associated singleton had negligible effect ( not shown ) . To keep the model as simple as possible , we did not account for any potential stabilization of TOG-bound singletons at the microtubule end . | Dynamic filaments of proteins , called microtubules , have several important roles inside cells . Microtubules provide structural support for the cell; they help to pull chromosomes apart during cell division; and they guide the trafficking of proteins and molecules across the cell . The building blocks of microtubules are proteins called αβ-tubulin , which are continually added to and removed from the ends of a microtubule , causing it to grow and shrink . Other proteins that interact with the microtubules can help to speed up these construction and deconstruction processes . Ayaz et al . took a closer look at the structure of one particular family of proteins that make it easier for the microtubules to grow , using a technique called X-ray crystallography . The resulting images show two sites—called TOG1 and TOG2—on the enzymes that attach to the αβ-tubulin proteins . Ayaz et al . found that this binding can only occur when αβ-tubulin has a curved shape , which only happens when the tubulins are not included in , or are only bound weakly to the end of , a microtubule . Previous research suggested that the two binding sites might work together to provide ‘scaffolding’ that stabilizes the microtubule . However , genetic experiments by Ayaz et al . show that microtubules will grow even if one of the binding sites is missing . Both TOG1 and TOG2 bind to αβ-tubulin in the same way , and by using computer simulations Ayaz et al . found that this helps to speed up the growth of microtubules . This is because the enzyme's two sites concentrate the individual tubulin building blocks at the ends of the filament . For example , TOG2 could bind to the end of the microtubule , while TOG1 holds an αβ-tubulin protein nearby and ready to bind to the filament's end . This tethering allows the microtubules to be assembled more efficiently . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2014 | A tethered delivery mechanism explains the catalytic action of a microtubule polymerase |
Amyloid protein aggregates are associated with dozens of devastating diseases including Alzheimer’s , Parkinson’s , ALS , and diabetes type 2 . While structure-based discovery of compounds has been effective in combating numerous infectious and metabolic diseases , ignorance of amyloid structure has hindered similar approaches to amyloid disease . Here we show that knowledge of the atomic structure of one of the adhesive , steric-zipper segments of the amyloid-beta ( Aβ ) protein of Alzheimer’s disease , when coupled with computational methods , identifies eight diverse but mainly flat compounds and three compound derivatives that reduce Aβ cytotoxicity against mammalian cells by up to 90% . Although these compounds bind to Aβ fibers , they do not reduce fiber formation of Aβ . Structure-activity relationship studies of the fiber-binding compounds and their derivatives suggest that compound binding increases fiber stability and decreases fiber toxicity , perhaps by shifting the equilibrium of Aβ from oligomers to fibers .
Protein aggregates , both amyloid fibers and smaller amyloid oligomers , have been implicated in the pathology of Alzheimer’s and other neurodegeneration diseases ( Chiti and Dobson , 2006; Eisenberg and Jucker , 2012 ) . The increasing prevalence of Alzheimer’s disease in our aging societies , the associated tragedy for patients and their families , and the mounting economic burden for governments have all stimulated intense research into chemical interventions for this condition . Much work has been focused on screening compounds that prevent aggregation and the associated cytotoxicity of the amyloid β-peptide ( Aβ ) ( reviews by Sacchettini and Kelly , 2002; Bartolini and Andrisano , 2010; Hard and Lendel , 2012 ) . Screens have often focused on natural products from plants and lichens . These include polyphenols , such as epigallocatechin gallate ( EGCG ) from green tea ( Ehrnhoefer et al . , 2008 ) and curcumin from the spice turmeric ( Yang et al . , 2005 ) . These natural polyphenolic compounds show inhibition on the fibrillation of a variety of amyloid proteins , including Aβ40 as well as α-synuclein , IAPP and PrP ( Porat et al . , 2006; Dasilva et al . , 2010; Ono et al . , 2012 ) . Several dyes have also been found to ameliorate amyloid toxicity . Orcein from lichens appears to diminish toxic oligomers and enhance fiber formation ( Bieschke et al . , 2011 ) . Congo red , thioflavin T and their analogs , commonly used as staining reagents for amyloid detection , exhibit ameliorative effects on neurodegenerative disorders , such as Alzheimer’s , Parkinson’s , Huntington’s , and prion diseases ( Frid et al . , 2007; Alavez et al . , 2011 ) , however their application is limited by significant side effects ( Klunk et al . , 2004 ) . Additional screens have identified a variety of molecules , including proteins ( Evans et al . , 2006 ) , antibodies ( Kayed et al . , 2003; Ladiwala et al . , 2012 ) , synthetic peptide mimetics ( Findeis , 2002; Kokkoni et al . , 2006; Takahashi and Mihara , 2008; Cheng et al . , 2012 ) and small molecules ( Wood et al . , 1996; Williams et al . , 2005; McLaurin et al . , 2006; Necula et al . , 2007; Bartolini and Andrisano , 2010; De Felice et al . , 2001; Ladiwala et al . , 2011; Hard and Lendel , 2012; Kroth et al . , 2012 ) , that inhibit Aβ fibrillogenesis and/or Aβ-associated cytotoxicity in vitro . While most efforts have targeted the deposition of Aβ fibers as the hallmark of Alzheimer’s , smaller amyloid oligomers are now receiving greater attention as the possible toxic entities in Alzheimer’s and other neurodegenerative diseases ( Hartley et al . , 1999; Cleary et al . , 2005; Silveira et al . , 2005 ) . Furthermore , emerging evidence suggests that mature , end-stage amyloid fibers may serve as a reservoir , prone to releasing toxic oligomer ( Xue et al . , 2009; Cremades et al . , 2012; Krishnan et al . , 2012; Shahnawaz and Soto , 2012 ) . Recent screens have identified compounds that reduce Aβ cytotoxicity , without interfering with Aβ fibrillation ( Chen et al . , 2010 ) or promoting the formation of stable Aβ aggregates ( Bieschke et al . , 2011 ) . Structural information about protein targets often aids drug development , so here we take a structure-based approach , combined with computational screening , to discover amyloid interacting compounds that reduce amyloid toxicity . This approach has been enabled by the determination of atomic structures of the adhesive segments of amyloid fibers , termed steric zippers ( Nelson et al . , 2005 ) , and of solid state NMR-based structures of amyloid fibers ( such as full-length Aβ fibers [Luhrs et al . , 2005; Petkova et al . , 2005] and the HET-s prion domain complexed with Congo Red [Schutz et al . , 2011] ) . The steric zipper structures reveal a common motif for the spine of amyloid fibers , in which a pair of fibrillar β-sheets is held together by the side-chain interdigitation ( Sawaya et al . , 2007 ) . We focus on Aβ , a peptide of 39–42 residues cleaved from the Amyloid precursor protein ( APP ) associated with Alzheimer’s , as a target for inhibitor discovery . The segment Aβ16–21 with the sequence KLVFFA is an amyloid-forming peptide , which packs in a steric zipper form , and has been identified as the spine of the full-length Aβ fiber ( Luhrs et al . , 2005; Petkova et al . , 2006; Colletier et al . , 2011 ) . Co-crystal structures have been determined for small molecules in complex with the fibrillar β-sheets of Aβ16–21 ( Landau et al . , 2011 ) . One of these structures—Aβ16–21 with the dye Orange G—reveals the specific pattern of hydrogen bonds and apolar interactions between orange G and the steric zipper: the negatively charged dye binds specifically to lysine side chains of adjacent sheets , and its planar aromatic portion packs against apolar residues ( phenylalanine and valine ) of adjacent sheets . By creating a tight , low energy interface across several β-strands within fiber core , this fiber-binding molecule appears to stabilize the fiber structure . With this atomic structure as a basis , we are able to screen for small molecular compounds that bind to amyloid fibers , stabilizing them and possibly reducing amyloid toxicity . Applying our structure-based screening procedure , we screen computationally for compounds that bind to Aβ fibers , termed BAFs ( Binders of Amyloid Fibers ) and then experimentally test their effects on Aβ aggregation and cytotoxicity .
We have devised a structure-based procedure for the identification of small molecules that bind to amyloid and affect amyloid toxicity ( Figure 1 ) . The procedure starts from a co-crystal structure of a ligand bound to an amyloidogenic segment of Aβ ( Landau et al . , 2011 ) , the dye orange G bound to the fiber-like crystal structure of KLVFFA ( Aβ16–21 ) segment . This structure reveals the chemical environment or ‘pharmacophore’ presented by the ligand binding site of this Aβ segment , that is , orange G binds to stacked β-sheets of Aβ . Knowledge of the amyloid pharmacophore ( Figure 1A ) permitted us to screen for compounds that could be expected to bind in this chemical environment , possibly stabilizing amyloid fibers . 10 . 7554/eLife . 00857 . 003Figure 1 . Structure-based identification of small compound inhibitors of Aβ toxicity . In step ( A ) the crystal structure ( Landau et al . , 2011 ) is determined of a complex of an amyloidogenic segment of Aβ ( in this case residues 16-KLVFFA-21 of the spine of the Aβ fiber ) with an amyloid-binding Ligand X ( in this case orange G ) , revealing aspects of the pharmacophore for Ligand X . Prior to step ( B ) a large library of available compounds is selected for computational docking ( ∼18 , 000 purchasable compounds in this case ) . In step ( B ) computational docking is applied to test the compatibility of each member of the library for the pharmacophore of the amyloidogenic segment defined in step ( A ) . In step ( C ) , the top scoring members of the library are tested for compatibility of binding within a full-length Aβ fiber ( in this case the 400 top scoring members were tested on a solid state NMR-derived model of an Aβ fiber , pdb entry 2LMO ) ( Petkova et al . , 2006 ) . The representative models from steps B and C are shown in Figure 1—figure supplements 1 and 2 . In step ( D ) , the compounds are ranked by tightest binding energy and best shape complementarity for the pharmacophore . In step ( E ) , the top-ranking compounds ( 25 in this case ) are selected for experimental characterization and validation , including NMR assessment of binding , EM assays of their effects on fiber formation , and cell viability assays for their effects on Aβ cytotoxicity . In step ( F ) , new compounds ( 9 in this case ) and compound derivatives ( 17 in this case ) are selected for an additional cycle of computational and experimental testing , based on their similarity to the lead compounds from the initial cycle . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 00310 . 7554/eLife . 00857 . 004Figure 1—figure supplement 1 . Structural models of the representative BAFs and orange G docked to the side of the KLVFFA ( Aβ16–21 ) fiber . In step ( B ) ( Figure 1 ) , a large library of ∼18 thousand commercially available compounds were docked onto the Aβ16–21 fiber , and ∼400 top ranking compounds , whose binding energy and shape complementary score are better than the control molecule orange G , were selected for the next docking step . The models of representative BAFs docked on single beta-sheet of Aβ16–21 fiber are compared to that of orange G . ( A ) . A side view of the compound BAF1 ( in green sticks ) docked on the Aβ16–21 fiber ( in a grey color ) with a predicted binding energy of −8 . 4 kcal/mol . ( B ) . A side view of BAF8 ( in cyan sticks ) docked on the Aβ16–21 fiber with a predicted binding energy of −12 kcal/mol . ( C ) . A side view of orange G ( in orange sticks ) docked on the Aβ16–21 fiber with a predicted binding energy of −8 . 0 kcal/mol . The charge interactions between the compounds and Lysine residues of Aβ16–21 fiber are highlighted by black lines . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 00410 . 7554/eLife . 00857 . 005Figure 1—figure supplement 2 . Structural models of the representative BAFs and orange G docked onto the full-length Aβ fiber . In step ( C ) ( Figure 1 ) , the top-ranking compounds after the first docking step were further filtered by docking onto full-length Aβ fiber model ( pdb entry 2LMO ) ( Petkova et al . , 2006 ) . The models of representative BAFs docked onto Aβ fiber are compared to that of orange G . ( A–C ) . A top view of the compounds ( BAF1 , BAF8 and orange G ) docked onto Aβ fiber ( in a light yellow color ) . ( D–F ) . A side view of the same compounds docked onto Aβ fiber . ( A and D ) . BAF1 ( in a green color ) binds to the side of Aβ fiber ( in a light yellow color ) with a predicted binding energy of −10 kcal/mol . ( B and E ) . BAF8 ( in a cyan color ) binds to the side of Aβ with a predicted binding energy of −12 kcal/mol . ( C and F ) . Orange G ( in an orange color ) binds to the side of Aβ fiber with a predicted binding energy of −9 kcal/mol . The charge interactions between the compounds and Lysine residues are highlighted by black lines . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 00510 . 7554/eLife . 00857 . 006Figure 1—figure supplement 3 . Alternative binding modes of BAF1 with the Aβ full-length fibers . When identifying BAFs by two steps of computational docking ( Figure 2A as well as step B and C in Figure 1 ) , most models of the second docking step ( docking to full-length Aβ fiber in step ( C ) retained their binding modes found in the previous docking step ( docking to Aβ16–21 fiber in step ( B ) . Interestingly , docking of BAF1 onto full-length Aβ fiber not only recapitulated the initial binding mode found in previous Aβ16–21 docking step but also revealed the different binding mode with comparable binding energies . Two examples of those alternative binding modes are shown in ( A and B ) . In both modes , BAF1 tends to use its polar ( hydroxyl ) group to interact with the charged residues Glu22 of Aβ and use its non-polar ( aromatic ) portion to pack against the hydrophobic residues Phe20 of Aβ full fibers . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 006 For assembling the compounds in our screening library , we sought three characteristics: ( a ) commercially available compounds since we intended to follow the in silico screening with experimental validation; ( b ) compounds with known three-dimensional structures such that our screening would be as realistic as possible; ( c ) generally flat compounds able to bind to the β-sheets of the steric zipper , as does orange G . Some ∼11 , 000 compounds having the first two characteristics ( CSD-ZINC set ) were selected as the intersection of molecules found both in the Cambridge Structure Database ( http://www . ccdc . cam . ac . uk ) and the Zinc Database of purchasable compounds ( http://zinc . docking . org/ ) ( Irwin and Shoichet , 2005 ) . This CSD-ZINC set spans a variety of structural shapes and molecular properties . A second set of ∼7000 compounds , the Flat Compound Set , was gathered from the ZINC database to include molecules expected to bind to the flat surface of a steric zipper . The members of this set contain multiple aromatic rings or one aromatic ring with additional planar groups . Computational screening was carried out with the RosettaLigand program ( Davis and Baker , 2009 ) , after adapting its docking approach to carry out high-throughput screening ( Figure 2 ) . The conformational flexibilities of ligand and protein side chains are in a ‘near-native’ perturbation fashion , meaning that the fine sampling of conformations was restrained to be close to the starting conformation . A balance was achieved between extensive sampling and the speed required for screening a large compound library by fine sampling of side chain and ligand torsion angles only around their starting conformations , as illustrated by sticks in Figure 2C . 10 . 7554/eLife . 00857 . 007Figure 2 . Computational screening for fiber-binding compounds . ( A ) . Outline of our procedure for structure-based screening . We prepare two sets of compounds ( shown in the upper left ) for screening against both types of fibers shown in the upper right . Compound Set 1 is the intersection of the ZINC Database of purchasable compounds with the Cambridge structural database ( CSD ) of known structures . Set 2 consists of other flat aromatic and multiple conjugated compounds found in the ZINC Database . The full description of each computational step is in ‘Materials and methods’ . ( B ) . Distribution of calculated binding energies for the compound libraries of Sets 1 and 2 . Those top-ranking compounds have better predicted binding energy than orange G . Structural comparison of docked models of such compound BAF8 and orange-G is discussed in the Figure 2—figure supplement 1 . Notice the starred bins which suggest that some members of Set 2 , containing flat compounds , tend to be among the top scoring compounds , presumably having the tightest binding to the flat fiber surface . ( C ) . The conformational ensemble of a compound representative shown docked onto the Aβ16–21 fiber structure . ( D ) . A model of BAF8 docked onto an NMR-derived model of full-length Aβ fiber . Notice that the apolar ring structure of the compound binds to the relatively flat apolar ( gray ) surface of the fiber , and the polar moieties of the compound ( red ) form hydrogen bonds to the polar groups of the fiber ( yellow ) . The stereo view of BAF8 model is shown in Figure 2—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 00710 . 7554/eLife . 00857 . 008Figure 2—figure supplement 1 . Structural comparison between docked models of BAF8 and orange G . BAF8 has a chemical structure similar to orange G ( top panels ) . The comparison of the shape complimentary at binding interfaces reveals that BAF8 binds more tightly to the side of fibers than orange G . ( A ) . A top view of the docked model of BAF8 ( in a cyan color ) with the predicted binding energy of −12 kcal/mol highlights the tight shape complementary at the fiber-ligand interface . ( B ) . A top view of the docked model of orange G ( in an orange color ) with the predicted binding energy of −8 kcal/mol shows a poorly packed interface with cavities . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 00810 . 7554/eLife . 00857 . 009Figure 2—figure supplement 2 . Stereo view of the structural model of BAF8 with Aβ fiber . A wall-eyed stereo view of BAF8 ( Figure 2D ) ( in cyan sticks ) docked to the side of an Aβ16–21 fiber ( light yellow ) reveals good non-polar and polar interaction across binding interfaces . The hydrophobic binding site for the aromatic portion of BAF8 is indicated by grey mesh surfaces to highlight the good shape complementary . The polar interaction of hydrogen bonds between the charged residues Lys 16 of Aβ and the polar portion of BAF8 are indicated by black thick lines . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 009 In the screening steps of computational docking ( Figure 2A ) , a library of ∼18 , 000 purchasable compounds ( Sets 1 and 2 ) was scanned computationally for structural compatibility with the pharmacophore ( ligand binding site ) presented by a single sheet of the Aβ16–21 steric zipper . Structural compatibility was assessed by a combination of binding energy ( Meiler and Baker , 2006 ) and steric complementarity ( Lawrence and Colman , 1993 ) . After computational docking , the distribution of calculated binding energies suggests that , statistically the flat compounds from Set 2 fit more snugly on the flat surfaces of Aβ16–21 fibers than those with diverse shapes in Set 1 ( Figure 2B ) . The best scoring compounds were screened further by requiring that each is also structurally compatible with the solid-state NMR-derived model of the Aβ full-length fiber structure ( Petkova et al . , 2006 ) ( Figure 1C and Figure 1—figure supplement 3 ) . After in silico screening of a library of ∼18 , 000 purchasable compounds , twenty-five of the top-ranking compounds all with better scores for binding energy and steric complementarity than orange G ( Figure 1D , Figure 2—figure supplement 1 ) , were selected for experimental validation . First these 25 compounds were tested for their ability to protect mammalian cells from Aβ toxicity ( Figure 1E , Tables 1 and 2 ) , and five of them were found to reduce the toxic effects of Aβ . These five were tested for binding to both Aβ1–42 and Aβ16–21 fibers by NMR . Two were found to have tighter binding than orange G , and the others gave insufficient NMR signals for detection . To expand this set of the five compounds , a second cycle of inhibitor discovery was performed . From the computed positions of the five compounds , a refined pharmacophore was inferred ( Figure 1F ) , and used in the next cycle of screening . Added to the compound set were nine additional compounds apparently related to the five lead compounds from the initial cycle , plus 17 chemical derivatives of compounds ( Tables 1 and 3 ) . The second cycle produced three additional compounds and three compound derivatives that also protected the mammalian cells from Aβ fibers . One of these compounds was confirmed by NMR to bind to Aβ fibers . The detailed description of those experimental results is as follows . 10 . 7554/eLife . 00857 . 010Table 1 . List of all tested BAF compoundsDOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 010CompoundMolecular formulaMolecular weight*Sources/purchasingRescuing percentage ( % ) ZINC entryBAF1C20H8Br4O5648Sigma-Aldrich44 ± 7ZINC04261875BAF2C19H14O5S354Sigma-Aldrich4 ± 3ZINC03860918BAF3C16H13NO3267Ryan Scientific4 ± 5ZINC04289063BAF4C24H16N2O6428Aldrich88 ± 22ZINC13346907BAF5C16H7Na3O10S3524Sigma-Aldrich11 ± 7ZINC03594314BAF6C26H20N2360Alfa-Aesar5 ± 7ZINC08078162BAF7C18H12N6312Alfa-Aesar2 ± 2ZINC00039221BAF8C17H14N2O5S358Sigma-Aldrich23 ± 11ZINC12358966BAF9C19H13N3O4S379NCI plated 2007†−3 ± 22ZINC03954432BAF10C17H13NO3279NCI plated 20073 ± 5ZINC00105108BAF11C20H13N2O5S393NCI plated 200748 ± 12ZINC04521479BAF12C13H8Br3NO434NCI plated 200738 ± 6ZINC12428965BAF13C19H16ClNO4358Sigma-Aldrich0 ± 2ZINC00601283BAF14C10H6S2O8318Sigma-Aldrich3 ± 3ZINC01532215BAF15C23H28O8432Sigma-Aldrich13 ± 4ZINC00630328BAF16C19H19NO5341Sigma-Aldrich5 ± 8ZINC28616347BAF17C23H25N5O2404Sigma-Aldrich6 ± 3ZINC00579168BAF18C24H16O2336ChemDiv6 ± 2ZINC02168932BAF19C18H14N2O6354ChemDiv3 ± 4ZINC01507439BAF20C25H19N5OS438ChemDiv8 ± 4ZINC15859747BAF21C19H14Br2O418ChemDiv6 ± 3ZINC38206526BAF22C21H16N2O3S2408Life Chemicals3 ± 5ZINC04496365BAF23C16H11ClO5S351Enamine Ltd3 ± 5ZINC02649996BAF24C23H19NO3357Sigma-Aldrich16 ± 5ZINC03953119BAF25C14H8Cl2N4303Sigma-Aldrich4 ± 3ZINC00403224BAF26C17H10O4278Aldrich46 ± 23ZINC05770717BAF27C21H16BrN3O6486ChemBridge4 ± 1ZINC01208856BAF28C17H12N2O3292ChemBridge2 ± 4ZINC00061083BAF29C22H10N4O2362ChemBridge1 ± 5ZINC00639061BAF30C14H8O5256Aldrich18 ± 13ZINC03870461BAF31C19H21NO3311Sigma84 ± 12ZINC00011665BAF32C15H14O7306Sigma-Aldrich15 ± 9ZINC03870336BAF33C27H33N3O8528Sigma-Aldrich7 ± 2SIGMA-R2253§BAF34C30H16N4O14S4785Aldrich‡ALDRICH-S432830§orange GC16H12N2O7S2408Sigma-Aldrich−2 ± 8ZINC04261935The 25 compounds ( BAF1-25 ) are from the first round , and the nine compounds ( BAF26-34 ) are from the second round . Another set of the 17 derivatives of the BAFs are shown in Table 3 . *Molecular weight ( anhydrous basis ) excluding the salt and water molecules . †National Cancer Institute ( NCI ) free compound library ( http://dtp . nci . nih . gov/ ) . ‡Toxicity results of BAF34 were not consistent among several independent replica experiments , possibly due to impurity and the high molecular weight of the compound . §ZINC entry of the compound is not applicable , and the catalog number from Sigma-Aldrich is provided . 10 . 7554/eLife . 00857 . 011Table 2 . Detailed list of the active BAF compoundsDOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 011CompoundMolecular formulaMolecular weight*Sources/companiesPurityRescuing percentage§ ( % ) ZINC entry code¶SMILES stringPC12HelaBAF1C20H8Br4O5647 . 9Sigma-Aldrich∼99%38 ± 1144 ± 7ZINC04261875c1ccc2c ( c1 ) C ( =O ) OC23c4ccc ( c ( c4Oc5c3ccc ( c5Br ) O ) Br ) OBAF4C24H16N2O6428 . 4Aldrich≥95%85 ± 1888 ± 22ZINC13346907c1cc ( c ( cc1O ) O ) c2cc3c ( cc2N ) oc-4cc ( =O ) c ( cc4n3 ) c5ccc ( cc5O ) OBAF8C17H14N2O5S358 . 4Sigma-Aldrich≥90%26 ± 1223 ± 11ZINC12358966Cc1ccc ( c ( c1 ) /N=N/c2c3ccccc3c ( cc2O ) S ( =O ) ( =O ) [O-] ) OBAF11C20H13N2O5S393 . 5NCI plated 2007†51 ± 1148 ± 12ZINC04521479c1ccc2c ( c1 ) ccc ( c2O ) /N=N/c3c4ccccc4c ( cc3O ) S ( =O ) ( =O ) [O-]BAF12C13H8Br3NO433 . 9NCI plated 2007†19 ± 638 ± 6ZINC12428965c1cc ( ccc1/N=C/c2cc ( cc ( c2O ) Br ) Br ) BrBAF26C17H10O4278 . 3Aldrich‡60 ± 2146 ± 23ZINC05770717c12c ( cc ( cc1 ) C ( =O ) C=O ) Cc1c2ccc ( c1 ) C ( =O ) C=OBAF30C14H8O5256 . 2Aldrich‡37 ± 1818 ± 13ZINC03870461c1cc2c ( cc1O ) C ( =O ) c3c ( ccc ( c3O ) O ) C2=OBAF31C19H21NO3311 . 4Sigma≥98%92 ± 2284 ± 12ZINC03874841CCCN1CCC2=C3C1CC4=C ( C3=CC ( =C2 ) O ) C ( =C ( C=C4 ) O ) OBAFs 1 , 4 , 8 , 11 , 12 are from the first round . BAFs 26 , 30 , 31 are from the second round . *Molecular weight ( anhydrous basis ) excluding the salt and water molecules . †With the standard of NCI free compound library . ‡Analytical data for AldrichCPR products are not available . §Rescue percentage is a scaled cell survival rate . ¶Entry code for the ZINC database ( http://zinc . docking . org ) . 10 . 7554/eLife . 00857 . 012Table 3 . List of the representative BAFs 11 , 30 , 31 and their derivativesDOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 012CompoundMolecular formulaMolecular weightDescriptionToxicity inhibition ( % ) ZINC entry/catalog no . BAF31C19H21NO331184 ± 12ZINC03874841 BAF31ΔOHC19H21NO2295remove one hydroxyl ( OH ) 15 ± 2ZINC03874841BAF30C14H8O525618 ± 13ZINC03870461 BAF30αRC22H20O13492add additional R group away from binding interface20 ± 10ZINC28095922 BAF30σOHAαOHC14H8O6272change one OH ( A ) position; add another OH9 ± 9ZINC03874832BAF30σOHAΔOHBαCOOC15H8O6284move one OH ( A ) position; delete an OH from loc B; add a carboxyl9 ± 3ZINC04098704 BAF30σOHABαCH3C15H10O5270move two OH ( AB ) positions; add a methyl6 ± 3ZINC03824868BAF11C20H13N2O5S39348 ± 12ZINC04521479 BAF11ISOC20H13N2O5S393isomer form of BAF1133 ± 5ZINC12405071 BAF11σR1C20H14N4O8S2502change the aromatic group35 ± 9ZINC25558261 BAF11σR2 ( BAF8 ) C17H14N2O5S358change the aromatic group22 ± 11ZINC12358966 BAF11σR3C16H12N2O6S360change the aromatic group28 ± 4ZINC04900892 BAF11αNO2-C20H12N3O7S438add charged group ( nitro ) 15 ± 6ZINC16218542 BAF11ISOαCOO-C21H12N2O7S436BAF11 isomer; add charged group ( carboxyl ) 6 ± 5ZINC03861030 BAF11ISOαSO3-C20H11N2O11S3552BAF11 isomer; add charged group ( sulfate ) 2 ± 5SIGMA-33936 BAF11ΔOHσRC20H14N2O4S378remove an OH;change the position of the aromatic group15 ± 6ZINC04803992 BAF11ΔOHαSO3−C20H14N2O7S2458remove an OH; add sulfate group12 ± 3ZINC03954029 BAF11ΔOHαRC20H18N4O5S426remove an OH; add additional group to the aromatic ring12 ± 6ZINC04416667 BAF11σOHαR1C24H20N4O4S461swap the position of the OH and aromatics5 ± 5ZINC04804174 BAF11σOHαR2C16H19N3O5S365swap the position of the OH and aromatics4 ± 6ZINC17378758 Having identified compounds that bind Aβ fibers , by a structure-based procedure , we tested their effects on the cytotoxicity of Aβ1–42 fiber against two mammalian cell lines: PC12 and HeLa ( Figure 3 ) . Five BAFs—1 , 4 , 8 , 11 , and 12—in the initial cycle and three additional BAFs—26 , 30 , and 31—from the second cycle , with diversified chemical structures shown in Figure 4 , significantly increased both PC12 and HeLa cell survival after 24 hr incubation with Aβ1–42 ( 0 . 5 µM ) at concentration of 2 . 5 µM , while the BAFs alone had little or no effect on cell survival ( Figure 3—figure supplement 1 ) . Three BAFs—11 , 26 , and 31—showed clear dose-response profiles in their protection of both PC12 and HeLa cells ( Figure 3B ) . Among them , the two best BAFs—26 and 31—were tested and did not affect the cytotoxicity of amyloid fibers other than Aβ ( Figure 3—figure supplement 2 ) . Although all of these BAFs provide protection against Aβ toxicity , none diminish the amount of Aβ fibers in electron micrographs ( Figure 3C ) . 10 . 7554/eLife . 00857 . 013Figure 3 . Experimental characterization of compounds that bind to amyloid fibers . Our newly discovered BAFs diminish Aβ1–42 toxicity without significantly reducing Aβ1–42 fibrillation . ( A ) . Eight BAFs reduce Aβ toxicity in mammalian cell lines ( PC12 in orange; HeLa in green ) . These identified compounds with diversified chemical structures are quite different from orange G , whose co-crystal structure with an amyloid segment is the basis of our approach ( Figure 4 and Table 2 ) . For each compound , 2 to 4 repeats of each independent experiment were performed . For each experimental repeat , four replicates per sample per concentration were tested . The symbol * indicates a p<0 . 1; the symbol ** indicates a p<0 . 01 and the symbol *** indicates a p<0 . 001 . The student’s t-test and p-value analysis are in Table 4 . ( B ) . The representative BAFs—31 , 26 , and 11—inhibit Aβ cyto-toxicity in a dose-dependent manner . ( C ) . Transmission electron microscopy ( TEM ) images of Aβ fibers alone and Aβ fibers with the BAFs , the same samples prepared for cell viability assay . All 8 BAFs that diminish Aβ toxicity do not noticeably diminish Aβ fibrillation . Scale bars indicate 200 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 01310 . 7554/eLife . 00857 . 014Figure 3—figure supplement 1 . The BAFs alone exhibit little or no toxicity on mammalian cell lines . Incubating cells with or without BAFs for 24 hours caused little or no change for cell viability of both PC12 and HeLa . The error bars are calculated from four experiment replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 01410 . 7554/eLife . 00857 . 015Figure 3—figure supplement 2 . BAFs cannot reduce the cytotoxicity of amyloid fibers formed by IAPP and α-synuclein , as much as those fibers formed by Aβ . The final concentration of IAPP is 1 µM and α-synuclein is 2 µM . The molar ratio of amyloid fibers and BAFs is 1:1 . BAFs ( 26 and 31 ) , which significantly reduces Aβ toxicity ( Figure 3 ) , cannot rescue the toxicity of IAPP and α-synuclein , suggesting that the toxicity alleviating effect of BAFs are specific to the fibers for which they were designed . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 01510 . 7554/eLife . 00857 . 016Table 4 . Student’s t-test and p value analysis suggests that BAFs reduce the cytotoxicity of Aβ fibers significantlyDOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 016Average of cell viability ( n = 4 ) SD ( σ ) Comparison to Aβ fiber alonet valuep valueHeLa cell line Aβ fiber alone0 . 400 . 05// BAF10 . 660 . 048 . 45E-05 BAF40 . 930 . 137 . 41E-4 BAF80 . 540 . 063 . 31E-2 BAF110 . 690 . 076 . 62E-04 BAF120 . 630 . 047 . 61E-04 BAF260 . 680 . 143 . 85E-3 BAF300 . 510 . 082 . 34E-2 BAF310 . 910 . 0711 . 57E-06PC12 cell line Aβ fiber alone0 . 370 . 07// BAF10 . 610 . 074 . 91E-3 BAF40 . 900 . 118 . 07E-05 BAF80 . 530 . 073 . 21E-2 BAF110 . 690 . 076 . 52E-4 BAF120 . 490 . 042 . 92E-2 BAF260 . 740 . 135 . 01E-3 BAF300 . 600 . 113 . 58E-3 BAF310 . 950 . 147 . 41E-4The Student’s T-test and p-value are based on the comparison to Aβ fiber alone . 10 . 7554/eLife . 00857 . 017Figure 4 . Diversified chemical structures of 8 active BAF compounds that reduce Aβ toxicity . Orange G in an orange box is also displayed for comparison . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 017 Promising candidate binders from in silico screening and toxicity tests were validated by titration of Aβ fibers into solutions of each compound , as monitored by NMR signals of aromatic protons of the compound ( Figure 5 ) . The proton resonances of the freely rotating compounds disappear as the compound binds to the fibers . By increasing the amount of fibers , an apparent Kd for compound binding can be estimated . From in silico screening , all tested BAF compounds are calculated to bind more tightly to Aβ fibers than orange G . In NMR studies , the apparent Kd of orange G binding to Aβ16–21 fibers was found to be 43 ± 21 µM , whereas the apparent Kd of BAF1 binding to Aβ16–21 fibers is 12 ± 7 µM . BAFs were found to bind to both Aβ16–21 fibers and Aβ1–42 fibers . Figure 5F shows a notable correlation between the calculated binding energies and the reduction in NMR peak areas upon Aβ binding . That is , all BAFs with predicted binding energy better than orange G also reduce NMR peak areas more than orange G . On the other hand , BAF31ΔOH , a derivative of BAF31 by removal of a key hydroxyl group essential for binding , exhibits both a worse calculated binding energy and a diminished reduction of NMR peak upon titration of Aβ1–42 fibers . 10 . 7554/eLife . 00857 . 018Figure 5 . NMR evidence for binding of compounds to both Aβ16–21 and Aβ1–42 fibers . NMR binding experiments were performed on BAF compounds and the dye orange G . By monitoring the aromatic regions of the 1H NMR spectra of BAFs 1 , 8 , and 31 , these compounds were shown to bind to both Aβ16–21 and Aβ1–42 fibers more tightly than does orange G . As shown in ( A and B ) , BAF1 binds to Aβ16–21 fibers with affinity stronger than orange G . The determination of binding parameters for Aβ16–21 fibers is detailed in Table 5 and Figure 5—figure supplements 1 and 3 . In panel ( A ) , the 1H NMR spectrum of compound BAF1 ( at 100 μM ) is shown as a function of increasing concentration of Aβ16–21 fibers ( 0–500 μM , as monomer ) . The insert shows the area decrease of BAF1 NMR peaks as a function of Aβ16–21 concentration , and the red curve fitting the data defines an apparent Kd of 12 ± 7 µM . In panel ( B ) , the NMR spectrum of orange G ( 50 μM ) is plotted against increasing concentration of Aβ16–21 fibers ( 0–950 μM ) , giving an apparent Kd of 43 ± 21 µM . In ( C , D and E ) , BAFs 1 and 8 both bind to Aβ1–42 fibers more strongly than orange G . Notice that the molar ratio of BAFs to Aβ1–42 fibers is comparable to that used in cell toxicity assays ( Figure 3 ) . ( F ) . The calculated binding energies of BAFs—1 , 8 , and 31—to Aβ1–42 fibers are compared to the decreases in NMR peak of these compounds upon their binding to full-length Aβ fibers . These three BAFs have higher affinities and a larger NMR peak reduction than orange G while the ‘knock-out’ derivative with removal of key interactions ( BAF31ΔOH ) discussed below has a weaker calculated affinity and a smaller NMR peak reduction than orange G . We observe good correlation between computed energies and experimental data from NMR . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 01810 . 7554/eLife . 00857 . 019Figure 5—figure supplement 1 . NMR peak assignment of BAF1 with Aβ16–21 fiber . The 1D 1H NMR spectrum shows the aromatic proton regions of BAF1 upon the titration of Aβ16–21 fibers shown in Figure 5A . The insert is the chemical structure of BAF1 with the color-labeled aromatic proton observed in the NMR spectrum . The arrows with different colors indicate the proton assignment for NMR peaks . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 01910 . 7554/eLife . 00857 . 020Figure 5—figure supplement 2 . NMR peak assignment of the control compound orange G with Aβ16–21 fiber . The 1D 1H NMR spectrum shows the aromatic proton regions of orange G against the increasing concentrations of Aβ16–21 fibers shown in Figure 5B . The insert is the chemical structure of orange G with the highlighted label of the aromatic proton shown in the NMR spectrum . The arrows with different colors indicate the proton assignment for NMR peaks . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 02010 . 7554/eLife . 00857 . 021Figure 5—figure supplement 3 . NMR titration of BAF8 with Aβ16–21 fibers . To validate our computation methods , NMR titration experiments were performed . ( A ) One representative peak of aromatic protons of the 1D 1H NMR spectra of the compound BAF8 ( at 100µM ) upon Aβ16–21 fibers titration ( 0–500 µM , monomer equivalent ) . ( B ) Fitting curve upon the area decrease of BAF8 NMR peaks as a function of fiber concentration . The apparent Kd of BAF8 ( 24 ± 5 µM ) is lower than that of orange-G ( Figure 4B ) , indicating the tighter binding affinity of BAF8 to Aβ16–21 fibers . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 02110 . 7554/eLife . 00857 . 022Table 5 . Predicted binding energy and experimental measurement of the binding of two BAFs and orange G against both Aβ16–21 ( KLVFFA ) and full-length Aβ fibersDOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 022Binding to KLVFFA fiberBinding to Aβ fiberPredicted binding energy ( kcal/mol ) NMR Kd ( µM ) Predicted binding energy ( kcal/mol ) NMR peak reduction ( % ) BAF1−812−108BAF8−1224−1213orange G−843−96The determination of the binding parameters with KLVFFA fiber is detailed in Table 6 . 10 . 7554/eLife . 00857 . 023Table 6 . Comparison of the measured binding parameters of the representative BAFs with orange G by NMR titrationsDOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 023CompoundPredicted binding energy ( kcal/mol ) fmaxKd ( µM ) BAF1−80 . 47 ± 0 . 0412 ± 7BAF8−120 . 82 ± 0 . 0424 ± 5Orange-G−80 . 46 ± 0 . 0643 ± 21The second column lists the predicted binding energy for each top docked model of BAF compounds with KLVFFA fiber , and the binding energy of Orange-G with KLVFFA fiber were also calculated for comparison . Our computational method identified the BAF with better fit to the binding interface than Orange-G . We then used NMR titration to determine the binding affinity . Our previous mass spectrometric analyses of the crystal of the Orange-G with KLVFFA fibers have suggested a binding ratio of compound:fiber with the range of 1:1 to 1:10 ( Landau et al . , 2011 ) . Together with our structural models and single binding site assumption , we estimated the binding ratio to be 1:3 . Accordingly , calculated NMR binding parameters are listed in the table . The third column fmax is the maximum fraction of NMR signal decrease of compound upon binding saturation ( ‘Materials and methods’ ) . Based on the lead compounds found in the initial cycle of the procedure , we carried out a second cycle to expand our understanding of the Aβ pharmacorphore . BAF11 ( Figure 6A ) , one of the lead compounds in the initial cycle , was used to perform structure-activity relationship studies . Twelve derivatives of BAF11 were scanned to pinpoint the essential apolar and polar interactions for the pharmacorphore refinement ( Figure 6B , Figure 6—figure supplement 1 ) . These derivatives are grouped in five classes , whose effects on Aβ toxicity have been tested ( Figure 6C ) . Classes I and II assess the polar region of BAF11 , which makes hydrogen bonds to charged Lys16 ladders of the Aβ fiber: the deletion of the hydroxyl group ( Class I ) significantly decreased the inhibition of toxicity; the swapping of the hydroxyl group with the aromatic tail ( Class II ) almost abolished inhibition of toxicity . Classes III , IV , and V focused on the aromatic moieties of BAF11: altering the sizes of aromatic groups ( Class III ) showed little change in inhibition of toxicity while adding charged or polar groups within aromatic region ( Classes IV and V ) resulted in a significant decrease of inhibition of toxicity . These differences among BAF11 derivatives in inhibition of toxicity ( Figure 6C ) further validated our structure-based approach and provided guidelines for the refinement of Aβ pharmacophore . 10 . 7554/eLife . 00857 . 024Figure 6 . Refinement of the Aβ pharmacorphore based on studies of BAF11 . ( A ) Atomic model of BAF11 from the initial cycle docked on the full-length Aβ fiber , viewed in perpendicular to the fiber axis ( left panel ) and down the fiber axis ( right panel ) . BAF11 is shown as a cyan stick model , whose polar groups form hydrogen bonds ( green thick lines ) to Lys16 of Aβ . The extensive non-polar interactions arise from the flat aromatic rings of BAF11 packing against the hydrophobic surface formed by Val18 and Phe20 of Aβ . ( B ) Schematic representation of the polar and nonpolar interactions of BAF11 and its derivatives modeled on the Aβ fiber ( in orange and light brown ) . In the process of the Aβ pharmacophore refinement , five different classes ( I–V ) of BAF11 derivatives were introduced into the second cycle of screening , to expand the BAF set and to assess the specificity of the compounds identified in the initial cycle . The full description and chemical structure of each derivative are in Table 3 and Figure 5—figure supplement 1 . ( C ) Comparison of the toxicity inhibition ( defined in ‘Materials and methods’ ) among five types of BAF11 derivatives after 24 hr incubation with Aβ ( 0 . 5 µM ) . Notice that all changes to BAF11 which remove binding groups diminish its effectiveness as an inhibitor of toxicity . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 02410 . 7554/eLife . 00857 . 025Figure 6—figure supplement 1 . Chemical structures of the lead compound BAF11 and its derivatives . 12 derivatives of the lead compound BAF11 were included to expand the set during the refinement of amyloid pharmacophore ( Figure 6C ) . ( A ) Chemical structures of BAF11 derivatives . A magenta open circle indicates the deletion of the important hydroxyl group . A green open circle indicates the missing of aromatic atoms in hydrophobic region of BAF11 . The red color in chemical structures indicates the addition of atoms or groups to BAF11 . The full description of each derivative is in Table 3 . ( B ) Comparison of toxicity inhibition among BAF11 derivatives after 24 hour incubation with Aβ fibers . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 025 In the second cycle , nine new compounds were derived from the refined pharmacophore ( Figure 7 ) . Three of them detoxified Aβ in cell survival assay . BAF31 , the best inhibitor which protected mammalian cells from Aβ toxicity in the second cycle , increased cell survival from the 40% induced by Aβ alone to >90% ( Figure 3 ) . A derivative of BAF31 , BAF31ΔOH , lacking the hydroxyl group believed to bind to the Lys residue of the Aβ fiber ( shown by the magenta oval in Figure 8B ) , is calculated no longer to bind to the Aβ fiber . NMR and cell viability assessments indicated that BAF31ΔOH binds much less strongly to Aβ fibers than BAF31 itself and shows significantly reduced power to inhibit toxicity ( Figure 8E ) . Similarly , the detoxifying profile of derivatives of another inhibitor , BAF30 , validated the key interactions of BAF30 across the binding interface ( Figure 9 ) . Our conclusion is that the NMR binding and toxicity results for the BAF derivatives studied are consistent with our model for the pharmacophore of Aβ ( Figure 10 ) . 10 . 7554/eLife . 00857 . 026Figure 7 . New BAFs derived from the refined amyloid pharmacophore . ( A ) . Amyloid pharmacophore based on the structural overlay of active BAFs and derivatives . The overlay of the lead compounds from the initial round ( BAF4 , BAF8 , and BAF11 ) elucidated the consensus of polar and nonpolar interactions at fiber binding interfaces , which sheds light on the amyloid pharmacophore . The amyloid pharmacophore was further refined by iterative approaches of computational docking and experimental testing . The derivatives of those lead compounds were tested to explore the essential role of those consensus interactions , and the differences of binding patterns and toxicity inhibition effects of the BAF derivatives can provide a guideline for the further refinement of amyloid pharmacophore . ( B ) . New BAFs were ‘designed’ based on the refined pharmacophore . One successful example , BAF31 ( green sticks ) derived from the pharmacophore ( grey sticks ) , showed the enhanced capability of inhibiting Aβ toxicity ( Figure 8C ) . The success of developing enhanced binder from pre-defined pharmacophore highlights the important role of iterative docking/test approach in structure-based drug development . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 02610 . 7554/eLife . 00857 . 027Figure 8 . Elimination of one key hydrogen bond from BAF31 causes both the loss of NMR binding to Aβ fibers and the decrease in inhibition of Aβ cyto-toxicity . ( A ) Atomic model of the new inhibitor BAF31 ( our most tightly binding BAF ) derived from the refined pharmacophore ( Figure 7 , Figure 1F ) in the second cycle , viewed perpendicular to the fiber axis on the left and down the fiber axis on the right . In panel ( B ) , one important hydroxyl group forming hydrogen bonds to Lys16 residue of Aβ is highlighted by a magenta circle . ( C ) A representative NMR band ( left panel ) of mixture of Aβ fiber with the compound BAF31 compares with that ( right panel ) of Aβ fiber the derivative BAF31ΔOH which omits that important hydroxyl group . Their full NMR spectrums showing the same trend are shown in Figure 8—figure supplement 1 . ( D ) Cell survival rates after 24 hr incubation with Aβ ( 0 . 5 µM ) , the molar ratio ( 1:5 ) of Aβ and the compound is comparable with the ratio in NMR binding experiment ( C ) . ( E ) Notably , the elimination of one hydrogen bond from BAF31 ( the derivative BAF31ΔOH ) causes both the marked decrease in inhibition of Aβ toxicity to HeLa cells ( D ) and the loss of NMR binding to Aβ fibers ( C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 02710 . 7554/eLife . 00857 . 028Figure 8—figure supplement 1 . NMR titration of BAF31 and its derivative with the Aβ1–42 fiber . ( A ) . 1D 1H NMR spectrum of BAF31 ( 100 µM ) without ( in black ) and with Aβ1–42 fiber ( 12 . 5 µM monomer equivalent , in a green color ) . The magnified peaks are shown in the right panel to highlight the peak differences . ( B ) . NMR spectrum of its derivative BAF31ΔOH ( 100 µM ) when BAF31 is modified by the removal of a key hydroxyl group , without or with Aβ1–42 fiber ( 0 µM , 12 . 5 µM ) . The significant difference in NMR signal reduction between the BAF31 and BAF31ΔOH further validates the model of BAF31docked onto Aβ fibers . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 02810 . 7554/eLife . 00857 . 029Figure 9 . Analysis of the lead compound BAF30 and its derivatives . Structural models of BAF30 ( green sticks ) docked on Aβ fiber structure ( in a light yellow color ) are shown in ( A and B ) . The important polar ( black hydrogen bonds ) interaction between BAF30 and single β-sheet of Aβ fiber , as well as shape complimentary between the aromatic rings of BAF30 and the hydrophobic patches of Aβ fiber are highlighted respectively . Schematic representation of the polar and nonpolar interactions of BAF30 with Aβ fiber is shown in panel ( C ) . The magenta circles highlight two important hydroxyl groups which are absent in BAF30 derivatives . ( D ) . The chemical structure of each derivative is listed . The dark blue open circles indicate the deletion of the important hydroxyl group . The red color in chemical structures indicates the addition of atoms or groups to BAF30 . ( E ) . HeLa cell survival rates in the presence of Aβ ( 0 . 5 µM monomer equivalent ) and BAF30 or the derivatives are compared . The hydrogen bonds between BAF30 and Lys16 residues of Aβ fiber are important for binding of Aβ fiber and inhibition of Aβ toxicity . With additional groups at the opposite side of hydrogen binding sites , the derivative BAF30αR showed little change in toxicity inhibition . However , two BAF30 derivatives ( σOHAαOH and σOHAΔOHBαCOO ) , which alter or delete the two important hydroxyl groups ( magenta circles in panel C ) of BAF30 that form hydrogen bonds to Lys16 , showed a significant decrease in the toxicity inhibition . Furthermore , when BAF30 was modified by shifting both hydroxyl groups ( A and B ) to their neighboring positions , the derivative BAF30σOHABαCH3 almost lost the inhibition of Aβ toxicity . The rescuing percentage ( % ) is defined in ‘Materials and methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 02910 . 7554/eLife . 00857 . 030Figure 10 . General rule of the essential interactions between BAFs and Aβ fiber can be derived from structure-based screening of Aβ toxicity inhibitor . The carbonyl group is used to represent the H-bond acceptor ( or negative charge ) of BAFs , and the naphthalene ring is used to represent the planar aromatic portion of BAFs . Based on the rounds of computing search and experimental test , the detailed description about essential interactions and geometrical parameters are in ‘Materials and methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 030
Amyloid fibers differ fundamentally in structure from the enzymes and signaling proteins that are the traditional targets in structure based design of binding compounds , and thus their pharmacophores might be expected to differ fundamentally as would the types of compounds that bind . In general , the binding sites of the traditional targets are often concave pockets; in contrast , the surfaces of amyloid fibers are flat and repetitive along the fiber axis , without well-defined surface cavities . The widely used ligand-docking software , such as DOCK ( Ewing et al . , 2001 ) , or AutoDock ( Morris et al . , 2009 ) , is intended to fit well-defined protein pockets rather than shallow grooves at flat fiber surfaces . Consequently we have adapted the RosettaLigand program ( Davis and Baker , 2009 ) for docking a library of commercially available compounds onto the flat surface of amyloid fibers . Similarly to other software packages , RosettaLigand scores each candidate compound for its energetic fit to its binding site . The initial site is chosen near that occupied by a bound compound , as determined in a crystal structure . The conformational flexibilities of ligand and protein side chains are modeled in a ‘near-native’ perturbation fashion ( ‘Materials and methods’ ) , meaning that the fine sampling of conformations was restrained to be close to the starting conformation . To find the position along the flat fibrillar surface of greatest binding energy for each candidate compound , our screening approach leverages the rotamer repacking algorithm ( Leaver-Fay et al . , 2011 ) and Rosetta energy function ( Kuhlman and Baker , 2000 ) to account for flexibility of protein side chains and ligand , which is critical in modeling of such shallow grooves on the fiber surface . Our procedure identified 34 BAF compounds predicted to bind to Aβ fibers , among which eight BAFs diminish the toxicity of the fibers in mammalian cells . We suggest that the same procedure can be used to discover other compounds that reduce the toxicity of Aβ fibers , starting from other co-crystal structures of Aβ segments with other bound ligands . Similarly , the same procedure can be applied to the discovery of compounds that bind to other amyloid proteins , for use as either toxicity inhibitors or imaging agents for amyloid diagnosis . Our observation is that our tightest binding BAFs all diminish the toxicity of Aβ fibers , and yet do not substantially diminish the amount of fibers . Further study will be required to understand the molecular mechanism underlying the inhibition of Aβ toxicity , but here we offer the following hypothesis . Emerging evidence suggests that amyloid oligomers , rather than amyloid fibers , are toxic entities ( Hartley et al . , 1999; Cleary et al . , 2005; Silveira et al . , 2005 ) , and that perhaps toxic oligomers can be released from amyloid fibers ( Xue et al . , 2009; Cremades et al . , 2012; Krishnan et al . , 2012; Shahnawaz and Soto , 2012 ) . By binding to fibers , BAFs stabilize them , thereby shifting the equilibrium of Aβ molecules from smaller , toxic entities towards the fibrillar state . The BAF compounds in their computationally docked sites on Aβ fibers contact several ( as few as three and as many as six ) adjacent β-strands of the fiber . By creating a low energy binding interface across several fiber strands , the BAFs apparently stabilize the Aβ fibers from breaking into smaller entities . From previous studies , we expect BAFs to bind to amyloid fibers rather than oligomers . In recent work ( Laganowsky et al . , 2012; Liu et al . , 2012 ) , we proposed that amyloid forming proteins can enter either of two distinct aggregation pathways , which are separated by an energy barrier . One pathway leads to in-register fibers in which every β-strand lies directly above or below an identical strand in the fiber . The other pathway leads to out-of-register oligomers in which antiparallel β-strands are sheared relative to one another and roll into a β-barrel . We found that three out-of-register amyloid-like structures exhibit cytotoxicity ( Laganowsky et al . , 2012; Liu et al . , 2012 ) , which tend to be transient , equilibrating eventually into in-register fibers . In our approach , we search for BAFs based on in-register β-sheets rather than out-of-register β-strands found in toxic oligomeric structures , to which our BAFs are not expected to bind ( Figure 11 ) . We speculate that BAFs stabilize the in-register fibers revealed by our steric zippers , relative to out-of-register toxic oligomers , thereby shifting the equilibrium from toxic oligomers towards fibers ( Figure 12 ) . Supporting this is our result that diminished toxicity accompanies compound binding . 10 . 7554/eLife . 00857 . 031Figure 11 . BAFs are designed to bind to in-register β-sheets , rather than out-of-register β-sheets . As illustrated in ( A ) , BAFs bind to in-register β-sheets . Our structure-based approach searches for BAFs based on in-register β-sheets in Aβ fibers . These BAFs are predicted to bind along the flat hydrophobic surfaces of the fibers and are anchored by polar sidechains of Lysine residues . The Cβ distances between the Lys residues interacting with the BAFs are ∼9 . 6 Å following the stacked arrangement of in-register β-sheets . Orange G , as well as screened BAFs , favorably interact with the in-register fiber and are compatible with the geometry of the Lys residues aligned in in-register β-sheets . As illustrated in ( B ) , BAFs cannot bind to out-of-register β-sheets . The estimation of Cβ distance between the lysine residues , based on three out-of-register β-sheets structures previously determined ( Laganowsky et al . , 2012; Liu et al . , 2012 ) , ranges from 11 Å to 14 Å , quite different from the ∼9 . 6 Å measured in in-register β-sheet . We speculate that the BAFs are unable to bind to out-of-register β-sheets , and this difference accounts for the diminished toxicity that accompanies compound binding . Supporting this is our in vitro cell toxicity tests ( Table 7 and Figure 11—figure supplement 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 03110 . 7554/eLife . 00857 . 032Figure 11—figure supplement 1 . Active BAFs show no or little effects on the cyto-toxicity of pre-formed Aβ oligomers . To assess if BAFs inhibit Aβ toxicity by directly interfering with toxic Aβ oligomers , four BAFs —1 , 11 , 26 , 31— , showing the inhibition to Aβ toxicity , were incubated with pre-formed Aβ oligomer and then tested by MTT cell viability assay using HeLa cell line . None of the BAFs significantly reduces toxicity of pre-formed Aβ oligomer . Aβ oligomer was prepared by incubating purified Aβ1–42 in PBS for 4 hr at 37°C at the concentration of 5 µM without agitation . Pre-formed Aβ oligomer was mixed with four different BAFs ( Aβ1–42/BAFs = 1:1 molar ratio ) and further incubated for 15 min to allow potential binding of BAFs to pre-formed oligomer . The final concentration of Aβ oligomer as monomer is 0 . 5 µM , the same as what we test Aβ toxicity in MTT cell viability assay . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 03210 . 7554/eLife . 00857 . 033Table 7 . BAFs reduce Aβ cyto-toxicity by targeting fibers rather than oligomers . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 033CompoundInhibition to the cyto-toxicity of Abeta oligomers ( % ) Inhibition to the cyto-toxicity of Abeta fibers ( % ) BAF1−4 ± 636 ± 9BAF11−9 ± 77 ± 7BAF26−6 ± 626 ± 7BAF31−17 ± 1558 ± 7The BAF inhibitions of toxicity from either Aβ oligomer or fibers are compared . Four BAFs , which reduce the toxicity of Aβ fibers , show no inhibitory effects to Aβ oligomer toxicity at the equal molar ratio of BAF to Aβ . The inhibition ( % ) are calculated using the same method defined in ‘Materials and methods’ . The toxicity assay of Aβ oligomer is described in Figure 11—figure supplement 1 . The toxicity assay of Aβ fiber is the same as that described in Figure 3 . 10 . 7554/eLife . 00857 . 034Figure 12 . Proposed mechanism of how compound binding increases fiber stability and decreases fiber toxicity . BAFs ( green ) bind to the side of amyloid fibers , stabilizing the fiber , and shifting the equilibrium from smaller and more toxic oligomers towards fibers . This shift in equilibrium reduces amyloid toxicity . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 034 The identification of BAFs starts with the atomic structure of orange G bound within the fiber-like crystals of Aβ16–21 , because as yet there is no high-resolution atomic structure available for ligands bound to full-length Aβ fibers . Nevertheless , we found that BAFs diminish toxicity of full-length Aβ fibers . This finding suggests that the steric zipper structure of Aβ16–21 fibers recapitulates some of the essential structural features of full-length Aβ fibers . We are currently attempting cocrystallization of BAFs with Aβ16–21 and other steric zipper structures . We speculate that coupled with computational methods , other steric zipper structures could enable the discovery of the lead compounds for inhibitors of other toxic amyloid entities .
Each molecule in our two compound libraries was prepared for the docking simulations . Hydrogen atoms of each molecule were added for the compounds lacking modeled hydrogens using the program Omega ( v . 2 . 3 . 2 , OpenEye ) ( Bostrom et al . , 2003 ) . Ligand atoms were represented by the most similar Rosetta atom type , their coordinates were re-centered to the origin , and their partial charges were assigned by OpenEye’s AM1-BCC implementation . We then generated the ligand perturbation ensemble near the crystal conformation ( CSD set ) or starting conformation ( FC set ) of each molecule . For each rotatable bond of the ligand , a small degree torsion angle deviation ( ±5° ) was applied . K-mean clustering method was used to generate the ligand perturbation ensemble and similar/redundant conformations ( rmsd to the selected conformation is less than 0 . 5 Å ) were omitted . Finally , up to 100 conformations for each ligand were generated and made available for Rosetta LigandDock . We adopted the docking algorithm based on the method previously described in the RosettaLigand docking paper ( Meiler and Baker , 2006; Davis and Baker , 2009 ) . In general , the algorithm includes three stages: coarse-grained stage , Monte Carlo minimization ( MCM ) stage and gradient-based minimization stage . Whereas the original RosettaLigand method performed a full sampling of torsional degrees of freedom in the internal ligands and protein side-chains , we made modifications to enable the fast run time required by the screening method . Specially , we sampled the ligand and protein side-chain torsion angles in near-‘native’ perturbation fashion , where only the near-‘native’ conformation of side-chain and ligand rotamers were allowed and any conformation far away from the starting conformation was omitted . For each protein side-chain , the deviations ( ±0 . 33 , 0 . 67 , 1 SD ) around each input torsion were applied based on the standard deviation value of the same torsion bin from the backbone-dependent Dunbrack rotamer library . For each internal torsional angle of the ligand , the deviations ( ±5° ) around the input torsion were applied as described above . To optimize possible interactions ( H-bonding or packing ) between compound and fiber , we carried out random perturbations to the TS rigid-body degrees of freedom ( 5 Å for translational degrees of freedom; 360° for full rotational degrees of freedom ) to explore different rigid body arrangements . For each rigid-body perturbation , different conformations of fiber sidechains , and compounds were explored to maximize the binding interactions . We next carried out simultaneous quasi-Newton optimization of the compound rigid body orientation and the sidechain torsion angles , and in some cases , the torsion angles of the compound and the backbone torsion angles in the binding site , using the complete Rosetta energy function . The structure of KLVFFA fiber was taken from the co-crystal structure of KLVFFA with orange G ( pdb entry: 3OVJ ) ( Landau et al . , 2011 ) . After removing orange G , the sidechain torsion of KLVFFA was optimized to correct any conformational bias from the presence of orange G , and then the optimized structure were inspected to ensure that sidechain torsions are still within the original conformation of the co-crystal structure . The Aβ fibrillar structure was from ssNMR fiber structure of full-length Aβ ( pdb entry: 2LMO ) 40 . The same optimization step was applied before docking . The comparison of docking onto both KLVFFA and Aβ fibrillar structure are discussed in Figure 13 . 10 . 7554/eLife . 00857 . 035Figure 13 . BAFs bind to in-register β-sheets and are compatible to both parallel and antiparallel amyloid β-sheets . A subtlety of our procedure for compound discovery is that it involves both parallel ( A ) and antiparallel ( B ) amyloid β-sheets . In the X-ray structure of orange G bound to the segment Aβ16–21 ( KLVFFA ) of Aβ , the sheets are antiparallel ( B ) . The library of compounds is initially selected based on docking to the antiparallel β-sheet of Aβ16–21 . In the next step of our procedure , each compound is further screened against the solid-state-NMR-derived model of full-length Aβ fiber , which is a parallel sheet ( A ) . The structure models of orange G docked onto Aβ16–21 structure and full-length Aβ model are shown in Figure 13—figure supplement 1 . As simplified here in ( A and B ) , sulfate ions ( red ) of orange G are respectively hydrogen bonded to two lysine residues ( light brown ) , which keep nearly identical geometry ( the same ∼9 . 6 Å distance between the two lysine residues ) in either parallel or antiparallel sheet . Evidence that orange G , as well as BAF compounds identified by our procedure , all bind to both antiparallel and parallel sheets is given by the NMR experiments summarized in Figure 5 , where orange G and BAFs are shown to bind to both Aβ16–21 and full-length Aβ fibers . Apparently both parallel and antiparallel amyloid β-sheets are effective in binding to the same compounds . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 03510 . 7554/eLife . 00857 . 036Figure 13—figure supplement 1 . Structural models of orange G docked onto the antiparallel Aβ16–21 ( A ) and parallel full-length Aβ ( B ) fiber . ( A ) . The side view of orange G ( in an orange color ) docked on the Aβ16–21 fiber ( in a grey color ) with a predicted binding energy of—8 kcal/mol . ( B ) side view of orange G ( in an orange color ) docked on the Aβ full fiber ( in a light pink color ) with a predicted binding energy of—9 kcal/mol . The charge interactions between the orange G and Lysine 16 are highlighted by dark grey lines . DOI: http://dx . doi . org/10 . 7554/eLife . 00857 . 036 The docked compounds were filtered based on the following criteria: ( 1 ) The docking models with a compound-fiber van der Waals attractive energy > −7 . 0 kcal/mol were removed; ( 2 ) The docking models with a compound-fiber hydrogen-binding energy >−0 . 2 kcal/mol were eliminated . The remaining docked compounds were then ranked according to the energy of binding of compound to fiber . We used not only the total binding energy but also on each of the energy components separately ( Lennard-Jones interactions , solvation , hydrogen bonding , and electrostatics ) ( Lazaridis and Karplus , 1999; Kuhlman and Baker , 2000; Kortemme et al . , 2003 ) for ranking . The compounds ranked in the top 40% according to all of these measures were selected . Finally , the compounds were ranked by tightest binding energy ( Meiler and Baker , 2006 ) and best shape complementarity ( Lawrence and Colman , 1993 ) . Based on the rounds of computing search and experimental test , general rules of the essential interactions of BAF binding to Aβ fibers are summarized here . As illustrated in Figure 10 , the geometrical parameters of those key interactions are specified as followings:H-bond acceptor ( or negative charge ) of the inhibitor makes either hydrogen bond or salt bridge to sidechain nitrogen atoms ( NZ ) of at least two Lysine residues from adjacent Aβ strands along the fiber axis . Our data suggest that the BAFs need to have good contacts across 2 to 4 adjacent Aβ strands , in order to effectively bind to Aβ fiber and reduce Aβ toxicity . The hydrogen bond or salt bridge described in 1 ) follows the general rule of H-bond geometry , which are:Distance ( d1 , as shown in the figure ) between the NZ atom of Lys16 and H-bond acceptor atoms of BAFs: 2 . 8∼3 . 5 angstrom;Angle ( Θ1 ) at BAF H-bond acceptor atoms:100∼150°;Angle ( Θ2 ) at the NZ atom of Lys16: 130∼180° . Hydrophobic interactions between the apolar residues ( phenylalanine18 and valine 20 ) and the planar aromatic portion of the compounds . The aromatic portion of compounds should be planar or semi-planar to pack against the flat surface of Aβ which spans across at least two adjacent Aβ strands . The hydrophobic interactions described in 3 ) follow the pi-pi stacking geometry , which are:Distance ( d2 ) between the center of the apolar sidechains and the center of BAF aromatic rings: 4 . 0∼5 . 0 angstrom;Dihedral angle ( Φ ) between the surface plane defined by Phe18 and Val20 and the aromatic ring of the BAFs: 0∼40° . | Alzheimer’s disease is the most common form of dementia , estimated to affect roughly five million people in the United States , and its incidence is steadily increasing as the population ages . A pathological hallmark of Alzheimer’s disease is the presence in the brain of aggregates of two proteins: tangles of a protein called tau; and fibers and smaller units ( oligomers ) of a peptide called amyloid beta . Many attempts have been made to screen libraries of natural and synthetic compounds to identify substances that might prevent the aggregation and toxicity of amyloid . Such studies revealed that polyphenols found in green tea and in the spice turmeric can inhibit the formation of amyloid fibrils . Moreover , a number of dyes reduce the toxic effects of amyloid on cells , although significant side effects prevent these from being used as drugs . Structure-based drug design , in which the structure of a target protein is used to help identify compounds that will interact with it , has been used to generate therapeutic agents for a number of diseases . Here , Jiang et al . report the first application of this technique in the hunt for compounds that inhibit the cytotoxicity of amyloid beta . Using the known atomic structure of the protein in complex with a dye , Jiang et al . performed a computational screen of 18 , 000 compounds in search of those that are likely to bind effectively . The compounds that showed the strongest predicted binding were then tested for their ability to interfere with the aggregation of amyloid beta and to protect cells grown in culture from its toxic effects . Compounds that reduced toxicity did not reduce the abundance of protein aggregates , but they appear to increase the stability of fibrils . This is consistent with other evidence suggesting that small , soluble forms ( oligomers ) of amyloid beta that break free from the fibrils may be the toxic agent in Alzheimer’s disease , rather than the fibrils themselves . In addition to uncovering compounds with therapeutic potential in Alzheimer’s disease , this work presents a new approach for identifying proteins that bind to amyloid fibrils . Given that amyloid accumulation is a feature of many other diseases , including Parkinson’s disease , Huntington’s disease and type 2 diabetes , the approach could have broad therapeutic applications . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2013 | Structure-based discovery of fiber-binding compounds that reduce the cytotoxicity of amyloid beta |
Protein kinases have evolved diverse specificities to enable cellular information processing . To gain insight into the mechanisms underlying kinase diversification , we studied the CMGC protein kinases using ancestral reconstruction . Within this group , the cyclin dependent kinases ( CDKs ) and mitogen activated protein kinases ( MAPKs ) require proline at the +1 position of their substrates , while Ime2 prefers arginine . The resurrected common ancestor of CDKs , MAPKs , and Ime2 could phosphorylate substrates with +1 proline or arginine , with preference for proline . This specificity changed to a strong preference for +1 arginine in the lineage leading to Ime2 via an intermediate with equal specificity for proline and arginine . Mutant analysis revealed that a variable residue within the kinase catalytic cleft , DFGx , modulates +1 specificity . Expansion of Ime2 kinase specificity by mutation of this residue did not cause dominant deleterious effects in vivo . Tolerance of cells to new specificities likely enabled the evolutionary divergence of kinases .
Phosphorylation networks coordinate many cellular processes . Their importance is underscored by the prevalence of kinases: the human genome encodes >500 kinases ( Manning et al . , 2002b ) and over 100 , 000 phosphorylation sites have been identified ( Hornbeck et al . , 2012 ) . The number and diversity of kinases expanded with increasing numbers of cell types during the evolution of metazoa ( Manning et al . , 2002a ) . The addition of new kinase families with new specificities presumably increases the information processing capacity of the cell , thus enabling the emergence of more complex biological processes ( Beltrao et al . , 2009; Lim and Pawson , 2010 ) . To achieve precise regulation , kinases have evolved mechanisms to selectively phosphorylate specific substrates . This specificity is encoded at multiple levels . The active site of some kinases is optimized to bind to a defined peptide sequence , referred to as the primary specificity . Kinases may have additional peptide interaction surfaces that bind to docking motifs on the substrate distinct from the site of phosphate transfer . Non-substrate proteins called scaffolds can also form tertiary complexes to direct the interaction between kinase and substrate . The sub-cellular localization of a kinase can limit access to a subset of proteins . Finally , systems-level effects such as substrate competition and the opposing activities of phosphatases all affect the degree to which substrates are phosphorylated in the context of the cell ( Reviewed in Remenyi et al . , 2005; Ubersax and Ferrell , 2007 ) . Phosphoregulatory networks are well suited to rapid information processing because phosphorylation reactions act on time scales of minutes ( Olsen et al . , 2006 ) . For this reason , kinase networks are crucial for processes that require a high degree of temporal control , such as the cellular division programs , mitosis , and meiosis . Taking mitosis as an example , kinase networks control a wide range of cell sizes ( 2 μm to several mm ) and cell biology ( from a single to a thousand chromosomes , closed vs open mitosis ) . Thus the phosphorylation networks that underlie these processes must adapt to enable these changes in cell biology . There has been considerable progress in the understanding of transcription-factor network evolution in recent years , and these studies have helped understand the generation of morphological diversity ( Carroll , 2008 ) and key principles of transcriptional rewiring ( Tuch et al . , 2008 ) . Despite recent progress ( Holt et al . , 2009; Tan et al . , 2009; Alexander et al . , 2011; Cross et al . , 2011; Freschi et al . , 2011; Pearlman et al . , 2011; Capra et al . , 2012; Lee et al . , 2012; Beltrao et al . , 2013; Coyle et al . , 2013; Goldman et al . , 2014 ) , there is still relatively little known about the evolution of kinase signaling networks . Phosphoregulatory networks evolve by the gain or loss of protein–protein interactions , either by changes to substrates or by changes to kinase specificity . Within substrates , the gain or loss of kinase interaction motifs and phosphorylation sites have occurred relatively rapidly ( changes occur within a few millions of years of divergence , Holt et al . , 2009; Beltrao et al . , 2009 ) . These substrate mutations affect only one protein at a time; therefore detrimental pleiotropic effects are avoided . Alternatively , networks can evolve by changing kinase specificity . Kinases act as hubs of phosphoregulatory networks and can coordinate the activities of hundreds or even thousands of substrates ( Manning et al . , 2002a; Matsuoka et al . , 2007; Holt et al . , 2009; Hornbeck et al . , 2012 ) . Changing the specificity of a kinase , therefore , can destroy many network connections , while also potentially creating a large number of new connections . It might be expected that there is a strong negative selection pressure against such drastic remodeling of kinase networks , but it is nevertheless clear that kinases have evolved diverse specificities , particularly following gene duplication ( Mok et al . , 2010 ) . The mechanisms underlying this diversification are poorly understood , and it is unknown how kinases successfully evolve significant changes to their biochemistry and network biology . To learn how kinase specificity evolves , we studied the evolutionary history of the CMGC ( Cyclin Dependent Kinase [CDK] , Mitogen Activated Protein Kinase [MAPK] , Glycogen Synthase Kinase [GSK] , and Casein Kinase [CK] ) group of kinases . The CMGC group also contains the CDK-Like kinases ( CDKL ) , SR-kinases , Homeodomain-Interacting Kinases ( HIPKs ) , CDC-Like Kinases ( CLKs ) , Dual-Specificity Tyrosine Regulated Kinases ( DYRKs ) , and a paralogous superfamily of kinases including LF4 , the mammalian RCK kinases ( ICK , MOK , and MAK ) , and the fungal IME2 kinases . CMGC kinases coordinate a wide range of cellular functions in different species . CDKs are the major coordinators of cell division in both mitosis and meiosis ( Morgan , 2007 ) . MAPKs are crucial for many cellular decisions , including proliferation , differentiation , and stress responses ( Chen and Thorner , 2007; Morrison , 2012 ) . The Ime2 kinase is crucial for meiosis in S . cerevisiae ( Dirick et al . , 1998; Benjamin , 2003 ) , while its orthologs in other Ascomycetes control distinct processes including mating ( Sherwood et al . , 2014 ) , differentiation ( Hutchison and Glass , 2010 ) , and response to light ( Bayram et al . , 2009 , for a review see Irniger , 2011 ) . The Ime2 paralogs in mammals ( the RCK kinases ) control diverse processes including spermatogenesis and control of retinal cilia-length ( MAK ) , as well as intestinal cell biology , control of cell proliferation , organogenesis , and cellular differentiation ( MOK and ICK ) ( Fu , 2012 ) . Within the evolutionary history of CMGC kinases , gene duplications followed by diversification resulted in multiple paralogous kinases with distinct specificities that coordinate diverse biological functions . For example , the specificities of Cdk1 and Ime2 are mostly non-overlapping ( Holt et al . , 2007 ) . In addition to acquiring distinct modes of regulation , it is likely that the divergence of the biological functions of this kinase family is , in part , due to evolution of their primary specificities . Therefore , understanding the mechanisms that drive specificity change and the consequences of these changes is crucial to rationalize the structures of modern phosphoregulatory networks . The shared evolutionary history of CMGC kinases , combined with their diverse specificities , make them an ideal gene family for studying the evolution of kinase specificity . In this study , we determined the primary substrate specificity of eight extant kinases from the IME2/RCK/LF4 group of kinases and found variation in the amino acid that is preferred immediately C-terminal to the phosphoacceptor ( the +1 position ) . To determine the mechanisms by which these specificities evolved , we used maximum likelihood phylogenetic models to reconstruct sequences for all ancestors of the CMGC kinases . We then resurrected seven ancestral kinases in the lineage starting with AncCMGI , which is the last common ancestor of the CDK , CDKL , MAPK , GSK , CLK , and IME2/RCK/LF4 kinases , up until the modern LF4 , RCK , and IME2 kinases . Biochemical characterization of these resurrected kinases allowed us to trace the evolution of primary specificity in this lineage . In addition , we determined a key residue that modulates primary specificity at the +1 position . By mutating this residue in modern IME2 we showed that , at least in some circumstances , the cell can readily tolerate changes that expand kinase specificity .
To understand how kinase specificity changes over a long evolutionary timescale , we determined the phosphorylation site specificities of eight kinases from the superfamily of kinase paralogs that includes fungal Ime2 , the mammalian RCK kinases ( ICK , MOK and MAK ) , and the LF4 kinases in algae and protists . This superfamily controls diverse biological processes , and we hypothesized that differences in primary specificity may underlie some of this functional divergence . In addition , previous work has shown that S . cerevisiae Ime2 and mouse ICK differ in their +1 specificities ( Fu et al . , 2006; Holt et al . , 2007 ) . We used a positional scanning peptide library ( PSPL , Hutti et al . , 2004 ) to characterize the full primary specificity of these kinases ( Figure 1A ) . Briefly , we used a set of 182 peptide mixtures , in which a central phosphoacceptor position ( an equal mixture of serine and threonine ) was surrounded by random sequence . Within each mixture , one of nine positions was fixed to a single amino acid residue ( see schematic , Figure 1A , top ) . Peptides were subjected in parallel to a radiolabelled kinase assay , and the extent of radiolabel incorporation indicates which residues are preferred or disallowed by the kinase at each position within the peptide sequence . 10 . 7554/eLife . 04126 . 003Figure 1 . The IME2/RCK/LF4 superfamily of kinases has variable specificity at the +1 position . ( A ) Positional scanning peptide libraries were used to profile the specificity of various kinases: left , M . musculus MOK; middle , N . gruberi LF4; right , S . cerevisiae Ime2 . Yellow indicates preference for a given amino acid while blue indicates counter selection . A schematic of the peptide library is shown above ( see text for details ) . Data show the average of two replicates for each kinase . Raw data for these kinases and four other superfamily members are shown in Figure 1—figure supplement 1 . Data shown here exclude peptides containing fixed Ser and Thr residues that typically produce an artificially increased signal due to the presence of an additional phosphoacceptor residue; heat maps of full peptide array results for all extant kinases are shown in Figure 1—figure supplement 2 . ( B ) Michaelis–Menten plots for individual 8-mer IME2/RCK/LF4 consensus peptides ( schematic , below ) in which the +1 position is varied to be either proline ( red ) or arginine ( blue ) . ( C ) Positional scanning peptide library data showing the average +1 preference of 8 kinases from the IME2/RCK/LF4 subgroup of CMGC kinases . DOI: http://dx . doi . org/10 . 7554/eLife . 04126 . 00310 . 7554/eLife . 04126 . 004Figure 1—figure supplement 1 . Raw data for positional scanning peptide library arrays . A set of 182 peptides with the general sequence shown at bottom , but with the indicated position in the array fixed to the indicated amino acid residue , were phosphorylated with the indicated kinase with radiolabeled ATP . Aliquots of each reaction were transferred to a streptavidin-coated membrane , which was washed and dried for autoradiography . Spot intensities reflect the extent of radiolabel incorporation into the indicated peptide mixture . Note: Peptide mixtures having fixed Ser and Thr residues generally give spuriously higher signals than other components of the library due to the presence of an additional phosphoacceptor site . This artifact is particularly evident at the +4 position for most kinases ( red asterisks ) as an artifact that arises from the nature of the linker sequence that follows the +4 position . This linker contains two residues ( +1 Ala and +2 Gly relative to the +4 S/T ) that are fixed to preferred amino acids , and so the +4 S/T gets heavily phosphorylated . We have previously verified that there is actually no preference for S/T at the +4 position ( Holt et al . , 2007 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04126 . 00410 . 7554/eLife . 04126 . 005Figure 1—figure supplement 2 . Quantified positional scanning peptide library data for all extant kinases analyzed . Spot intensities shown in Figure 1—figure supplement 1 and a replicate run of the same kinase were quantified ( using BioRad QuantityOne software ) and normalized so that the average signal in a given row was equal to unity . Data from the two replicates were averaged , log2 transformed , and used to generate heat maps in Microsoft Excel . Positively selected residues are shown in yellow and negatively selected residues are shown in blue according to the scale at bottom right . Data for ICK were previously published ( Fu et al . , 2006 ) and are shown here in heat map form for comparison . DOI: http://dx . doi . org/10 . 7554/eLife . 04126 . 005 As previously reported for S . cerevisiae Ime2 and mouse ICK , PSPL analysis revealed that all kinases assayed share a strong preference for arginine at the −3 position and proline at the −2 position ( Figure 1—figure supplements 1 and 2 ) . However , we found that selectivity for residues C-terminal to the phosphoacceptor was more variable . Specifically , the preferred residue at the +1 position varied between arginine ( +1R ) for the S . cerevisiae , Candida glabrata , and Yarrowia lipolytica Ime2 homologs and proline ( +1P ) for the three mammalian RCK kinases . Neurospora crassa Ime2 and Naegleria gruberi LF4 phosphorylated peptides with +1R and +1P relatively equally ( Figure 1C ) . All kinases also tolerated alanine ( +1A ) relatively equally . Additional biochemical characterization using four consensus peptides that were varied at the phosphoacceptor ( 0 ) and +1 positions ( acetyl-R-P-R-S/T-R/P-R-amide ) revealed differences in steady-state kinetics underlying the +1 specificity switch . Figure 1B shows Michaelis–Menten curves for a pair of peptide substrates with identical sequence except for having either proline ( red ) or arginine ( blue ) at the +1 position . S . cerevisiae Ime2 phosphorylated the +1R peptide with 65-fold greater efficiency ( kcat/KM ) than the corresponding +1P peptide , and these differences were attributable to differences in both the kcat and the KM values . As anticipated , the N . gruberi homolog LF4 phosphorylated the +1R and +1P peptides with similar kinetics , while mouse MOK showed a twofold preference for the +1P peptide ( Table 1 ) . 10 . 7554/eLife . 04126 . 006Table 1 . Michaelis-Menten kinetic parameters for three members of the IME2/RCK/LF4 subgroup of kinasesDOI: http://dx . doi . org/10 . 7554/eLife . 04126 . 006Kinase+1 residueKM ( μM ) kcat ( min-1 ) Kcat/KM ( min-1μM-1 ) S . cerevisiae Ime2R2 . 2 ± 0 . 310 . 8 ± 0 . 44 . 84 ± 0 . 67n = 4P44 . 5 ± 12 . 23 . 3 ± 0 . 30 . 07 ± 0 . 02n = 4N . gruberi LF4R1 . 2 ± 0 . 45 . 8 ± 0 . 54 . 76 ± 1 . 6n = 3P2 . 7 ± 0 . 67 . 8 ± 0 . 52 . 95 ± 0 . 69n = 3M . musculus MokR41 . 9 ± 10 . 85 . 9 ± 0 . 50 . 14 ± 0 . 04n = 4P37 ± 2 . 711 . 7 ± 0 . 30 . 32 ± 0 . 03n = 4 In summary , primary specificity at the +1 position is relatively plastic among IME2/RCK/LF4 kinases , while the other major specificity determinants remained strongly conserved . Taken together with previous characterization of other CMGC kinases ( Songyang et al . , 1996 , 1994; Fu et al . , 2006; Holt et al . , 2007; Sheridan et al . , 2008; Mok et al . , 2010; Alexander et al . , 2011; Bullock et al . , 2009; Kettenbach et al . , 2012 ) , our results suggest two evolutionary hypotheses . One possibility is that the ancestor of modern CMGC kinases had dual specificity for arginine and proline and then lost either proline or arginine to specialize extant kinases . Alternatively , the +1 specificity for arginine could have evolved as a switch from proline to arginine . We sought to reconstruct the evolutionary events that led to the modern diversity of IME2/RCK/LF4 specificities . To achieve this goal , we curated a library of 329 amino acid sequences sampled broadly from acrross the CMGC group and then reconstructed their evolutionary history using maximum likelihood phylogenetic methods ( Thornton , 2004; see ‘Materials and methods’ ) . The resulting phylogeny and reconstructed ancestral sequences were strongly supported by the evolutionary model ( Figure 2 , Figure 2—figure supplement 1 ) . The full library of CMGC kinase ancestors is available at http://104 . 131 . 121 . 138/cmgc . 10 . 2013/ . 10 . 7554/eLife . 04126 . 007Figure 2 . The common ancestor of CMGI kinases had a slight preference for proline at the +1 position of the substrate peptide . ( A ) Summary of current knowledge about CMGC group kinase specificity in the context of the maximum likelihood phylogeny of protein sequences . Major groups , such as IME2 , MAK , ICK , etc , have been collapsed for simplicity . Branch lengths express the number of amino acid substitutions per protein sequence site . Branch support values are approximate likelihood ratios ( aLRs ) , expressing the ratio of the likelihood of the maximum likelihood phylogeny to the next-best phylogeny lacking the indicated branch . For example , an aLR value of 10 indicates that the branch is 10 times more likely than the next-best phylogenetic hypothesis . The position of the common ancestor of CDK , MAPK , CDKL , GSK , and the IME2/LF4/RCK superfamily ( AncCMGI ) , is indicated by a purple circle . ( B ) Position scanning peptide libraries were used to determine the primary specificity of the maximum likelihood resurrected AncCMGI kinase . Raw peptide data are shown in Figure 2—figure supplement 1 . A complete repeat of the PSPL for a Bayesian sampled alternative reconstruction of AncCMGI ( AncCMGI-B2 ) is shown in Figure 2—figure supplement 2 . A structural model of AncCMGI is shown in Figure 2—figure supplement 3 in phylogenetic context along with structures and models for extant kinases that were derived from this ancestor . DOI: http://dx . doi . org/10 . 7554/eLife . 04126 . 00710 . 7554/eLife . 04126 . 008Figure 2—figure supplement 1 . Support for reconstrutctions . We characterized the support for each ancestor by binning their posterior probabilities of states into 5%-sized bins and counting the proportion of ancestral sites within each bin . ( A ) Posterior probability of each maximum likelihood amino acid ( P ( ML ) ) is shown as a function of position within the kinase primary sequence . Sites with lower support generally correspond to loop regions . ( B ) Histogram showing distribution of posterior probabilities of maximum likelihood amino acids . Mean ( x ) and standard deviation ( σ2 ) values are indicated . Complete reconstruction data with complete probabilities at every position is available at http://www . phylobox . com/cmgc . 10 . 2013/ . DOI: http://dx . doi . org/10 . 7554/eLife . 04126 . 00810 . 7554/eLife . 04126 . 009Figure 2—figure supplement 2 . Raw data and selectivity values for a positional scanning peptide library array of an alternate reconstruction of AncCMGI . ( A ) Raw PSPL result for an alternative reconstruction of AncCMGI . ( B ) Averaged quantified data from ( A ) and a replicate analysis with AncCMGI . Data were collected and quantified as in Figure 1—figure supplements 1 and 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 04126 . 00910 . 7554/eLife . 04126 . 010Figure 2—figure supplement 3 . Structural evolution in the CMGC kinase group . A model of AncCMGI was generated from the CDK2 structure using Phyre2 . This structure is compared to the structures of extant CMGC kinases in their phylogenetic context . The core structure present in AncCMGI is colored grey . Additional domains in modern kinases are indicated in various colors: e . g . C-terminal extension of CDKL ( cyan ) ; C-terminal extension of MAPK ( red ) ; Cyclin ( beige ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04126 . 010 We synthesized DNA encoding the maximum likelihood common ancestor of CDK , CDKL , MAPK , GSK , and IME2/RCK/LF4 ( referred to as AncCMGI ) . We expressed and purified the kinase from both S . cerevisiae and E . coli . We obtained active kinase in both cases . We employed positional scanning arrays to determine the primary specificity of AncCMGI , as described above and in Figure 1 . Many kinases have fairly degenerate primary specificities ( Mok et al . , 2010 ) and rely on docking , scaffolding , and localization to discriminate their correct substrates ( Ubersax and Ferrell , 2007 ) . We therefore reasoned that an ancestral kinase might have lower specificity than the extant enzymes derived from this ancestor . Indeed , there have been several studies in which enzymes have become sub-specialized from broad-specificity ancestors following gene duplication during evolution ( Thornton , 2003 ) . However , we observed that AncCMGI had a well-defined primary specificity , with a strong preference for arginine at the −3 position and proline at the −2 position ( Figure 2B ) . These N-terminal determinants correspond to the conserved motif found in the IME2/RCK/LF4 kinases as well as the DYRK kinases ( Figure 2A ) . Interestingly , AncCMGI could phosphorylate peptides having either a proline or an arginine residue at the +1 position , though it displayed a 5 . 6-fold preference for proline ( Figure 2B ) . Thus AncCMGI appeared to have a modest +1 proline preference , in contrast to the more stringent proline requirement of the extant CDK and MAPK families ( Himpel et al . , 2000; Fu et al . , 2006; Holt et al . , 2007; Mok et al . , 2010; Alexander et al . , 2011 ) . Thus , the specificity of AncCMGI contains elements of the diverged specificities of many major sub-families of the CMGC group . Notably , the domain architecture of AncCMGI is most similar to IME2/RCK/LF4 kinases . That is , AncCMGI contains the canonical CMGC insert loop , but it lacks any C-terminal extension ( found in MAPKs ) . Furthermore , AncCMGI does not appear to require cyclin for activity ( as with CDKs ) : we observed no significant co-purifying proteins , and E . coli does not encode any cyclin orthologs ( Figure 2 , Figure 2—figure supplement 3 ) . These data indicate that the cyclin dependence of CDKs and the requirement for an additional C-terminal extension to stabilize the Cα helix in MAPKs are characteristics that arose later during evolution ( E . coli does not encode any cyclin orthologs [Figure 2 , Figure 2—figure supplement 3] ) . In addition , AncCMGI contains a MAPK-like TXY motif in the activation loop . Phosphorylation of this motif is required for the activation of MAPKs . Because E . coli lack endogenous kinases capable of phosphorylating this TXY motif , AncCMGI is likely activated through autophosphorylation , similar to extant mammalian DYRK and GSK family kinases ( Cole et al . , 2004; Lochhead et al . , 2005 ) . In order to learn the trajectory by which kinase specificity at the +1 position evolved , we reconstructed ancestral kinases within the CMGC group at multiple evolutionary time points before and after the +1 specificity change from proline to arginine was presumed to have occurred ( Figure 3A ) . These kinases were assayed using consensus peptide substrates with identical sequence except for having either proline ( red ) or arginine ( blue ) at the +1 position . The log-ratio of arginine/proline preference from this assay is plotted in Figure 3B . Because the +1 specificity could be dependent on the surrounding sequence context present in the consensus peptide substrates , we also characterized the full primary specificities of AncLF4 and AncICK kinases by PSPL arrays ( Figure 3 , Figure 3—figure supplement 1 ) . While the specificity for arginine at the −3 position and proline at the −2 position was conserved among these ancestors , we observed significant variation in their relative preference for +1R vs +1P ( Figure 3B , C; Figure 3 , Figure 3—figure supplement 1 ) . As described above , AncCMGI appears to have preferred +1P substrates . On the phylogenetic branch leading to the common ancestor of the IME2/RCK/LF4 group ( i . e . , AncNgru ) , the +1 specificity relaxed to equally accommodate both +1R and +1P ( +1PR ) . This hybrid specificity was conserved in the LF4 lineage and also on the branch leading to AncICK . Evolution after AncICK , however , proceeded along two divergent evolutionary paths . Namely , the specificity reverted to the ancestral +1P-preferring state along the branch leading to the mammalian RCK kinases ( ICK and MAK ) . In contrast , the specificity shifted to +1R in the fungal lineage leading to the ancestor of the IME2 kinases ( i . e . , AncIME2 ) . AncIME2 had a moderate preference for +1R , and this preference is maintained in other fungal IME2 ancestors , becoming more pronounced in the ancestor of Yarrowia lipolytica ( AncYlip ) . These results are summarized in their phylogenetic context in Figure 3A and are robust to statistical uncertainties about the reconstructed ancestral sequences , although the degree of +1 proline selectivity was slightly lower in AncCMGI-B1 and higher in AncCMGI-B2 ( see ‘Materials and methods’ , Figure 3 , Figure 3—figure supplement 2 ) . 10 . 7554/eLife . 04126 . 011Figure 3 . The substrate peptide +1 specificity evolved from proline in AncCMGI to arginine in S . cerevisiae Ime2 via an expanded specificity intermediate . ( A ) Phylogenetic tree for the IME2/LF4/RCK superfamily , also showing the positions of other major CMGC group families . The positions of ancestral nodes resurrected in this study are indicated by circles . The tree is color-coded: red indicates +1 proline preference , blue indicates +1 arginine preference , and purple indicates equal tolerance of both Arg and Pro at the +1 position . ( B ) Seven resurrected kinases were incubated with 45 μM peptide ( see schematic , below ) . Bars show the log2 ratio of +1R and +1P initial velocities ( V0R/V0P ) . Black and white bars indicate Ser and Thr respectively as phosphoacceptor . Error bars are standard error of three assays . ( C ) Peptide phosphorylation rates for the same resurrected kinases shown in ( B ) using the peptides from the positional scanning peptide library having the indicated residue fixed at the +1 position . Data show the average of two replicates and are normalized and log2 transformed so that the average value for a given kinase is zero . The heat map follows the same color scheme as in Figure 1A . DOI: http://dx . doi . org/10 . 7554/eLife . 04126 . 01110 . 7554/eLife . 04126 . 012Figure 3—figure supplement 1 . Raw data and selectivity values for full positional scanning peptide arrays of AncLF4 and AncICK . ( A ) Raw PSPL result for maximum likelihood reconstructions of AncNgru and AncICK . ( B ) Averaged , quantified selectivity values for two replicate runs of the kinases shown in ( A ) . Data were collected and processed as in Figure 1—figure supplements 1 and 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 04126 . 01210 . 7554/eLife . 04126 . 013Figure 3—figure supplement 2 . The substrate peptide +1 specificity evolved from proline in AncCMGI to arginine in S . cerevisiae Ime2 via an expanded specificity intermediate – robustness to uncertainty in reconstructions . ( A ) Alternative reconstructions of each of the seven resurrected kinases ( ‘Materials and methods’ ) were incubated with 45 μM peptide ( see schematic , bottom ) . Bars show the log2 ratio of +1R and +1P initial velocities ( V0R/V0P ) . Black and white bars indicate Ser and Thr respectively as phosphoacceptor . Error bars are standard error of three assays . ( B ) Peptide phosphorylation rates for the same resurrected kinases shown in ( A ) , using the peptides from the positional scanning peptide library having the indicated residue fixed at the +1 position . Data show the average of two to three replicates and are normalized and log2 transformed so that the average value for a given kinase is zero . The heat map follows the same color scheme as in Figure 1A . DOI: http://dx . doi . org/10 . 7554/eLife . 04126 . 013 In initial results comparing +1 specificities obtained from our positional scanning peptide library arrays to those from our ratiometric peptide assays , we noticed that ICK was an outlier among the mammalian RCKs: this kinase phosphorylated peptides with +1R and +1P equally in our ratiometric assay , while the arrays showed a clear +1P preference ( albeit with detectable phosphorylation of the +1R peptide mixture , Figure 4 , Figure 4—figure supplement 1 ) . We had initially used peptides with only serine as a phosphoacceptor in our ratiometric assay , while the PSPL array peptides contained equal mixtures of serine and threonine . We therefore reasoned that the nature of the phosphoacceptor might influence the +1 specificity of kinases . To test this hypothesis , we analyzed additional peptide sets with equal mixes of serine and threonine , or with only threonine as the phosphoacceptor in our ratiometric assay . From these experiments , we found that , indeed , the phosphoacceptor affects +1 specificity: serine causes a shift towards +1R preference , while threonine causes a shift towards +1P preference ( Figure 4 , Figure 4—figure supplement 1 ) . This dependence of +1 specificity on the phosphoacceptor is present in AncCMGI and is maintained in all ancestors and extant members of the IME2/RCK/LF4 family ( Figures 3B and 4C , Figure 3—figure supplement 2 ) . To understand how kinase phosphorylation site specificity changes in evolution from a structural standpoint , we sought to identify specific residues in the kinase catalytic domain that mediate +1 specificity . Kinase substrate co-crystallography and biochemical analysis of large numbers of kinases have revealed some of the rules connecting kinase sequence to specificity ( Zhu et al . , 2005; Goldsmith et al . , 2007; Mok et al . , 2010 ) . The peptide-binding groove is formed from a number of structural elements within the kinase catalytic domain . One key point of kinase-substrate interaction is the activation loop , a conformationally flexible region that extends between two highly conserved amino acid motifs , DFG , and APE , connecting the N- and C-terminal kinase lobes ( Huse and Kuriyan , 2002 ) . Previous work revealed that the amino acid immediately C-terminal to the conserved DFG motif contributes to preference for serine vs threonine at the phosphoacceptor ( Chen et al . , 2014 ) . This residue was previously referred to as the DFG+1 residue , but here we will refer to this amino acid as the DFGx residue to avoid confusion with the +1 amino acid position of the substrate peptide . Since the DFGx residue communicates with the phosphoacceptor , and the phosphoacceptor in turn influences +1 specificity , we hypothesized that the identity of the DFGx residue may be a determinant for +1P vs +1R specificity . This hypothesis is consistent with X-ray crystal structures of kinase peptide complexes , in which the +1 residue in the substrate is in close proximity to the DFGx residue ( Soundararajan et al . , 2013 ) ( Figure 4A ) . Our ancestral reconstructions indicate that the DFGx residue was a leucine in AncCMGI and then mutated to serine multiple times in the CMGC family ( Figure 4B ) . Further , the presence of leucine vs serine at DFGx in present-day kinases correlates with specificity for +1P vs +1R . Taken together , this suggests that Leu vs Ser at DFGx could affect kinase specificity at the +1 site . To test this hypothesis , we examined the effect of mutating Leu to Ser in mammalian ICK ( L146S ) . This single mutation at DFGx made ICK approximately threefold more proline specific , such that its specificity resembled that of MOK ( Figure 4C ) . Conversely , mutating the DFGx residue in the opposite direction in MOK ( S148L ) had the reverse effect of converting MOK from +1P preference to a non-selective kinase ( +1PR ) . We also mutated the DFGx residue in S . cerevisae Ime2 and in the ancestors AncLF4 ( S152L ) , AncICK ( L152S ) , and AncCMGI ( L151S ) . In all cases , we observed that mutation from leucine to serine at DFGx shifted kinase function towards +1P specificity , while mutation from serine to leucine at DFGx shifted the kinases towards +1R specificity ( Figure 4C ) . We note that though DFGx mutation was sufficient to substantially shift the +1 residue preference , other residues must also be important for +1 specificity , since proline to arginine selectivity shifts occur in the evolution of CMGC kinases without a DFGx mutation . 10 . 7554/eLife . 04126 . 014Figure 4 . The DFGx amino acid and the phosphoacceptor influence the +1 specificity of extant and ancestral kinases . ( A ) Structural model of AncCMGI in complex with a consensus peptide substrate . The box shows the active site with the position of the DFGx amino acid highlighted in orange . ATP is blue and the substrate peptide is red . For clarity , sidechains are only shown for residues discussed in the text . ( B ) Phylogenetic tree indicating the identity of the DFGx amino acid and the transitions that occurred in the evolution of the CMGC group . Numbers indicate support for ancestral reconstructions . ( C ) Kinases were incubated with 45 μM peptide and initial velocities measured . In general , L to S mutations shift substrate preference towards +1P while S to L mutations shift preference towards +1 R . Bars show the log2 ratio of +1R and +1P initial velocities ( V0R/V0P ) . Black and white bars indicate wild type or maximum likelihood kinases incubated with peptides that contain serine and threonine respectively as phosphoacceptor . Dark and light orange indicate DFGx mutant kinases incubated with peptides that contain serine and threonine respectively as phosphoacceptor . Error bars are standard error of three assays . Figure 4—figure supplement 1 shows data for ICK compared to PSPL results . Figure 4—figure supplement 2 shows full Michaelis–Menten curves for selected kinases and DFGx mutants . DOI: http://dx . doi . org/10 . 7554/eLife . 04126 . 01410 . 7554/eLife . 04126 . 015Figure 4—figure supplement 1 . The phosphoacceptor affects the +1 specificity of ICK . ICK kinase was incubated with 45 μM peptide and initial velocities measured . Bars show the log2 ratio of +1R/+1P initial velocities ( V0R/V0P ) . Black and white bars indicate that ICK was incubated with peptides that contain serine and threonine respectively as phosphoacceptor . The gray bar indicates an equal mix of serine and threonine as phosphoacceptor . The lower gray bar is ratio data taken from the PSPL array in Figure 1—figure supplement 1 i . e . a ratio of phosphate incorporation into Y-A-X-X-X-X-X-S/T-R-X-X-X-A-G-K-K-biotin vs Y-A-X-X-X-X-X-S/T-P-X-X-X-A-G-K-K-biotin peptides , where S/T indicates an equal mixture of serine or threonine as phosphoacceptor and X indicates an equal mixture of all amino acids except C , S , or T at all other positions except +1 and the terminal linker amino acids . DOI: http://dx . doi . org/10 . 7554/eLife . 04126 . 01510 . 7554/eLife . 04126 . 016Figure 4—figure supplement 2 . Michaelis–Menten curves for selected kinases . Saccharomyces cerevisiae Ime2 , Saccharomyces cerevisiae Ime2-DFGx ( L231S ) , Mus musculus ICK , and Mus musculus ICK-DFGx ( L146S ) were incubated with varying peptide concentrations ( x-axis ) , and initial velocities for the reaction are plotted on the y-axis . The curves on the left contain serine as phosphoacceptor ( S-peptides ) while those on the right contain threonine ( T-peptides ) . Peptides with Pro at the +1 position are colored red and marked with triangles , peptides with Arg at the +1 position are colored blue and marked with circles ( see schematic , top ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04126 . 01610 . 7554/eLife . 04126 . 017Figure 4—figure supplement 3 . Variable phosphoacceptor preference for extant IME2/RCK/LF4 kinases . Kinases were incubated with 45 μM peptide and initial velocities measured . Bars show the log2 ratio of initial velocities with serine or threonine as phosphoacceptor ( V0S/V0T ) . Red and blue bars indicate the presence of arginine and proline , respectively , at the +1 substrate position . Error bars are standard deviations of three replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 04126 . 01710 . 7554/eLife . 04126 . 018Figure 4—figure supplement 4 . Phosphoacceptor preference shows a general shift from threonine towards threonine during evolution in the IME2/RCK/LF4 lineage . Kinases were incubated with 45 μM peptide and initial velocities measured . Bars show the log2 ratio of initial velocities with serine or threonine as phosphoacceptor ( V0S/V0T ) . Red and blue bars indicate the presence of arginine and proline , respectively , at the +1 substrate position . Error bars are standard deviations of three replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 04126 . 018 In addition to affecting kinase selectivity at the +1 position , as anticipated mutation of the DFGx residue also impacted phosphoacceptor preference ( see Figure 4 , Figure 4—figure supplement 2 ) . However , the identity of the DFGx residue appeared to only modestly affect the phosphoacceptor specificity in comparison to specificity at the +1 position , and did not follow a clear pattern . These results are in keeping with previous observations that CMGC kinases are generally less specific than other kinase groups for the phosphoacceptor residue ( Chen et al . , 2014; Figure 4 , Figure 4—figure supplement 3 ) . We also examined phosphoacceptor preference in our resurrected CMGC ancestors . A general trend is observed in which AncCMGI has a slight preference for serine and this shifts towards a slight threonine preference in the lineage leading to Ime2 ( Figure 4 , Figure 4—figure supplement 4 ) . In S . cerevisiae , Ime2 is expressed exclusively during meiosis and is required for all stages of this process , including meiotic initiation , S-phase , the meiotic divisions , and gametogenesis ( Yoshida et al . , 1990; Dirick et al . , 1998; Benjamin , 2003; Holt et al . , 2007; McDonald et al . , 2009 ) . This meiotic exclusive expression allows us to engineer allelic replacements of IME2 without any impact on vegetative cells , and then assess the ability of strains to complete various aspects of the meiotic program . The Ime2 DFGx mutant shifts from a strong +1R preference to an expanded specificity that tolerates both proline and arginine at the +1 position . This mutant has a comparable turnover to the wild-type kinase ( Figure 4 , Figure 4—figure supplement 2 ) and therefore can be used to meaningfully test the effect of changing primary specificity on phosphoregulatory networks in vivo . To this end , we replaced the endogenous IME2 gene with the ime2-DFGx allele and assayed the ability of cells to undergo meiosis . As reported previously ( Benjamin , 2003 ) , a kinase-dead version of Ime2 ( ime2-K97R ) failed to support meiosis ( not shown ) . However , cells with both copies of IME2 replaced with a DFGx mutant ( ime2-L231S ) completed meiosis , but with a reduction of sporulation efficiency thus indicating that the ime2-DFGx allele has significant activity in vivo ( Figure 5A ) . The homozygous ime2-DFGx cells that did correctly form tetrads had reduced spore viability ( Figure 5B , Figure 5—figure supplement 1 ) , and initiated S-phase ( Figure 5C ) and the meiotic divisions ( Figure 5D ) with a 2–3 hr delay . These defects may be caused by a weakening of network connections due to a reduced preference for arginine at the +1 position in the DFGx mutant . 10 . 7554/eLife . 04126 . 019Figure 5 . The S . cerevisiae meiotic phosphoregulatory network tolerates an expanded specificity DFGx mutant . ( A ) Sporulation efficiency with various IME2 alleles: wild-type IME2 ( WT ) , an ime2- ( L231S ) heterozygote ( WT/DFGx ) , or an ime2- ( L231S ) homozygote ( DFGx/DFGx ) . ( B ) Fraction of tetrad spores that , when dissected , gave rise to colonies ( representative images shown in Figure 5—figure supplement 1 ) . ( C ) Synchronous meiosis was induced and DNA content analyzed by SYTOX-Green staining and flow cytometry ( representative raw data shown in Figure 5—figure supplement 2 ) . ( D ) Synchronous meiosis was induced and DNA segregation events were scored by fluorescence microscopy . Error bars represent standard error of three or more biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 04126 . 01910 . 7554/eLife . 04126 . 020Figure 5—figure supplement 1 . Representative tetrad dissections . Synchronous meiosis was induced for wild-type IME2 ( WT ) , an ime2- ( L231S ) heterozygote ( WT/DFGx ) , or an ime2- ( L231S ) homozygote ( DFGx/DFGx ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04126 . 02010 . 7554/eLife . 04126 . 021Figure 5—figure supplement 2 . Representative cytometry data from meiosis experiments . Synchronous meiosis was induced and DNA content analyzed by SYTOX-Green staining and flow cytometry . Approximate positions of 2c and 4c DNA content are indicated . Beyond 8 hr , gametogenesis occurs , leading to packaging of DNA into spores and resulting in additional peaks . DOI: http://dx . doi . org/10 . 7554/eLife . 04126 . 021 The defects in the homozygous ime2-DFGx ( L231S ) strain could be due to either a weakening of existing network connections , a gain of new network connections that interfere with meiotic processes or a combination of both . We reasoned that the loss of important phosphorylation events would be recessive and could be compensated for by the presence of a wild-type IME2 allele , while the gain of new phosphorylation sites would be dominant . To test for dominant effects in the DFGx mutant , we generated heterozygous IME2/ime2-DFGx strains . IME2/ime2-DFGx strains had no noticeable defects in meiosis or sporulation . DNA replication and divisions occurred on schedule , spore formation was normal , and spore viability was high ( Figure 5 ) . Therefore , we conclude that the defects seen in the homozygous DFGx mutant strains are most likely due to weakening of network connections , rather than the creation of new deleterious phosphorylation sites . This experiment is analogous to gene duplication followed by paralog divergence in evolution and indicates that the meiotic phosphoregulatory network can tolerate divergence of specificity in a second copy of the IME2 gene .
During the process of submission the authors have learned that the DFGx specificity mutation described in this study is also observed in multiple cancer kinases including Aurora A , PKCgamma , Haspin , DDR1 , ITK , TRKA , IRAK3 and BRAF ( Creixell et al . unpublished ) , suggesting that amino acid changes that occur during evolution are resampled in cancer . It will be interesting to see if these mutations cause specificity changes in the context of oncogenesis .
Kinases were either cloned by PCR from respective organisms or from gifts: N . crassa Ime2 was a gift from Louise Glass , ICK and MOK cDNA were gifts from Tom Sturgill and Zheng Fu ( Fu et al . , 2006 ) , MAK cDNA was a gift from Alex Bullock , N . gruberi genomic DNA was a gift from Lillian Fritz-Laylin ( Fritz-Laylin et al . , 2010 ) . Ancestral kinases were synthesized either from gBlock gene fragments ( IDT ) or by Genscript . All plasmids were assembled by Gibson isothermal assembly ( Gibson et al . , 2009 ) , cloned in E . coli XL1-blue strains , and prepared by miniprep ( Qiagen , Redwood City , CA ) . The list of plasmids used in this study is presented in Supplementary file 1 . The RCK family of kinase is identified by various synonyms in the literature . Therefore , to avoid confusion , the mammalian RCK kinases used in this study are: Yeast strains were generated by standard transformation and crossing protocols . Protein purification was performed from W303 strains . We initially performed meiotic experiments in SK1 strains derived from the Herskowitz collection ( Benjamin , 2003 ) , but later switched to SK1 strains from Angelika Amon ( a gift from Elçin Ünal ) . Both SK1 strains gave similar results but the Amon background was more consistent . All yeast strains were generated by standard LiAc transformation ( Amberg et al . , 2005 ) . SK1 and W303 strains were heat shocked at 42°C for 15 and 40 min respectively . Point mutations of IME2 in the SK1 background were generated by 2-step loop-in , loop-out gene replacement technique using selection and counter-selection of the URA3 marker at the IME2 genomic locus . The list of yeast strains used in this study is presented in Supplementary file 2 . Liquid sporulation was conducted at 30°C as follows: strains were thawed on YP + 3% glycerol plates overnight , then patched on YPD plates , and grown overnight . 2 ml YPD liquid cultures were inoculated from patches and grown to saturation by shaking at 30°C , 250 rpm for 21–23 hr ( OD600 ≈ 7 . 0 ) . Cultures were diluted in 50 mL of YP + 1% KOAc to OD600 = 0 . 25 and grown overnight by shaking at 30°C , 250 rpm for 15–16 hr . Cells were pelleted and washed once with sterile water , then resuspended in 1% KOAc sporulation media to OD600 ≈ 2 . 5 . Sporulation cultures were shaken at 30°C , 250 rpm and samples were collected every hour for DNA staining and flow cytometry . Cells were fixed by mixing 0 . 5 ml of sporulation culture with 1 ml of EtOH . Fixed cells were pelleted and resuspended in water , then sonicated on a Branson Sonifier Model 450 at 10% amplitude for 3 s to break up cell clumps . Cells were pelleted , then resuspended in 100 µl of 40 µg/ml RNase A + 0 . 05% NP-40 + 50 mM Tris pH 7 . 4 , and incubated at 37°C for 1 hr . Finally , 50 µl of 40 µg/ml proteinase K + 1 µM SYTOX Green + 50 mM Tris pH 7 . 4 was added to each sample and incubated at 55°C for 1 hr prior to analysis by cytometry . Progression of meiotic divisions was measured by epifluorescence of SYTOX Green-stained cells . 100–200 cells were counted per sample . To measure sporulation efficiency , we counted the proportion of tetrad , dyad , and monad or unsporulated cells from synchronous sporulation cultures after 24–48 hr . W303 S . cerevisiae strains containing a 2 μm PGAL1-kinase-TAP plasmid ( pRSAB1234 backbone , originally a gift from Erin O'Shea ) were grown overnight to log phase in SC-URA media containing 2% raffinose ( Sigma ) , and then expression of N-terminal kinase domains was induced by addition of 2% galactose ( Sigma ) for 4 hr at 30°C . Cells were harvested by centrifugation at 8000 rpm , cell pellet washed and resuspended in 1× cell volume of lysis buffer containing 25 mM HEPES pH 8 . 0 , 300 mM NaCl , 0 . 1% NP-40 , 30 mM EGTA , 1 mM EDTA , and a protease/phosphatase inhibitor set was added immediately prior to harvest including 80 mM β-glycerophosphate , 50 mM NaF , 1 mM DTT , 1 mM Na3O4V , and 1 mM PMSF . The cell slurry was slowly dripped into liquid nitrogen to produce frozen pellets . These pellets were then pulverized in a cryogenic ball mill ( Retsch MM301 with 50 ml stainless steel grinding jars ) by five rounds of agitation at 15 Hz for 2 min , re-cooling the grinding jars in liquid N2 after each cycle . The grindate was then thawed and cell debris was cleared by centrifugation at 8000 rpm for 30 min followed by sequential filtration through 2 . 7 and 1 . 6 μm Whatman GD/X filters ( GE ) . C-terminally TAP-tagged kinases were immobilized on IgG Sepharose 6 Fast Flow beads ( GE ) . These beads ( ∼500 μl slurry per 1 l culture ) had been pre-equilibrated in lysis buffer with inhibitors and were then incubated with lysate for 1 hr at 4°C . Bound beads were then loaded into a disposable Bio-Spin column ( cat . #732-6008; BioRad ) by pipette and washed with 20 ml total wash buffer ( lysis buffer + 10% glycerol , 1 mM DTT ) at 4°C . The column was then rotated for 20 min at 23°C in 700 μl wash buffer as a final wash to mimic elution conditions . The bound protein was then cleaved from the IgG beads by TEV protease in 600 μl elution buffer ( 0 . 21 mg/ml TEV protease [QB3 MacroLab , UC Berkeley] , 25 mM HEPES pH 8 . 0 , 310 mM NaCl , 0 . 09% NP-40 , 26 . 9 mM EGTA , 0 . 9 mM EDTA , 1 mM DTT , and 10% glycerol ) for 1 hr at 23°C . For bacterial purification , proteins were expressed with an N-terminal 6× HIS tag in Rosetta2 DE3 pLysS competent cells ( QB3 MacroLab , UC Berkeley ) by induction with 100 µM IPTG for 18 hr at 16°C . The cell pellet was resuspended in an equal volume of bacterial wash buffer ( 50 mM Tris-Cl pH 8 . 0 , 300 mM NaCl , 10% glycerol , and 0 . 1% Triton X-100 ) , freeze-thawed once with liquid nitrogen , then lysed by 3× french press ( American Instrument Company ) cycles in the presence of 0 . 05 mg/ml DNAse I , 2 mM MgCl2 , 5 mM β-mercaptoethanol , and 1 mM PMSF . Lysate was cleared by centrifugation and filtration as above . NiNTA-Sepharose beads ( GE ) ( ∼660 μl slurry per 1 l culture ) were equilibrated in bacterial wash buffer containing 20 mM imidazole and combined at 1:1 ratio with lysate before adding a further 2 . 5 mM βME , 1 mM Na3O4V , 80 mM β-glycerophosphate , 50 mM NaF , and 0 . 5 mM PMSF . This lysate was incubated with beads for batch binding overnight at 4°C . Bound beads were loaded onto a disposable column ( as above ) and rinsed with the remaining unbound fraction . The column was washed 3 × with 10 ml bacterial wash buffer containing 10 mM imidazole and 2 mM βME , 3 × with 10 ml bacterial wash buffer containing 20 mM imidazole and 2 mM βME , then one final time with 20 ml bacterial wash buffer containing 20 mM imidazole , 2 mM βME , and 2 mM ATP . The column was rotated in previous ATP solution for 15 min at 4°C in an attempt to remove chaperones , before rinsing once more with 5 ml bacterial wash buffer containing 20 mM imidazole and 2 mM βME to remove ATP . The protein was released by incubation with 300 μl of the above buffer containing 250 mM imidazole twice , followed by incubation with 300 μl of the above buffer containing 500 mM imidazole twice . All purification samples were analyzed by SDS-PAGE with Coomassie Brilliant Blue R-250 stain . The library consisted of 182 peptide mixtures having the general sequence Y-A-x-x-x-x-x-S/T-x-x-x-x-A-G-K-K ( biotin ) , where X represents an equimolar mixture of the 17 proteogenic amino acid residues ( excluding Ser , Thr , and Cys ) , S/T indicates an even mixture of Ser and Thr , and biotin is conjugated through an aminohexanoic acid spacer to the C-terminal Lys residue . In each mixture , a single residue at one of the 9 ‘x’ positions was fixed as one of the 20 amino acids . In addition , we included two peptide mixtures in which all ‘x’ positions were degenerated , but the phosphoacceptor residue was fixed as either Ser or Thr . Peptides were arrayed in 1536 well plates to a final concentration of 50 μM in 2 μl reaction buffer ( 50 mM HEPES , pH 7 . 4 , 150 mM NaCl , 10 mM MgCl2 , 0 . 1% Tween 20 ) per well . Reactions were initiated by adding kinase mixed with ATP ( final concentration 50 μM with 0 . 03 μCi/μl [γ-33P]ATP ) . Plates were incubated at 30°C for 2 hr , and then 200 nl aliquots were transferred to streptavidin-coated membrane ( Promega ) , which was quenched by immersion in 0 . 1% SDS , 10 mM Tris–HCl , pH 7 . 5 , 140 mM NaCl . Membranes were then washed twice with the same solution , twice with 2 M NaCl , and twice with 1% H3PO4 , 2 M NaCl . After briefly rinsing with ddH2O , membranes were air-dried and exposed to a phosphor imager screen . Following scanning on a phosphor imager ( BioRad ) , radiolabel incorporation was quantified using QuantityOne software . Data were normalized so that the average signal for a given peptide position was 1 . For visualization normalized data from two separate runs were averaged , log transformed , and used to generate heat maps in Microsoft Excel using the color scheme shown in the figures . Ratiometric specificities were profiled in buffer containing 77 . 5 mM HEPES pH 7 . 5 , 77 . 5 mM NaCl , 15 . 5 mM MgCl2 , 250 μM ATP , 0 . 45 mg/ml BSA , 4 . 5% glycerol , and 0 . 2 μCi [γ-32P]ATP . Minimal kinase concentrations sufficient for signal were determined empirically and ranged from 5 to 50 nM . Peptides obtained from Tufts University Core Facility ( http://www . tucf . com ) were added to a final concentration of 45 μM to start the reaction . Comparative peptide assays were always performed in parallel . Reaction assays were aliquoted onto Whatman P81 phosphocellulose ( GE ) strips , which were then quenched and washed 5 × in 75 mM phosphoric acid to remove free [γ-32P] ATP . Samples were dried on a slab gel dryer ( Model 1125B; BIORAD ) and exposed to a phosphor screen ( Molecular Dynamics ) to determine the rate of [γ-32P] ATP incorporation . Phosphor screens were analyzed with a Typhoon 9400 scanner ( Amersham ) using ImageQuant software ( GE ) . Final Image quantification was performed using ImageJ ( http://imagej . nih . gov/ij/ ) . Michaelis–Menten curves were generated in a similar manner , except the buffer contained 50 mM HEPES pH 7 . 5 , 50 mM NaCl , 10 mM MgCl2 , 500 µM ATP , 83 . 3 μg/ml BSA , 0 . 833% glycerol , and 0 . 2 μCi [γ-32P]ATP . In this case , substrate concentration was varied for each kinase peptide combination . Data were fit by nonlinear regression to the Michaelis–Menten model V0 = Vmax*[S]/KM+[S] using Prism ( GraphPad software ) and Matlab ( MathWorks ) and this fit was used to determine values for Vmax and KM . Orthologs of the CMGC gene family were identified by a BLAST search based on the amino acid sequence of S . cerevisiae IME2 and H . sapiens CDK1 , using the NCBI BLAST tool ( Altschul et al . , 1990 ) . To eliminate false positives , hit sequences were reverse BLAST queried , and we eliminated any hits that did not have either IME2 or CDK1 as a result with at least 50% sequence identity . Using the remaining 329 amino acid sequences , a multiple sequence alignment was inferred using MSAProbs with default settings ( Liu et al . , 2010 ) . This alignment was best-fit by the LG model ( Le and Gascuel , 2008 ) with a gamma-distributed set of evolutionary rates ( Yang and Kumar , 1996 ) , according to the Akaike Information Criterion as implemented in PROTTEST ( Abascal et al . , 2005 ) . Using LG+G , we used a maximum likelihood ( ML ) algorithm ( Yang et al . , 1995 ) to infer the ancestral amino sequences with the highest probability of producing all the extant sequence data . Specifically , we used RAxML version 7 . 2 . 8 to infer the ML topology , branch lengths , and evolutionary rates ( Stamatakis , 2006 ) . We exported this ML phylogeny to another software package , PhyML ( Guindon et al . , 2010 ) , in order to calculate statistical support for branches as approximate likelihood ratios . We next reconstructed ML ancestral states at each site for all ancestral nodes using the software package Lazarus ( Hanson-Smith et al . , 2010 ) . We used sequences from the CK family as the outgroup to root the tree . We placed ancestral insertion/deletion characters according to Fitch's parsimony ( Fitch , 1971 ) , in which each indel character was treated independently . We extracted the ancestral sequences from the phylogenetic nodes corresponding to AncCMGI , AncLF4 , AncNgru , AncICK , AncCneo , AncNcra , and AncYlip . We characterized the support for these ancestors by binning their posterior probabilities of states into 5%-sized bins and counting the proportion of ancestral sites within each bin ( Figure 2—figure supplement 1 ) . We also generated alternate versions of these ancestral sequences by randomly sampling from their posterior distributions to generate between 2 and 3 alternate ancestors for every node , as described in Williams et al . ( 2006 ) . Ancestral sequence reconstruction is a probabilistic method and involves uncertainties in the amino acid identities . Even for relatively well-conserved protein families like kinases , these uncertainties become more pronounced when attempting to reconstruct deep ancestors . A summary of the statistical supports for each of the resurrected kinases is presented in Figure 2—figure supplement 1 . For each of the seven resurrected kinases between 60 and 90 alternative amino acids were sampled by a Bayesian method to address the impact of uncertainties on our results . In all cases , both the overall primary specificities and the +1 specificities as determined by individual peptides were robust to uncertainties in ancestral sequence reconstruction . That is , kinases that we resurrected with amino acid substitutions that explored different possible amino acids had very similar activity levels and specificities as the maximum likelihood ancestors presented in the main figures . The degree of preference for proline varies slightly in the deepest alternative reconstruction , but the general trend of evolution from +1P preference , to expanded specificity , and finally to +1R preference , is maintained . The data from alternative reconstructions is presented in Figure 2—figure supplement 2 and Figure 3—figure supplement 2 . Together , these results give us confidence in the validity of our approach . | All living things have enzymes called protein kinases that transfer chemical tags called phosphates onto other proteins . Adding a phosphate to a protein can change the protein's activity—for example , by switching it on or off—and many biological processes involve large networks of kinases that phosphorylate hundreds of proteins . Humans have approximately 500 different protein kinases , which can each phosphorylate many proteins , and so human cells are regulated by tens of thousands of different phosphorylation sites . This raises a number of questions: how have these different kinases evolved over evolutionary history ? And how have they come to recognize , and phosphorylate , so many different sites ? Howard , Hanson-Smith et al . studied members of a large family of protein kinases called the CMGC group . These enzymes are found in all organisms that have a cell nucleus , including animals , plants , and fungi . All proteins , including kinases , are built up of a chain of smaller molecules called amino acids , and the ability of a kinase to phosphorylate a protein depends on the kinase recognizing a short string of amino acids known as a motif . The phosphate is added in the middle of this motif at the so-called ‘0’ position . Howard , Hanson-Smith et al . found that all of the CMGC protein kinases tested ‘preferred’ an arginine or proline amino acid at the ‘+1’ position of this motif . However , some kinases preferred motifs with an arginine amino acid at this position , and others preferred a proline instead . Howard , Hanson-Smith et al . predicted how the ancestors of a number of CMGC protein kinases might have looked and then ‘resurrected’ them by producing them in yeast cells . When the preference of these ancestral enzymes was tested , the oldest ancestor was found to slightly prefer motifs that had a proline amino acid at the +1 position . Testing six more recent ancestors showed that , over a billion years of evolution , this amino acid preference became broader to include both proline and arginine—and that some modern protein kinases subsequently evolved and specialized to prefer arginine at the +1 position , thus creating a new specificity . Kinases are sometimes likened to microchips in complex electronic networks . In this analogy , expanding the specificity of a kinase could be like creating many ‘loose wires’ and cause short-circuits . From their evolutionary analysis , Howard , Hanson-Smith et al . were able to identify a structural change in the enzyme that causes an expansion of kinase specificity , which allowed them to directly test this idea in cells . Expanding the specificity of a protein kinase that controls sexual cell division in yeast cells did not stop the yeast from dividing to produce spores , suggesting that these changes are more readily tolerated than was expected . Howard , Hanson-Smith et al . suggest that this unexpected robustness of cellular circuits enabled the evolution of the wide variety of protein kinases seen today . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology",
"biochemistry",
"and",
"chemical",
"biology"
] | 2014 | Ancestral resurrection reveals evolutionary mechanisms of kinase plasticity |
Phosphorylation of the α-subunit of initiation factor 2 ( eIF2 ) controls protein synthesis by a conserved mechanism . In metazoa , distinct stress conditions activate different eIF2α kinases ( PERK , PKR , GCN2 , and HRI ) that converge on phosphorylating a unique serine in eIF2α . This collection of signaling pathways is termed the ‘integrated stress response’ ( ISR ) . eIF2α phosphorylation diminishes protein synthesis , while allowing preferential translation of some mRNAs . Starting with a cell-based screen for inhibitors of PERK signaling , we identified a small molecule , named ISRIB , that potently ( IC50 = 5 nM ) reverses the effects of eIF2α phosphorylation . ISRIB reduces the viability of cells subjected to PERK-activation by chronic endoplasmic reticulum stress . eIF2α phosphorylation is implicated in memory consolidation . Remarkably , ISRIB-treated mice display significant enhancement in spatial and fear-associated learning . Thus , memory consolidation is inherently limited by the ISR , and ISRIB releases this brake . As such , ISRIB promises to contribute to our understanding and treatment of cognitive disorders .
In metazoa , diverse stress signals converge at a single phosphorylation event at serine 51 of a common effector , the translation initiation factor eIF2α . This step is carried out by four eIF2α kinases in mammalian cells: PERK , which responds to an accumulation of unfolded proteins in the endoplasmic reticulum ( ER ) , GCN2 to amino acid starvation and UV light , PKR to viral infection , and HRI to heme deficiency . This collection of signaling pathways has been termed the ‘integrated stress response’ ( ISR ) , as they converge on the same molecular event . eIF2α phosphorylation results in an attenuation of translation with consequences that allow cells to cope with the varied stresses ( Wek et al . , 2006 ) . eIF2 ( which is comprised of three subunits , α , β and γ ) binds GTP and the initiator Met-tRNA to form the ternary complex ( eIF2-GTP-Met-tRNAi ) , which , in turn , associates with the 40S ribosomal subunit forming the 43S pre-initiation complex ( PIC ) that scans the 5′UTR of mRNAs to select the initiating AUG codon . Upon phosphorylation of its α-subunit , eIF2 becomes a competitive inhibitor of its guanine nucleotide exchange factor ( GEF ) , eIF2B ( Hinnebusch and Lorsch , 2012 ) . The tight and non-productive binding of phosphorylated eIF2 to eIF2B prevents loading of the eIF2 complex with GTP thus blocking ternary complex formation and reducing translation initiation ( Krishnamoorthy et al . , 2001 ) . Because eIF2B is less abundant than eIF2 , phosphorylation of only a small fraction of the total eIF2 has a dramatic impact on eIF2B activity in cells . Paradoxically , under conditions of reduced protein synthesis , a small group of mRNAs that contain upstream open reading frames ( uORFs ) in their 5′UTR are translationally up-regulated ( Hinnebusch , 2005; Jackson et al . , 2010 ) . These include mammalian ATF4 ( a cAMP element binding [CREB] transcription factor ) and CHOP ( a pro-apoptotic transcription factor ) ( Harding et al . , 2000; Vattem and Wek , 2004; Palam et al . , 2011 ) . ATF4 regulates the expression of many genes involved in metabolism and nutrient uptake and additional transcription factors , such as CHOP , which is under both translational and transcriptional control ( Ma et al . , 2002 ) . Phosphorylation of eIF2α thus leads to preferential translation of key regulatory molecules and directs diverse changes in the transcriptome of cells upon cellular stress . One of the eIF2α kinases , PERK , lies at the intersection of the ISR and the unfolded protein response ( UPR ) that maintains homeostasis of protein folding in the ER ( Pavitt and Ron , 2012 ) . The UPR is activated by unfolded or misfolded proteins that accumulate in the ER lumen because of an imbalance between protein folding load and protein folding capacity , a condition known as ‘ER stress’ . In mammals , the UPR is comprised of three signaling branches mediated by ER-localized transmembrane sensors , PERK , IRE1 , and ATF6 . These sensor proteins detect the accumulation of unfolded protein in the ER and transmit the information across the ER membrane , initiating unique signaling pathways that converge in the activation of an extensive transcriptional response , which ultimately results in ER expansion ( Ron and Walter , 2007 ) . The lumenal stress-sensing domains of PERK and IRE1 are homologous and likely activated in analogous ways by direct binding to unfolded peptides ( Gardner and Walter , 2011 ) . This binding event leads to oligomerization and trans-autophosphorylation of their cytosolic kinase domains , and , for PERK , phosphorylation of its only known substrate , eIF2α . In this way , PERK activation results in a quick reduction in the load of newly synthesized proteins that are translocated into the ER-lumen ( Harding et al . , 2000 ) . Upon ER stress , both the transcription factor XBP1s , produced as the consequence of a non-conventional mRNA splicing reaction initiated by IRE1 , and the transcription factor ATF6 , produced by proteolysis and release from the ER membrane , collaborate with ATF4 to induce the vast UPR transcriptional response . Transcriptional targets of the UPR include the ER protein folding machinery , the ER-associated degradation machinery , and many other components functioning in the secretory pathway ( Walter and Ron , 2011 ) . Although the UPR initially mitigates ER stress and as such confers cytoprotection , persistent and severe ER stress leads to activation of apoptosis that eliminates damaged cells ( Shore et al . , 2011; Tabas and Ron , 2011 ) . By interrogating a large chemical library for small molecules that block PERK signaling , we identified ISRIB as a potent ISR inhibitor , functioning downstream of all eIF2α kinases . ISRIB proves a powerful tool to explore the consequences of acute inhibition of the ISR in cells and animals .
To identify inhibitors of PERK signaling , we engineered a reporter that allows monitoring of PERK activation in living cells . To this end , we constructed a retroviral vector containing the open-reading frame of firefly luciferase fused to the 5′UTR of ATF4 mRNA ( Figure 1A ) , which contains two short open-reading frames ( uORFs ) that control ATF4 translation in a stress-dependent manner . After infection , we established a HEK293T cell line harboring the stably integrated reporter . We used thapsigargin , a potent ER stressor that inhibits the ER calcium pump , to activate PERK and induce eIF2α phosphorylation . Thapsigargin treatment resulted in a 4 . 9-fold induction in luciferase activity in a 384 well format with a Z factor of 0 . 5 ( Figure 1B ) . This format was used to screen 106 , 281 compounds covering a wide chemical space . We identified 460 hits ( 0 . 43% ) ( Figure 1C ) , which were further validated in an 8-point dose-response assay using the same reporter . We further triaged the compounds by discarding inhibitors that also affected the IRE1 branch of the UPR using a luciferase-based XBP1 splicing reporter . Less than half ( 187 hits ) of our initial hits proved unique to the PERK branch . We next used an orthogonal secondary screen that employed a different reporter ( bi-cistronic ATF4-dGFP-IRES-mCherry ) stably integrated into a different cell line ( U2OS cells ) . The read-out of this latter screen was microscopy-based , which allowed us to simultaneously assess acute toxicity by cell counting , further reducing the number of viable hits to 77 ( data not shown ) . As a tertiary screen , we tested compounds for their ability to inhibit ER stress-elicited induction of endogenous ATF4 by Western blot analysis . Twenty-eight compounds passed this test and were analyzed further . 10 . 7554/eLife . 00498 . 003Figure 1 . High-throughput cell-based screen for inhibitors of PERK signaling . ( A ) Schematic representation of the ATF4 luciferase reporter used in the primary screen . The 5′ UTR of human ATF4 containing the uORFs 1 and 2 was fused to firefly luciferase and inserted into a retroviral expression system . ( B ) Primary screen optimization . HEK293T stably expressing the ATF4 luciferase reporter were plated in 384-well plates and treated for 6 hr with 100 nM thapsigargin ( Tg ) or DMSO as a no ER stress control . Luciferase production was measured at the end point after 6 hr ( mean ± SD ) . The Z′ was calculated as 1− ( 3 [σ Tg + σ DMSO]/[μ Tg–μ DMSO] ) . ( C ) Primary screen results . The ATF4 luciferase reporter cell line was treated for 6 hr with 100 nM thapsigargin and library compounds ( 10 µM ) . Inhibition of the luciferase activity reporter was calculated as the percent reduction in relative luminescence normalized to thapsigargin treatment ( 0% inhibition ) and the no-ER stress control ( 100% inhibition ) . Compounds were considered hits if they lied beyond three standard deviations ( SD ) from the thapsigargin treatment mean ( red line ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00498 . 003 One of the 28 compounds was of particular interest because of its high potency in cells ( library compound IC50 = 40 nM ) . This compound ( henceforth referred to as ‘ISRIB’ for Integrated Stress Response inhibitor ) is a symmetric bis-glycolamide , containing a central bi-substituted cyclohexane , and can exist as two diastereomers , cis and trans ( Figure 2A ) . We synthesized both isomers and tested their ability to inhibit the ATF4-luciferase reporter ( Figure 2B ) . Trans-ISRIB proved 100-fold more potent ( IC50 = 5 nM ) than cis-ISRIB ( IC50 = 600 nM ) , indicating that the compound’s interaction with its cellular target is stereospecific . Given the two-order-of-magnitude difference in activity in this assay , the measured activity of cis-ISRIB may be an over-estimate , as we cannot exclude a small contamination with trans-ISRIB , which is far more potent . The lower IC50 of trans-ISRIB relative to the compound in the small molecule library indicates that the library likely contains a mixture of the two stereoisomers . All further experiments in this study were carried out with the synthesized trans-isomer of ISRIB . 10 . 7554/eLife . 00498 . 004Figure 2 . Identification of ISRIB as a potent cell-based inhibitor of PERK signaling . ( A ) Structures of ISRIB isosteromers . ( B ) Inhibition of the ATF4 luciferase reporter in HEK293T cells by ISRIB stereoisomers . Inhibition is plotted in relation to the concentration of either the cis or trans isomer of ISRIB . Cells were treated with 2 µg/ml of tunicamycin to induce ER stress and different concentrations of the inhibitors for 7 hr ( N = 2 , mean ± SD ) . ( C ) Effect of ISRIB on production of endogenous ATF4 , PERK phosphorylation , and XBP1s production . An immunoblot analysis of PERK , ATF4 and XBP1s in HEK293T cells treated with different ER stress inducers ( 2 . 5 µg/ml tunicamycin [Tm] or 100 nM thapsigargin [Tg] ) with or without 200 nM ISRIB for 3 hr is shown . The arrowhead marks the XBP1s specific band . ( D ) Effect of ISRIB on XBP1 mRNA splicing . Taqman assays for XBP1unspliced ( XBP1u ) and XBP1spliced ( XBP1s ) on cDNA synthesized from total RNA extracted from U2OS cells treated with 2 µg/ml of tunicamycin in the presence or absence of 200 nM ISRIB for the indicated times are shown . Percent splicing was calculated as the ratio of XBP1s over total XBP1 mRNA ( XBP1u + XBP1s ) ( mean ± SD ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00498 . 004 We next determined at which step ISRIB blocks ATF4 production . To this end , we first probed the phosphorylation status of PERK by Western blotting . PERK phosphorylation is indicative of its activation by autophosphorylation and can be recognized by reduced mobility on SDS-polyacrylamide gels . Notably , ISRIB did not inhibit the mobility shift of PERK observed in ER-stressed cells ( Figure 2C ) . Rather , we observed an exaggerated mobility shift , indicative of increased phosphorylation of PERK upon ER stress , induced by either thapsigargin or tunicamycin ( an inhibitor of N-linked glycosylation ) . In each case , in the absence of ISRIB , ATF4 and XBP1s were produced upon ER stress induction . In agreement with the behavior of the reporters described above , ISRIB blocked production of endogenous ATF4 , whereas XBP1 mRNA splicing ( Figure 2D ) and XBP1s production persisted ( Figure 2C and Figure 3—figure supplement 1 ) . As shown below ( Figure 5D ) , ISRIB also did not affect activation of the ATF6-branch of the UPR . We conclude that ISRIB specifically blocks signaling of the PERK-branch of the UPR . Given that PERK phosphorylation was not diminished in ISRIB-treated , ER-stressed cells , we next directly assessed eIF2α phosphorylation . We measured the levels of phosphorylated eIF2α using an antiphospho-eIF2α antibody-based assay to quantify phosphorylation at serine 51 ( see ‘Materials and methods’ ) . Upon induction of ER stress by tunicamycin or thapsigargin , phosphorylation of eIF2α increased over time , reaching a fourfold and sevenfold increase after 120 min respectively ( Figure 3A ) . Unexpectedly , ISRIB did not block eIF2α phosphorylation under either of these ER stress-inducing conditions . On the contrary , 120 min after tunicamycin addition , ISRIB further increased the level of eIF2α phosphorylation , approaching that obtained with thapsigargin . ISRIB alone had no effect on eIF2α phosphorylation . These results indicate that ISRIB blocks effects downstream of PERK and eIF2α phosphorylation . 10 . 7554/eLife . 00498 . 005Figure 3 . ISRIB makes cells resistant to eIF2α phosphorylation . ( A ) ISRIB does not block eIF2α phosphorylation upon ER stress . eIF2α phosphorylation was measured using an alpha-screen Surefire eIF2α p-S51 assay ( see ‘Materials and methods’ ) . U2OS cells were plated in 96-well plates and treated with 2 µg/ml tunicamycin or 100 nM thapsigargin in the presence or absence of 100 nM ISRIB for the indicated times or with ISRIB alone for 120 m ( N = 4 , mean ± SD ) . See Figure 3—figure supplement 1 for supporting Western blot analysis of eIF2α phosphorylation . ( B ) ISRIB blocks global translational attenuation observed after eIF2α phosphorylation during ER stress . HEK293T cells were treated with 100 nM thapsigargin and 200 nM ISRIB for either 1 or 3 hr prior to a 20 min pulse with 35S methionine before lysis . Equal amounts of lysate were loaded on an SDS-PAGE gel and quantification of radiolabeled methionine incorporation of lysates was done by gel densitometry ( N = 2 , SD ) using ImageJ . ( see Figure 3—figure supplement 2 for SDS-PAGE ) . ( C ) Polysome gradient analysis showing the block in global translational attenuation upon addition of ISRIB on ER-stressed cells . MEFs were grown in the presence or absence of 2 µg/ml of tunicamycin with or without 200 nM ISRIB for 1 hr . Cell lysates were loaded on a 10–50% sucrose gradient , centrifuged at 150 , 000×g for 2 . 4 hr and absorbance at 254 nm was measured across the gradient ( see Figure 3—figure supplement 3 for quantitation of polysome profile ) . A representative experiment is shown ( N = 3 ) . See Figure 3—figure supplement 4 for a close-up of the disome and trisome peaks . ( D ) Cells treated with ISRIB are resistant to the global translational attenuation exerted by forced expression of eIF2α ( S51D ) . HEK293Trex cells were transduced with a tetracycline inducible phospho-mimetic ( S51D ) allele of eIF2α . Transgene expression was induced by addition of 25 nM doxycycline for 14 hr in the presence or absence of 200 nM ISRIB . Lysates were collected and analyzed as described in panel ( C ) ( see Figure 3—figure supplement 6 for quantitation of polysome profile ) . A representative experiment is shown ( N = 2 ) . ( E ) ISRIB does not reverse global translational attenuation exerted through inhibition of CAP-dependent initiation . Wild-type MEFs were treated with 750 nM Torin-1 in the presence or absence of 200 nM ISRIB for 2 hr . Lysates were collected and analyzed as described in panel ( C ) . A representative experiment is shown ( N = 2 ) . ( F ) ISRIB blocks production of ATF4 upon GCN2 or HRI activation . An immunoblot analysis of PERK , ATF4 and total eIF2α in HEK293T cells starved for cysteine and methionine or treated with an HRI activator ( 6 µM ) for 5 hr in the presence or absence of 200 nM ISRIB is shown . Tunicamycin was used as a positive control for induction of ATF4 and the shift in PERK mobility . Under amino acid starvation we consistently observe a partial reduction of ATF4 production by ISRIB by Western blot analysis but observe a complete block in induction of the ATF4 luciferase reporter ( see Figure 3—figure supplement 7 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00498 . 00510 . 7554/eLife . 00498 . 006Figure 3—figure supplement 1 . ISRIB does not inhibit eIF2α phosphorylation or XBP1s production . Western blot analysis of PERK , ATF4 , XBP1s , phospho S51-eIF2α , total eIF2α , phospho S539-eIF2Bε and total eIF2Bε in HEK293T cells treated with or without 2 µg/ml of tunicamycin or 100 nM thapsigargin in the presence or absence of 200 nM ISRIB for the indicated times . DOI: http://dx . doi . org/10 . 7554/eLife . 00498 . 00610 . 7554/eLife . 00498 . 007Figure 3—figure supplement 2 . ISRIB blocks translational attenuation upon ER stress . Autoradiogram ( left ) and total protein ( right ) obtained from HEK293T cells that were treated with 100 nM thapsigargin with or without 200 nM ISRIB for either 1 or 3 hr prior to a 20 min pulse with 35S-methionine before lysis . Equal amounts of lysate were loaded on an SDS-PAGE gel . DOI: http://dx . doi . org/10 . 7554/eLife . 00498 . 00710 . 7554/eLife . 00498 . 008Figure 3—figure supplement 3 . ISRIB blocks translational attenuation upon ER stress . The polysome profile in Figure 3C was quantitated by calculating the area under the curve corresponding to the monosome peak ( 80S ) , or the area under the curve corresponding to the trace covering the polysome region and then plotted as a ratio over the area under the curve corresponding to the peak of the 60S subunit . DOI: http://dx . doi . org/10 . 7554/eLife . 00498 . 00810 . 7554/eLife . 00498 . 009Figure 3—figure supplement 4 . ISRIB partially restores the halfmer population in ER stressed cells . Wildtype MEFs were grown in the presence or absence of 2 µg/ml of tunicamycin with or without 200 nM ISRIB for 1 hr . This graph is a close up of the disome and trisome peaks of the polysome gradients in Figure 3C . DOI: http://dx . doi . org/10 . 7554/eLife . 00498 . 00910 . 7554/eLife . 00498 . 010Figure 3—figure supplement 5 . Disappearance of the halfmer peaks upon ER-stress is dependent on eIF2α phosphorylation . eIF2αS51A/S51A ( Eif2s1S51A/S51A ) MEFs were grown in the presence or absence of 2 µg/ml of tunicamycin with or without 200 nM ISRIB for 1 hr and polysomes gradients were processed and analyzed as described in Figure 3C . DOI: http://dx . doi . org/10 . 7554/eLife . 00498 . 01010 . 7554/eLife . 00498 . 011Figure 3—figure supplement 6 . ISRIB sustains translation upon expression of eIF2α ( S51D ) . The polysome profile in Figure 3D was quantitated as described in Figure 3—figure supplement 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 00498 . 01110 . 7554/eLife . 00498 . 012Figure 3—figure supplement 7 . ISRIB blocks induction of the ATF4 luciferase translational reporter upon HRI and GCN2 activation . HEK293T carrying the ATF4 luciferase reporter were treated with 2 µg/ml of tunicamycin to induce ER stress , 6 µM of the HRI activator or grown in media lacking cysteine and methionine for 7 hr in the presence or absence of 200 nM ISRIB ( N = 4 ) . The relative luciferase units are normalized to the no treatment control . Using this reporter we observe a smaller fold change in production of luciferase by amino acid starvation that is fully blocked by addition of ISRIB . DOI: http://dx . doi . org/10 . 7554/eLife . 00498 . 012 One way of explaining why ISRIB blocks ATF4 production yet leaves eIF2α phosphorylation intact is by rendering cells insensitive to the effects of this phosphorylation event . In agreement with this notion , ISRIB sustained global translation ( as monitored by 35S-methionine incorporation into newly synthesized polypeptides ) even in the presence of ER stress ( Figure 3B ) . After thapsigargin treatment , cells experienced a 40% drop in translation , which was abolished by ISRIB . As predicted by this result , extracts prepared from mouse embryonic fibroblasts ( MEFs ) experiencing ER stress showed a pronounced increase in the 80S monosomes at the expense of polyribosomes ( Figure 3C ) , which was reversed ( at least partially ) by addition of ISRIB . We chose MEFs for this analysis because they show stronger translational inhibition in response to ER stress than HEK293T cells . ISRIB was the only molecule in our collection of 28 hits that reversed translational attenuation upon ER-stress . Under normal growth conditions , an abundance of 43S pre-initiation complexes ( PICs ) leads to mRNAs loaded with a small ribosomal subunit in addition to fully assembled ribosomes . The presence of PICs on an mRNA can be detected as ‘halfmer’ peaks on polysome gradients . In the gradients shown in Figure 3C , addition of a PIC to disomes and trisomes was well resolved ( enlarged in Figure 3—figure supplement 4 ) . Upon eIF2α phosphorylation in ER-stressed cells , the reduction in PIC resulted in disappearance of the halfmer population . As expected , the disappearance of the halfmer peak upon ER-stress was dependent on eIF2α phosphorylation as no reduction was observed in MEFs that solely express non-phosphorylatable eIF2α ( S51A ) ( Figure 3—figure supplement 5 ) . Importantly , ISRIB partially restored the halfmer population in ER-stressed cells , providing support to the notion that it helps maintain high PIC levels even when eIF2α is phosphorylated ( Figure 3—figure supplement 4 ) . These data indicate that ISRIB exerts its function by maintaining elevated ternary complex levels . To further ascertain that cells treated with ISRIB are resistant to the effects of eIF2α phosphorylation , we transduced an inducible phospho-mimetic allele of eIF2α in which serine 51 was changed to an aspartic acid ( S51D ) into HEK293T cells . Expression of this allele upon doxycycline addition induced translational attenuation ( Figure 3D ) as seen by an increase in the 80S peak and a decrease in the polysome population . ISRIB rescued translation returning it to the levels observed in non-induced cells . In conclusion , ISRIB restores translation in cells containing either phospho-eIF2α or eIF2α ( S51D ) , thereby excluding any pleiotropic effects that might have been caused by the reagents used to activate ER stress . To rule out that ISRIB exerts non-specific effects on translation independent of eIF2α phosphorylation , we tested whether ISRIB reverses a translational block in CAP-mediated initiation . To this end we used Torin-1 , an inhibitor of mTOR that blocks phosphorylation of 4E-BP1 and S6K1 , and leads to translational attenuation ( Thoreen et al . , 2012 ) . Addition of Torin-1 to MEFs led to an increase in the 80S peak and reduction in the polysome population to a similar extent as shown above in cells treated with ER stressors or expressing eIF2α ( S51D ) ( Figure 3E , compare with Figure 3C , D ) . In contrast to these treatments , addition of ISRIB did not reverse the effect of Torin-1 on translation . Therefore , the ability of ISRIB to block translational attenuation is specific to eIF2α phosphorylation . If ISRIB makes cells insensitive to eIF2α phosphorylation , it should not matter which kinase phosphorylates eIF2α . To test this prediction , we subjected cells to amino acid starvation , which activates the eIF2α kinase GCN2 and leads to ATF4 production . In addition , we used a recently identified small molecule activator to induce eIF2α phosphorylation by activating HRI , another eIF2α kinase ( Chen et al . , 2011 ) . As expected , ISRIB blocked ATF4 induction after activation of either GCN2 or HRI ( Figure 3F ) . Under both conditions , PERK was not activated as shown by a lack of mobility shift . These data suggest that ISRIB is a bona fide ISR inhibitor that blocks signaling downstream of all eIF2α kinases . Both DDIT3 ( the gene encoding CHOP ) and PPP1R15A ( the gene encoding GADD34 ) are transcriptional targets of ATF4 . Thus , blocking ATF4 accumulation with ISRIB should result in a reduction in the transcriptional induction of the mRNAs encoding these targets . As shown in Figure 4A , GADD34 and CHOP mRNAs accumulated in ER-stressed U2OS cells , and ISRIB significantly reduced their induction . In agreement , we observed no CHOP accumulation after induction of ER stress in ISRIB-treated cells ( Figure 4B ) . Thus ISRIB impairs the transcriptional network governed by ATF4 during the ISR . 10 . 7554/eLife . 00498 . 013Figure 4 . ISRIB impairs induction of the transcriptional network controlled by ATF4 . ( A ) ER-Stress dependent induction of CHOP and GADD34 mRNA is impaired in cells treated with ISRIB . qPCR analysis of total RNA extracted from U2OS cells treated with 2 µg/ml of tunicamycin in the presence or absence of 200 nM ISRIB for the indicated times . mRNA levels for each sample were normalized to GAPDH ( N = 4 , mean ± SD ) . p values are derived from a one-tail Student’s t-test for unpaired samples . Statistical significance: CHOP , *p=0 . 0006; GADD34 , *p=0 . 0008 . ( B ) ISRIB blocks CHOP production during ER stress . An immunofluorescence analysis of U2OS cells treated with 100 nM thapsigargin for 2 or 4 hr in the presence or absence of 200 nM ISRIB is shown . A secondary Alexa Dye 488 anti-mouse antibody and rhodamine-phalloidin were used to visualize CHOP and actin , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 00498 . 013 As previously shown , cells homozygous for non-phosphorylatable eIF2α , eIF2α ( S51A ) , are unable to cope with ER stress properly , leading to reduced viability ( Lu et al . , 2004 ) . This indicates that events downstream of eIF2α phosphorylation are required to resolve the stress . As shown in Figure 5A , ISRIB treatment of wild-type cells had similar consequences . Importantly , addition of ISRIB alone did not affect cell viability , as judged by the number of colonies that form after acute treatment . By contrast , ISRIB addition caused a strong synergistic effect on ER-stressed cells , reducing colony number and size significantly more than ER-stress alone . This reduction in cell survival resulted from activation of apoptosis as the activity of the executioner caspases 3 and/or 7 was significantly induced under these conditions ( Figure 5B; Salvesen and Ashkenazi , 2011 ) . 10 . 7554/eLife . 00498 . 014Figure 5 . ISRIB impairs adaptation to ER-stress prolonging activation of the UPR sensors . ( A ) ISRIB sensitizes cells to acute ER stress . HEK293T cells were subjected with an acute dose of tunicamycin ( 2 µg/ml ) , ISRIB ( 200 nM ) or a combination of both for 24 hr . The treated cells were equally diluted to a concentration that would allow single cell clonal expansion and re-seeded onto six-well plates in a threefold dilution series . Clonal colonies were visualized by Crystal Violet stain . ( B ) ISRIB synergizes with ER stress to activate caspase 3/7 . Hela cells were plated in 96-well plates and treated with 5 µg/ml of tunicamycin or 500 nM thapsigargin with or without 25 nM ISRIB for the indicated times . Caspase3/7 activation was measured using Cellplayer kinetic caspase 3/7 reagent and cells were imaged in an IncuCyte system . Green object count/mm2 representing caspace-3/7 activation was measured at 2 hr intervals ( See Figure 5—figure supplement 1 for endpoint quantitation of % cells with activated caspase 3/7 ) . ( C ) IRE1 oligomers are sustained on ER-stressed cells treated with ISRIB . Confocal microscopy micrographs of HEK293Trex cells carrying an inducible GFP-tagged IRE1 allele were treated with 10 nM doxycycline for 24 hr to induce the transgene , followed by treatment with 5 µg/ml of tunicamycin in the presence or absence of 200 nM ISRIB for the indicated times . ( See Figure 5—figure supplement 2 for corresponding XBP1 mRNA splicing data ) . ( D ) ATF6 cleavage is sustained in ER-stressed cells treated with ISRIB . Immunoblot analysis of ATF6 processing in HEK293Trex cells carrying an inducible FLAG epitope-tagged ATF6 . Cells were treated with 50 nM doxycycline for 18 hr to induce the transgene followed by treatment with 100 nM thapsigargin in the presence or absence of 200 nM ISRIB for the indicated times . Total eIF2α is used as a loading control . DOI: http://dx . doi . org/10 . 7554/eLife . 00498 . 01410 . 7554/eLife . 00498 . 015Figure 5—figure supplement 1 . ISRIB synergizes with ER-stress to induce caspase 3/7 . Green object count/mm2 representing caspase-3/7 activation depicted in Figure 5A was normalized to the total number of cells at two different endpoints . In order to quantify the total number of cells , Vybrant DyeCycle Green staining solution ( 1 µM ) was added directly to the well immediately after the caspase-3/7 scan and incubated for 1 hr prior to acquiring final images at both 46 and 72 hr . Data is presented as % cells with activated caspase 3/7 at these two endpoints . Note that by 72 hr the ER-stress inducing conditions used in this experiment are so detrimental that they diminish the synergistic effects observed by addition of ISRIB . The synergy was clearly seen at the 46 hr time-point . DOI: http://dx . doi . org/10 . 7554/eLife . 00498 . 01510 . 7554/eLife . 00498 . 016Figure 5—figure supplement 2 . XBP1 splicing is sustained in ER-stressed cells upon addition of ISRIB . HEK293T cells were treated with tunicamycin ( 2 µg/ml ) for the indicated times in the presence or absence of 200 nM ISRIB . RNA was isolated from the cells and reverse transcribed followed by PCR with oligos that amplify both the unspliced and spliced versions of XBP1 mRNA or GAPDH . The DNA was electrophoresed in a 2 . 5% agarose gel . The asterix ( * ) denotes a hybrid PCR product . DOI: http://dx . doi . org/10 . 7554/eLife . 00498 . 016 The notion that ER stress remains unmitigated in ISRIB-treated cells is supported by sustained activation of all three UPR sensors . First , as shown in Figure 2C , PERK was hyper-phosphorylated . Second , cells expressing an IRE1-GFP fusion protein showed prolonged foci formation ( Figure 5C ) , indicative of IRE1 oligomerization . Third , we observed prolonged ER stress-induced proteolytic processing of ATF6 ( Figure 5D ) . Importantly , in the absence of ER stress ISRIB treatment alone did not induce any of these sensors ( Figure 3—figure supplement 1; Figure 5C and data not shown ) . eIF2α+/S51A ( Eif2s1+/S51A ) heterozygote mice display enhanced memory , while induction of the eIF2α kinase PKR in brain pyramidal cells impairs memory ( Costa-Mattioli et al . , 2007; Jiang et al . , 2010 ) . Based on these observations , we wondered whether treatment of mice with ISRIB would affect memory . ISRIB showed favorable properties in pharmacokinetic profiling experiments indicating sufficient bioavailability for in vivo studies . ISRIB displayed a half-life in plasma of 8 hr ( Figure 6A ) and readily crossed the blood-brain barrier , quickly equilibrating with the central nervous system ( Figure 6B ) . After a single intraperitoneal injection , we detected ISRIB in the brain of mice at concentrations several fold higher than its IC50 ( 24 hr after injection , the ISRIB concentration in the brain was approximately 60 nM ) . To explore ISRIB's effects on memory , we injected mice intraperitoneally with ISRIB and tested hippocampus-dependent spatial learning . To this end , we trained mice in a Morris water maze , in which animals learn to associate visual cues with the location of a submerged hidden platform . Because we were looking for memory enhancement , we used a weak training protocol . As shown in Figure 6C , ISRIB-treated mice reached the hidden platform significantly faster ( escape latency after 5 days of training = 16 . 4 ± 4 . 8 s ) compared to vehicle treated controls ( 68 . 1 ± 20 s , p<0 . 05 ) . The difference was already pronounced by days 3 and 4 . In agreement with these results , ISRIB-treated mice significantly preferred the target quadrant in a ‘probe test’ conducted at the end of the training sessions , in which the platform was removed from the pool ( p<0 . 05; Figure 6D ) and showed increased crossing of the platform location ( p<0 . 05; Figure 6E ) . 10 . 7554/eLife . 00498 . 017Figure 6 . ISRIB enhances spatial and fear-associated learning in rodents . ( A ) Plasma concentration ( ng/ml ) of ISRIB after a single intraperitoneal injection ( 5 mg/kg ) . Plasma was collected at the indicated times and the concentration was determined by mass spectrometry ( mean ± SEM , N = 3 ) . ( B ) Brain ( ng/g tissue ) and plasma concentrations ( ng/ml ) of ISRIB after a single intraperitoneal injection ( 2 . 5 mg/kg ) . Data ( mean ± SEM , N = 3 ) were obtained at the indicated times . ( C ) Escape latencies are significantly shorter in mice treated with ISRIB . Data ( mean ± SEM ) were obtained in a weak 5 days-long training session in the hidden platform version of the Morris water maze ( 1 trial per day ) . Mean escape latencies were plotted as a function of training days in mice treated with ISRIB ( closed squares , N = 8 ) or vehicle ( open circles N = 8 ) ( *p<0 . 05 ) . Mice were injected daily with ISRIB immediately after training . ( D ) After completion of training in the study shown in panel ( A ) , mice treated with ISRIB ( black column ) showed a significant preference for the target quadrant ( *p<0 . 05 ) . The probe test was performed 24 hr after the last training session . p values are derived from a two-tailed Student’s t test for unpaired samples . ( E ) After completion of training in the study shown in panel a , mice treated with ISRIB ( black column ) increased the number of times they crossed the platform location as compared to the vehicle-treated mice ( grey column ) ( *p<0 . 05 ) . p values are derived from a two-tailed Student’s t test for unpaired samples . ( F ) Systemic administration of ISRIB ( intraperitoneally after training ) enhances long-term contextual fear memory ( right bars , 24 hr ) , while it does not affect short-term memory ( left bars , 1 hr ) ( N = 10 per group , *p<0 . 05 ) . Data are presented as mean ± SEM . ( G ) Auditory fear conditioning is enhanced in rats treated with ISRIB . Freezing in response to a tone was assessed 3 hr ( short-term memory , STM , left panel ) and 24 hr ( long-term memory , LTM , right panel ) after training ( vehicle-treated N = 8 , and ISRIB-treated N = 7 ) after tone presentation ( CS ) and before tone presentation ( pre-CS ) . For these experiments vehicle or ISRIB was infused directly by cannula into the amygdala after training . ISRIB-infused rats show increase freezing at 24 hr ( *p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00498 . 017 We next tested contextual fear conditioning , which represents a different kind of hippocampus-dependent learning in which eIF2α phosphorylation has also been implicated to play a role ( Costa-Mattioli et al . , 2007 ) . In these experiments , we paired a particular environmental context ( a different cage ) with a foot shock . In this case the context acts as the ‘conditioned stimulus , CS’ and is associated with the foot shock , the ‘unconditioned stimulus , US’ . ISRIB-treated mice showed increased freezing upon presentation of the conditioned environment 24 hr after training as compared to vehicle treated mice ( p<0 . 05; Figure 6F ) . No differences were observed in short-term memory ( 1 hr ) between these two treatments . Taken together , we conclude that treatment with ISRIB enhances both hippocampus-dependent spatial learning and hippocampus-dependent contextual fear conditioning . To test learning associated with a different region of the brain , we explored the effects of ISRIB on auditory fear conditioning , which depends on the amygdala . In this type of learning a tone ( CS ) is paired with a foot shock ( US ) . In these experiments , we injected ISRIB or vehicle directly into the amygdala of rats via cannulation . ISRIB-treated rats showed a significant increase over vehicle-injected rats in the level of freezing when presented with the tone ( CS ) at 24 hr ( long-term memory , p<0 . 05; Figure 6G ) . By contrast , we observed no difference between ISRIB- and vehicle-treated rats at 3 hr ( short-term memory ) . As expected , both ISRIB- and vehicle-treated rats showed similar freezing responses prior to training ( pre-CS ) . Taken together , these data suggest that long-term memory is selectively enhanced in ISRIB-treated animals .
The structure of ISRIB provides no insights as to its target in cells . To date , we have synthesized and assayed more than 75 analogs , which demonstrate a tractable structure-activity relationship ( to be published elsewhere ) . The analyses have identified sites on the molecule where affinity tags and/or crosslinking moieties can be added , which promise to aid in target identification . Based on previous insights on how cells can become resistant to eIF2α phosphorylation , we consider two likely scenarios by which ISRIB could act:First , ISRIB could weaken the effects of the non-productive interaction of phospho-eIF2α with eIF2B , thereby increasing the available eIF2α-GEF activity in the cell , restoring the concentration of ternary complex that can engage in translation initiation . Precedence for this possibility derives from genetic studies in Saccharomyces cerevisiae , where the molecular mechanism of regulation by eIF2α phosphorylation was first discovered . As in mammalian cells , amino acid starvation in yeast leads to GCN2 activation and eIF2α phosphorylation , resulting in overall translational down-regulation and translational induction of a transcriptional activator , GCN4 , mediated by uORFs in the 5′UTR of its mRNA ( Hinnebusch , 2005 ) . eIF2B is a conserved protein complex comprised of five different subunits , two of which form the catalytic core , and the remaining three have regulatory roles . Mutations in different eIF2B subunits can elicit a phospho-eIF2α resistant phenotype ( Vazquez de Aldana and Hinnebusch , 1994; Pavitt et al . , 1997 , 1998 ) . These mutations have been proposed to act either by weakening the interaction of phospho-eIF2α with eIF2B , reducing its ability to outcompete non-phosphorylated eIF2 for binding , or by allowing binding of phospho-eIF2α to the mutant eIF2B in a manner that is conducive to nucleotide exchange . ISRIB could be altering the affinity of phospho-eIF2α for eIF2B or overcoming the nonproductive interaction that blocks GTP loading , mimicking the effect of these mutations . Second , ISRIB could increase the activity of eIF2B , so that the residual amount not engaged with phospho-eIF2α is sufficient to sustain normal levels of ternary complex . Precedence for this possibility derives from studies in macrophages , where engagement of toll-like receptor ( TLR ) 4 results in activation of the catalytic activity of eIF2B ( Woo et al . , 2012 ) . This activation results from engagement of the TLR-signaling pathway that induces a phosphatase removing a constitutively present inhibitory phosphate from the eIF2B ε-subunit ( S539 ) . Pathogens utilize this mechanism to circumvent translational attenuation and CHOP production under prolonged stress-inducing conditions ( Woo et al . , 2009 ) . ISRIB did not reduce phosphorylation of S539 in the eIF2B ε-subunit ( Figure 3—figure supplement 1 ) , indicating that it does not utilize this particular regulatory phosphorylation to increase GEF activity . However , ISRIB may modulate other post-translational modifications that impinge on the activity of eIF2B . As a signaling network with interconnected signaling branches , the UPR exhibits both cytoprotective and pro-apoptotic functions . When faced with ER stress , PERK-mediated translational attenuation contributes to adaptation by reducing the load of newly synthesized proteins that are translocated into the ER ( Harding et al . , 2000 ) . In addition , induction of the transcription regulator ATF4 upregulates many genes that increase the protein folding capacity in the ER . Both of these activities serve to reestablish homeostasis , balancing the protein folding load and protein folding capacity in the ER lumen . This reasoning is supported by the increased sensitivity to ER stress exhibited by MEFs that lack PERK or ATF4 , as well as MEFs that carry a non-phosphorylatable knock-in allele of eIF2α ( S51A ) ( Harding et al . , 2000; Harding et al . , 2003; Lu et al . , 2004 ) . In agreement , we show that ISRIB decreases the viability of cells that are subjected to ER-stress . In these cells , ISRIB sustains IRE1 and ATF6 activation , indicating that ER stress remains unmitigated in the absence of PERK signaling . As some cancer cells sustain an activated UPR to aid in their survival , ISRIB could provide a new therapeutic approach to cancer chemotherapy . In agreement , a PERK-specific inhibitor demonstrates antitumor activity in a human pancreatic tumor xenograft model ( Axten et al . , 2012 ) . The deleterious synergistic effect between ER-stress and ISRIB may be generally advantageous to kill cancer cells , especially those derived from secretory lineages that have increased secretory load and increased basal levels of ER stress ( including myelomas , and pancreatic and breast cancers ) . Importantly , by acting downstream of eIF2α phosphorylation , ISRIB blocks multiple stress effectors ( i . e . , all eIF2α kinases ) . During tumor growth , hypoxic conditions and a lack of nutrients can activate both PERK and GCN2 , and PERK−/− or GCN2−/− MEFs give rise to significantly smaller tumors in mouse xenograft models than their wild-type counterparts ( Bi et al . , 2005; Ye et al . , 2010 ) . Hence both kinases have pro-survival roles in tumor development . By blocking signaling by both kinases , ISRIB displays unique properties that may be beneficial in reducing cellular fitness of tumor cells . The importance of eIF2/eIF2B function in the human brain is underscored by familial diseases caused by mutations in these factors . One example is Childhood Ataxia with CNS Hypomyelination ( CACH ) , also known as Vanishing White Matter disease ( VWM ) , which has been mapped to mutations in different subunits of eIF2B ( Li and Wang , 2004 ) . A second example links a familial intellectual disability syndrome to a mutation in the γ-subunit of eIF2 complex ( Borck et al . , 2012 ) . Several lines of genetic evidence in mice suggest that phosphorylation-dependent regulation of eIF2α phosphorylation is a critical hub for the control of synaptic plasticity ( as assessed by late long-term potentiation [L-LTP] in brain slices ) and memory consolidation ( as assessed in behavioral tasks in animals ) . In particular , the threshold for induction of L-LTP is reduced and memory consolidation is enhanced in mice lacking GCN2 or PKR and in mice heterozygous for non-phosphorylatable eIF2α ( S51A ) , which have reduced levels of eIF2α phosphorylation ( Costa-Mattioli et al . , 2005; Costa-Mattioli et al . , 2007; Zhu et al . , 2011 ) . As we show here , ISRIB pharmacologically phenocopies these genetic manipulations in behavioral tasks by rendering cells insensitive to eIF2α phosphorylation . In agreement , treatment of mice with a PKR inhibitor was reported to enhance memory consolidation , and , conversely , treatment with salubrinal , an inhibitor that prolongs eIF2α phosphorylation , to block L-LTP and memory consolidation ( Costa-Mattioli et al . , 2007; Zhu et al . , 2011 ) . eIF2α phosphorylation results in a dual effect on gene expression: a global translational diminution and translational upregulation of select mRNA , including ATF4 mRNA . Both may be important to explain the observed effects on L-LTP and memory . It has long been appreciated that new protein synthesis is required for memory consolidation and that ATF4 represses CREB-mediated transcription of ‘memory genes’ ( Klann and Dever , 2004; Sutton and Schuman , 2006 ) . Indeed , this latter function of ATF4 in memory consolidation is evolutionarily conserved from Aplysia to rodents ( Yin et al . , 1994; Bartsch et al . , 1995; Chen et al . , 2003 ) . Because a small physiological increase in the level of eIF2α phosphorylation that does not significantly alter overall translation is sufficient to induce ATF4 , production of this transcription factor can be finely tuned in neuronal cells by perhaps selective activation of different eIF2α kinases . The observed effects of ISRIB may therefore result from overcoming effects caused by a relatively small level of regulatory phosphorylation that is distinct from the high level resulting from ER stress-inducing agents . In light of this reasoning , a therapeutic window may exist in which ISRIB’s effects as memory enhancer can be exploited without encountering long-term toxic consequences . ISRIB increases memory consolidation , allowing pharmacological enhancement of the brain’s ability to learn . Evolution therefore did not arrive at a maximally optimized process , imposing a brake ( via eIF2α phosphorylation ) on memory consolidation . This mechanism may underscore the importance of filtering memories before committing them to long-term storage . Indeed , eIF2α phosphorylation also plays a role in dynamic restructuring of memory , as indicated by studies showing that ablation of PERK in the brain impairs behavioral flexibility ( Trinh et al . , 2012 ) . Our findings raise the possibility that ISRIB or compounds with related activities could serve as invaluable tools in deciphering these higher order brain functions and perhaps be beneficial as a therapeutic agent effecting memory improvement in diseases associated with memory impairment .
HEK293T , TREx293 , U2OS , Hela , and mouse embryonic fibroblasts ( MEFs ) were maintained at 37°C , 5% CO2 in DMEM media supplemented with 10% FBS , L-glutamine and antibiotics ( penicillin and streptomycin ) . Tunicamycin was obtained from Calbiochem EMB Bioscience , Billerica , MA . Thapsigargin was obtained from Sigma-Aldrich , St Louis , MO . Torin-1 was obtained from Tocris , MN . The HRI activator was purchased from Maybridge ( KM09748 ) , Cornwall , UK . ATF4 reporters were constructed by fusing the human full-length ATF4 5′-UTR ( NCBI Accession BC022088 . 2 ) in front of the firefly luciferase ( FLuc ) or a destabilized eGFP ( dEGFP ) coding sequences lacking the initiator methionine . The ATF4-FLuc reporter was generated by cloning a PCR-product containing the ATF4 full-length 5′-UTR ( from +1 position a the transcription start site down to one nucleotide after the terminator codon of the second uORF ) flanked by KpnI/XhoI and BglII sites at the 5′ and 3′ ends , respectively , into the KpnI-BglII sites of pCAX-F-XBP1-Luc ( kind gift of Takao Iwawaki , RIKEN , Hirosawa , Japan ) . The resulting construct , pCAX-ATF4-FLuc , was then digested with BamHI , blunted with T4 DNA polymerase , and then digested with XhoI . The resulting fragment was then subcloned into the retroviral expression vector pLPCX ( Clontech , Mountain View , CA ) after digesting it with HindIII , blunting with T4 DNA polymerase and then digesting with XhoI to generate pLPCX-ATF4-FLuc ( DAA-312 ) . DAA-312 was used to produce recombinant retroviruses using standard methods and the resulting viral supernatant was used to transduce HEK293T cells , which were then subsequently selected with puromycin to generate a stable cell line employed in the primary screen . The ATF4-dEGFP reporter was generated using a PCR fusion-based approach . A PCR product containing the ATF4 full-length 5′ leader sequence ( from +1 position a the transcription start site ) fused to the eGFP coding sequence 1 nucleotide downstream of the terminator codon of the second uORF , and flanked by BamHI and EcoRI , was cloned into the cognate sites of pEGFP-N3 ( Clontech , Mountain View , CA ) to generate pCMV-ATF4-eGFP . To destabilize the eGFP fusion protein and increase the dynamic range of the reporter , residues 422–461 of mouse ornithine decarboxylase ( mODC1 ) , corresponding to its PEST sequence ( Li et al . , 1998 ) , were fused to the C-terminus of the ATF-eGFP fusion protein . To such end , the corresponding mODC1 coding sequence was amplified by PCR and cloned into the BstXI and EcoRI sites of pCMV-ATF4-eGFP . The resulting construct was designated pCMV-ATF4-d2EGFP . To further destabilize the ATF4-d1EGFP fusion protein , alanine substitutions E428A , E430A , E431A ( Li et al . , 1998 ) were introduced in the ODC1 PEST sequence to generate pCMV-ATF4-d1EGFP . The ATF4-d1EGFP coding sequence was then excised from the expression vector using BamHI and EcoRI and subcloned into the BglII-EcoRI sites of the retroviral expression vector pLPCX ( Clontech , Mountain View , CA ) to generate pLPCX-ATF4-d2EGFP . Lastly , a fusion PCR product containing the encephalomyocarditis virus internal ribosomal entry site ( EMCV-IRES ) upstream of the monomeric cherry ( mCherry ) coding sequence and flanked by EcoRI and NotI recognition sites was subcloned into the cognate sites of pLPCX-ATF4-d1EGFP , thereby generating pLPCX-ATF4-d1EGFP-IRES-mCherry ( DAA-361 ) . DAA-361 was used to produce recombinant retroviruses using standard methods and the resulting viral supernatant was used to transduce U2OS cells , which were then subsequently selected with puromycin to generate a stable cell line employed in the secondary screen . The coding sequences of wild-type mouse eIF2α , phosphomimetic ( S51D ) mutant was amplified by PCR from a mammalian expression vector ( kind gift of David Ron ) . BamHI and EcoRI recognition sites were engineered into the primers . In addition a Kozak consensus sequence and a N-terminal FLAG epitope tag were engineered in the forward primer . The resulting PCR products were subcloned into the cognate sites of the tetracycline-inducible retroviral expression vector pRetroX-Tight-Pur-GOI ( Clontech , Mountain View , CA ) . 293T target cells stably expressing the reverse tetracycline transactivator ( rtTA ) were generated by standard retroviral transduction using VSV-G pseudotyped retroviruses encoding rtTA ( pRetroX-Tet-On Advanced; Clontech , Mountain View , CA ) and selected with G418 . These cells were subsequently transduced with a VSV-G pseudotyped retrovirus , encoding the eIF2α ( S51D ) ( DAA-A681 ) mutant allele , resulting in a puromycin-selected , tetracycline inducible , stable cell line . 6xHis-3xFLAG-hsATF6-alpha was generated by PCR from p3xFLAGCMV7 . 1-ATF6 ( Shen et al . , 2002 ) and cloned into pcDNA5/FRT/TO . pcDNA5/FRT/TO-6xHis-3xFLAG-hsATF6-alpha was co-transfected with pOG44 into Flp-In TRex cells ( Cohen and Panning , 2007 ) according to manufacturers instructions ( Invitrogen , Carlsbad , CA ) . After selection with 100 µg/ml Hygromycin B ( Gold Biotechnology , St Louis , MO ) single colonies were isolated , expanded and tested for expression of tagged ATF6 . HEK293T cells carrying the ATF4 luciferase reporter were plated on poly-lysine coated 384-well plates ( Greiner Bio-one , Monroe , NC ) at 30 , 000 cells per well . Cells were treated the next day with 100 nM thapsigargin and 10 µM of the library compounds ( diversity library of 106 , 281 compounds ) for 6 hr . Luminescence was measured using One Glo ( Promega , Madison , WI ) as specified by the manufacturer . The primary screen had a Z′ = 0 . 5 and its hit rate was 0 . 6% ( compounds were considered a hit if their luciferase readouts were beyond three standard deviations of the mean luminescence intensity of thapsigargin treated cells , which corresponded to 54% inhibition ) . Of these , only 187 compounds did not hit a luciferase-based XBP1 splicing reporter used as proxy to measure activation of the IRE1 branch of the UPR . Thus , these were considered unique to the PERK branch and were cherry-picked for further analysis . U2OS cells carrying the ATF4-dGFP-IRES-Cherry reporter were plated in 96 well plates and treated with 100 nM Thapsigargin and 10 µM of the cherry-picked compounds for 8 hr . Cells were stained with Hoechst 33 , 258 and were visualized using an automated microscope ( InCell Analyzer 2000; GE Healthcare , Waukesha , WI ) . Data acquisition and image analyses were performed with the INCell Developer Toolbox Software , version 1 . 9 ( GE Healthcare , Waukesha , WI ) . Compounds that blocked induction of the ATF4-dGFP reporter , did not block the accumulation of mCherry downstream of the IRES , and were deemed non-toxic as determined by cell number measured by counting nuclei , were repurchased for further analyses . Cells were lysed in SDS-PAGE loading buffer ( 1% SDS , 62 . 5 mM Tris-HCl pH 6 . 8 , 10% glycerol ) . Lysates were sonicated and equal amounts were loaded on SDS-PAGE gels ( BioRad , Hercules , CA ) . Proteins were transferred onto nitrocellulose and probed with primary antibodies diluted in Tris-buffered saline supplemented with 0 . 1% Tween 20 and 5% bovine serum albumin . The following antibodies were used: CREB-2 ( C-20 ) ( 1:800 ) , eIF2Bε ( B-7 ) ( 1/500 ) ( Santa Cruz Biotechnologies , Dallas , TX ) ; PERK ( D11A8 ) ( 1:1000 ) , PERK ( C33E10 ) ( 1:1000 ) , eIF2α ( 9722 ) ( 1:1000 ) , phospho-eIF2α ( Ser51 ) ( D9G8 ) XP ( 3398 ) ( 1:1000 ) ( Cell Signaling Technology , Danvers , MA ) ; XBP1s ( C-terminus ) ( 1:500 ) ( BioLegend , San Diego , CA ) ; phospho-S539 eIF2Bε ( 1/1000 ) ( Abcam ) ; M2 Flag ( 1:1000 ) ( Sigma , St Louis , MO ) . An HRP-conjugated secondary antibody ( Amersham , Piscataway , NJ ) was employed to detect immune-reactive bands using enhanced chemiluminescence ( SuperSignal; Thermo Scientific , Waltham , MA ) . U2OS cells were seeded on Slide Flasks ( Thermo Scientific , Waltham , MA ) 18 hr prior to processing for immunofluorescence . Cells ( 60% confluent ) were fixed with 4% paraformaldehyde in PBS for 15 min . The cells were then rinsed three times with PBS and permeabilized with 0 . 3% Triton X-100 . The fixed cells were rinsed three times with PBS and blocked for 1 hr at room temperature in PBS supplemented with 0 . 1% Triton X-100 and 5% normal goat serum . The cells were then incubated overnight at 4°C with an anti-CHOP mouse monoclonal antibody ( Cell Signaling Technology L63F7 , Danvers , MA ) at a 1:1000 dilution in blocking buffer . The next morning the slides were washed three times ( 5 min each time ) with PBST ( PBS-0 . 1% Triton X-100 ) . The slides were then incubated for 1 hr at room temperature in a 1:500 dilution ( in blocking buffer ) of secondary anti-mouse antibody labeled with Alexa Dye 488 ( Molecular Probes , Invitrogen , Carlsbad , CA ) . The slides were then washed three additional times with PBST . The cells were then counterstained with rhodamine-phalloidin ( 1:1000 in PBS ) for 10 min at room temperature to reveal the actin cytoskeleton . Lastly , the slides were mounted using Vectashield ( Vector , Burlingame , CA ) mounting medium and imaged using a Zeiss Axiovert 200M epifluorescence microscope . Mouse embryonic fibroblasts ( MEFs ) or TREx-293 cells expressing eIF2α ( S51D ) were seeded on 150-mm plates and allowed to grow to 80% confluence . Cells were then induced with 25 nM doxycycline for 14 hr and subsequently treated with the appropriate compounds for the indicated times . 100 µg/ml of cycloheximide was added to the cells for 1 min before lysis . Cells were washed twice with PBS supplemented with 100 µg/ml cycloheximide and subsequently lysed in 20 mM Tris pH 7 . 4 , 200 mM NaCl , 15 mM MgCl , 1 mM DTT , 8% Glycerol , 100 µg/ml cycloheximide , 1% Triton X-100 and EDTA-free protease inhibitor tablets ( Roche , South San Francisco , CA ) . Cells were scraped , collected , triturated with a 257/8 gauge needle , and the homogenate was centrifuged for 10 min at 10 , 000×g . The supernatant was loaded on a 10–50% sucrose gradient and sedimented in a SW40 rotor at 150 , 000×g for 2 . 4 hr . The gradients were fractionated using a piston gradient fractionator ( BioComp Instruments , Fredericton , NB , Canada ) and UV absorbance at 254 nm was monitored using a UV-Monitor ( BioRad , Hercules , CA ) . U2OS cells were plated on 96-well plates and left to recover overnight . Cells were treated with either with 2 µg/ml tunicamycin or 100 nM thapsigargin in the presence or absence of 100 nM ISRIB or with ISRIB alone for the indicated and the level of eIF2α phosphorylation was determined using the AlphaScreen SureFire eIF2α ( p-Ser51 ) Assay kit ( Perkin Elmer , Waltham , MA ) following the manufacturer’s recommendations . Plates were read in an Envision Xcite Multilabel Reader using the standard Alpha Screen settings . HEK293T cells were seeded on 12-well plates , allowed to recover overnight and treated for the indicated times with the indicated compounds . The cells were subsequently switched to media lacking methionine and cysteine supplemented with the indicated compounds and 50 µCi of 35S-methionine ( Perkin Elmer , Waltham , MA ) for 20 min . Cells were lysed by addition of SDS-PAGE loading buffer . Lysates were sonicated and equal amounts were loaded on SDS-PAGE gels ( BioRad , Hercules , CA ) . The gel was dried and radioactive methionine incorporation was detected by exposure to a phosphor-screen and visualized with a Typhoon 9400 Variable Mode Imager ( GE Healthcare Waukesha , WI ) . T-REx293 cells carrying GFP-IRE1 were imaged as described in Li et al . , PNAS ( Li et al . , 2010 ) . Hela cells were plated in 96-well Corning plates at 0 . 4 × 104 cells per well , 24 hr prior to imaging . On the day of experiment , DMEM media was replaced with F12 media with appropriate concentration of inhibitors and ER stress inducers and caspase 3/7 reagent at 1:1000 dilution ( Essen Bioscience No . 4440 , Ann Arbor , MI ) . Cells were imaged in the IncuCyte FLR live cell imaging system at 2 hr intervals for 70 hr . In order to quantify the total number of cells , Vybrant DyeCycle Green staining solution ( 1 µM ) was added directly to the well immediately after the final Caspase-3/7 scan and incubated for 1 hr prior to acquiring final images . Data was analyzed using IncuCyte analysis software . U2OS cells were plated on 96-well plates and allowed to recover overnight . Cells were treated for the indicated times with the indicated compounds , lysed and cDNA was synthesized using the PowerSYBR Green Cells-to-CT kit ( Ambion , Invitrogen , Carlsbad , CA ) following the manufacturer’s recommendations . The reactions were ran in an Opticon 2 thermal cycler ( BioRad , Hercules , CA ) and analyzed with the Opticon Monitor v3 software ( BioRad , Hercules , CA ) . The following oligonucleotides were used for the amplification reaction: Human GADD34: 5′-GTAGCCTGATGGGGTGCTT -3′ and 5′- TGAGGCAGCCGGAGATAC -3′; Human CHOP: 5′- AGCCAAAATCAGAGCTGGAA -3′ and 5′-TGGATCAGTCTGGAAAAGCA -3′; Human GAPDH: 5′-TGGAAGATGGTGATGGGATT -3′ and 5′- AGCCACATCGCTCAGACAC -3′ . cDNA obtained with the PowerSYBR Green Cells-to-CT kit ( Ambion , Invitrogen , Carlsbad , CA ) as described above was used for the Taqman Assay . TaqMan assays were set up using iQ Supermix ( BioRad , Hercules , CA ) , 250 nM of each outer primer , 200 nM FAM-XBP1U probe , or 100 nM HEX-XBP1S probe . The reactions were then run on a real-time DNA Engine Opticon 2 PCR thermal cycler ( BioRad ) and analyzed with the Opticon Monitor v3 software ( BioRad ) . The outer primers employed for the human XBP1unspliced/spliced ( u/s ) TaqMan assay were: 5′-GAAGCCAAGGGGAATGAAGT-3′ , and 5′-GAGATGTTCTGGAGGGGTGA-3′ . TaqMan probes specific for human XBP1s or XBP1u were: 5′-FAM-CAGCACTCAGACTACGTGCACCTCTG-BHQ1-3′ , and 5′-HEX-TCTGCTGAGTCCGCAGCAGGTGCA-BHQ1-3′ . Total RNA from treated or untreated HEK293T cells was extracted using TRIzol ( Invitrogen , Carlsbad , CA ) following the manufacturer’s recommendations . 500 ng of total RNA were reverse transcribed using the SuperScriptVilo cDNA Synthesis kit ( Invitrogen ) . The cDNA was diluted 1 in 10 in TE ( pH = 8 ) and 1% of the total reaction was used as a template for the PCR amplification reactions . The XBP1 primers flank the 26-nucleotide intron and produce both spliced ( 222 bp ) and unspliced ( 248 bp ) amplicons . The PCR products were resolved in 2 . 5% agarose . The following oligonucleotides were used for the amplification reaction: for human XBP1 , 5′-ACTGGGTCCAAGTTGTCCAG -3′ and 5′- GGAGTTAAGACAGCGCTTGG -3′; for human GAPDH 5′- TGGAAGATGGTGATGGGATT -3′ and 5′-AGCCACATCGCTCAGACAC -3′ . Intra-peritoneal ( ip ) route of administration was performed on 6–7 wk old female CD-1 mice ( Harlan Laboratories , Indianapolis , IN ) . Animals received a single , 5 mg/kg dose in groups of three mice/compound/route of administration . ISRIB was dissolved in DMSO then diluted 1:1 in Super-Refined PEG 400 ( Croda , Edison , NJ ) . Blood ( 80 μl ) was collected from the saphenous vein at intervals post-dosing ( 20 min , 1 hr , 3 hr , 8 hr , 24 hr ) in EDTA containing collection tubes ( Sarstadt CB300 ) and plasma was prepared for analysis . Compounds were detected by time-of-flight mass spectroscopy . Intra-peritoneal ( ip ) route of administration was performed at a single dose of 2 . 5 mg/kg in groups of three for each time-point ( 2 , 6 , 24 and 36 hr ) . Brain tissue samples were individually homogenized with a Tissue Tearor ( Model 985-370 type2 , BioSpec Products Inc , Bartlesville , OK ) . Approximately 300 mg of tissue was placed in 5-ml polypropylene tube , and four volumes of water were then added to mix . The speed scale of Tissue Tearor was set at 3 for 2 min . After homogenization , the supernatant was analyzed by LC-MS/MS to determine their brain concentration . Plasma samples were collected prior to extraction of brain samples . Eight to ten-week-old male C57BL/6J mice were used for behavioral experiments . Food and water were provided ad libitum , and mice were kept on a 12:12 hr light/dark cycle ( lights on at 08:00 hr ) . All procedures complied with Canadian Council on Animal Care guidelines . Mice were trained in a water pool of 100 cm diameter with a hidden platform of 10 cm diameter . Mice were handled daily for 3 days before the experiment , and the training protocol consisted of one swimming trial per day . Each mouse swam until it found the hidden platform or 120 s , when it was gently guided to the platform and stayed there for 10 s before being returned to the cage . Immediately after the swimming trial the mice were injected intraperitoneally with ISRIB ( 0 . 25 mg/kg in saline , 1% DMSO ) . For the probe test , the platform was removed and each mouse was allowed to swim for 60 s , while its swimming trajectory was monitored with a video tracking system ( HVS Image , Buckingham ) . Mice were trained with a protocol that consisted of a 2-min period of context exploration , followed by a single foot shock of 0 . 35 mA for 1 s . Mice received a single injection of ISRIB ( 2 . 5 mg/kg in 50% DMSO , 50% PEG 400 , IP ) immediately after training and were returned to their home cage . One and 24 hr after training , the mice were tested for contextual fear memory by placing the animals in the conditioning context for a 4-min period . The incidence of freezing was scored in 5-s intervals as either ‘freezing’ or ‘not freezing’ . Percent of freezing indicates the number of intervals in which freezing was observed divided by total number of 5-s intervals . Statistical analyses were done by Student’s t tests and one-way ANOVA followed by between-group comparisons using Tukey’s posthoc test . Male Sprague Dawley rats ( 275–350 g ) were used for cannulation as described in Migues et al . , 2010 ( Migues et al . , 2010 ) . ISRIB ( 0 . 05 mg/ml , 0 . 5 μl ) was infused bilaterally into the amygdala immediately after auditory fear conditioning training . The infusion was performed with a microinjector ( 28 gauge ) connected to a Hamilton syringe with plastic tubing at a rate of 0 . 25 μl/min . To allow for the solution containing ISRIB to diffuse from the tip of the cannula into the tissue , the microinjector stayed in the cannula for one additional minute . Training protocol for auditory fear conditioning consisted of a 2-min period of context A exploration , followed by one pairing of a tone ( 5000 Hz , 75 dB , 30 s ) with a co-terminating foot shock ( 0 . 75 mA , 1 s ) . Rats were returned to their home cage 1 min after the shock . Test for auditory fear memory consisted of a 2 min acclimatizing period to the context B ( pre-CS ) , followed by tone presentation ( CS ) ( 2800 Hz , 85 dB , 30 s ) . Freezing time was measured and percent of freezing was calculated . At the end of the experiment , cannula placement was checked by examining 50 μm brain sections stained with formal-thionin under a light microscope . trans-ISRIB: 2- ( 4-Chlorophenoxy ) -N-[ ( 1r , 4r ) -4-[2- ( 4-chlorophenoxy ) acetamido] cyclohexyl] acetamide To a mixture of ( 1r , 4r ) -cyclohexane-1 , 4-diamine ( 20 mg , 0 . 18 mmol ) in tetrahydrofuran:water ( 1:1 , 1 ml ) were sequentially added potassium carbonate ( 73 mg , 0 . 53 mmol ) and 4-chlorophenoxyacetyl chloride ( 56 µl , 0 . 36 mmol ) . Upon addition of the acid chloride , a white solid immediately formed . The reaction mixture was vigorously stirred at ambient temperature for 30 min . Water ( 2 . 5 ml ) was added . The mixture was vigorously vortexed then centrifuged , and the water was decanted . This washing protocol was repeated with potassium bisulfate ( 1% aq , 2 . 5 ml ) , water ( 2 . 5 ml ) , and diethyl ether ( 2 × 2 . 5 ml ) . The resulting wet white solid was dried by partially dissolving in dichloromethane/methanol ( 10/1 , 10 ml ) and gravity filtering through an Autochem 4 . 5-ml reaction tube . The residual undissolved product was extracted from the wet filter cake by adding dichloromethane ( 4 × 4 . 5 ml ) and gravity filtering . The combined filtrate was concentrated using rotary evaporation to afford 51 mg ( 65% ) of the title compound as a white solid . 1H NMR ( 400 MHz , DMSO-d6 ) δ 7 . 91 ( d , J = 8 . 1 Hz , 2H ) , 7 . 31 ( d , J = 9 . 0 Hz , 4H ) , 6 . 94 ( d , J = 9 . 0 Hz , 4H ) , 4 . 42 ( s , 4H ) , 3 . 55 ( br . s . , 2H ) , 1 . 73 ( br . d , J = 5 . 9 Hz , 4H ) , 1 . 30 ( quin , J = 10 . 5 Hz , 4H ) ; LC-MS: m/z = 451 [M+H , 35Cl x 2]+ , 453 [M+H , 35Cl , 37Cl]+ . cis-ISRIB: 2- ( 4-chlorophenoxy ) -N-[ ( 1s , 4s ) -4-[2- ( 4-chlorophenoxy ) acetamido] cyclohexyl]acetamide To a mixture of ( 1s , 4s ) -cyclohexane-1 , 4-diamine ( 21 µl , 20 mg , 0 . 18 mmol ) in tetrahydrofuran:water ( 1:1 , 1 ml ) were sequentially added potassium carbonate ( 73 mg , 0 . 53 mmol ) and 4-chlorophenoxyacetyl chloride ( 56 µl , 0 . 36 mmol ) . The reaction mixture was vigorously stirred at ambient temperature for 1 . 5 hr then partitioned between 30 ml of 1:1 dichloromethane:KHSO4 ( 10% aq . ) . After separating the organic layer , it was sequentially washed with water ( 1 × 10 ml ) and brine ( 1 × 10 ml ) then dried by gravity filtration using an Autochem 4 . 5-ml reaction tube . The filtrate was concentrated and loaded onto a Silicycle 4g SiO2 column using a minimal amount of dichloromethane ( ∼2 ml ) . The product was eluted with acetone in dichloromethane ( 0–50% ) . Product-containing fractions were combined and concentrated to afford 56 mg ( 71% ) of the title compound as a white solid . 1H NMR ( 400 MHz , DMSO-d6 ) δ 7 . 76 ( d , J = 7 . 0 Hz , 2H ) , 7 . 32 ( d , J = 9 . 0 Hz , 4H ) , 6 . 94 ( d , J = 9 . 0 Hz , 4H ) , 4 . 47 ( s , 4H ) , 3 . 70 ( br . s . , 2H ) , 1 . 44 − 1 . 67 ( m , 8H ) ; LC-MS: m/z = 451 [M + H , 35Cl x 2]+ , 453 [M + H , 35Cl , 37Cl]+ . | The synthesis of proteins is an essential step in many biological processes , including memory , and drugs that inhibit protein synthesis are known to impair memory in rodents . It is thought that the brain needs these proteins to convert short-term memories into long-term memories through a process known as consolidation . A protein called EIF2α has a key role in the regulation of protein synthesis , and has also been implicated in memory . EIF2α can be activated as a result of being phosphorylated by any of four protein kinases: these are in turn activated by processes that subject cells to stress , such as viral infection , UV light or—in the case of a kinase known as PERK—the accumulation of unfolded proteins in a cellular organelle called the endoplasmic reticulum . Activation of EIF2α downregulates most protein synthesis inside the cell , but upregulates the production of a small number of key regulatory molecules: these changes help cells to cope with whatever stressful event they have just experienced . To obtain further insight into the cellular stress response , Sidrauski et al . screened a large library of compounds in search of one that inhibits PERK . They identified a molecule—known as ISRIB—which acts downstream of all four protein kinases by reversing the effects of EIF2α phosphorylation . ISRIB is the first molecule shown to have this effect , and thus represents an important tool for investigating the stress response inside cells . When Sidrauski et al . injected ISRIB into mice , the animals showed improved memory: for example , they learnt to locate a hidden platform in a water maze more rapidly than controls . This suggests that ISRIB could be used to explore the mechanisms that underlie memory consolidation , and possibly even as a memory enhancer . Moreover , given that many tumor cells exploit the cellular stress response to aid their own growth , ISRIB may have potential as a novel chemotherapeutic agent . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2013 | Pharmacological brake-release of mRNA translation enhances cognitive memory |
Acquired therapeutic resistance by tumors is a substantial impediment to reducing the morbidity and mortality that are attributable to human malignancies . The mechanisms responsible for the dramatic shift between chemosensitivity and chemoresistance in colorectal carcinoma have not been defined . Here , we report that LRP16 selectively interacts and activates double-stranded RNA-dependent kinase ( PKR ) , and also acts as scaffolds to assist the formation of a ternary complex of PKR and IKKβ , prolonging the polymers of ADP-ribose ( PAR ) -dependent nuclear factor kappa B ( NF-κB ) transactivation caused by DNA-damaging agents and confers acquired chemoresistance . We also identified a small molecule , MRS2578 , which strikingly abrogated the binding of LRP16 to PKR and IKKβ , converting LRP16 into a death molecule and forestalling colon tumorigenesis . Inclusion of MRS2578 with etoposide , versus each drug alone , exhibited synergistic antitumor cytotoxicity in xenografts . Our combinatorial approach introduces a strategy to enhance the efficacy of genotoxicity therapies for the treatment of tumors .
Death in patients afflicted with cancer is mainly due to uncontrollable recurrence of the primary tumor following failure of current therapies . Colorectal cancer ( CRC ) is one such cancer type wherein aggressive treatment strategies including surgery , ionizing radiation ( IR ) , and chemotherapy provide only palliation ( Benson et al . , 2004; de Gramont et al . , 2000 ) . The most important factor limiting the success of systemic anticancer therapy in achieving cure or prolonged overall survival has been drug resistance , either because the initial tumor fails to respond to therapy or because it acquires resistance during relapse ( Holohan et al . , 2013; Kuczynski et al . , 2013 ) . The anticancer activity of most chemotherapy drugs relies on the induction of DNA damage in rapidly cycling tumor cells with inadequate DNA repair . However , damage induced by drugs is not invariably lethal but instead actively triggers damage responses , and it is these responses that determine the eventual fate of the cell ( Johnstone et al . , 2002 ) . Given the adaptability of tumor cells , it seems likely that drug resistance will continue to be an important clinical problem , even in the age of targeted therapeutics and tailored treatment regimes ( Lowe et al . , 2004 ) . Thus , novel therapeutic strategies that circumvent the mechanisms of resistance are urgently required to improve the survivorship of cancer patients . Our understanding of the mechanisms that protect tumor cells from the cytotoxicity induced by drugs must be improved . The nuclear factor kappa B ( NF-κB ) signaling pathway , which is activated by many stimuli , is a crucial transcription factor in a variety of pathophysiological conditions ( Harhaj and Dixit , 2012; Hayden and Ghosh , 2008; Smale , 2011; Wan and Lenardo , 2010 ) . NF-κB responds to genotoxic threats via the activation of the inhibitor of NF-κB kinase ( IKK ) and NF-κB liberation from IκB proteins , similar to the canonical pathway activated by external stimuli ( Janssens et al . , 2005; Perkins , 2007; Wu et al . , 2006 ) . The NF-κB signaling pathway has emerged as an important mediator for cellular responses to DNA damage , in particular NF-κB-conferred anti-apoptotic transcription facilitates the escape of cells from the lethal effects of DNA damage ( Janssens et al . , 2005; Perkins , 2007; Wu et al . , 2006 ) , and initiates cell cycle checkpoint control to promote cellular recovery from damage ( McCool and Miyamoto , 2012; Miyamoto , 2011 ) , thereby contributing to acquired resistance to DNA-damaging cytotoxic therapies . Besides NEMO ( also named IKKγ ) and ataxia telangiectasia mutated ( ATM ) , two known crucial regulators of the genotoxic stress-activated NF-κB signaling pathway ( Miyamoto , 2011 ) , poly ( ADP-ribose ) polymerase 1 ( PARP1 ) was recently revealed to be indispensable for the signaling cascade that links nuclear DNA damage recognition to cytoplasmic IKK activation ( Stilmann et al . , 2009 ) . Sequential posttranslational modifications ( PTMs ) , including phosphorylation , ubiquitination and SUMOylation , of these signaling regulators are critical for NF-κB activation following DNA damage ( Huang et al . , 2003; Mabb et al . , 2006; Wu et al . , 2006 ) . In this regard , PARP1 is an abundant nuclear protein that senses and contributes to the repair of DNA single-strand breaks ( SSBs ) and of DNA double-strand breaks ( DSBs ) , working by catalyzing poly ( ADP-ribosyl ) ation ( PARylation ) of itself , histones , and other target proteins ( Gibson and Kraus , 2012 ) . In particular , PARP1-catalyzed PARylation has emerged as a vital means for the rapid assembly of signaling complexes that are critical for DNA damage-initiated NF-κB activation ( Mabb et al . , 2006; Stilmann et al . , 2009 ) . However , the cooperative function of NF-κB with other key stress elements in cellular resistance to DNA-damaging therapies remains to be clarified . How DSBs trigger the activities of such a large number of factors with such precise spatiotemporal coordination also remains unclear . Leukemia-Related Protein 16 ( LRP16 ) , a member of the macro domain family , was identified as a PAR-binding protein in genotoxic threat-treated cells and a putative substrate of PARP1 , which suggests that LRP16 could be an important molecule in the cellular response to DNA damage ( Han et al . , 2011; Timinszky et al . , 2009 ) . We have previously established a critical role for LRP16 in DSB-induced activation of the NF-κB signal transduction cascade , which counteracts apoptosis and allows cells to escape the lethal effects of DNA damage ( Wu et al . , 2015 ) . Mechanistically , LRP16 facilitates the lesion-specific recruitment of PARP-1 and NEMO through its constitutive interactions with these two proteins , and then ultimately facilitates the concomitant recruitment of ATM and protein inhibitor of activated STAT Y ( PIASy ) to NEMO to ensure the activation of NF-κB after the induction of DSBs ( Wu et al . , 2015 ) . Although these studies have considerably advanced our understanding of the cellular response to DNA damage , the genotoxic stress-initiated nucleoplasmic NF-κB signaling pathway remains poorly understood , in particular , the early signaling networks linking DNA lesion recognition in the nucleus to subsequent activation of IKK and liberation of NF-κB in the cytoplasm . The double-stranded-RNA ( dsRNA ) -activated protein kinase PKR , a ubiquitously expressed serine/threonine kinase , has been implicated in the regulation or modulation of cell growth through multiple signaling pathways and has also been described as a signal integrator in the translational and transcriptional control pathways ( Bennett et al . , 2006; Dar et al . , 2005; Donzé et al . , 2004; Liu et al . , 2013 ) . In addition to its functional regulatory function , PKR has a role in signal transduction and transcriptional control through the IκB/NF-κB pathways ( Gil et al . , 2000 , Gil et al . , 2004 ) . PKR has also been implicated in different stress-induced signaling pathways including dsRNA signaling to NF-κB activation , although the precise function of PKR in these signaling pathways remains controversial . Not only is PKR an effector molecule in the cellular response to dsRNA , but it also integrates signals in response to the activation of Toll-like receptors , growth factors , and diverse cellular stresses ( Hsu et al . , 2004; Nakamura et al . , 2010 ) . PKR is involved in multiple pro- and anti-apoptotic pathways in normal and cancer cells . Investigators have shown that PKR specifically interacts with STATs , FADD , and IKK ( Williams , 2001 ) , and researchers have also described a novel role for PKR as a mediator of IR resistance , modulated partly by the protective effects of NF-κB activation ( von Holzen et al . , 2007 ) . Although some models have been proposed , the precise molecular mechanism by which PKR activates IKK/NF-κB pathways remains to be determined . Moreover , PKR is also involved in many cellular pathways , exerting various functions on cell growth and tumorigenesis ( Marchal et al . , 2014 ) . However , the exact role of PKR in cancer biology remains controversial . Chemotherapeutic drugs such as doxorubicin , etoposide , and 5-fluorouracil are able to induce and activate PKR protein , triggering apoptosis ( García et al . , 2011; Peidis et al . , 2011; Yoon et al . , 2009 ) . Conversely , PKR has been suggested to be involved in the neoplastic process of the proliferative transcription factor NF-κB ( Delgado André and De Lucca , 2007 ) . Curiously , different expression patterns of PKR/eIF2α/NF-κB activity , even in the same type of cancer , such as different cholangiocarcinoma cell lines , point to the complexity of the role of PKR in cancer ( Kunkeaw et al . , 2013 ) . In the current study , we report how DNA-damage-induced nuclear events are linked to the activation of cytoplasmic IKK kinase , thereby activating NF-κB . LRP16 selectively interacts and activates PKR , and it also acts as a scaffold to assist the formation of a ternary complex of PKR and IKKβ , prolonging PAR-dependent NF-κB transactivation caused by genotoxic threats . We also discovered a small molecule MRS2578 that could profoundly abolish these interactions , converting LRP16 into a death molecule and forestalling colon tumorigenesis and acquired resistance to DNA-damaging therapies driven by the oncogenes of the NF-κB pathway . This study takes CRC as a model to understand mechanisms that account for a limited response of genotoxic therapies in solid tumors and to seek combination solutions . Inclusion of etoposide plus MRS2578 , versus each drug alone , exhibits synergistic tumor cytotoxicity both ex vivo and in vivo . Our combinatorial approach introduces a strategy to enhance the efficacy of DNA-damaging cytotoxic therapies for the treatment of cancer . Thus , targeting the biological function of LRP16 will provide the proof of principle for two understudied concepts in cancer therapy: ( 1 ) blocking subsignals , rather than total signals , as a means of impeding oncogenic NF-κB signaling and ( 2 ) targeting regulatory protein–protein interactions as a way to produce effective antitumor agents and to sensitize tumor cells to DNA-damaging cytotoxic therapies .
To clarify the clinicopathological relevance of LRP16 in patients with CRC , we used immunohistochemistry ( IHC ) to examine the LRP16 protein levels in a human tissue array containing 202 CRC clinical specimens with paired adjacent normal colon tissues from patients with CRC . An analysis using the Image-Pro Plus software showed that the level of LRP16 expression was significantly higher in the CRC tissue samples than the adjacent normal tissues ( Figure 1A–B ) . LRP16 was highly elevated in primary CRC tumors compared with their adjacent normal tissues as determined by reverse transcription polymerase chain reaction ( RT-PCR ) and Western blot analysis ( Figure 1—figure supplement 1A ) . These CRC samples were staged according to the system developed by the American Joint Committee on Cancer ( AJCC ) , also known as the Tumor-Node-Metastasis ( TNM ) system . Notably , we found that the level of LRP16 expression positively correlated with the histological grades of the tumors and was also strongly associated with a higher tumor stage ( Figure 1C–D and Figure 1—figure supplement 1A and Figure 1—source data 1 ) , suggesting that the level of LRP16 expression is progressively elevated during the progression of the patients with CRC . A Kaplan–Meier survival analysis of the expression of LRP16 and the clinical behavior of CRC further showed that a low level of LRP16 was associated with better overall survival ( OS ) in CRC patients ( p=0 . 0224 ) ( Figure 1E ) . In this context , LRP16 expression was an independent prognostic factor , with a hazard ratio ( HR ) of 0 . 58 ( 95% confidence interval [CI] 0 . 36–0 . 90 ) in a multivariate analysis ( tumor grades HR 0 . 46 [95% CI 0 . 26–0 . 82]; tumor stages HR 0 . 44 [95% CI 0 . 21–0 . 93] ) , which is similar to our previous findings ( Xi et al . , 2010 ) . Interrogation of the most comprehensive public database , The Cancer Genome Atlas ( TCGA ) ( Cancer Genome Atlas Network , 2012 ) , also supported the notion that the expression of LRP16 transcripts is profoundly elevated in CRC . Interestingly , this database also showed that when CRC samples were further stratified according to their tumor stage , the level of LRP16 expression also positively correlated with the clinical tumor stages ( TNM stage ) of these patients with CRC ( Figure 1F–G ) . Further analysis of our data and the TCGA database showed that the level of LRP16 expression also positively correlated with the histological grades of these CRC patients and that , remarkably , high levels of LRP16 expression in CRC strongly correlated with lymph-node positivity and metastasis in the cohort of patients with CRC ( Figure 1H–I and Figure 1—source data 1 ) . NF-κB is constitutively activated in many malignancies , including CRC ( Sakamoto et al . , 2009 ) , but the molecular mechanism underlying the constitutive activation of NF-κB in tumors remains to be defined . NF-κB activation was defined as the detection of p65 nuclear staining in over 50% of the tumor cells in the CRC tissues . Constitutive activation of NF-κB was observed in 28 . 7% ( 58 of 202 ) of the cohort of patients with CRC ( Figure 1J ) . The total p65 levels slightly differed among the CRC samples , but they were significantly elevated in CRC samples compared with the adjacent normal tissues ( Figure 1J ) , and also positively correlated with the histological grades of the tumors ( Figure 1—figure supplement 1B ) . An analysis of consecutive tissue sections showed that the level of LRP16 expression positively correlated with the level of p65 expression ( Figure 1K ) . The correlation between LRP16 and p65 was further confirmed in a larger scale array of 202 samples . Specifically , approximately 66% of the samples with high LRP16 expression displayed high p65 expression , whereas approximately 68% of the low LRP16 samples displayed low p65 expression ( Figure 1L ) . Similar to the expression of LRP16 , the phosphorylation of p65 ( phospho-p65 ) at Ser536 , which represents the activated form of NF-κB , was significantly higher in CRC tissues . Analysis of phospho-p65 protein expression by Western blot analysis mirrored that of LRP16 , with the highest expression observed in CRC samples . Most importantly , high expression levles of phospho-p65 correlated positively with high expression levels of LRP16 in CRC clinical specimens ( Figure 1—figure supplement 1B ) . Informatively , XIAP ( a target of NF-κB ) expression was significantly elevated in all 202 CRC tissues compared with the adjacent normal tissues ( Figure 1—figure supplement 1C ) . However , we did not observe a significant correlation between LRP16 and XIAP expression in these CRC samples ( Figure 1—figure supplement 1D ) . Moreover , analysis of the TCGA CRC RNA-seq dataset revealed that LRP16 expression was also not significantly correlated with that of BCL2L1 ( anti-apoptotic transcriptional target of NF-κB ) , but it was inversely correlated to that of XIAP ( Figure 1—figure supplement 1D ) . A different trend in the relevance of LRP16 and XIAP expression in our and the TCGA cohorts of patients with CRC might be attributable to the dynamic expression patterns of NF-κB target genes , which did not follow a standardized protocol and differed in the number of CRC samples analyzed and in the patient inclusion criteria , making it difficult to compare results between groups . Taken together , these results provide overwhelming clinical evidence that LRP16 is overexpressed in human CRC samples compared with adjacent normal samples , and also confirms the critical role of LRP16 in promoting CRC tumorigenesis . To explore the biological role of LRP16 in CRC , we screened a panel of CRC cell lines for their endogenous LRP16 levels ( Figure 2A ) . We also observed that the level of phospho-p65 expression was positively associated with the level of LRP16 expression ( Figure 2—figure supplement 1A ) . Next , we investigated the cytotoxic and cytostatic effects of chemotherapeutic drugs on CRC cell lines in culture models . In these cells , we compared the cytotoxic and cytostatic effects of etoposide to those of 5-fluorouracil and oxaliplatin , as reflected by their half-maximal growth inhibitory concentrations ( IC50 ) . The results indicated that CRC cell lines ( i . e . LS180 , HCT116 , and RKO ) expressing relatively higher levels of endogenous LRP16 were less sensitive to etoposide . In contrast , CRC cell lines expressing relatively lower or undetectable levels of endogenous LRP16 ( i . e . SW480 , CaCO2 , and LoVo ) were more sensitive to etoposide ( Figure 2—figure supplement 1B ) . Based on our findings , it is reasonable to speculate that the molecular mechanisms underlying the therapeutic response difference to the cytotoxic and cytostatic effects of etoposide might be dependent on the different levels of LRP16 expression in CRC cell lines . We further examined whether expression of exogenous LRP16 could restore the resistance of CRC cell lines to a basic low level of LRP16 expression in response to chemotherapeutic drugs . As expected , the forced overexpression of LRP16 in SW480 cells with a basic low level of LRP16 rendered tumor cells more resistant to etoposide than to the other two drugs , as evidenced by increased cell viability and clonogenicity ( Figure 2B and Figure 2—figure supplement 2A–B ) . Increasing the expression of LRP16 substantially rescued etoposide-induced apoptosis , as evidenced by reduced caspase three cleavage ( Figure 2C ) . However , no significant protective effects were observed on the invasive and metastatic capacities of the CRC cell line ex vivo ( Figure 2—figure supplement 2C ) . Next , we assessed how LRP16-mediated anti-apoptotic signaling accounted for the limited response to etoposide . A conceivable possibility is that NF-κB activation caused by LRP16 in response to etoposide could account for this protective effect ( Figure 2D ) . We found that elevated phospho-p65 levels , which were associated with substantially enhanced NF-κB transcriptional activity , occurred in the most etoposide non-responsive cell lines . Conversely , in some etoposide-responsive cells , upregulation of the phospho-p65 level or NF-κB transcriptional activity did not remarkably occur ( Figure 2—figure supplement 1B and Figure 2—figure supplement 2D ) . The Western blot results also showed that with a gradually increase in LRP16 expression significantly increased the phosphorylated forms , but not the total forms , of the upstream regulators of the NF-κB pathway , IKKα and IKKβ , and also increased the levels of phospho-p65 ( Figure 2E ) , compared with those in controls . Similar results in clonogenicity experiments demonstrated that the gradual re-expression of exogenous LRP16 in SW480 cells restored their resistance to etoposide ( Figure 2—figure supplement 3A ) . Hence , we wondered whether NF-κB activation enhanced by LRP16 potentially contribute to the limited response to etoposide . To achieve this goal , we assessed the impact of the NF-κB inhibitor , BAY 11–7082 , and the IκBα super-repressor , IκBαSR , a dominant-negative mutant of IκBα ( Wu et al . , 2015 ) , which blocks the effects of NF-κB activation on the viability and clonogenicity of CRC cells . Blocking NF-κB activation with either BAY 11–7082 or IκBαSR in SW480 cells exogenously expressing LRP16 or the vector control profoundly reduced their viability and clonogenicity ( Figure 2F–G and Figure 2—figure supplement 3B ) . These results suggest that LRP16 confers DNA damage-triggered NF-κB-mediated expression of anti-apoptotic signaling molecules and is involved in restraining the response to DNA-damaging cytotoxic therapies in CRC cells . To determine whether LRP16 is the relevant target at the cellular level and whether cell killing is truly dependent on this particular mechanism , LRP16 was specifically knocked down using two different LRP16 small hairpin RNAs ( shRNAs , siRNA_374 and _668 ) , as described previously ( Wu et al . , 2015 ) . Cells stably expressing LRP16 siRNA_374 , siRNA_668 , or control_siRNA were treated with etoposide for 72 hr . Notably , depletion of endogenous LRP16 resulted in significantly reduced cell viability and clonogenicity and profoundly enhanced the sensitivity of cells to etoposide ( Figure 2H and Figure 2—figure supplement 3C ) . Beyond its indispensable role in DNA repair ( Gibson and Kraus , 2012 ) , emerging evidence reveals that PARP1-mediated PARylation is one of the most crucial PTMs orchestrating DNA-damage-initiated NF-κB signaling ( Stilmann et al . , 2009 ) . The high affinity of LRP16 for PAR led us to further ascertain whether LRP16 could facilitate PAR-dependent NF-κB signaling/transactivation of anti-apoptotic genes to counter intrinsic DNA damage in CRC cells and to protect against cell death induced by genotoxic agents . Of note , ectopic expression of LRP16 mutants ( D160A or I161A ) , which was sufficient to significantly reduce its affinity for PAR , but not LRP16 wild-type ( WT ) , in SW480 cells , dramatically sensitized the tumor cells to genotoxic stress-induced apoptosis , as conveyed by reduced cell viability and clonogenicity , in line with the indispensible role of the affinity of LRP16 for PAR in NF-κB activation and anti-apoptotic transcription ( Figure 2—figure supplement 3D ) . Consistently , PARP inhibition by PJ-34 or 3-AB treatment also significantly reduced CRC cells survival and dramatically sensitized cancer cells to etoposide-induced cell death , as evidenced by the reduced cell viability and clonogenicity , in LRP16-upregulated SW480 cells ( Figure 2—figure supplement 3D ) . Taken together , these data suggest that LRP16 plays a critical role in DNA-damage-initiated and PAR-dependent NF-κB transactivation , which could account for the limited or lack of response to the cytotoxic and cytostatic effects of etoposide and protection against cell death induced by genotoxic agents . To further assess whether LRP16 is an essential component of NF-κB signaling , we used an immunoblotting assay to evaluate genotoxicity-induced NF-κB transcriptional activity . The results indicated that the re-expression of LRP16 in cells exposed to etoposide profoundly increased the phosphorylated forms , but not the total forms , of the upstream regulators of the NF-κB pathway , IKKα , IKKβ and IκBα , in a dose- and time-dependent manner ( Figure 3A ) . Consistent with this phenomenon , exposure to IR , together with exogenous LRP16 expression , also largely prolonged the IR-stimulated activation of NF-κB compared with that in cells transfected with control vector ( Figure 3B ) . Similar results were repeated in another CRC cell line , LoVo cells ( Figure 3—figure supplement 1A–B ) . Moreover , compared with control cells , etoposide treatment triggered the marked nuclear translocation of NF-κB/p65 in SW480 cells exogenously expressing LRP16 ( Figure 3—figure supplement 1C ) . Conversely , LRP16 deficiency introduced by its siRNAs in cells exposed to etoposide reduced the phosphorylated forms , but not the total forms , of IKKα , IKKβ , and IκBα , in a dose- and time-dependent manner ( Figure 3C ) . Consistent with these data , IR triggered the phosphorylation of IKKα , IKKβ , and p65 , and IκBα was substantially reduced in SW620 cells expressing LRP16-specific siRNAs compared with those transfected with scrambled nonspecific siRNAs ( Figure 3D ) . Moreover , the nuclear accumulation of NF-κB/p65 was also substantially reduced in the LRP16-specific siRNAs-transfected cell lysates compared with that in the controls ( Figure 3—figure supplement 1D ) . Taken together , these results support the critical function of LRP16 in controlling the DNA damage-stimulated activation of NF-κB . To control for potential off-target effects of the LRP16 siRNA and to confirm that LRP16 deficiency alone impairs the genotoxicity-induced activation of NF-κB , we generated an LRP16-expressing vector that contained silent mutations in the sequences that were targeted by the LRP16-directed siRNAs . Our results showed that the etoposide-triggered phosphorylation of IKKα , IKKβ , p65 , and IκBα was substantially reduced in the LRP16 siRNA_374-transfected cell lysates . Conversely , exogenous expression of LRP16 with a silent mutation at the shRNA target site rescued this effect ( Figure 3E ) . However , our results also indicated that , compared with the control , the exogenous expression of LRP16 had almost no effect on the activation of NF-κB1 ( noncanonical NF-κB signaling pathway ) induced by etoposide ( Figure 3—figure supplement 2A ) . These data suggest that LRP16 is a key regulator of the activating phosphorylation of the activation loops of IKKs and hence of IKK function in the physiological context . Exactly how any specific signaling pathway targets the IKK complex remains contentious . TAK1 is considered to be the immediate upstream activator of IKK and an essential component of both the nuclear and receptor-mediated activation of NF-κB , phosphorylating the activation loops of the IKKs ( Perkins , 2007 ) . Next , we asked whether LRP16 regulates the kinase activity of TAK1 upon DNA damage and mediates activation of the IKK complex . Unexpectedly , the exogenous expression of LRP16 in SW480 cells did not induce significant phosphorylation of TAK1 compared with that in the control ( Figure 3—figure supplement 2B ) , raising the additional possibility that other upstream kinases could contribute to the activation of IKK mediated by LRP16 within the context of etoposide-induced NF-κB signal transduction . Next , we sought to evaluate whether the function of LRP16 in genotoxic stresses-induced NF-κB activation was dependent on its PAR binding ability . Ectopic expression of LRP16 WT , but not its mutants ( D160A or I161A ) , which were sufficient to significantly reduce its affinity to PAR ( Wu et al . , 2015 ) , significantly enhanced NF-κB activation after etoposide treatment , as conveyed by the phosphorylated forms of IKKα and IKKβ , p65 , and IκBα ( Figure 3F ) . A luciferase assay using an NF-κB-responsive element demonstrated that NF-κB transcriptional activity is altered in response to the modulation of LRP16 upon stimulation of DNA damage . Ectopic of LRP16 remarkably enhanced the levels of NF-κB-dependent luciferase reporter gene activity in both SW480 and LoVo cells following either etoposide or IR treatment ( Figure 3—figure supplement 3A ) . Conversely , LRP16 deficiency introduced by two siRNAs in both HCT116 and SW620 cells considerably diminished the NF-κB transcriptional activity in response to etoposide or IR ( Figure 3—figure supplement 3B ) . Of note , ectopic expression of LRP16 WT , but not its mutants ( D160A or I161A ) dramatically enhanced the NF-κB transcriptional activity in response to etoposide or IR ( Figure 3—figure supplement 3A ) , thus supporting the critical function of LRP16 in controlling DNA damage-initiated and PAR-dependent NF-κB signaling activation . NF-κB-mediated transcription of a panel of anti-apoptotic molecules is an important factor for cell fate determination after DNA damage . To further gain support for the notion that LRP16 regulates distinct target genes of NF-κB , an exploratory microarray analysis was performed . The results indicate that cells exogenously expressing LRP16 in response to etoposide develop alterations of the global transcription profile ( Figure 3—figure supplement 4A ) . Further pathway analysis has revealed several pathways that are highly affected by LRP16 , including protein-K63-linked ubiquitination , IκB phosphorylation , and the DNA damage response ( Figure 3—figure supplement 4B ) . To further confirm that LRP16 is involved in the expression of NF-κB-dependent genes , human NF-κB signaling pathway PCR array was used . The results indicate that after exposure to etoposide , the expression of apoptosis-related genes was markedly downregulated , whereas the expression of anti-apoptosis-related genes was strikingly increased in cells with exogenously expression of LRP16 , compared with the control cells ( Figure 3—figure supplement 4C ) . These results suggest that LRP16 is functionally involved in the NF-κB-dependent gene expression induced by etoposide stimulation and might account for the compromised therapeutic efficacy of etoposide . We hypothesized that a unique LRP16-containing protein complex forms under genotoxic threat conditions and that the complex component ( s ) affect NF-κB activity . Thus , LRP16 interactors were purified by immunoaffinity purification ( Nakatani and Ogryzko , 2003 ) , and we identified the LRP16-associated proteins by exhaustive rounds of ‘shotgun’ liquid chromatography and high-throughput mass spectrometry ( LC–MS/MS ) ( Figure 4A ) . An in-depth bioinformatics analysis of the LC–MS/MS data indicated that PKR with the matching peptide potentially interacted with LRP16 . Importantly , the interaction of endogenous LRP16 with PKR protein was confirmed by coimmunoprecipitation ( co-IP ) in SW480 cells after etoposide treatment ( Figure 4B ) . To exclude the possibility that the interaction between LRP16 and the PKR protein was indirect and mediated by DNA or chromatin , ethidium bromide ( EB ) was added to the cell lysates during co-IP . As expected , ethidium bromide did not markedly impair the etoposide-induced interaction between LRP16 and PKR , suggesting that this interaction was also DNA-independent ( Figure 4C ) . To identify the region within PKR responsible for its interaction with LRP16 , we generated several PKR deletions and mutant constructs , and co-expressed them with LRP16 in SW480 cells . A deletion analysis demonstrated that the C-terminal catalytic domain of PKR mediated its physical interaction with LRP16 ( Figure 4D ) . To map the regions in LRP16 that interacted with PKR , we generated a series of LRP16 deletion mutant constructs and co-expressed them with PKR in SW480 cells . Figure 4E shows that the macrodomain of LRP16 retained is ability to interact with PKR . Next , we performed a glutathione S-transferase ( GST ) pull-down assay using GST-fused LRP16 and in vitro-transcribed and -translated components of WT PKR and PKR deletions . Our pull-down assays using the recombinant proteins demonstrated a direct LRP16–PKR interaction ( Figure 4F ) . Similarly , in vitro GST pull-down assays indicated that almost all PKR fragments , except ΔPKc , directly bound to LRP16 . In contrast to the strong association between LRP16 and full-length PKR , the ΔPKc-truncated PKR protein interacted negligibly with LRP16 ( Figure 4G ) , further supporting the critical role of the C-terminus of PKR in the LRP16–PKR interaction . Intriguingly , according to our LC–MS/MS data , IKKβ with two matching peptides potentially interacted with LRP16 ( Figure 4A ) . Consistent with this finding , previous studies have shown that PKR binds specifically to the IKKβ subunit of the IKK complex and mediates the activation of NF-κB ( Bonnet et al . , 2000; Zamanian-Daryoush et al . , 2000 ) . To investigate whether the physical associations between LRP16 and PKR or IKKβ reflected a capacity of LRP16 to interact with both the PKR and IKKβ proteins simultaneously , or whether LRP16 interacted with either PKR or IKKβ in different cellular environments , a GST pull-down assay was performed using GST-fused LRP16 and in vitro-transcribed and -translated components of PKR and IKKβ . The assay indicated that LRP16 was capable of interacting directly with both PKR and IKKβ , but independently with each ( Figure 4H ) . Together , these experiments not only showed the molecular details involved in the interaction between LRP16 and PKR , but also provided additional support for the physical associations among LRP16 , PKR , and IKKβ in vitro . We further identified the contribution of LRP16 to the assembly of this PKR–IKK kinase complex . Notably , depletion of LRP16 introduced by its siRNAs resulted in the significant attenuation of the PKR–IKKβ interaction , both at baseline and after stimulation with etoposide ( Figure 4I ) , suggesting that LRP16 was required for the formation of a ternary complex with PKR and IKKβ . The link between PKR and TP53 has been described most frequently in cancer cells , in which there is a bidirectional and complex regulatory relationship between the two proteins ( Cuddihy et al . , 1999 ) . Unexpectedly , the interaction of endogenous LRP16 with the PKR protein was confirmed after etoposide treatment in TP53-null cells ( Figure 4—figure supplement 1A ) , suggesting that the interaction between the two proteins did not depend on the presence of p53 under these conditions . To understand how LRP16 drives NF-κB activity in association with the cytoplasmic proteins IKKβ and PKR , we used confocal microscopy to observe the intracellular distribution of LRP16 during etoposide , camptothecin , doxorubicin or IR stimulation . After stimulation , both endogenous and exogenous LRP16 dispersed from the nucleus to the cytosol ( Figure 4—figure supplement 1B–C ) . We also prepared nuclear and cytoplasmic fractions from cells treated with etoposide or IR for various times , and examined each for the presence of LRP16 . These cells also revealed the dynamic dispersion of LRP16 from the nucleus to the cytosol ( Figure 4—figure supplement 1B–C ) . More importantly , the translocation of LRP16 was also detected after genotoxic stress in other tumor cells , including breast cancer and lung cancer cells ( Figure 4—figure supplement 1D ) . The kinetics of LRP16 translocation were consistent with the dynamic changes in both IKKα and KKβ phosphorylation in response to etoposide stimulation , suggesting that LRP16 was possibly required for the assembly and activation of the IKK complexes triggered by DNA damage . To test this possibility , we prepared nuclear and cytoplasmic fractions from cells treated with etoposide for different times and used co-IP to pull down both endogenous and exogenous LRP16 from the control cells and cells expressing FLAG-tagged LRP16 in the presence or absence of etoposide . As shown in Figure 4J and K , in the etoposide-treated cells , the interactions of both endogenous and exogenous LRP16 with PKR and IKKβ were observed and confirmed by co-IP only in cytoplasmic fractions , but they were negligible in the nuclear fractions . Notably , the overexpression of FLAG-tagged LRP16 in SW480 cells upon exposure to etoposide increased the interaction between PKR and IKKβ detected with co-IP , but only in the cytoplasmic fractions , with negligible effects in the nuclear fractions ( Figure 4L ) . This finding suggests that LRP16 plays an important role in the recruitment of PKR and IKKβ into a physical complex that facilitates the downstream signaling of NF-κB . We then asked whether the intracellular trafficking of LRP16 occurs in response to stimulation with TNFα . In contrast to the etoposide-induced translocation of LRP16 to the cytoplasm , the translocation of LRP16 was completely abrogated by stimulation with TNFα ( Figure 4—figure supplement 1E ) . The IR and/or etoposide-induced shift of LRP16 into the cytoplasmic complex fractions was completely blocked both by PARP inhibitors ( 3-AB or PJ-34 ) and by the PAR-binding-deficient mutant LRP16_I161A ( Figure 4—figure supplement 1F ) , suggesting that this process strictly depends on the presence of PAR and the intact PAR-binding motifs in LRP16 . Our data suggest that upon the detection of DNA lesions and PAR , LRP16 assembles the PKR and IKKβ complex in the cytoplasm . This complex could be formed by direct protein–PAR interactions , as well as by protein–protein interactions . The PAR chains bound to LRP16 act as an interaction platform , which is required for PKR-mediated IKK activation and the activation of the NF-κB pathway in response to DNA damage . We were particularly interested in how LRP16 drives the activation of the IKK kinases by PKR . We found that PKR is ubiquitously expressed in CRC cell lines ( Figure 5—figure supplement 1A ) . However , the exact role of PKR in cancer biology remains controversial , and thus we ascertained the PKR protein levels with IHC using a human tissue array containing 202 CRC samples with paired adjacent normal colon tissues . Analysis of these data showed that the level of PKR expression was significantly elevated in the carcinoma tissues relative to that in the adjacent tissues ( Figure 5A ) . Moreover , PKR was highly elevated in primary CRC tumors compared with their adjacent normal tissues as determined by RT-PCR and Western blot analysis and , noticeably , also positively correlated with the histological grades of the CRC patients ( Figure 5—figure supplement 1A ) , which is consistent with the previous study ( Kim et al . , 2002 ) . Further analysis of consecutive tissue sections showed that LRP16 expression positively correlated with PKR expression . Specifically , approximately 75% of the samples with high LRP16 expression also displayed high PKR expression , and approximately 58% of the low LRP16 samples displayed low PKR expression ( Figure 5B ) . Chemotherapeutic drugs , especially etoposide , activated PKR in CRC cell lines ( Figure 5—figure supplement 1B–C ) , although the molecular mechanism that underlies PKR activation remains to be clarified . We then asked whether LRP16 affects the PKR activity induced by etoposide . Cells ectopically expressing LRP16 profoundly enhanced the PKR activity induced by etoposide in a time- and dose dependent manner ( Figure 5C ) . Conversely , LRP16 depletion introduced by its siRNAs dramatically blocked the etoposide-induced activation of PKR ( Figure 5D ) . We then examined whether the PAR-binding ability of LRP16 is also critical for PKR activation induced by DNA damage . Upon exposure to etoposide , the introduction of LRP16 WT , but not its mutants LRP16_D160A or LRP16_I161A , profoundly increased the activation of PKR , as evidenced by the level of its autophosphorylation , without affecting its total protein level ( Figure 5E ) , suggesting that the PAR-binding activity of LRP16 is also required for the DNA-damage-induced PKR activity . Next , we hypothesized that the interaction between LRP16 and IKKβ and NF-κB activity is required for or dependent on the activation of PKR . Western blot analysis showed that knockdown of PKR introduced by its siRNAs reduced the phosphorylated forms , but not the total forms , of the upstream regulators of the NF-κB pathway , IKKα and IKKβ , which are induced by LRP16 upon exposure to etoposide ( Figure 5F ) , and also impaired the interaction between LRP16 and IKKβ after etoposide treatment ( Figure 5—figure supplement 1D ) , suggesting that PKR acts as an adaptor protein and may be critical for the LRP16-mediated activation of IKK in response to DNA damage . Similarly , upon IR stimulation , depletion of PKR significantly abrogated the activation of NF-κB mediated by LRP16 ( Figure 5G ) . To determine whether both LRP16- and PKR-dependent signaling to NF-κB was channeled by IKK , we used siRNAs targeting either LRP16 or PKR to repress their expression , and performed a kinase array analysis in vitro . Noticeably , the etoposide-induced activation of IKK kinases was significantly reduced in the absence of either LRP16 or PKR expression ( Figure 5H ) , suggesting that the formation and integration of the ternary complex containing LRP16 , PKR , and IKK kinases plays a crucial role in the DNA-damage-induced transactivation of NF-κB . To investigate the physiological role and the influence of PKR on the cellular behavior of CRC cells , we analyzed the effect of a loss-of-function PKR on cell survival and proliferation after etoposide treatment . The results of cell viability and clonogenicity assays showed that the effect of LRP16 overexpression on the resistance of CRC cells to etoposide was offset , at least in part , when PKR was simultaneously depleted ( Figure 5I and Figure 5—figure supplement 1E ) . Similarly , knocking down PKR in cells expressing LRP16 re-sensitized the cells to etoposide-induced DNA damage , when measured by Annexin-V/PI binding assay . Meanwhile , this phenomenon was accompanied by elevated level of caspase three cleavage ( Figure 5J and Figure 5—figure supplement 1F ) . Whether NF-κB activation requires the catalytic activity of PKR is still contentious ( Bonnet et al . , 2000; Ishii et al . , 2001; Zamanian-Daryoush et al . , 2000 ) . Thus , we next ascertained whether the catalytic activity of PKR is required for the LRP16-mediated activation of IKK complexes during genotoxic stress . Notably , the inhibition of PKR catalytic activity with the chemical inhibitor 2-aminopurine ( 2-AP ) had no significant effect on the activation of IKK induced by LRP16 after DNA damage ( e . g . etoposide or IR ) ( Figure 5—figure supplement 1G–H ) , but it markedly inhibited the phosphorylation of IκBα ( Figure 5—figure supplement 1G ) , which is consistent with a previous study that suggested that PKR directly phosphorylates IκBs ( Kumar et al . , 1994 ) . These data together demonstrate that PKR activates IKK via a direct protein–protein interaction , possibly via the IKKβ subunit , rather than by its kinase activity . Thus , LRP16 activates the NF-κB signaling pathway via its interaction with and activation of PKR , which acts as an adaptor protein and activates IKK complexes in response to DNA-damaging cytotoxic therapies . Our aforementioned results favor a model in which LRP16-mediated PKR/NF-κB activation via protein interactions limits the response of CRC cells to etoposide . We were particularly interested in the implications of this for the development of strategies to optimize the therapeutic benefits of DNA damage in CRC . Therefore , we identified small molecules that can inhibit the LRP16-mediated activation of the PKR/NF-κB pathway and small molecule inhibitors that suppress the proliferation of CRC both ex vivo and in vivo . Several chemical libraries containing hundreds of compounds with different structures were screened , and their ability to dock with the structural pocket of the macro domain region of LRP16 ( amino acids [aa] 153–319; PDB ID code: 2 × 47 ) was examined using the University of California , San Francisco ( UCSF ) DOCK 6 . 1 program suite ( Figure 6—figure supplement 1A ) . The small molecules were ranked according to their energy scores . Among these small molecules , two compounds , MRS2578 ( C20H20N5S4 , molecular weight [MW] 472 . 67 ) , a potent P2Y6 receptor antagonist ( Syhr et al . , 2014 ) , and NECA ( C12H16N6O4 , MW 308 . 29 ) , a potent adenosine receptor agonist ( Ye et al . , 2016 ) , were identified as competitive inhibitors of the affinity of LRP16 for PAR ( Figure 6—figure supplement 1B ) . In this context , we have previously shown that the PAR-binding ability of LRP16 plays an essential role as a spatial regulator of PKR/NF-κB signals by orchestrating the extranuclear signaling of IKK . Accordingly , we found that IKKβ kinase activity was markedly attenuated in IR-stimulated cells pretreated with these two compounds or the PARP inhibitors 3-AB and PJ-34 ( used as positive controls ) compared with that in IR-stimulated control cells . However , the inhibition of IKKβ kinase activity by NECA was less efficient than MRS2578 ( Figure 6—figure supplement 1B ) . Therefore , the small molecule MRS2578 conformed to our requirements and limited the activation of NF-κB induced by DNA damage . To investigate the antitumor effect of MRS2578 in CRC cells ex vivo , CCK-8 assays were conducted to assess the growth of six CRC cell lines ( RKO , LS180 , HCT116 , SW1116 , SW480 , and LoVo ) after treatment with MRS2578 . As indicated in Figure 6A , MRS2578 treatment for 24 hr markedly and dose-dependently reduced the viability of the CRC cell lines . Consistent with this finding , a significantly lower percentage of 5-ethynyl-2’-deoxyuridine ( EdU ) -incorporated ( proliferating ) cells was observed in the MRS2578-treated cells than in the control cells ( Figure 6B ) . MRS2578 also markedly suppressed the proliferation in the CRC cells , as evidenced by their reduced clonogenic survival ( Figure 6C and Figure 6—figure supplement 1C ) . Collectively , these data show that MRS2578 inhibits the survival and proliferation of CRC cells in vitro . We next wondered whether MRS2578 could kill CRC cells by inactivating the LRP16-mediated activation of NF-κB induced by etoposide , and thus sensitizing the tumor cells to DNA-damaging cytotoxic therapies . Of note , etoposide-induced the activation of NF-κB , as judged by the phosphorylation forms of IKKα and IKKβ , p65 , and IκBα , but not the total form , was significantly and dose-dependently inhibited when cells were pretreated with MRS2578 ( Figure 6D and Figure 6—figure supplement 1D ) . Taken together , these results suggested that CRC pretreated with MRS2578 induces cell killing and abrogates LRP16-mediated NF-κB activation in response to etoposide , leading to conversion of LRP16 from a survival into a killer molecule . To evaluate whether MRS2578 increased the sensitivity of CRC cells to etoposide , we first assessed the responses of CRC cells to a combination of etoposide and MRS2578 or each drug alone . At an unfixed ratio , the concentrations of MRS2578 and etoposide spanning the IC50 of each cell line were selected for a combination study in which we calculated the combination index ( CI ) values with the ComboSyn software ( CI <1 , synergistic; CI = 1 , additive; and CI >1 , antagonism ) ( Chou , 2010 ) . To more accurately analyze the degree of synergy between etoposide and MRS2578 , a CI value was calculated . Of note , both SW480 and HCT116 cells treated with the combination of the two drugs showed synergistic cytotoxicity when assessed with the CalcuSyn model ( Figure 6E ) . Cells stably overexpressing LRP16 displayed less sensitivity to etoposide alone than control cells , but these cells were even more sensitive to a combination of etoposide and MRS2578 ( Figure 6F–G ) . To further assess whether the synergistic inhibition of cell proliferation by MRS 2578 and etoposide was modulated by enhanced apoptosis , we performed an Annexin-V/PI binding assay . Notably , the protection against apoptosis afforded by LRP16 overexpression in response to etoposide was offset , at least partially , when MRS2578 was used stimultaneously ( Figure 6H ) . Additionally , results from cell-cycle profiling revealed that both cells stably expressing LRP16 and the control pretreated with MRS2578 displayed significant reductions in the proportions of cells in G0/G1 phase and showed significantly increased in the percentages of the cells in G2 phase ( Figure 6I ) . These findings reveal that MRS2578 increases the sensitivity of CRC cells to etoposide by enhancing their apoptosis and inhibiting cell growth . To further investigate whether MRS2578 increased the sensitivity of CRC cells to IR , we selected two cell lines ( DLD1 and HT-29 ) and exposed them to IR over time until resistance emerged , to generate the two cell lines DLD1_R ( parental DLD1 ) and HT-29_R ( parental HT-29 ) . Compared with their parental cells , both the DLD1_R and HT-29_R cells were substantially sensitized to MRS2578 ( Figure 6J ) , indicating that the cytostatic effect of MRS2578 ( by inducing apoptosis and reducing cell growth ) contributed to increasing the sensitivity or reducing the resistance of the cells to IR , and that the cytostatic effect was also independent of the IR resistance mechanism . Our next step was to understand how MRS2578 blunts genotoxicity-induced NF-κB activation in CRC and significantly sensitizes tumor cells to DNA-damaging cytotoxic therapies . We have shown that LRP16 , interacting with PKR , plays a role in extranuclear NF-κB signaling , but not in nuclear NF-κB signaling , during the DNA damage response . Using LRP16 mutants ( D160A and I161A ) , which function in a dominant inhibitory fashion by impeding the PAR-binding ability of LRP16 , significantly reduced etoposide-induced NF-κB activity ( Figure 3F ) . Thus , we evaluated the effect of MRS2578 on genotoxicity-induced NF-κB signaling and it is possible that this inhibitor behaves in similar fashion by impeding the PAR-binding ability of LRP16 to impede NF-κB activity . Of note , cells pretreated with MRS2578 significantly impeded the phosphorylation of IKKα and IKKβ induced by etoposide , similar to the phenomenon detected in cells stably expressing D160A and I161A ( Figure 6D ) . Pretreating cells with MRS2578 profoundly reduced both the etoposide-stimulated phosphorylation of IKKα and IKKβ and the etoposide-stimulated autophosphorylation of PKR ( Figure 6K ) . Furthermore , it also inhibited the etoposide-stimulated interaction among LRP16 , PKR , and IKKβ and thus prevented the formation of a ternary complex , which might eventually impede the PKR activation of IKK ( Figure 6—figure supplement 1E ) . Taken together , these results suggested that inhibiting the binding of LRP16 to PAR or pretreating cells with MRS 2578 is as effective as inhibiting PKR activity in blocking the biochemical effects of PKR in the cytoplasm and in inhibiting the NF-κB activity induced by LRP16 in response to DNA damage . In light of our ex vivo findings , we examined the effect of LRP16 knockdown in vivo . Stable knockdown of LRP16 introduced by its shRNAs in SW620 cells suppressed tumor growth in the subcutaneous xenograft model . The mean weight of LRP16-silenced tumors was also reduced in SW620 cells compared with controls ( Figure 7A and Figure 7—figure supplement 1A–B ) . Of note , compared with controls , stable knockdown of LRP16 was not only associated with a significant reduction in the growth of the primary SW620 tumors , but also with a marked increase in the sensitivity of the xenograft tumors to etoposide treatment ( Figure 7A and Figure 7—figure supplement 1B ) . Silencing of LRP16 at mRNA levels in SW620 were verified by qPCR ( Figure 7—figure supplement 1C ) . Body weight measurements made during the study indicated that etoposide treatment was tolerated by the animals ( Figure 7—figure supplement 1D ) . Knockdown of LRP16 significantly inhibited cell proliferation and sensitized tumor cells to etoposide in SW620 cell xenograft models , as determined by Ki-67 staining . SW620 xenografts stably expressing shLRP16 showed the induction of apoptosis following etoposide treatment , as evidenced by the increased expression of terminal deoxynucleotidyl transferase dUTP nick-end labeling ( TUNEL ) ( Figure 7B ) . Similar results were also obtained after IR treatment of the mouse xenograft tumors ( Figure 7—figure supplement 2A–E ) , indicating that the expression of LRP16 is not only required , but also sufficient to promote radio-resistance in a CRC tumor model . Collectively , knockdown of LRP16 is synthetically lethal in CRC in vivo through suppressing cell growth and sensitizing the tumor to genotoxicity therapies . Based on our findings obtained in cell culture models ex vivo , we next evaluated the anti-tumor effects of MRS2578 in mouse xenograft models in vivo . We first analyzed the toxicity of MRS2578 in mice . To define the appropriate doses for the in vivo experiments , mice were administered intraperitoneal injections of MRS2578 at tolerable doses ( 10 mg/kg or 20 mg/kg ) , as previously described ( Syhr et al . , 2014 ) . At these doses , NF-κB activity could be inhibited in our cell culture models ( Figure 6D and Figure 6—figure supplement 1D ) . The dose range between 10 and 20 mg/kg was tolerable , with no deaths recorded for up to 12 days of twice weekly administration . To test the potency of MRS2578 in vivo , CRC xenografts derived from SW620 cells were treated with increasing doses ( 0 , 10 , and 20 mg/kg , twice weekly ) of MRS2578 via the intraperitoneal route for 28 days . The results showed that treatment of mice with MRS2578 resulted in a dose-dependent repression of CRC in vivo ( Figure 7Ca–Cb ) . Hematoxylin and eosin ( H and E ) staining of the tumor tissues showed that the number of cells was substantially reduced in the tumors treated with MRS2578 , and that the tumors were filled with fibrosis-like tissue . Consistent with this finding , an IHC analysis indicated that the expression of Ki67 was profoundly reduced in the tumor cells treated with MRS2578 ( Figure 7Cc ) . No significant weight loss or organ toxicities were observed in the mice treated with MRS2578 ( Figure 7—figure supplement 3Aa–Ab ) . These findings suggest that doses between 10 and 20 mg/kg provide the optimal therapeutic index for MRS2578 for in vivo experimentation involving CRC xenografts . As further evidence for the important translational implications of the present studies , in in vivo therapy models in immune-deficient mice , our combinations of etoposide and MRS2578 yielded potent antitumor responses . First , for SW620 cells xenografts , as expected , mice treated with etoposide alone showed a significant reduction in tumor burden versus those treated with vehicle alone . However , the combination ( simultaneous drug administration ) treatment yielded a further tumor burden reduction compared with vehicle or etoposide alone ( Figure 7Da–Db and Figure 7—figure supplement 3Ba–Bb ) . The treatment was well tolerated , as no significant weight loss and no organ toxicities were observed in mice that received these treatments , including combined treatment , throughout the study ( Figure 7—figure supplement 3Ba–Bb ) . Of note , the combination of etoposide and MRS2578 resulted in significantly greater growth inhibition , as determined by Ki-67 staining ( Figure 7Dc ) . Lastly , we also checked the effects of the small molecule inhibitor MRS2578 on oncogenic NF-κB signaling in xenograft tumors . As expected , phospho-p65 levels , but not total forms , are clearly decreased in the compound MRS2578 treated groups ( Figure 7Dd ) . These results are consistent with the observations in the CRC cell experiments ex vivo . We also investigated the combined effects of etoposide and MRS2578 on CT26 colorectal adenocarcinomas ( BALB/c origin ) implanted into BALB/c mice . Both the therapeutic effect and safety profile were evaluated . As expected , treatment with the combination of etoposide and MRS2578 further reduced tumor growth , indicating that this combination was more active than either single agent alone in inhibiting CRC cell growth and significantly prolonged mouse survival ( Figure 7E–F and Figure 7—figure supplement 3C ) . There were no significant reductions in the mouse bodyweights , indicating that the toxicity of this combination is controllable ( Figure 7—figure supplement 3C ) . Taken together , these findings show that pharmacologically targeting LRP16/NF-κB signaling prevents tumorigenesis and suggest that the combination of MRS2578 and etoposide offers therapeutic opportunities for CRC and may warrant an immediate clinical trial .
As our understanding of the molecular basis of cancer has improved , a number of dysregulated signaling pathways responsible for driving disease progression have been identified . Efforts have been made to exploit these pathways as targets for therapeutic intervention , with the expectation that drugs capable of modulating them would deliver previously unachievable efficacy . However , the genetic instability and hypermutation rates of cancer , coupled with the redundancy often built into biological systems , have undermined the importance of singular targets . In the presence of this confluence of factors , resistant mutations arise and compensatory signaling pathways become upregulated , limiting the utility of specific inhibitors ( Ribic et al . , 2003 ) . These observations suggest that while the development of new targets is critical , priority should be given to those that have the potential to act in synergy with and increase the therapeutic index of established treatment modalities . LRP16 , as a cofactor for multiple nuclear receptors , fits this description ( Han et al . , 2011 ) . LRP16 initially attracted attention because of the foundamental roles it plays in driving tumor progression; however more recently , relationships between the cofactor and other disease pathways have emerged . The intersection of these two features has thus made LRP16 an ideal oncology target . A key pathway linking DNA damage with apoptosis , senescence and DNA repair mechanisms involves activating the NF-κB complex ( Bernard et al . , 2004; Hayden and Ghosh , 2008; McCool and Miyamoto , 2012; Wan and Lenardo , 2010 ) . It is widely accepted that genotoxicity-induced NF-κB activation is initiated by both ATM and PARP1 , which trigger the phosphorylation and PIASy-mediated SUMOylation of nuclear NEMO . Our previous and present studies favor two possible roles of LRP16 in activating NF-κB ( Figure 8 ) : first , LRP16 functions not only by providing the lesion specificity during the cellular response to DNA damage by its unique interactions with Ku70/Ku80 , but also ensures the successful PARP1/PAR-dependent recruitment of both ATM and PIASy , together with NEMO . These results underscore the function of LRP16 , a versatile protein that preferentially occurs in the nucleus , in the early nuclear signaling cascade following DNA damage; second , when LRP16 detects DNA lesions and PAR , it assembles the PKR and IKKβ complex , the key players in NF-κB activation , in the cytoplasm . This complex can be formed by a direct protein–PAR interaction , as well as by protein–protein interactions . The PAR chains bound to LRP16 act as an interaction platform , which is required for PKR-mediated IKK activation and then activation of the NF-κB pathway . Thus , LRP16 appears to have a dual function , which might at least partially explain how these DNA-damage-initiated nuclear events are linked to the activation of cytoplasmic IKK . Optimizing radiotherapy and chemotherapy for the treatment of malignant neoplasms has relied on the iterative development and testing of models involving tumor growth dynamics , mutation rates and cell-killing kinetics . However , the most theoretically effective tumoricidal strategies must usually be tempered because of detrimental effects to the host . This reality has led to the development of regimens in which therapies are administered at intervals or cycles to avoid irreparable damage to vital host functions . However , the recovery and repopulation of tumor cells between treatment cycles is a major cause of treatment failure ( Kim and Tannock , 2005 ) . Interestingly , rates of tumor cell repopulation have been shown to accelerate in the intervals between successive courses of treatment and solid tumors commonly show initial responses followed by rapid regrowth and subsequent resistance to further chemotherapy . Understanding the resistance mechanisms involved may open new therapeutic opportunities . Extensive previous research has demonstrated that cancer cells develop resistance to platinum drugs in a variety of ways , including the following: first , pre-target resistance , for instance , through the reduced accumulation or increased extrusion of etoposide by transporters; second , on-target resistance , caused by DNA repair; third , post-target resistance , through the modulation of DNA damage recognition , the damage response , and apoptosis; and fourth , off-target resistance , for example , through compensatory pro-survival signals or nonspecific adaptive responses that are not directly activated by drugs ( Martin et al . , 2008 ) . The apoptotic capacities of chemotherapeutic agents have been widely used to determine the responses of cancer cells to them ( Johnstone et al . , 2002 ) . Cellular apoptosis is a tightly regulated process that is controlled by numerous signal transduction pathways , including the NF-κB pathway ( Johnstone et al . , 2002 ) . The NF-κB signaling pathway remains a very attractive target for pharmacological intervention because it has crucial functions in human health and disease , particularly in inflammatory diseases and cancers ( Hayden and Ghosh , 2008; Perkins , 2007; Smale , 2011 ) . Collectively , these studies support several conclusions: first , the outcomes of genotoxic exposures to any specific benign or neoplastic cell depend on the integration of innate damage response capabilities and the context that is dictated by the composition of the tumor microenvironment; second , although intrinsic drug resistance is clearly operative in some cancers , acquired resistance can also occur without alterations in intrinsic cellular chemosensitivity ( Davis and Tannock , 2000; Kim and Tannock , 2005 ) ; and third , specific tumor microenvironment that promote therapy resistance are attractive targets for augmenting responses to more general genotoxic therapeutics . However , the complexity of the damage response program also supports strategies that are focused on inhibiting upstream master regulators , such as NF-κB , which may be more efficient and effective adjuncts to cytotoxic therapies , provided their side effects are tolerable . Clinically , PARP1 inhibitors are a very exciting spectrum of drugs that are currently used in cancer management . Blocking the catalytic activity of PARP1 has been shown to inhibit base-excision repair ( BER ) , resulting in the accumulation of SSBs , as well as DSBs , during DNA replication , and this damage , in turn , activates homologous recombination ( HR ) ( O'Connor , 2015 ) . Recent studies have shown that the disruption of any HR-related pathway , such as by BRCA mutations , and disruption of Fanconi anemia , and ATM genes , can predict the sensitivity of tumors to the inhibition of PARP1 by small-molecule inhibitors and the associated cytotoxicity ( D'Andrea , 2010; Mateo et al . , 2015; O'Connor , 2015; Pommier et al . , 2016 ) . Advances in understanding how and where , at a molecular level , these agents function optimally as cytotoxic agents and the recent progress in developing the best reagents are particularly important for the future use of PARP1 inhibitors in cancer therapy . The clinically available PARP1 inhibitors have shown considerable efficacy , especially in the treatment of breast and ovarian cancers , in patients with hereditary deletions of the HR BRCA1/2 genes ( Bryant et al . , 2005; Farmer et al . , 2005 ) . Cancers presenting with these mutations represent 5–10% of all triple-negative breast cancers ( estrogen , progesterone , and HER2 receptor-negative breast cancers; TNBCs ) ( Bryant et al . , 2005; Farmer et al . , 2005 ) . However , the responses to PARP inhibitor therapy , even in BRCA-mutant breast cancers , have not been highly persistent . Furthermore , PARP inhibitors have failed to show impressive clinical benefits in patients with sporadic TNBCs and/or other cancers , suggesting that new strategies must be developed to maximize the efficacy of PARP inhibitors ( Lord and Ashworth , 2013 ) . In summary , we discovered a model in which LRP16 selectively interacts and activates PKR and also acts as a scaffold to assist the formation of a ternary complex of PKR and IKKβ , prolonging the PAR-dependent NF-κB activation caused by genotoxic threats . Our preclinical data in CRC cell lines and mouse xenografts , as outlined in this study , suggest the potential for improving the clinical efficacies of DNA-damaging cytotoxic therapies by combining MRS2578 for patients with CRC . Our studies convincingly demonstrate that using small molecules to interfere with the binding of LRP16 to PKR and IKKβ , leading to the inactivation of its downstream NF-κB signaling pathways , is feasible with regard to the suppression of the proliferation of human CRC cancer cells . Thus , our results imply that LRP16 would not classically be viewed as ‘undruggable’ , and open up an avenue that a small molecule targeting LRP16 would be a promising strategy to combat CRC in future therapies .
Human Subjects: The human patient study was approved by the Ethics Committee of the Chinese PLA General Hospital ( Beijing , China ) , and informed consent was obtained from all patients ( S2016-127-01 ) . All samples were obtained in accordance with the Health Insurance Portability and Accountability Act ( HIPAA ) . Animal experimentation: This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health , and the protocol was approved by the Institutional Animal Care and Treatment Committee of the Chinese PLA General Hospital ( IACTC-CPGH-062 ) . All surgery were performed under sodium pentobarbital anesthesia , and every effort was made to minimize suffering . The samples of carcinomas and adjacent normal tissues were obtained from surgical specimens from patients with CRC for whom information on their clinicopathological characteristics were available and were approved by the department of pathology at the Chinese PLA General Hospital and the department of pathology at the Xiyuan Hospital of China Academy of Chinese Medical Sciences . The samples were frozen in liquid nitrogen immediately after their surgical removal until analysis . Colon tissue arrays were prepared and subjected to IHC analysis with the standard 3 , 3′-diaminobenzidine ( DAB ) staining protocol . All experiments were approved by the Ethics Committee of the Chinese PLA General Hospital ( Beijing , China ) , and informed consent was obtained from all patients . All human CRC cell lines HCT116 ( ATCC Cat# CCL-247 , RRID:CVCL_0291 ) , DLD1 ( ATCC Cat# CCL-221 , RRID:CVCL_0248 ) , LoVo ( ATCC Cat# CCL-229 , RRID:CVCL_0399 ) , LS180 ( ATCC Cat# CL-187 , RRID:CVCL_0397 ) , RKO ( ATCC Cat# CRL-2577 , RRID:CVCL_0504 ) , SW48 ( ATCC Cat# CCL-231 , RRID:CVCL_1724 ) , CACO2 ( ATCC Cat# HTB-37 , RRID:CVCL_0025 ) , HT-29 ( ATCC Cat# HTB-38 , RRID:CVCL_0320 ) , SW1116 ( ATCC Cat# CCL-233 , RRID:CVCL_0544 ) , SW480 ( ATCC Cat# CCL-228 , RRID:CVCL_0546 ) , and SW620 ( ATCC Cat# CCL-227 , RRID:CVCL_0547 ) were obtained from the American Type Culture Collection ( ATCC ) ( Manassas , VA ) and the identities have been authenticated by short tandem repeat DNA profiling . All cells described above were regularly tested for mycoplasma contamination . Cells were cultured in RPMI 1640 medium containing 10% fetal calf serum , 2 M glutamine , 100 U/ml each of penicillin and streptomycin . The following antibodies were used in this study for immunoblotting and immunoprecipitation: antibodies directed against p-TAK1 ( Thr184/187 , Cat# 4508S , RRID:AB_561317 ) ( 1:1000 ) , TAK1 ( Cat# 5206S , RRID:AB_10694079 ) ( 1:1000 ) , p-IKKα/IKKβ ( Ser176/Ser177 , Cat# 2078S , RRID:AB_2079379 ) ( 1:1000 ) , IKKβ ( Cat# 2370S , RRID:AB_2122154 ) ( 1:1000 ) , NEMO ( Cat# 2695S , RRID:AB_10695250 ) ( 1:1000 ) , p-p65 ( Ser536 , Cat# 3033S , RRID:AB_331284 ) ( 1:1000 ) , p53 ( 7F5 , Cat# 2527S , RRID:AB_10695803 ) ( 1:1000 ) , p-IκBα ( Ser32/Ser36 , Cat# 9246S , RRID:AB_2151442 ) ( 1:1000 ) , PARP1 ( 46D11 , Cat# 9532S , RRID:AB_10695538 ) ( 1:1000 ) , Caspase 3 ( 8G10 , Cat# 9665S , RRID:AB_2069872 ) ( 1:1000 ) , PKR ( D7F7 , Cat# 12297 , AB_2665515 ) ( 1:1000 ) , and NF-κB1 p105/p50 ( D4P4D , Cat# 13586 , AB_2665516 ) ( 1:1000 ) were obtained from Cell Signaling Technology ( Beverly , MA , USA ) ; antibodies directed against p65 ( Cat# sc-372 , RRID:AB_632037 ) ( 1:1000 ) , p50 ( Cat# sc-114 , RRID:AB_632034 ) ( 1:1000 ) , p-JNK ( Cat# sc-81502 , RRID:AB_1127391 ) ( 1:1000 ) , JNK ( Cat# sc-7345 , RRID:AB_675864 ) ( 1:1000 ) , GST ( Cat# sc-80998 , RRID:AB_1124757 ) , Sp1 ( Cat# sc-17824 , RRID:AB_628272 ) ( 1:2500 ) , β-tubulin ( Cat# sc-53140 , RRID:AB_793543 ) ( 1:2000 ) , β-actin ( Cat# sc-69879 , RRID:AB_1119529 ) ( 1:2000 ) , and glyceraldehde-3-phosphate dehydrogenase GAPDH ( Cat# sc-166545 , RRID:AB_2107299 ) ( 1:3000 ) were obtained from Santa Cruz Biotechnology ( Santa Cruz , CA , USA ) ; antibody directed against p-PKR ( Thr446 ) ( Cat# 11280 , AB_2665517 ) ( 1:1000 ) was obtained from Signaling Antibody ( SAB , Baltimore , MD , USA ) ; ant-FLAG antibody ( Cat# SAB4200071 , RRID:AB_10603396 ) was obtained from Sigma; Alexa-Fluor-488-conjugated goat anti-rabbit IgG antibody and Alexa-Fluor-594-conjugated goat anti-mouse IgG antibody were obtained from Life Technologies . The following reagents were used in this study . Etoposide , 3-aminobenzamide ( 3-AB ) , PJ-34 , benzamide ( BEN ) , ethidium bromide ( EB ) , and human recombinant TNF-α were purchased from Sigma-Aldrich ( St . Louis , MO , USA ) . Doxorubicin and camptothecin were obtained from KeyGEN Biotech ( Nanjing , China ) . Lipofectamine 3000 and Superfect Transfection Reagent were obtained from Invitrogen ( Carlsbad , CA , USA ) and Qiagen ( Chatsworth , CA , USA ) , respectively . Protein A agarose was obtained from the Millipore Corporation ( Bedford , MA , USA ) . PhosSTOP Inhibitor Cocktail Tablets and Complete Protease Inhibitor Cocktail Tablets were obtained from Roche Applied Science ( Mannheim , Germany ) . Biotin-NAD+ , NAD , human PARP enzyme , activated DNA , and PARP buffer were purchased from Trevigen ( Gaithersburg , MD , USA ) . GST Bind Resin was obtained from Novagen ( Madison , WI , USA ) . CCK-8 was obtained from Dojindo ( Tokyo , Japan ) . Apoptosis Detection Kit was obtained from BD Bioscience ( San Jose , CA , USA ) . The constructs expressing wild-type LRP16 and LRP16 mutants have been described in our previous report ( Han et al . , 2003 , Han et al . , 2007; Wu et al . , 2015; Yang et al . , 2009 ) . Wild-type PKR and its mutants were cloned by PCR amplification into pcDNA3–FLAG . siRNA-resistant LRP16 was cloned by PCR amplification into pcDNA3–FLAG . The dominant-negative mutant IκBm ( IκBSR ) , as described previously ( Wu et al . , 2015 ) , in which serines 32 and 36 were mutated to alanine , was inserted into the pcDNA3–FLAG vector . The identities of all the constructs used in this study were verified by DNA sequencing . Cells were transfected with the indicated plasmids using Superfect Reagent , as described previously ( Wu et al . , 2011 , Wu et al . , 2015 ) . For siRNA transfection , cells were plated at 30–60% confluence in Opti-MEM serum-free medium and transfected with a specific siRNA duplex using Lipofectamine RNAiMAX Transfection Reagent ( Life Technologies , Paisley , UK ) , according to the manufacturer’s instructions for 48 hr . siRNAs were ordered as reverse phase ( RP ) -HPLC-purified duplexes from GenePharma ( Shanghai , China ) . The sequences were as follows: LRP16-374 , 5′-GCAGCGGGAGGAACAUUAC-3′ , LRP16-668 , 5′-GACUGGCAAGGCCAAGAUC-3′ , siPKR-1 , 5′-GCAGGGAGUAGUACUUAAA-3′ , siPKR-2 , 5′-GCAUGGGCCAGAAGGAUUU −3′ , siPKR-3 , 5′-GCAGAUACAUCAGAGAUAA −3′ , and siPKR-4 , 5′-CCUGAGACCAGUGAUGAUU −3′ . Cells were plated and treated for 48 hr with the indicated reagents; 1 × 106 cells were washed with cold PBS and resuspended in 100 μl of binding buffer . An apoptosis assay was performed with the Annexin-V and propidium iodide using the protocol provided with the apoptosis detection kit ( BD Biosciences ) . Cells without Annexin V or propidium iodide were used to detect autofluorescence . EdU staining was performed using the Click-iT EdU Alexa Fluor 488 Imaging Kit ( Invitrogen ) , according to the manufacturer’s protocol . EdU was added directly to the culture medium at a final concentration of 10 μM and incubated for 2 hr . The cells were fixed with 4% paraformaldehyde in PBS for 15 min , and then permeabilized with 0 . 5% Triton X-100 for 20 min . Nuclei were counterstained with Hoechst 33342 . Three biological replicates were prepared for each treatment and each assay was performed in triplicate . RNA from the cell lines was isolated with TRIzol Reagent and purified with the RNeasy Mini Kit according to the manufacturer’s protocol ( Qiagen ) . Aliquots of 1 μg of total RNA were reverse-transcribed with the ThermoScript RT–PCR System ( Invitrogen ) . RT–qPCR was performed according to the instruction for SYBR Green PCR Master Mix ( Applied Biosystems , Foster City , CA , USA ) with the ViiA 7 Real-Time PCR System ( Applied Biosystems ) . Relative expression levels were calculated using the 2−ΔΔCT method . β-Actin was used as the housekeeping gene for normalization . The results for these experiments were all reproducible and therefore only one of each experiment is presented . Cells were seeded in 96-well plates overnight and treated with the indicated drugs . A CCK8 assay was performed after incubation for 72 hr . Untreated cells were used to indicate 100% cell viability . The absorbance ( optical density , OD ) was read at a wavelength of 450 nm on an enzyme-linked immunosorbent assay ( ELISA ) plate reader . The cell viability rate was calculated as follows: ( OD treated/OD control ) ×100% . The IC50 values were calculated with GraphPad Prism version 6 . 0 ( GraphPad Software , La Jolla , CA , USA ) . The cells were trypsinized to generate a single-cell suspension , and seeded in six-well plates at 104 cells per well , and treated with or without the indicated drugs . At 2–3 weeks after seeding , the colonies were stained with crystal violet as described previously ( Li et al . , 2013 ) . Three biological replicates were prepared for each treatment and each assay was performed in triplicate . The cells were immunostained as described previously ( Wu et al . , 2011 , Wu et al . , 2015 ) . Briefly , the cells were fixed in 4% paraformaldehyde ( PFA ) in PBS . After blocking and permeabilization with 0 . 3% Triton X-100% and 1% bovine serum albumin in PBS , the cells were probed with the indicated antibodies . Alexa-Fluor-488- and Alexa-Fluor-594-labeled secondary antibodies were used to visualize the immunofluorescent signals . Representative fluorescent images were acquired with a confocal laser scanning microscope ( FV1000 , Olympus , Tokyo , Japan ) . Cells were washed with ice-cold PBS and lysed on ice for 30 min with RIPA lysis buffer supplemented with protease inhibitors and PhosSTOP inhibitors . Protein concentrations were determined with the Bradford Protein Assay Kit ( Bio-Rad , Carlsbad , CA , USA ) and a calibration standard curve created with BSA . The samples were prepared for loading by adding 4 × sample buffer ( Invitrogen ) and heating the samples at 90°C for 10 min . The total proteins were separated with SDS-PAGE . The proteins in the gel were electrophoretically transferred to a PVDF membrane ( Millipore ) , and then the membrane was blocked in 5% milk or BSA with Tris-buffered saline/Triton X-100 buffer ( 100 mM Tris-HCl [pH 7 . 4] , 500 mM NaCl , 0 . 1% Triton X-100 ) ( TBS-T 0 . 1% ) . The membranes were incubated overnight at 4°C with primary antibodies in 5% milk or BSA in 0 . 1% TBS-T . Horseradish peroxidase ( HRP ) -conjugated secondary antibody was added and incubated for 1 hr at room temperature in 0 . 1% TBS-T , and the signals were visualized with an enhanced chemiluminescence system . Immunoprecipitation ( IP ) experiments were performed , essentially as described previously ( Wu et al . , 2011 , Wu et al . , 2015; Yang et al . , 2009 ) . Briefly , for conventional IP experiments , cells were cultured to 70% confluence in 10 cm culture dishes . The cells were then either harvested without additional treatments or after treatment with either IR or a specific genotoxic agent . The cells were then lysed with IP buffer ( 50 mM Tris-HCl [pH 7 . 4] , with 150 mM NaCl , 1 mM EDTA , and 1% Triton X-100 ) in the presence of a protease inhibitor mixture . After a preclearing step with protein A/G-agarose beads ( Upstate ) , the protein lysate ( 1 mg ) was immunoprecipitated with the agarose-immobilized antibody overnight at 4°C . The blot was incubated with HRP-conjugated secondary antibody , and the signals were visualized with an enhanced chemiluminescence system as described by the manufacturer . Similar results were obtained from two independent sets of experiments , and the result of one experiment is presented . Cells were seeded in 12-well plates . Experiments were set up as described previously ( Wu et al . , 2011 , Wu et al . , 2015 ) . Briefly , the cells were then transfected with a luciferase reporter plasmid containing 3 × κB–luc motifs and the indicated plasmids and/or siRNAs together with pRL-SV40 . The total amount of input DNA for each treatment was kept constant by supplementing it with pcDNA3 . 1 . At 48 hr after transfection , the cells were either left untreated or treated with genotoxic stress . At the time of harvest , the promoter activity was assessed with a dual-luciferase assay kit . Briefly , the feeding medium was removed from the wells , and the cells were washed once with ice cold PBS and lysed with 100 μl of ice-cold reporter lysis buffer . The cell lysate ( 10 μl ) was then added to 50 μl of luciferase substrate solution , after which 50 μl of Stop and Glow Buffer was added to visualize the luciferase signal . The bioluminescence generated was measured with a luminometer ( Berthold Detection System , Pforzheim , Germany ) . The luminescence readings obtained were normalized to the protein concentration of the corresponding cell lysate and presented as the fold difference relative to the control setup . For bacterial expression , IκBα cDNA ( encoding aa 1–66 ) was cloned into pGEX-6p-1 . The recombinant fusion protein expressed by the pGEX-6p-1–LRP16 plasmid has been described in our previous report ( Wu et al . , 2011 , Wu et al . , 2015 ) . The procedure used to obtain the purified recombinant proteins from Escherichia coli cells has been described previously ( Wu et al . , 2015 ) . Escherichia coli BL21 cells were transformed with pGEX-6p-1 , pGEX-6p-1-LRP16 , or pGEX-6p-1-IκBα ( aa 1–66 ) plasmid , and at OD600 = 0 . 5–1 . 0 , the cultures were induced with 100 mM IPTG for 10 hr at 20°C . The crude bacterial lysates were prepared by sonication ( nine rounds of 20 s bursts on ice ) in lysis buffer ( 40 mM Tris-HCl [pH 7 . 5] , 150 mM NaCl , 1 mM EDTA , 0 . 5% NP-40 , and 10% [v/v] glycerol ) in the presence of a protease inhibitor mixture . The GST-tagged recombinant proteins were then purified using GST Bind Resin according to the manufacturer’s protocol . The GST-tagged proteins that were associated with the GST Bind Resin were incubated with various proteins of interest . The GST pull-down assay was performed as described previously ( Wu et al . , 2011 , Wu et al . , 2015 ) . The total proteins that were associated with the beads were then analyzed by Western blotting analysis using the appropriate antibodies . Each GST pull-down sample was analyzed three times and reproducible results were obtained . The bead-associated FLAG-tagged proteins of interest were eluted by incubation with the 3 × FLAG Peptide antigen ( Sigma ) . The eluted products were fractionated by an SDS-PAGE . The gel was then subjected to silver staining , and the visible bands were excised and subjected to a mass-spectrometric analysis . A 100 μl reaction mixture containing 3 μl of human PARP1 enzyme , 3 μl of 20 mM NAD or 250 μM biotin–NAD , 10 μl of 10 × activated DNA , and 5 μl of 20 × PARP buffer was incubated at 25°C for 1 hr . The reaction mixture was then used directly in a PAR-binding assay . The recombinant proteins were incubated for 30 min at 32°C in 40 mM Tris-HCl ( pH 7 . 5 ) , 150 mM NaCl , 1 mM EDTA , 0 . 5% NP40 , 10% ( v/v ) glycerol , biotin-labeled PAR and/or small molecules . The reaction mixtures were applied to nitrocellulose and washed for 30 min with TBS-T containing 100 nM NaCl . After incubation with streptavidin–HRP , the bound biotin-labeled PAR was detected with a DAB Horseradish Peroxidase Color Development Kit . Each sample was detected three times . Each experiment was performed three times , and reproducible results were obtained . Cells were plated on 96-well plates and treated with various concentrations of each drug alone or in combination , at a constant ratio . Following daily treatments for 4 days , the cells were assayed with CellTiter 96 Aqueous One Solution Cell Proliferation Reagent ( Promega , Madison , Wi , USA ) . The absorbance values were used to determine the fraction of cells affected by each treatment and to determine the combination indices ( CIs ) according to the Chou–Talalay method , using CompuSyn software: CI <1 indicates synergism , CI = 1 indicates an additive effect , and CI >1 indicates an antagonistic effect . The total RNA from colon cancer cells stably transfected with the control vector or LRP16-expressing plasmid , and treated with or without etoposide ( 50 μM ) for the indicated periods , was isolated and purified with an RNeasy Kit ( Qiagen , Hilden , Germany ) . The integrity of the RNA was assessed with an Agilent BioAnalyzer 2100 ( Agilent Technologies , Palo Alto , CA , USA ) . The samples were processed and hybridized to an Affymetrix GeneChip Human Transcriptome Array ( HTA ) 2 . 0 , which contains >6 . 0 million probes covering the exons of >65 , 000 coding and noncoding transcripts . The gene expression microarray data were preprocessed and normalized with the Affymetrix Expression Console software ( V1 . 3 . 1 ) using the RMA sketch method . Several sources of gene sets were selected , including gene ontology ( GO ) and Kyoto Encyclopedia of Genes and Genomes ( KEGG ) . The sets included groups based on molecular function , cellular localization , biological processes , and signaling pathways . The resulting lists were examined for their enrichment in terms of the GO ( biological process ) and KEGG pathways . For the latter , pathways associated with diseases were filtered out as reported ( Li et al . , 2013 , Li et al . , 2014 ) . The enrichment analysis was based on a hypergeometric test . p-Values were adjusted using Benjamini–Hochberg’s false discovery rate ( FDR ) ; only FDRs <0 . 1 were considered . A correction for genes in overlapping clusters was applied . Tumor tissues were fixed in 4% formaldehyde solution and processed routinely for paraffin embedding . Sections were cut to a thickness of approximately 4 μm , placed on glass slides , and counterstained with H and E . For IHC , formalin-fixed paraffin-embedded tissue sections were dewaxed , hydrated , heated for 2 min in a conventional pressure cooker , treated with 3% H2O2 for 20 min , and then incubated with normal goat serum for 30 min . The sections were then incubated with antibodies overnight . After washing , the sections were incubated with biotin-labeled secondary antibody for 20 min at 37°C . The slides were then rinsed and incubated with streptavidin–biotin–peroxidase for 20 min . The color reaction was developed with 3 , 3′-diaminobenzidine tetrahydrochloride . The IHC specimens were independently assessed by two pathologists who were blinded to the origin of the samples , including the clinicopathological data . The staining intensity and extent of the stained area were graded according to the German semiquantitative scoring system: staining intensity of the nucleus , cytoplasm , and/or membrane ( no staining = 0; weak staining = 1; moderate staining = 2; strong staining = 3 ) ; the extent of stained cells ( 0% = 0 , 1%–24% = 1 , 25%–49% = 2 , 50%–74% = 3 , 75%–100% = 4 ) . The final immunoreactive score ( 0 to 12 ) was determined by multiplying the intensity score by the extent of stained cells . Based on the QS , LRP16 expression was graded as low ( 0–1 ) or high ( 2–3 ) . SW620 cells were transfected with either pooled LRP16 siRNAs or pooled PKR siRNAs for knockdown . After 36 hr , the cells were lysed with IP buffer , and the kinase complex was prepared with an anti-NEMO antibody . GST–IκBα ( aa 1–66 ) was purified on a glutathione–agarose column and used as the substrate . The reaction was performed with a mixture of the kinase complex , 0 . 5 μg of GST–IκBα ( aa 1–66 ) , and 100 μM ATP in kinase buffer ( 25 mM Tris [pH 7 . 5] , 10 mM MgCl2 , 2 mM EGTA , 1 mM dithiothreitol , and 1 mM sodium orthovandate ) + inhibitors ( 0 . 5 mM PMSF , 10 mM β-glycerol phosphate ( BGP ) , 300 μM sodium othovanadate , 1 μg/ml leupeptin , 1 μg/ml aprotinin , 10 mM sodium fluoride , 10 mM p-nitrophenyl phosphate ) at 30°C for 30 min . The reactions were stopped by the addition of 10 mM EDTA and stored at −20°C until analysis by immunoblotting . The purified IKK complex and GST–IκBα ( aa 1–66 ) were analyzed by Western blotting with the appropriate antibodies . Each sample was detected two times . Each experiment was performed two times and reproducible results were obtained . Tumor cell xenografts were generated as previously described ( Li et al . , 2013 ) . Female 6–8 week-old athymic nu/nu nude mice ( RRID:IMSR_CRL:088 ) and BALB/c mice ( RRID:IMSR_CRL:28 ) were purchased from HFK Bioscience Co . , Ltd ( Beijing , China ) . Mice CT26 ( ATCC Cat# CRL-2638 , RRID:CVCL_7256 ) was also obtained from ATCC ( Manassas , VA ) and the identities have been authenticated by short tandem repeat DNA profiling . Cells described above were regularly tested for mycoplasma contamination . All studies were approved by the Institutional Animal Care and Treatment Committee of the Chinese PLA General Hospital . Tumor cells ( 3 × 106 ) were injected subcutaneously into the lateral flanks of mice and allowed to develop for 10–15 days . The tumor-bearing mice were randomized into groups and began treatment when the average tumor volume reached 100–150 mm3 . The mice with colon cancer xenografts were treated with etoposide , MRS 2578 , or their combination , given intraperitoneally , at the indicated dose or with IR . For the combination treatment , both drugs were given concurrently . During treatment , the tumor volume ( V ) was measured with a Vernier caliper 2–3 times per week and calculated with the formula V = ( L × W2 ) /2 , where L is the length and W is the width , as previously described ( Li et al . , 2013 ) . The volumes of the tumors treated with the vehicle control or the compounds were compared , and the p values were determined with a two-tailed Student’s t test . At the completion of the study , the mice were killed and necropsied , and the tumor tissues were removed for further analysis . All statistical analyses were performed with GraphPad Prism version 6 . 0 ( GraphPad Software , La Jolla , CA , USA ) . Two-tailed Student’s t test , two-way ANOVA , or one-way ANOVA with Dunnett’s multiple comparisons test was used to compare the statistical differences between the relevant groups . The synergistic effect in the animal study was analyzed with two-way ANOVA . A p value < 0 . 05 was considered statistically significant . The half maximal inhibitory concentration ( IC50 ) values were calculated with GraphPad Prism version 6 . 0 . For all in vivo experiments , we used the GraphPad Prism software ( ANOVA/Mann–Whitney test ) to calculate the statistical significance of differences . Microarray data were submitted to the NCBI Gene Expression Omnibus under accession number GSE93625 . | Most chemotherapy drugs kill cancer cells by damaging their DNA . The cells have systems to combat this damage and help them to survive , and in some cells these systems work effectively enough to make the cancer effectively resistant to the treatment . For example , a protein called NF-κB can turn on various genes that help to repair damaged DNA . However , DNA is contained the cell nucleus , whereas the inactive form of NF-κB is found outside the cell nucleus . So how does the damaged DNA communicate with – and activate – NF-κB ? Previous research had found that another protein called LRP16 , which resides in the cell nucleus , plays a crucial role in the repair process that NF-κB is involved in . Li , Wu , An et al . have now studied bowel cancer cells taken from human tissue samples and found that the cancerous cells contained higher levels of LRP16 than cells from the surrounding tissue . Patients with cancers containing very high levels of LRP16 were more severely affected by cancer . Further investigation revealed that when DNA is damaged , LRP16 moves out of the cell nucleus and stabilises how NF-κB interacts with two other proteins; this stabilisation activates NF-κB . LRP16 therefore appears to regulate the signal that travels out of the nucleus from the damaged DNA to activate NF-κB . Further experiments showed that anti-cancer treatments worked best on cancer cells that lacked LRP16 . Thus it appears that LRP16 helps cancer cells to respond to and resist the DNA damage caused by chemotherapy . Li , Wu , An et al . went on to identify a drug that prevented the activation of NF-κB by blocking the effects of LRP16 . Using this drug alongside chemotherapy drugs made the cells more likely to self-destruct . More work is now needed to develop therapies based on the newly identified drug and to establish how DNA damage activates LRP16 . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"cancer",
"biology"
] | 2017 | Blockade of the LRP16-PKR-NF-κB signaling axis sensitizes colorectal carcinoma cells to DNA-damaging cytotoxic therapy |
The striatum is an input structure of the basal ganglia implicated in several time-dependent functions including reinforcement learning , decision making , and interval timing . To determine whether striatal ensembles drive subjects' judgments of duration , we manipulated and recorded from striatal neurons in rats performing a duration categorization psychophysical task . We found that the dynamics of striatal neurons predicted duration judgments , and that simultaneously recorded ensembles could judge duration as well as the animal . Furthermore , striatal neurons were necessary for duration judgments , as muscimol infusions produced a specific impairment in animals' duration sensitivity . Lastly , we show that time as encoded by striatal populations ran faster or slower when rats judged a duration as longer or shorter , respectively . These results demonstrate that the speed with which striatal population state changes supports the fundamental ability of animals to judge the passage of time .
Time , like space , is a fundamental dimension of the environment , yet how it is processed in the brain is poorly understood . A number of recent studies have identified dynamics that allow for robust representation of time by populations of neurons in multiple areas including the hippocampus ( Pastalkova et al . , 2008; MacDonald et al . , 2011 ) , prefrontal ( Kim et al . , 2013; Xu et al . , 2014 ) , parietal ( Leon and Shadlen , 2003; Janssen and Shadlen , 2005 ) and motor ( Lebedev et al . , 2008 ) cortices , cerebellum ( Mauk and Buonomano , 2004 ) , and the striatum ( Matell et al . , 2003; Jin et al . , 2009; Adler , 2012; Mello et al . , 2015 ) . However , any dynamics that result in a continuously-evolving and non-repeating population state can be used to encode time ( Buonomano , 2014 ) , and it is not known whether such temporal representations would inform subjects’ judgments of duration or merely covary with elapsing time . The striatum , a brain structure known to be involved in reinforcement learning and decision making ( Lau and Glimcher , 2008; Samejima et al . , 2005; Lee et al . , 2015 ) , has been implicated in interval timing by several lines of evidence ( Hinton and Meck , 2004; Harrington et al . , 2009; Wencil et al . , 2010; Malapani , 1998; Meck , 2006 ) . However , whether dynamics in striatal activity can explain the perceptual performance of behaving subjects is unknown . To determine whether striatal ensembles drive subjects’ judgments of duration , we manipulated and recorded from striatal neurons in rats performing a duration categorization psychophysical task .
To measure the duration sensitivity of subjects’ timing judgments , we trained rats to judge whether time intervals belonged to a long or short category ( Gouvêa et al . , 2014 ) ( see Materials and methods; Figure 1a ) . At each self-initiated trial , two brief tones ( interval onset , offset ) were presented separated in time by an interval randomly selected from the set I = {0 . 6 , 1 . 05 , 1 . 26 , 1 . 38 , 1 . 62 , 1 . 74 , 1 . 95 , 2 . 4} seconds . Judgments about interval duration were reported at two laterally located nose ports: choosing the left side was reinforced with water after intervals longer than 1 . 5 seconds ( long stimuli ) , and the right side otherwise ( short stimuli , Figure 1b ) . Animals were required to withhold choice until interval offset . Animals made virtually no errors when categorizing the easiest ( i . e . shortest and longest ) intervals , but categorization performance declined as intervals approached the 1 . 5 second categorical boundary ( Figure 1c ) . 10 . 7554/eLife . 11386 . 003Figure 1 . Rats judged interval durations as either long or short . ( a ) Rats triggered interval stimuli ( i . e . two brief auditory tones separated by a silent interval of random duration ) by inserting their snout into a central port . Following interval offset , animals reported their long vs short judgment at two lateral choice ports . Correct trials yielded a water reward , while incorrect or premature responses produced a white noise sound and a time out . Top view , high-speed video was acquired throughout task performance . ( b ) Reward contingency . ( c ) Averaged psychometric curves following bilateral muscimol or saline injections in dorsal striatum ( mean ± standard deviation across session means , and logistic fit; n = 3 rats , 4 sessions each ) . Inset: slope of psychometric curves on consecutive saline and muscimol sessions . All raw data for Figure 1 can be found in Figure 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 11386 . 00310 . 7554/eLife . 11386 . 004Figure 1—source data 1 . The . txt file contains trial by trials stimulus ( Interval ) , choice ( choiceLong ) , animal ( Name ) , treatment ( MuscimolDose ) and session ( Date ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11386 . 00410 . 7554/eLife . 11386 . 005Figure 1—figure supplement 1 . Histological confirmation of cannula placements . Histology slices and schematic representation of the location of saline and muscimol injections . Coronal slices at intermediate anterior posterior ( AP ) positions are shown for reference at +0 . 84mm ( left , rat Albert ) , +1 . 68 mm ( center , rat Yuri ) and + 0 . 60 mm ( right , rat Zack ) from Bregma . Vertical grey bars represent the location of the cannula placements . Yellow asterisks ( * ) show the approximate dorsal-ventral ( DV ) position from where the injectors extended 1 . 5 mm bellow the cannulae . DOI: http://dx . doi . org/10 . 7554/eLife . 11386 . 005 We recorded action potentials during task performance ( see Materials and methods ) , from populations of single striatal neurons targeting dorsal-central striatum , an area where manipulations produced timing deficits ( Meck , 2006 ) ( Figure 2a . For a reconstruction of striatal recording sites see Figure 2—figure supplement 1 . ) . We observed that striatal neurons displayed diverse firing patterns , with different units firing at different times within the interval period ( Figure 2b–d ) . Can such firing patterns support duration judgments ? To determine whether and the degree to which individual neurons could contribute to duration judgments , for each trial , we counted spikes in the last 500 ms of the interval period and compared spike count distributions of short vs long stimulus trials using a receiver operating characteristic ( ROC ) analysis ( see Materials and methods ) . We found that the majority of neurons ( ~57% ) preferred either short or long stimuli ( Figure 2e; short-preferring: n = 159/433 , 36 . 7%; long-preferring: n = 87/433 , 20 . 1%; permutation test , p<0 . 05 ) . As expected , short-preferring neurons displayed higher firing on average prior to the 1 . 5 s category boundary , after which long-preferring neurons displayed higher firing ( Figure 2f ) . These averaged activity patterns resemble the likelihood of receiving reward on a moment-by-moment by basis should the animal choose short or long ( compare with reward contingency in Figure 1b ) . Such signals , previously observed in the parietal cortex of monkeys performing a similar timing task ( Leon and Shadlen , 2003 ) and in the striatum in a value based decision task ( Lau and Glimcher , 2008 ) , are potentially useful for guiding choice . However , were animals’ judgments indeed guided by such signals , it should be possible to predict choices reported later in the trial using neural activity collected during interval stimuli . Indeed , in trials wherein a near boundary interval was judged as long , firing of the short ( long ) preferring subpopulation dropped ( rose ) faster , so that the two curves crossed before the 1 . 5 s boundary ( Figure 2g ) . Conversely , in trials wherein the same interval was judged as short , the two curves evolved more slowly so that at the time of interval offset the short preferring subpopulation was still firing at a higher level and a crossing point had not yet been reached ( Figure 2h ) . 10 . 7554/eLife . 11386 . 006Figure 2 . Dynamics of striatal subpopulations predict duration judgments . ( a ) Psychometric function for neural recording sessions ( mean ± standard deviation across sessions and logistic fit , n = 37 sessions from 3 rats ) . ( b , c ) Raster plot and peri-stimulus time histogram ( PSTH ) of two example cells for trials in which the longest stimulus interval ( 2 . 4 s ) was presented . Time = 0 corresponds to stimulus onset . ( d ) Normalized PSTHs of all neurons for trials in which the longest stimulus interval was presented . Arrowheads indicate cells shown in ( b , c ) . Blue and red ticks indicate cells with significant short and long preferences , respectively . ( e ) Histogram of preference indices . Blue and red outlines indicate subpopulations with significant short and long preferences , respectively . ( f ) Averaged , normalized PSTH of the two subpopulations outlined in ( e ) for trials in which the longest stimulus interval was presented ( mean ± SEM ) . ( g ) Same as in ( f ) , for trials in which a near-boundary stimulus interval ( 1 . 62 s ) was judged as long . For comparison , curves shown in ( f ) are reproduced as a watermark . ( h ) same as ( g ) for trials in which the stimulus was judged as short . For single subjects , see Figure 2—figure supplement 2 . Behavior and neural spike count data for Figure 2 and Figure 2—figure supplements 1 and 2 can be found in Figure 2—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 11386 . 00610 . 7554/eLife . 11386 . 007Figure 2—source data 1 . Folder with raw data for Figures 2–4 . Each file in the source data folder is named “‘<animal>_<session>"’ , and contains an 8x2 cell "‘spikeCounts"’ . For each of the eight stimuli and two choices , "‘spikeCounts"’ contains a matrix of spike counts with dimensions:number of neurons x number of trials x number of time steps . The stimuli ( delay to the second tone in seconds ) and choices are ordered as follows:STIM_SET = [0 . 60 1 . 05 1 . 26 1 . 38 1 . 62 1 . 74 1 . 95 2 . 4];CHOICES = ['incorrect' 'correct'];The spike counts are given from 500 ms prior to trial initiation up to 500 ms after the time of the second tone and were binned in 2 ms wide bins . This folder also includes underlying data for Figures 3 and 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 11386 . 00710 . 7554/eLife . 11386 . 008Figure 2—figure supplement 1 . Electrophysiological recordings in dorsal striatum . ( a ) Movable microwire bundle array ( Innovative Neurophysiology ) used for all neural recordings . ( b ) Histogram of firing rates for all selected cells ( bin size 1 spike/s ) . ( c ) Schematic representation of the striatal recording sites . Coronal slices at intermediate AP positions are shown for reference ( left to right , rats Bertrand , Edgar and Fernando ) . Colored rectangles show the approximate DV position of the wire bundles across recording sessions and horizontal black lines represent session by session recording sites , for 10 , 9 and 18 recording sessions , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 11386 . 00810 . 7554/eLife . 11386 . 009Figure 2—figure supplement 2 . Dynamics of striatal subpopulations predict duration judgments . ( a , f , k ) Psychometric functions for the recording sessions of rats Bertrand ( a ) , Edgar ( f ) and Fernando ( k ) ( mean ± standard deviation across sessions and logistic fit ) . ( b , g , l ) Histograms of preference indices for the same individual animals . Blue and red outlines indicate subpopulations with significant short and long preferences , respectively . ( c , h , m ) Averaged , normalized PSTHs of the two subpopulations outlined in ( b , g , l ) for trials in which the a near-boundary stimulus interval ( 1 . 62 s ) was judged as long ( mean ± SEM ) . ( d , i , n ) same as in ( c , h , m ) for short judgment trials . ( e , j , o ) Normalized PSTHs of all neurons for each animal for trials in which the longest stimulus interval was presented . Blue and red ticks indicate cells with significant short and long preferences , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 11386 . 009 The observation of large proportions of short- and long-preferring neurons whose dynamics predicted choice is evidence that duration judgments are guided by the state of striatal populations . Might the information afforded by ensembles of striatal neurons account for the pattern of subjects’ judgments across all stimuli ? To test this hypothesis , we compared session to session fluctuations in behavioral performance with the separability of activity states of simultaneously recorded ensembles at the offset of short as compared to long intervals . Briefly , for each trial in a session we characterized neural population state as a vector r = ( r1 , r2 , . . . , rN ) , where rn is the number of spikes fired by neuron n ∈ [1 , N] within the last 500 ms of the interval period . Next , for each session we found the linear discriminant that best separated population state vectors according to whether they came from a long or a short interval trial ( Figure 3a; see Materials and methods ) . A threshold placed along the linear discriminant was then used as a decision rule ( black line in Figure 3a ) to generate a ‘neural duration judgment’ for each trial . This procedure allowed us to obtain , for each session , a quantitative description of how well simultaneously recorded neurons could categorize stimuli , i . e . , a neurometric function comparable to the behavioral psychometric function ( Figure 3b ) . Consistent with duration information being encoded at the population level , we found that for sessions in which greater numbers of neurons were recorded simultaneously ( i . e . upper tercile of sessions with regard to population size ) psychometric and neurometric performances were similar and strongly correlated ( r2 = 0 . 76 , p<0 . 001; Figure 3c ) . These results demonstrate that a read out of stimulus category from even modestly-sized ensembles of striatal neurons was in many cases sufficient to explain the pattern of duration judgments produced by behaving subjects . 10 . 7554/eLife . 11386 . 010Figure 3 . Simultaneously recorded population state at interval offset can explain behavioral performance . ( a ) Low dimensional representation of population state at interval offset for one example session . Black line is the decision rule ( see text ) . ( b ) Example psychometric , neurometric and videometric curves for the same session as in ( a ) . ( c ) Slopes of psychometric and neurometric curves for all sessions . Color indicates terciles of population size . Highlighted data point corresponds to the session in ( a-b ) . Inset: regression slope of neurometric and videometric curves for sessions in the upper tercile . See Figure 3—figure supplement 2 for psychometric-videometric comparison at interval offset and choice . ( d ) Performance of an ideal observer analysis in predicting stimulus category , applied to neural ( orange ) and video ( blue ) data obtained at different times relative to interval offset . Thin lines corresponds to individual sessions . Thick lines are averages . ( e ) The orange and blue curves ( thin lines in panel ( d ) ) for corresponding sessions were regressed against each other at different time shifts . The regression R2 values for each session are shown in thin grey lines . The average over all sessions is shown in black . Sizes of black squares indicate the number of sessions with significant positive correlations ( largest squares at 200 and 100 ms correspond to 5 sessions and smallest one at -200 ms to 1 , out of a total of 8 sessions ) . ( f ) Psychometric curves constructed from trials separated according to whether the population state at stimulus offset had advanced more or less along the mean trajectory . Color indicates terciles shown in ( g ) . ( g ) Distributions of projection on normalized mean trajectory for all trials for each stimulus are shown ( stimuli color coded as in [a] ) . The equal area bins shown correspond to the groups of trials used for constructing the three psychometric curves shown in panel ( f ) . Data in f-g are from rat Bertrand . See Figure 3—figure supplement 3 for the remaining two subjects . DOI: http://dx . doi . org/10 . 7554/eLife . 11386 . 01010 . 7554/eLife . 11386 . 011Figure 3—figure supplement 1 . Image frames at the end of the neural analysis window do not show a consistent separation between short and long stimulus trials . ( a ) Example frames for a 0 . 60 and 2 . 40 s stimulus trials ( top and bottom , respectively ) at stimulus offset . ( b ) Superimposed frames ( thresholded and background subtracted ) for each stimulus type ( same color conventions as in Figure 3a ) . ( c ) sum of all frames shown in ( b ) . ( d ) Same as in ( c ) but excluding the easiest short ( 0 . 60 s ) and long ( 2 . 40 s ) stimulus trials . The data show are from the same session presented in Figure 3b . DOI: http://dx . doi . org/10 . 7554/eLife . 11386 . 01110 . 7554/eLife . 11386 . 012Figure 3—figure supplement 2 . Behavior at the end of the neural analysis window did not explain the categorization performance of neural populations . ( a ) Neurometric ( orange data points ) or videometric ( purple data points ) logistic slope plotted against the psychometric slope for each session in the upper tercile with respect to simultaneously recorded population size . ( b ) Videometric slope plotted against the psychometric slope where the videometric curve was built using image frames taken at the time that animals expressed their choice . DOI: http://dx . doi . org/10 . 7554/eLife . 11386 . 01210 . 7554/eLife . 11386 . 013Figure 3—figure supplement 3 . Population state at interval offset can explain behavioral performance . ( a ) Psychometric curves constructed from trials separated according to whether the population state at stimulus offset had advanced more or less along the mean trajectory for rats Edgar ( top ) and Fernando ( bottom ) ( b ) Distributions of projection on normalized mean trajectory for all trials for each stimulus are shown for rats Edgar ( left ) and Fernando ( right ) . The equal area bins shown correspond to the groups of trials used for constructing the three psychometric curves shown in ( a ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11386 . 013 It has been previously reported that duration judgments could be predicted by animals’ ongoing behavior during the interval period ( Gouvêa et al . , 2014; Matthews and Lerer , 1987; Killeen and Fetterman , 1988; Fetterman et al . , 1998; Machado , 1997; Machado and Keen , 2003 ) . In addition , it is well known that striatal neurons can fire around movements ( Alexander and Crutcher , 1990; Jin and Costa , 2010 ) . Could the categorization performance of striatal ensembles reflect activity related to movements the animal might be making during the task ? To test to what degree ongoing behavior could explain the categorization performance of striatal neural activity , we applied an analogous classification analysis to video images taken of the animal just before interval offset ( see Materials and methods ) . We found that our ability to categorize intervals using video frames was consistently poorer as compared to neural data collected at the analogous time periods during the task ( Figure 3b , inset in Figure 3c , Video 1 , Figure 3—figure supplement 1 , Figure 3—figure supplement 2a ) . In contrast , we were able to categorize stimuli as well as the animal using video frames taken at the point when animals expressed their choice at one of the reward ports ( Figure 3—figure supplement 2b ) . Furthermore , movement related responses in the striatum are known to occur both pre- and post-movement onset , much later than in other motor areas such as pre-motor and motor cortex ( Alexander and Crutcher , 1990 ) . Thus , if purely movement-related activity were responsible for the categorization performance of striatal ensembles , we would expect ensemble performance to display a similar time course to that of video frames . Applying the same analyses at multiple points in time ranging from 500 ms preceding to 500 ms following stimulus offset revealed a strikingly different profile of categorization performance for video frames as compared to neural ensembles ( Figure 3d–e ) . Specifically , the time course of duration categorization by neural ensembles was best correlated with the duration categorization by video frames when using spikes collected between 400 ms and 200 ms preceding a reference video frame . These indicate that the categorization performance of striatal neurons was not simply related to the immediate sensorimotor state of the animal , and instead likely reflects that striatal neurons encode an internal neural representation of the state of animals’ categorical decisions . 10 . 7554/eLife . 11386 . 014Video 1 . Video clips from the entire stimulus period do not show a consistent separation between short and long stimulus trials . Superimposed video clips ( thresholded and background subtracted ) for each stimulus type and for the corresponding stimulus duration ( same color conventions as in Figure 3a and Figure 3—figure supplement 1 ) . Red stop marks signal the end of each video and the corresponding stimulus duration . DOI: http://dx . doi . org/10 . 7554/eLife . 11386 . 014 We have shown thus far that categorical information about interval duration contained in the firing of striatal populations at the time of stimulus offset can explain the precision of animals’ judgments about duration . However , in the task employed here , categorical judgments must be derived from a continuously evolving decision variable that represents how much time has elapsed since the onset of the stimulus . As indicated by the diversity of firing patterns ( Figure 2d ) , the state of population activity evolved continuously during interval stimuli ( Figure 3g , Figure 4a , Figure 3—figure supplement 3b ) , a feature not captured by binary classification . Might trial to trial variations in population state predict the apparent speed of animals’ internal representation of elapsed time ? To test this possibility , we performed two additional analyses . 10 . 7554/eLife . 11386 . 015Figure 4 . Smoothly changing population state encodes elapsing time in accordance with perceptual report for a long stimulus . ( a ) Low dimensional representation of population state during entire interval period of correct trials . Line colors indicate interval duration ( warmer colors are longer intervals , as in Figure 3a ) . Dots are placed at the interval offset end , and their color indicates judgment ( blue: short; red: long ) . ( b–g ) Population state and decoded time for a single long , near boundary stimulus interval ( 1 . 62 s ) . ( b ) Yellow curve is same as in ( a ) . Red dots are 6 time points evenly spaced between interval onset and offset . Blue dots are projections of population state during short judgment trials . Grey lines link population states at equivalent time points . ( c ) Average cumulative distance travelled in full neural space along trajectory represented in ( b ) on long versus short judgment trials . ( d ) Posterior probability of time given population state at the time points indicated in ( b ) , averaged within trials of each judgment type . ( e , f ) Same as ( d ) for the entire interval period . ( g ) Difference between posteriors for long and short judgment trials . Arrowheads indicate same time points used in ( b , d ) . n = 433 neurons from 3 rats . See Figure 4—figure supplement 1 for a different near boundary stimulus , and Figure 4—figure supplement 2 for data from individual subjects . DOI: http://dx . doi . org/10 . 7554/eLife . 11386 . 01510 . 7554/eLife . 11386 . 016Figure 4—figure supplement 1 . Smoothly changing population state encodes elapsing time in accordance with perceptual report for a short stimulus . ( a ) Low dimensional representation of population state during entire interval period of correct trials . Line colors indicate interval duration ( same color code as in Figures 3 and 4 ) . Dots are placed at the interval offset end , and their color indicates judgment ( blue: short; red: long ) . ( b-g ) Population state and decoded time for a single short , near boundary stimulus interval ( 1 . 38 s ) . ( b ) Green curve is the population state trajectory for long judgment trials . Red dots are 6 time points evenly spaced between interval onset and offset . Blue dots are projections of population state during short judgment trials . Grey lines link population states at equivalent time points . ( c ) Average cumulative distance travelled in full neural space along trajectory represented in ( b ) on long versus short judgment trials . ( d ) Posterior probability of time given population state at the time points indicated in ( b ) , averaged within trials of each judgment type . ( e , f ) Same as ( d ) for the entire interval period . ( g ) Difference between posteriors for long and short judgment trials . Arrowheads indicate same time points used in ( b , d ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11386 . 01610 . 7554/eLife . 11386 . 017Figure 4—figure supplement 2 . Single subjects show smoothly changing population states that encode elapsing time in accordance with perceptual report . ( a , e , i ) Low dimensional representation of population state during entire interval period of correct trials of rats Bertrand ( a ) , Edgar ( e ) and Fernando ( i ) . Line colors indicate interval duration ( warmer colors are longer intervals , as in Figures 3 and 4 ) . Dots are placed at the interval offset end , and their color indicate choice ( blue: short; red: long ) . ( b , f , j ) Yellow/green line is same as in ( a , e , i ) for a single near boundary stimulus interval ( 1 . 62/1 . 38 s; stimulus of highest choice variance for each subject ) . Red dots are 6 time points evenly spaced between interval onset and offset . Blue dots are projections of population state during short judgment trials . Grey lines link population states at equivalent time points . ( c , g , k ) Average cumulative distance travelled in full neural space along trajectory represented in ( b , f , j ) on long versus short judgment trials . ( d , h , l ) Difference between posteriors for long and short judgment trials for rats Bertrand ( d ) , Edgar ( h ) and Fernando ( l ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11386 . 017 First , we projected the state of simultaneously recorded neuronal populations at stimulus offset in individual trials onto the mean trajectory within each session , noted the fraction of the mean trajectory traversed for each trial , and pooled the data for each stimulus over all sessions within a given subject . Indeed , when population state at stimulus offset had advanced relatively more or less along the mean trajectory , animals were more likely to judge intervals as long or short respectively ( Figure 3f–g , Figure 3—figure supplement 3 ) . This effect was observed most consistently for interval stimuli that were closer to the category boundary , and thus variations in projected population state led to horizontal shifts in the psychometric curves ( see Materials and methods ) . These data are consistent with striatal population state encoding a representation of elapsed time that is used by animals to determine categorical judgments . Indeed , such a pattern of population activity has been proposed as a suitable neural code for elapsing time ( Buonomano , 2014; Buonomano and Merzenich , 1995 ) . However , if such a representation encodes elapsed time , and not only subjects’ judgments in this task , neural activity should continuously traverse a non-repeating trajectory in state space in a manner that predicts duration judgments during presentation of particular stimuli . Indeed , even in a low dimensional projection of population activity , we found that network state ran ahead or behind depending on whether the animal judged a near boundary stimulus as long or short ( Figure 4b–c , Figure 4—figure supplement 1b , c , Figure 4—figure supplement 2b , f , j ) . The correspondence between population trajectory and duration judgments further suggests that striatal dynamics may form an internal representation of elapsed time that informed categorical decisions about duration . To directly test this hypothesis , we focused on stimuli near the category boundary and decoded time from the population using a naive Bayes decoder and asked whether such a representation correlated with animals’ judgments , exhibiting choice probability ( Britten et al . , 1996 ) . We found that decoded estimates of time ran faster or slower when animals judged a given stimulus as long or short , respectively ( Figure 4d–g , Figure 4—figure supplement 1d-g , cross validated naive Bayes decoder; see Materials and methods ) . This indicates that striatal activity provides information about elapsing time , the continuously varying decision variable necessary to inform judgments in the task . Furthermore , if this information were read out and used to guide judgments , those judgments would match those of the rats . If the striatal activity we describe above directly supported task performance , manipulating the striatum should modify duration judgments . To test whether this was the case , we bilaterally injected the GABAa receptor agonist muscimol ( see Materials and methods ) . As a result , the duration sensitivity of animals’ judgments dropped significantly as compared to interleaved saline control sessions ( Figure 1c; psychometric slope on saline sessions = [1 . 03 1 . 20] vs on muscimol sessions = [0 . 28 0 . 67]; 95% confidence intervals ) , yet animals otherwise performed normally . These results , by demonstrating that duration categorization in this task was dependent on a normally functioning striatum , suggest that the neural signals we observed directly supported duration judgments . However , the possibility that muscimol infusions changed other functions important for task performance such as reward processing or memory for the mapping between time and choice can not be ruled out .
Attempts to understand the neural mechanisms of time estimation have begun to focus on continuously evolving population dynamics as a general mechanism for time encoding across the brain ( MacDonald et al . , 2011; Mauk and Buonomano , 2004; Buonomano , 2014; Buonomano and Merzenich , 1995; Gershman et al . , 2013 ) . According to this view , time may be encoded by any reproducible pattern of activity across a population of neurons for as long as the pattern is continuously changing and non-repeating . However , no study to date has directly compared the speed of such “population clocks” with the duration judgments of the behaving subjects in which they are found . We show that as rats judged the duration of interval stimuli , striatal neurons displayed dynamics in firing rate that contained information about elapsed time . Furthermore , this information was sufficient to account for the animals’ perceptual decisions , and was not accompanied by systematic differences in outwardly expressed behavior over time . Combined with the observation that striatal inactivation caused a specific decrement in timing performance , these data suggest that striatal dynamics form a central neural representation of time that guides animals’ decisions about duration . Such a coding mechanism in the striatum is well situated to inform the appropriate selection of actions through downstream circuitry involving the globus pallidus , substantia nigra , and various extrinsic connections between the basal ganglia and brainstem , thalamic , and cortical motor areas ( Steiner and Tseng , 2010 ) . An intriguing question for future studies is how the striatal dynamics we observed during the interval discrimination task are generated . Neurons in multiple cortical layers spread across the entire cortical mantle , as well as thalamic , pallidal , and neuromodulatory populations provide input to the striatum . While time coding has been assessed in some of these populations , a careful analysis of simultaneously recorded populations that might uncover causal relationships between signals in multiple brain areas has not been carried out . Furthermore , local striatal circuitry may also play a role in shaping dynamics . However , the coding properties tested here could be tested in other brain areas where timing signals have been identified such as the hippocampus ( Pastalkova et al . , 2008; MacDonald et al . , 2011 ) , medial prefrontal ( Kim et al . , 2013; Xu et al . , 2014 ) , parietal ( Leon and Shadlen , 2003; Janssen and Shadlen , 2005 ) and motor ( Lebedev et al . , 2008 ) cortices , and the cerebellum ( Mauk and Buonomano , 2004 ) , among others . By comparing the signals recorded simultaneously in multiple brain areas during time estimation tasks , it should be possible to identify signatures of functional interaction between brain areas where they exist . Such an approach promises to elucidate where and how time information encoded at the population level is used by the brain to support the myriad time-dependent functions we and other organisms rely on for survival .
Six male Long-Evans hooded rats ( Rattus norvegicus ) between the ages of 6 and 24 months were used for this study . Three rats were used for neural recordings and three rats for pharmacological manipulations . All experiments were in accordance with the European Union Directive 86/609/EEC and approved by the Portuguese Veterinary General Board ( Direcção-Geral de Veterinária , project approval 014303 - 0420/000/000/2011 ) . Rats were trained to perform a previously described two-alternative forced-choice timing task ( Gouvêa et al . , 2014 ) . Briefly , animals had to categorize time intervals as either long or short by making left/right choices . For each session the animals were placed in a custom made behavioral box containing 3 nose ports and a speaker . Each trial was self-initiated by entry into the central nose port and was followed by a pair of brief auditory tones ( square pulses at 7 , 500 Hz , 150 ms ) separated by an interval selected randomly out of 8 possible durations ( 0 . 6 , 1 . 05 , 1 . 26 , 1 . 38 , 1 . 62 , 1 . 74 , 1 . 95 and 2 . 4 s ) . Judgments were reported at two laterally located nose ports . Left responses were reinforced with a drop of water ( solenoid valves , Lee Company ) after intervals longer than 1 . 5 seconds , and right responses otherwise . Incorrect responses were punished with a brief white noise burst ( 150 ms ) and a time out ( 10 s ) . High speed video ( 120 fps ) was collected from above during task performance . Psychometric functions were fitted using the following two-parameter logistic function f ( x ) =11+e-b ( x-c ) where b controls the slope and c is the inflection point of the curve . Rats were implanted with 32-channel tungsten microwire moveable array bundles ( Figure 2—figure supplement 1a , Innovative Neurophysiology ) under isoflurane anaesthesia . All recordings ( Figure 2—figure supplement 1 ) targeted dorsal striatum with coordinates centred at +0 . 2 mm AP and ± 3 mm ML ( rat Bertrand ) , and +0 . 84 mm AP and ± 2 . 5 mm ML ( rats Edgar and Fernando ) , from Bregma . Rats were given a week of post-surgical recovery and array placements were confirmed with histology ( Figure 2—figure supplement 1c ) . Neural signals were recorded at 30 kHz during behavior , amplified and band-pass filtered at 250–750 Hz ( Cerebus - Blackrock Microsystems ) . Each independent bundle was moved 50-100 μm after every recording session to ensure that independent neural populations were sampled across recording sessions . Waveforms corresponding to action potentials from single neurons were sorted offline using principal component analysis ( PCA ) ( offline sorter , Plexon ) . All remaining analysis were run in Matlab ( Mathworks ) software . We selected all isolated units with a mean session firing rate >0 . 5 Hz and from sessions with >70% correct performance ( averaged across all stimuli ) and a minimum of 250 trials ( n=433 cells , 37 recording sessions , 3 animals; rat Bertrand: 136 units , 10 sessions; rat Edgar: 163 units , 9 sessions; rat Fernando: 134 units , 18 sessions ) . The general result was found in all subjects . Sample size was not computed during study design . To build PSTHs , spikes were counted in 2-ms bins and convolved with a Gaussian kernel with 25-bin standard deviation . PSTHs in Figure 2d were ordered by angular position in the space formed by the first 2 principal components describing firing dynamics ( i . e . , dimensions are all time bins within interval period , samples are each neuron’s mean PSTH ) . This method ( Geffen et al . , 2009 ) orders cells with respect to their dynamics while taking into consideration the full response profile over the relevant temporal window , and not just a single response feature such as peak response time . We implanted 3-mm 20-gauge stainless steel guide cannulae ( Bilaney ) bilaterally into the striatum of 3 rats [+0 . 84 mm anterior-posterior ( AP ) , ± 2 . 5 mm medial-lateral ( ML ) , from Bregma , and -3mm dorsal-ventral ( DV , from cortex surface ) under isoflurane anesthesia . After one week of post-surgical recovery and 4 days of training , rats were injected with either vehicle ( saline , PBS 1x ) or muscimol ( GABA-A agonist , 100 mg/L ( rats Albert and Yuri ) and 300 mg/L ( rat Zack ) , SigmaTM ) solutions in four alternate days . Two 1-μL syringes ( Hamilton ) , attached to an injection pump ( Harvard Apparatus ) through 20-gauge internal cannulae that extended 1 . 5 mm bellow the guide cannulae , injected 0 . 6 μL of solution per site during 2 . 5 min . The internal cannulae were left in place for an additional 1 . 5 min and the rats were given a 45-min recovery period in their home-cage before starting the task . Cannula placements were confirmed by histology ( Figure 1—figure supplement 1 ) . The general result was found in all sessions of all subjects . Sample size was not computed during study design . We counted spikes during the last 500 ms of the stimulus period , and built two separate spike count distributions for short and long judgment trials . Next , we used a ROC analysis to measure the separation between distributions ( 95% bootstrap confidence interval , 1000 iterations ) . We then transformed the area under the ROC curve ( auROC ∈ [0 , 1] ) into a preference index ( k = 2*auROC - 1; k ∈ [−1 , 1] ) . We adopted the convention that neurons with positive preference indices fired preferentially for long stimuli ( Figure 2e ) . We refer to the vector describing instantaneous firing rates ( measured within 500-ms wide , 10-ms apart , overlapping time bins ) across a population of neurons as the population state . The population state vector is a high dimensional variable ( i . e . , it has as many dimensions as neurons ) . With the purpose of visualizing population state in 2d plots , we employed standard dimensionality reduction techniques . In Figure 3a , we chose to represent in the abscissa a direction that emphasizes the separability between short and long stimulus trials ( i . e . , the direction that maximizes variance between groups while minimizing variance within groups; Fisher’s linear discriminant; see below ) , and in the ordinate the axis of maximal variance that is also orthogonal to the abscissa ( i . e . first principal component calculated in the null space of the linear discriminant ) . In Figure 4a–b , population state is represented in the space formed by the first 2 principal components describing population state , calculated during presentation of the interval for which choice variance is maximal ( i . e . dimensions are neurons , samples are averaged spike counts for the time bins within that interval ) . For each trial in a session we characterized neural population state as a vector r = ( r1 , r2 , . . . , rN ) , where rn is the number of spikes fired by neuron n ∈ [1 , N] within the last 500 ms of the interval period in that trial . Next , for all trials but one from each session ( training set; leave-one-out cross-validation procedure ) , we found the linear discriminant that best separated population state vectors according to whether they came from long or short interval trials ( Fisher’s linear discriminant analysis , LDA ) . The linear discriminant is given by w =argmaxwTSBwwTSWw=SW-1 ( μ1-μ2 ) where w is the vector of coefficients for the linear discriminant , SB is the between class covariance , SW the within class covariance and µ1 and µ2 are the means of all points in class 1 and class 2 respectively . A threshold placed along the linear discriminant was then used as a decision rule applied to neural data from the remaining trial ( test set ) . Figure 3a depicts population vectors from an example session ( projection 1: linear discriminant , no cross-validation; projection 2: first principal component of the orthogonal subspace; black line: decision rule ) . We iterated over this procedure until all trials had been tested , thus obtaining for each trial a ‘neural duration judgment’ . In analogy with behavioral judgments , we used two parameter logistic fits to obtain a quantitative description of the performance of simultaneously recorded neurons in categorizing stimuli -the neurometric function ( Figure 3b , orange curve ) . Full session videos ( 256x192 pixels resolution ) were cut into 3-s long clips with Bonsai ( Lopes et al . , 2015 ) . Individual frames from approximately 75 ms before interval offset were used for this analysis ( Figure 3—figure supplement 1 ) . This buffer was added to ensure that all frames used preceded stimulus offset . Images were first represented as vectors composed of individual pixel luminance values . Given that image sequences tend to lie on curved low dimensional manifolds in pixel space ( Pless , 2003 ) , any slight differences in behavioral state reflected in images collected at the offset of short and long interval categories are not necessarily expected to be linearly separable . Thus , we employed isomap ( Tenenbaum and Silva , 2000 ) , a non-linear dimensionality reduction method , to obtain an information rich yet low dimensional representation of animals’ ongoing behavior . This approach has the advantage over tracking methods that it does not make assumptions as to what part of the animals’ movements might provide information about stimulus category . The neighborhood size , used to compute the shortest paths between data points , was set to 25 frames to minimize , on average , the dimensionality at which the reconstruction error elbow occurred . In analogy with the neurometric curves , for each stimulus type , we then trained a linear discriminant ( leave-one-out cross-validation procedure ) to classify frames into those that were recorded during trials where a ‘short’ or ‘long’ stimulus interval was presented . The classification was performed in the reduced space determined by isomap . As a positive control for the method , we repeated the same analysis for frames captured at the moment animals expressed their judgment by inserting their snout at one of the two choice ports ( Figure 3—figure supplement 2 ) . Here , the neighborhood size was chosen to be the minimum for which all frames ( from a single session ) could be included in a single embedding . This analysis was done for all usable videos ( 8 out of 11 ) of sessions in the upper tercile with regard to population size . To compare how the decoding performance using neural and video data evolved over time , the classification analyses described in Neurometric curves and Videometric curves was performed every 100 ms within a one second window centered around stimulus offset . Video frames at the each time point and neural data in a 200 ms time bin terminating at each time point were used for the analysis . This generated neural and video classification curves that described the ability of simultaneously recorded neural ensembles and video frames to correctly classify interval stimuli as long or short ( Figure 3d ) . To determine the relative timing of classification ability in neural ensembles and behavior , we regressed the neural classification curve against the video classification curve for shifts ranging from −300 ms to 300 ms in 100 ms steps ( Figure 3e ) . We projected neural activity ( of populations composed of simultaneously recorded neurons ) on individual trials in high dimensional neural space onto the mean trajectory of those neurons during the delay period for correct trials . Neural activity was defined as the vector of firing rates of the population obtained by convolving spike trains using a causal kernel given by gamma density function with parameters θ = 100 ms and k = 2 . We normalized these projections by the length of the mean trajectory of that group of neurons for the longest interval . Pooling normalized projections over all sessions for each animal , we plotted , for each stimulus , distributions of normalized projections at interval offset . To test whether distance traversed along the mean trajectory is predictive of animals' perceptual report , we separated the distribution of pooled projections for each stimulus into 3 bins . Psychometric curves were constructed using trials from each bin . To quantify the key differences between each of these psychometric curves , we performed model comparison using the following 4 parameter logistic function f ( x ) =d+a-d1+e-b ( x-c ) where b controls the slope , c is the inflection point and a and d are the maximum and minimum values of the curve respectively . For two of three animals ( Bertrand and Edgar ) , the model that best accounted for the differences between the three curves ( based on Bayesian Information Criterion ( BIC ) scores ) was one with only horizontal shifts between the curves . In the third animal ( Fernando ) , the model that best fit the data was one in which the fit to the three curves differed in both horizontal shift and slope . A trial's projection on the mean trajectory can be interpreted as a method for decoding time from neural state . Hence , trials that are outliers in the distribution of projections on the mean could potentially correspond to trials where the animal was disengaged . To remove such trials we defined a fraction ( 60% ) of normalized trajectory around the mode of the distribution of pooled projections for each stimulus and excluded trials with projections outside this window . We decoded elapsed time from striatal population activity using a cross validated , flat prior naive Bayes decoder . For each neuron n ∈ [1 , N] , spike counts rn were observed in 500-ms wide , 10-ms apart , overlapping time bins within the interval period ( time referring to the right edge of the bin ) . For a given rn , the probability that the current time is t was estimated as the likelihood of observing rn spikes at time t: P ( t|rn ) ∝P ( rn|t ) To obtain the likelihood term P ( rn | t ) , we estimated the joint distribution P ( rn , t ) by computing , for each time bin , a weighted histogram of spike counts across all correct trials . For trials in which stimulus interval i was presented , spike counts contributed to the histogram with weight wi defined as the normalized choice variance associated with that interval , wi=P ( CS|i ) P ( CL|i ) where CS and CL indicate short and long choices respectively . As a result , near boundary interval trials had a greater influence on the estimate of the joint distribution . Histograms were then smoothed using local linear regression ( lowess ) and normalized to unit area . When decoding from correct trials , leave-one-out cross validation was implemented by computing the joint distribution from all correct trials but the decoded one; incorrect trials were decoded using an estimate of the joint distribution computed from all correct trials . Multi-session population state vectors r = ( r1 , r2 , . . . , rN ) were obtained by concatenating together data from trials of same stimulus and choice type . By assuming statistical independence between spike counts of different neurons in r , we could compute population estimates of t as the product of single neuron estimates: P ( t|r ) ∝∏n=1NP ( rn|t ) Data presented is the average over 100 random concatenations . | You know someone is a good cook from their rice - grains must be well cooked , but not to the point of being mushy . Despite consistently using the same pot and stove , we , however , will sometimes overcook it . It is as if our inner sense of time itself is variable . What is it about the brain that explains this variability in time estimation and indeed our ability to estimate time in the first place ? One issue the brain must confront in order to estimate time is that individual brain cells typically fire in bursts that last for tens of milliseconds . So how does the brain use this short-lived activity to track minutes and hours ? One possibility is that individual neurons in a given brain region are programmed to fire at different points in time . The overall firing pattern of a group of neurons will therefore change in a predictable way as time passes . Gouvêa , Monteiro et al . found such predictably changing patterns of activity in the striatum of rats trained to estimate and categorize the duration of time intervals as longer or shorter than 1 . 5 seconds . Interestingly , when rats mistakenly categorized a short interval as a long one , population activity had travelled farther down its path than it would normally ( and vice-versa for long intervals incorrectly categorized as short ) , suggesting that variability in subjective estimates of the passage of time might arise from variability in the speed of a changing pattern of activity across groups of neurons . As further evidence for the involvement of the striatum , inactivating the structure impaired the rats’ ability to correctly classify even the longest and shortest interval durations . The next challenge is to determine exactly how the striatum generates these time-keeping signals , at which stage variability originates , and how the brain regions that the striatum signals to use them to control an animal’s behavior . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Striatal dynamics explain duration judgments |
In many species , males can make rapid adjustments to ejaculate performance in response to sperm competition risk; however , the mechanisms behind these changes are not understood . Here , we manipulate male social status in an externally fertilising fish , chinook salmon ( Oncorhynchus tshawytscha ) , and find that in less than 48 hr , males can upregulate sperm velocity when faced with an increased risk of sperm competition . Using a series of in vitro sperm manipulation and competition experiments , we show that rapid changes in sperm velocity are mediated by seminal fluid and the effect of seminal fluid on sperm velocity directly impacts paternity share and therefore reproductive success . These combined findings , completely consistent with sperm competition theory , provide unequivocal evidence that sperm competition risk drives plastic adjustment of ejaculate quality , that seminal fluid harbours the mechanism for the rapid adjustment of sperm velocity and that fitness benefits accrue to males from such adjustment .
Sperm competition ( Parker , 1970 ) occurs commonly across many invertebrate and vertebrate taxa and is a potent evolutionary force influencing male reproductive biology ( Birkhead and Møller , 1998; Birkhead and Pizzari , 2002; Simmons and Fitzpatrick , 2012 ) . Sperm competition theory predicts that males will trade-off between energy expended making high-quality ejaculates with obtaining mating opportunities and that males will invest differentially in ejaculates with respect to sperm competition risk ( Parker , 1990; Parker et al . , 1997; Parker , 1998; Wedell et al . , 2002; Birkhead et al . , 2009; Parker and Pizzari , 2010 ) . In agreement with these predictions , males of many species can make rapid adjustments to ejaculate quality within days ( Rudolfsen et al . , 2006; Pizzari et al . , 2007; Thomas and Simmons , 2007; Gasparini et al . , 2009; Smith and Ryan , 2011 ) , hours ( Cornwallis and Birkhead , 2007a ) and even minutes ( Kilgallon and Simmons , 2005; Joseph et al . , 2015 ) of exposure to a new social cue that signals changing sperm competition risk , such as the presence of a female , or a male competitor . For example , in fowl ( Gallus gallus ) , males of dominant social status strategically allocate sperm , ejaculating more and faster swimming sperm in initial copulations and to females of higher quality ( Pizzari et al . , 2003; Cornwallis and Birkhead , 2006; Cornwallis and Birkhead , 2007a; Cornwallis and Birkhead , 2007b ) , and alter their allocation strategy accordingly when changing social status ( Cornwallis and Birkhead , 2007a ) . While males of several vertebrate species ranging from fish ( Rudolfsen et al . , 2006; Gasparini et al . , 2009; Smith and Ryan , 2011 ) to humans ( Kilgallon and Simmons , 2005; Joseph et al . , 2015 ) can strategically alter the quality of their ejaculate in response to social cues , the mechanism behind such rapid adjustments is as yet unknown . A promising candidate mechanism for rapid adjustment of sperm velocity may be found in the non-sperm component ( seminal fluid and its constituents ) of the ejaculate , particularly if such adjustments occur more rapidly than spermatogenesis ( Cameron et al . , 2007; Perry et al . , 2013; Fitzpatrick and Lüpold , 2014 ) . Seminal fluid is a complex medium containing a great diversity of molecules ( Poiani , 2006; Juyena and Stelletta , 2012 ) and is known to influence sperm velocity and motility in vertebrates ( Lahnsteiner et al . , 1998; Lahnsteiner et al . , 1996; Poiani , 2006; Locatello et al . , 2013; González-Cadavid et al . , 2014 ) . For example , research using an externally fertilising fish , the grass goby ( Zosterisessor ophiocephalus ) , compared males for which sperm competition strategy is determined by age/size and found large males that adopt a guarding strategy have a greater concentration of the seminal fluid glycoprotein mucin ( Scaggiante et al . , 1999 ) . Furthermore , by separating and recombining seminal fluid and sperm from different males , research using the same species found seminal fluid had a tactic-specific effect on sperm velocity , with seminal fluid from sneak males decreasing the velocity of rival guard male sperm and seminal fluid from guard males increasing the velocity of sneak male sperm ( Locatello et al . , 2013 ) . However , only one study to date has investigated the role that seminal fluid plays as a mediator of short-term plastic sperm performance in a vertebrate species using fowl and the results were inconsistent with theoretical expectation: Cornwallis and O'Connor , 2009 found that while ejaculates produced by male fowl that were allocated to females of higher quality contained faster sperm , seminal fluid from those ejaculates reduced the velocity of sperm from the same male allocated to females of lower quality . To be consistent with the prediction that seminal fluid mediates changes in sperm velocity , seminal fluid from ejaculates allocated to higher quality females should increase , not decrease the speed of sperm isolated from ejaculates allocated to lower quality females . Thus , although there is evidence that seminal fluid can influence sperm velocity , evidence that seminal fluid mediates the rapid plastic adjustment of an ejaculate’s motile performance consistent with theoretical expectation is lacking . We use an ideal model species , chinook salmon ( Oncorhynchus tshawytscha ) , to examine patterns of ejaculate plasticity in response to changes in male social status and the reproductive consequences of these changes . In salmonids , fertilisation occurs externally and sperm competition occurs in the majority of spawnings ( Berejikian et al . , 2010; Sørum et al . , 2011 ) . Male chinook salmon adopt Alternative Reproductive Tactics ( ARTs ) situationally , as ‘hooknose’ males fight to establish social dominance ( Esteve , 2005 ) . Only dominant males guard spawning females thus obtaining priority in mating position , while subdominant males that lose contests attempt to sneak fertilisations by invading spawning pairs and releasing their sperm ( Esteve , 2005 ) . The social status of male salmon is subject to change over the course of a spawning season; for example , in coho salmon ( O . kisutch ) , 22% of observed contests between hooknose males resulted in displacement of the previous dominant male ( Healey and Prince , 1998 ) . Therefore , in this mating system , females mate with multiple males in a dynamic social environment that results in intense levels of fluctuating sperm competition risk . Previous research has shown that when males engage in sperm competition , sperm swimming speed is the primary predictor of fertilisation success in chinook salmon ( Evans et al . , 2013; Rosengrave et al . , 2016 ) and other salmonids ( Gage et al . , 2004; Liljedal et al . , 2008; Egeland et al . , 2015 ) . Sperm competition theory therefore predicts subdominant males , which have greater sperm competition risk , will invest in ejaculates with faster swimming sperm than dominant males and males changing social status should adjust their investment accordingly ( Parker , 1990; Parker et al . , 1997; Parker , 1998; Wedell et al . , 2002; Birkhead et al . , 2009; Parker and Pizzari , 2010 ) . Indeed , several studies that experimentally manipulated social status using Arctic charr ( Salvelinus alpinus ) have found that subdominant males produce ejaculates with more sperm and faster swimming sperm than dominant males ( Liljedal and Folstad , 2003; Rudolfsen et al . , 2006; Vaz Serrano et al . , 2006; Haugland et al . , 2009 ) . Furthermore , Rudolfsen et al . ( 2006 ) demonstrated that following a social challenge , both sperm concentration and velocity decreased over a 4-day period compared with pre-trial levels in dominant males , and observed an increase in sperm concentration but no change in sperm velocity for subdominant males . However , Rudolfsen et al . ( 2006 ) did not evaluate male social status prior to the social challenge , so it is unknown if these males actually changed or simply retained the same status through the course of the experiment . Recent research shows that ejaculates from subdominant Arctic charr sire the same number of eggs when in competition with ejaculates from dominant males if their sperm were released after the average delay observed under natural conditions ( Egeland et al . , 2015 ) . These results suggest that salmonid males in disfavoured mating positions can compensate by producing more competitive ejaculates than dominant males , but whether males changing social status adjust their sperm velocity , and if such adjustments to ejaculates are mediated by sperm or non-sperm components of the ejaculate , is yet to be determined . Here , we use a comprehensive experimental approach to determine if changes in sperm velocity observed in response to an individual’s social position are the result of alterations to the gametes or to seminal fluid and if such responses actually alter a male’s reproductive success against a sperm competitor . Specifically , we examine whether ejaculate quality is phenotypically plastic in response to changes in sperm competition risk over 48 hr periods , using a two-stage challenge to manipulate social status ( Cornwallis and Birkhead , 2007a; Pizzari et al . , 2007 ) and collected ejaculates at each stage of the experiment . In the second stage , males either retained or were forced to change their social status , creating four social phenotypes with varying sperm competition risk ( Figure 1 ) . We found that subdominant males , which have greater sperm competition risk , invest more in both sperm concentration and sperm velocity compared to socially dominant males . Additionally , we find males that change from dominant to subdominant social status , thus elevated their sperm competition risk , increased their sperm velocity as predicted by sperm competition theory ( Parker , 1990; Parker et al . , 1997; Parker , 1998; Wedell et al . , 2002; Birkhead et al . , 2009; Parker and Pizzari , 2010 ) . We also separated sperm from seminal fluid and created reciprocal combinations both within and between rival males , finding that males can make rapid adjustments to sperm velocity by producing seminal fluid that enhances sperm function . We then used in vitro fertilisation trials and found the seminal fluid effects on sperm swimming speed influences male reproductive success under sperm competition . Our combined experimental results provide compelling evidence that seminal fluid is the mediator of rapid strategic adjustment of sperm velocity , thus bringing us a critical step closer to identifying the underlying molecular mechanism that enables plasticity of ejaculate performance in dynamic social environments .
Subdominant ( S ) males had on average faster swimming sperm ( Average Path Velocity , or VAP ) than dominant ( D ) males . This difference was not significant when social status was initially determined in stage 1 ( Table 1; Figure 2a ) but was significant for stage 2 ( Table 1; Figure 2b ) . Overall , there was considerable variation in sperm swimming speeds among males , accounted for by the random predictor ‘male identity’ that was significant in both stages ( stage 1: χ2 ( 1 ) =105 . 11 , p<0 . 001; stage 2: χ2 ( 1 ) =70 . 02 , p<0 . 001 ) . Additionally , sperm concentration was significantly higher in S than in D males in stage 1 ( Table 1; Figure 3a ) , but not stage 2 ( Table 1; Figure 3b ) . However , sperm concentration for males that remained subdominant ( SS ) was significantly higher than for those males that remained socially dominant ( DD ) in stage 2 ( Table 1 ) . There was a significant increase in mean VAP for males that changed from dominant to subdominant social status ( DS; Table 2; Figure 4 ) . Throughout the social status experiment , there were no other significant changes in either VAP or sperm concentration for males of the other social phenotypes ( Table 2; Figure 4 ) . There was also a significant overall interaction effect between social phenotype and experimental stage on VAP ( χ2 ( 3 ) =11 . 8 , p=0 . 008 ) , with a significant interaction effect found only for males changing from dominant to subdominant status ( DS; p=0 . 02 , 95% CI = 2 . 9–34 . 9 ) . We found no significant interaction effects between social phenotype and experimental stage on sperm concentration ( χ2 ( 3 ) =3 . 0 , p=0 . 385 ) . Within each dyad , the social status of the rival male was a significant predictor of the difference in VAP between focal male’s sperm incubated in their own seminal fluid and the focal male’s sperm incubated in their rival’s seminal fluid ( Table 3 ) . Seminal fluid from subdominant males increased the sperm swimming speed of sperm from dominant males , conversely , the seminal fluid from dominant males decreased sperm swimming speed of the sperm from subdominant males ( Figure 5 ) . However , rival’s social status was no longer significant ( Table 3 ) when the difference in VAP between the focal male control and rival male control was added as a fixed predictor to the model , for which a significant positive linear relationship was detected ( Table 3 ) , with sperm in the seminal fluid of a rival that had faster VAP increasing sperm velocity and sperm in the seminal fluid of a rival that had slower VAP decreasing sperm velocity relative to VAP in their own seminal fluid ( Figure 6 ) . Male social status was a significant predictor of the proportion of eggs sired ( Table 4 ) , with subdominant males siring a greater proportion ( 0 . 54 ± 0 . 08 95% CI , n = 21 ) than dominant males ( 0 . 46 ± 0 . 06 95% CI , n = 21 ) . The social status of the seminal fluid donor when seminal fluid was swapped between males was also a significant predictor of the proportion of eggs sired ( Table 4 ) , with sperm incubated in the seminal fluid of subdominant males siring a greater proportion ( 0 . 6 ± 0 . 09 95% CI , n = 21 ) of eggs than sperm incubated in the seminal fluid of dominant males ( 0 . 4 ± 0 . 09 95% CI , n = 21 ) . The difference in sperm velocity between competitors was also a significant predictor of the proportion of eggs sired in both unmanipulated ( Table 4 ) and recombined ejaculate seminal fluid treatments ( Table 4 ) . The change in relative sperm velocity between males within the same male-male-female combinations across seminal fluid treatments was a significant predictor of the change in the proportion of eggs sired by that male’s sperm across treatments ( Table 4; Figure 7 ) .
In this study , we experimentally manipulated social status to produce four social phenotypes with differing levels of sperm competition risk , and in accordance with sperm competition theory ( Parker , 1990; Parker et al . , 1997; Parker , 1998; Wedell et al . , 2002; Birkhead et al . , 2009; Parker and Pizzari , 2010 ) , found males with the highest risk of sperm competition produced ejaculates with both higher sperm concentration and faster swimming sperm . We also found males can make rapid adjustments to sperm velocity in a strategic response to changes in social position that signal increased sperm competition risk . While seminal fluid is often implicated to harbour the unknown mechanism behind plastic sperm performance in vertebrates ( Perry et al . , 2013; Fitzpatrick and Lüpold , 2014 ) , our combined results for the first time , unequivocally demonstrate that seminal fluid acts as a mediator of rapid strategic adjustment to sperm velocity . Furthermore , we demonstrate strategic adjustments of sperm velocity mediated by seminal fluid directly impact male fitness , highlighting the adaptive significance of plastic ejaculate performance . Sperm competition theory predicts that males should strategically adjust ejaculates in response to changing sperm competition risk ( Wedell et al . , 2002; Parker and Pizzari , 2010 ) . In chinook salmon , relative sperm velocity among males is the primary determinant of fertilisation success ( Evans et al . , 2013; Rosengrave et al . , 2016 ) . We show males forced to change from dominant to subdominant social status , and therefore exposed to increased sperm competition risk , responded by increasing the quality of their ejaculate , in this case sperm velocity , within 48 hr ( Figure 4 ) . While we predict that males forced to change from subdominant to dominant social status , therefore exposed to decreased sperm competition risk , would respond by decreasing their ejaculate quality , we did not see a significant change in sperm velocity for these males . However , subdominant males that later became dominant had a relatively low mean sperm velocity that was more similar to dominant males than those from the other subdominant phenotype in the first stage of the experiment ( Figure 4 ) . In this case , these subdominant males may have attempted to adopt a guarding tactic even after losing in the first social challenge , as males that lose contests can either sneak or fight for dominance elsewhere ( Esteve , 2005 ) . Males should also strategically adjust sperm concentration in response to changing sperm competition risk ( Wedell et al . , 2002; Parker and Pizzari , 2010 ) . Accordingly , we found subdominant males produced ejaculates with greater sperm concentration than dominant males . However , our results show that there was no significant increase in sperm concentration for any of the social phenotypes over a 48-hr period . The exact time taken for spermatogenesis in salmonids is unknown; however , the process almost certainly takes more than 48 hr ( Billard , 1983a; Billard , 1983b; Schulz et al . , 2010 ) . Therefore , these results suggest that the observed changes in sperm velocity are mediated by a component of the ejaculate that modifies the competitiveness of existing sperm , rather than simply via the production of new sperm . Our results clearly demonstrate the observed plasticity of sperm velocity in chinook salmon , a key determinant of fertilisation success in several vertebrate species ( Birkhead et al . , 1999; Malo et al . , 2005; Gasparini et al . , 2010; Boschetto et al . , 2011 ) including salmonids ( Gage et al . , 2004; Liljedal et al . , 2008; Evans et al . , 2013; Egeland et al . , 2015; Rosengrave et al . , 2016 ) , is mediated by seminal fluid . We found sperm from the same male , when incubated in seminal fluid from different males , had significantly different sperm velocities , and the direction of this effect could be predicted by social status . For example , when sperm from dominant males were incubated in seminal fluid from subdominant males we found that on average their sperm velocity increased compared to the baseline measures in their own seminal fluid , and found the opposite effect when sperm from subdominant males were incubated in seminal fluid from dominant males ( Figure 5 ) . Contrary to Cornwallis and O'Connor , 2009 , for which seminal fluid from higher quality ejaculates decreased the velocity of sperm from lower quality ejaculates in fowl , our findings are consistent with the prediction that seminal fluid from ejaculates with faster swimming sperm will enhance the speed of sperm from ejaculates with slower sperm . The disparity between our findings and those in fowl ( Cornwallis and O'Connor , 2009 ) possibly reflect differences in the reproductive biology of these species; including internal and external modes of fertilisation and differences in the structure and formation of social hierarchies and associated sperm competition risk . Ejaculate allocation in fowl is also influenced by factors other than sperm competition risk , including female quality and the probability of future mating opportunities ( Pizzari et al . , 2003; Cornwallis and Birkhead , 2006; Cornwallis and Birkhead , 2007a; Cornwallis and Birkhead , 2007b ) ; whether such factors influence ejaculate allocation strategies in salmonids is unknown . It is also possible that seminal fluid in fowl has evolved to interact with sperm from rivals , as observed in some insect species ( den Boer et al . , 2010 ) and reported for the grass goby ( Zosterisessor ophiocephalus ) ( Locatello et al . , 2013 ) . Fertilisation occurs rapidly in salmonids , with the majority of eggs fertilised within 10 s post ejaculation ( Hoysak and Liley , 2001; Liley et al . , 2002; Yeates et al . , 2007 ) . Such rapid time frames may allow for little interaction between seminal fluid and sperm from different males during spawning . This is supported by research using Arctic charr that found the activation of sperm with a solution containing seminal fluid from another male had no effect on sperm velocity ( Rudolfsen et al . , 2015 ) . However , a recent experiment that separated and recombined ejaculates from precocious chinook salmon males ( obligate sneakers ) and adult hooknose males report similar results to those found in the grass goby , with seminal fluid from precocious males significantly decreasing the velocity of hooknose male sperm ( Lewis and Pitcher , 2017 ) . Our results suggest chinook salmon seminal fluid has not evolved a targeted effect on sperm from males adopting a different tactic within the same age/size class , as regardless of social status , males that have faster recorded sperm velocities produced seminal fluid that increases the velocity of sperm from other males with slower speeds , and likewise males with slower sperm velocity produced seminal fluid that decreases the velocity of sperm from males with faster speeds ( Figure 6 ) . In addition to demonstrating that seminal fluid influences sperm competitiveness , our in vitro sperm competition trials show the influence seminal fluid has on sperm velocity translates to having an effect on male fitness . We measured the fertilisation success within the same male x male x female combinations across trials , and compared those males across unmanipulated and recombined ejaculate treatments , finding changes in the relative sperm velocity between competitors were significantly correlated with the change in the proportion of eggs sired by each male ( Figure 7 ) . That is , the change in sperm velocity due to the seminal fluid in which sperm were incubated had a significant influence on the proportion of eggs sired by those sperm , in some cases completely reversing the ‘winner’ of sperm competition within the same male-female group . We now need further investigation to determine the component of seminal fluid that is strategically adjusted by males in response to sperm competition risk . Previous studies have found that natural variation in several seminal fluid metrics was not correlated with sperm velocity in chinook salmon , including pH , osmolality and ion composition ( Rosengrave et al . , 2009a; Flannery et al . , 2013 ) . It is possible that seminal fluid contains different levels of available nutrients therefore fuelling differential energy production in sperm . In the short term following activation of motility in salmonids , sperm utilise ATP as the energy source for flagellar movement ( Christen et al . , 1987 ) using both stored ATP reserves and increasing ATP production significantly via aerobic respiration ( Lahnsteiner et al . , 1993 , Lahnsteiner et al . , 1999 ) . Sperm ATP levels have been positively correlated with sperm velocity ( Lahnsteiner et al . , 1998; Bencic et al . , 1999; Burness et al . , 2004 ) and fertilisation success ( Zilli et al . , 2004; Vladić et al . , 2010 ) in external fertilisers . Exposure to different levels of exogenous nutrients in seminal fluid while sperm are immotile in the testis may influence energy metabolism , for example altering available energy reserves or stored nutrient reserves , influencing sperm velocity post activation ( Lahnsteiner et al . , 1999 ) . Alternatively , seminal fluid may contain peptide or RNA signalling molecules , that alter sperm behaviour . For example , chemotaxis in several marine invertebrates is controlled by signalling pathways that are initiated by chemoattractant peptides released by ova ( Kaupp et al . , 2003; Darszon et al . , 2008; Evans and Sherman , 2013 ) . Evidence is also accruing that proteins and RNAs in seminal fluid exosomes may play critical roles in regulating sperm development and fertilisation ( Vojtech et al . , 2014; Jodar et al . , 2016 ) . Several Seminal Fluid Proteins ( SFPs ) have been associated with sperm velocity in vertebrate species ( Lahnsteiner et al . , 1996 , 1998; Poiani , 2006 ) and are therefore likely candidates for modifying rapid adjustment of sperm velocity ( Simmons and Fitzpatrick , 2012 ) . Differences in SFP composition have been documented among males adopting different reproductive tactics in externally fertilising fish ( Scaggiante et al . , 1999; Gombar et al . , 2017 ) . Additionally , a growing body of empirical work has demonstrated that males can tailor SFP composition in response to sperm competition risk ( Wigby et al . , 2009; Fedorka et al . , 2011; Ramm et al . , 2015; Simmons and Lovegrove , 2017 ) and the mating status of females ( Sirot et al . , 2011 ) . The role of SFPs in sperm competition , with the exception for some insect species ( den Boer et al . , 2010; Avila et al . , 2011 ) and specific proteins in mammals ( Ramm et al . , 2008 ) , is generally poorly understood . While the activity of SFPs associated with energy metabolism and respiration have been correlated with sperm velocity in a Cyprinid species ( Lahnsteiner et al . , 1996 ) and rainbow trout ( O . mykiss ) ( Lahnsteiner et al . , 1998 ) , total protein concentration as well as the activity of lactate dehydrogenase , anti-trypsin and superoxide dismutase enzymes were not correlated with sperm velocity in chinook salmon ( Flannery et al . , 2013 ) . However , these SFPs represents only a small fraction of the enzymatic activity likely to occur in fish seminal fluid ( Gombar et al . , 2017; Nynca et al . , 2014 ) . The critical next step in determining the molecular mechanism ( s ) involved will be to link variation in seminal fluid components to sperm velocity , and confirm these results experimentally . In conclusion , as predicted by sperm competition theory ( Parker , 1990; Parker et al . , 1997; Parker , 1998; Wedell et al . , 2002; Birkhead et al . , 2009; Parker and Pizzari , 2010 ) , we find male chinook salmon can make rapid adjustment to sperm velocity in response to social cues that signal changing sperm competition risk and such changes have a significant impact on the outcome of sperm competition and therefore male fitness . We further demonstrate that seminal fluid , even in a species with external fertilisation , plays a key role in mediating the strategic rapid adjustment of sperm velocity and for the first time provide strong evidence the mechanism behind plasticity in sperm velocity lies within the non-sperm component of the ejaculate . Our results support plastic adjustment of ejaculate quality in response to changing sperm competition risk is an effective evolutionary strategy in systems with dynamic social environments and we show seminal fluid mediates such adjustments .
Wild chinook salmon were caught during their annual spawning runs in a trap located on the Kaiapoi River , a tributary of the Waimakariri River system , Canterbury , New Zealand ( Unwin et al . , 2000 ) . We studied a total of 17 sexually mature 3-year-old females and 44 sexually mature 3-year-old ‘hooknose’ males captured between 27 April and 30 May in 2013 , 2014 and 2015 . Sample size was informed by related empirical research in this system ( Rosengrave et al . , 2008 , 2009a ) . Fish were individually tagged and maintained in a natural river-water raceway ( 12 . 5–13°C ) at a hatchery ( Salmon Smolt NZ , Canterbury , New Zealand ) using standard husbandry procedures . All animals were collected and maintained according to the standards of the Animal Ethics Committee for the University of Otago , New Zealand . A total of 11 social status manipulation trials were conducted each using four males ( n = 44; Figure 1 ) . On day 1 , two male dyads were formed pairing males of similar size ( average fork length = 71 . 5 cm , 95% CI = 70 . 2–72 . 9 cm , n = 44 ) . Each dyad was then placed in a sectioned off part of a river-water raceway ( approx . 2 . 5 m x 2 m x 1 m ) . Social interactions between the two fish in each dyad were observed for the first day using a series of 10 min under-water video recordings ( GoPro Hero 3 ) , one taken each hour over a 5-hr period , with the first recording starting 15 min after introducing fish to the raceway . Male dominance was then determined by calculating a Dominance Index ( DI; Winberg et al . , 1991; Bailey et al . , 2000; see Behavioural observations ) using the number of aggressive interactions between males . The male with the higher DI was ranked as dominant ( D ) and the male with the lower DI as subdominant ( S , stage 1 - Figure 1 ) . On day 2 , male dyads were left undisturbed and male social status within each dyad established on day one typically remained unchanged ( Table 5 ) . On day 3 , male dyads were re-formed placing dominant with dominant and subdominant with subdominant , and a new social hierarchy developed with male social status assigned to each male as described for day one . This forced one fish of each original dyad to change his social status ( DS or SD ) , while the other retained their original status ( DD or SS , stage 2 - Figure 1 ) . On day 4 , the male dyads were left undisturbed , and the experiment was complete on day 5 . We determined social status after all the social challenges except in one case where no interaction between males was recorded in the second stage and thus these individuals were excluded from further analyses . A further four males were excluded from analyses due to males escaping from the raceway in the second stage of the experiment , giving a total sample sizes n = 44 in stage 1 and n = 38 in stage 2 . Dominance Index ( DI ) was calculated using the following equation: DI = Agg+ / ( Agg++Agg- ) , where Agg+ represents the total number of aggressive acts performed and Agg- the total number of aggressive acts received by the individual ( Zilli et al . , 2004; Bailey et al . , 2000 ) . Aggressive acts were scored using the following criteria: Ejaculates were obtained from males by gently applying pressure to the abdomen , taking care to avoid contaminating samples , and were held at 4°C for up to 4 hr . We depleted the ejaculate reserves of each male before the experiment so ejaculates collected later were produced during each 48-hr period . We collected ejaculates in a random order on day 3 at the end of stage 1 and after social status was manipulated on day 5 at the end of stage 2 so samples were collected 48 hr after social status was established in each stage . Sperm velocity measurements were performed in a random order and blind to the social status of each male . We measured sperm swimming speed twice for each male at 10 s post-activation using a CEROS sperm tracker ( v 1 . 2 , Hamilton-Thorne Research , Beverly , MA ) . Approximately 1 µl of milt was activated with river water or ovarian fluid ( diluted to 50% with river water ) onto a 20 µl Leja slide ( Leja Products B . V . , Nieuw-Vennep , The Netherlands ) on a temperature-controlled stage cooler ( TS-4 Thermal Microscope Stage , Physitemp ) set to 12 . 5°C to match the natural spawning water temperature . We used average path velocity ( VAP , µm s−1 ) as our measure of sperm swimming speed which estimates the average velocity of a sperm cell for 0 . 5 s over a smoothed path ( Rosengrave et al . , 2008 , 2009a , 2016; Figure 2—figure supplement 1 ) . Sperm concentration ( sperm/ml ) was determined using a Neubauer haemocytometer . To determine the relative roles of sperm and seminal fluid on sperm velocity we centrifugally separated and remixed sperm and seminal fluid of each male with those from the other male in each dyad ( n = 42 males in 39 dyads ) . To prepare recombined ejaculates , milt was centrifuged in 1 . 5 ml tubes at 4°C , 300 g for 10 min to separate sperm cells from seminal fluid . The seminal fluid was then transferred into a new tube after which 500 µl of artificial seminal fluid ( 80 mM NaCl , 40 mM KCl , 1 mM CaCl2 , 20 mM Tris-HCl ) was added to the sperm cells and this was centrifuged again at 4°C , 300 g for 10 min to wash any remaining seminal fluid from the sperm cells . The artificial seminal fluid was then discarded and recombined ejaculates were prepared using 10 µl of sperm resuspended in 90 µl of seminal fluid from the same male ( control ) or seminal fluid from their rival , incubated at 12°C for 20 min . In 2014 and 2015 , at both stages of the social status manipulation trials ( Figure 1 ) we conducted a total of 21 replicated in vitro fertilisation trials to determine the effects of ejaculate recombination ( seminal fluid ) on male fertilisation success . This involved 24 individual males and 17 females in which sperm from the dominant and subdominant male in each dyad competed to fertilise a female’s eggs . For each trial , we performed two seminal fluid treatments , using either unmanipulated or recombined ejaculates , in addition to non-competitive controls using sperm from each of the males individually . Haphazardly chosen female fish were killed with a stroke to the head , and their egg batch was expelled through an incision in the abdomen , into a clean bowl . Ovarian fluid was collected by carefully pipetting from each egg batch . Sperm density was adjusted prior to each fertilisation trial so that approximately the same number of sperm per male ( 107 spermatozoa ) were used in each trial . For each trial , we placed approximately 100 unfertilised ova from the focal female in a dry 2 l plastic beaker , then added ejaculate samples from each male simultaneously by injecting them separately into a steady stream of raceway water ( 250 ml at 12 . 5–13 °C ) . This technique simulated natural spawning conditions by facilitating the rapid mixing of eggs with sperm from both males ( Rosengrave et al . , 2016 ) . We added the ejaculate samples separately into the water to ensure the spermatozoa were activated before the ejaculate samples came into contact , minimising any effects of each male’s seminal fluid on the other male’s sperm function . The eggs were allowed to sit for 5 min undisturbed until water hardened and were then gently transferred to heath rack trays ( 12 . 5–13°C ) . We randomly sampled 24 alevins from each replicate fertilisation trial ( 40 days post fertilisation ) , placing them in 99% ethanol for DNA extraction and microsatellite genotyping to assess paternity . To assess paternity share for the males in each sperm competition trial , DNA was extracted from a fin clip for both adult males , the female and 24 offspring from each trial using Chelex100 resin ( Walsh et al . , 1991 ) . Three microsatellite loci ( Ots 100 , Ots 101 , Oki 3a; Table 6 ) were then amplified in a multiplex PCR and used to determine paternity by manually matching alleles between offspring , mother and either potential sire . A fourth locus ( Ots 104; Table 6 ) was amplified separately using a touchdown PCR protocol and employed when three loci were insufficient to determine paternity without certainty . The genotype of each offspring was always consistent with the expected genotype based on the alleles for the potential sires , i . e . in no offspring did we record unique alleles present for both potential sires . Multiplex PCRs were run in 10 μL volume reactions and included the following reagents: 1x PCR buffer ( Bioline ) , 2 mM MgCl2 , 0 . 3 mM dNTPs , 0 . 4 μM forward and reverse Ots 101 primers , 0 . 2 μM forward and reverse Ots 100 and Oki 3a primers , 0 . 5 U of Bioline Taq DNA polymerase , and 0 . 5 μL of DNA . The thermal cycling conditions for the multiplex protocol were: 12 min at 95°C followed by 10 cycles of 15 s at 94°C , 30 s at 60°C , and 30 s at 72°C , followed by 30 cycles of 15 s at 89°C , 30 s at 60°C , 30 s at 72°C , and a final extension period of 10 min at 72°C . PCRs for amplification of Ots 104 were run in 10 μL volume reactions and included the following reagents: 1x PCR buffer ( Bioline ) , 2 mM MgCl2 , 0 . 3 mM dNTPs , 0 . 5 μM forward and reverse Ots 104 primers , 0 . 5 U of Bioline Taq DNA polymerase , and 0 . 5 μL of DNA . The thermal cycling conditions for the touchdown protocol were: 2 min at 95°C followed by 10 cycles of 30 s at 95°C , 45 s at Ta°C , and 30 s at 72°C , where Ta starts at 55°C and drops by 0 . 5°C each cycle ( last cycle should be 50 . 5°C ) , followed by 20 cycles of 30 s at 95°C , 45 s at 50°C , 30 s at 72°C , and a final extension period of 10 min at 72°C . PCR samples were genotyped by adding 0 . 5 μL PCR product to 12 μL HiDi formamide and 0 . 3 μL Genescan LIZ500 size standard ( Applied Biosystems ) then run on an ABI3130 × 1 Genetic Analyser ( Applied Biosystems ) . Results were visualised using GeneMarker v 2 . 2 ( SoftGenetics , RRID:SCR_015661 ) and alleles were scored manually . All statistical analyses were performed using R v 3 . 1 . 3 ( R Core Team , 2016; RRID:SCR_001905 ) . To compare changes in ejaculate quality ( sperm velocity ( VAP ) or sperm concentration ) between D and S males , generalised linear mixed effects models ( GLMM ) were fitted using the package ‘lme4’ ( Bates et al . , 2015; RRID:SCR_015654 ) . GLMMs using a Gaussian error distribution were fitted using VAP as the response variable , while GLMMs with a Poisson error distribution were fitted using sperm concentration as the response variable . Each GLMM used male social status as a fixed predictor , for stage 1 two levels comparing D and S; and for stage 2 , separate models were run with either two levels comparing D and S males with data pooled together ( D = DD + SD and S = SS + DS ) , or four levels ( males that retained the same status DD and SS , and males that changed status SD and DS ) . Models with VAP as the response variable used both replicate measurements for each male and included male identity as a random predictor to account for repeated measures . To test whether males that change social status adjust ejaculate quality , we compared both VAP ( GLMMs using a Gaussian error distribution ) and sperm concentration ( GLMMs with a Poisson error distribution ) in the same males across the two stages of the experiment . Four separate models were run for each of the response variables , separately comparing males in each of the four social phenotypes ( DD , DS , SD , SS ) and each model used experimental stage ( factor with two levels ) as a fixed predictor . Additionally , we used an alternative analysis for each of the response variables to test for an interaction effect between social status and experimental stage , both models used social status ( factor with four levels; DD , DS , SD and SS ) , experimental stage ( factor with two levels ) and the interaction between social status and experimental stage as fixed predictors . Male identity was included as a random predictor to account for repeated measures from the same male . A linear mixed effects model ( GLMM ) was fit using the difference in VAP between focal male’s sperm recombined with his own seminal fluid and focal male’s sperm recombined with his rival male’s seminal fluid as the response variable , with difference in VAP between focal male’s sperm recombined with his own seminal fluid and rival male’s sperm recombined with his own seminal fluid , and social status of rival’s seminal fluid as fixed predictors . To fulfil the model’s assumption of normality a cube-root transformation was performed on the response variable . We used the random predictors focal male identity , rival male identity and each pairing to account for repeated measures . All VAP measures used were those activated in river water , not ovarian fluid , to avoid female effects on sperm velocity ( Rosengrave et al . , 2009b , Rosengrave et al . , 2016 ) that could mask the influence of seminal fluid . To assess the importance of sperm velocity as a predictor of fertilisation success , we used a GLMM that was fit using the difference in the number of offspring sired between the focal and rival male in each trial as the response variable , with the relative sperm velocity between males as a fixed predictor . To assess social status as a predictor of fertilisation success we used a binomial GLMM that was fit using the proportion of offspring sired by each male as the response variable , with male social status as a fixed predictor in unmanipulated milt trials and the social status of seminal fluid donor as a fixed predictor in swapped seminal fluid trials . In order to assess the influence of seminal fluid on male fertilisation success , we used a GLMM that was fit using the change in the proportion of eggs sired by each focal male across seminal fluid treatments ( within the same triad , i . e . within the same male-male-female combination ) as the response variable with the change in relative sperm velocity across treatments used as a fixed predictor . For all above models , we used the random predictors focal male identity , rival male identity , female identity and each unique triad to control for repeated measures . We tested for repeatability of replicate trials conducted for each triad ( supplementary material: Statistical analysis and R code ) , removing one triad for which the proportion of eggs sired differed significantly between replicates ( n = 20 ) . So that sperm velocity in our model reflected conditions during the fertilisation trials , all VAP measures used were those activated in ovarian fluid , as female effects on sperm velocity can influence the outcome of sperm competition in chinook salmon ( Rosengrave et al . , 2009b , Rosengrave et al . , 2016 ) . All mentioned models used the week during the spawning season when milt samples were collected as a random predictor to control for potential seasonal effects on milt quality ( Butts et al . , 2010; Hajirezaee et al . , 2010 ) , and the year fish were collected as a covariate ( Bolker , 2015 ) . To determine the significance of fixed effects , we present both 95% confidence intervals ( CI ) calculated using the Wald method , and p values calculated for linear mixed effects models with the package ‘lmerTest’ ( Kuznetsova et al . , 2016; RRID:SCR_015656 ) using Satterthwaite approximations to calculate degrees of freedom . Assumptions underlying parametric models were verified using residual plots and Shapiro tests . An alpha value of 0 . 05 was used to evaluate the significance of P-values and adjusted for multiple tests using the Bonferroni method . Refer to supplementary file: Statistical analysis and R code , for all R code used and output from analyses . Contains all R ( R Core Team , 2016; RRID:SCR_001905 ) code , including output and model diagnostics . The following packages were used: lme4 ( Bates et al . , 2015; RRID:SCR_015654 ) , lmerTest ( Kuznetsova et al . , 2016; RRID:SCR_015656 ) , nlme ( Pinheiro et al . , 2015; RRID:SCR_015655 ) , ggplot2 ( Wickham , 2009; RRID:SCR_014601 ) , lattice ( Sarkar , 2008; RRID:SCR_015662 ) , RVAideMemoire ( Hervé , 2016; RRID:SCR_015657 ) , LMERConvenienceFunctions ( Tremblay and Ransijn , 2015; RRID:SCR_015658 ) , and Deducer ( Fellows , 2012; RRID:SCR_015659 ) . | Males of many animal species fight to establish social dominance and control access to females so that they have more opportunities to reproduce than their competitors . Males with lower social status will struggle to directly compete for mates , thus they attempt to mate with females by stealth . This often leads to more than one male mating with the same female so that the sperm from each male end up competing to fertilise that female’s eggs , a phenomenon known as sperm competition . Males suspend their sperm in a fluid to make a mixture known as semen . It has been shown that , compared to high status males , low status males will produce higher quality semen that contains greater numbers of faster swimming sperm , giving them an advantage in sperm competition . Growing evidence from several species indicates that males can quickly adjust how fast their sperm swim in response to social cues that signal changing risks of sperm competition . However , how these rapid adjustments occur remains largely unknown , and whether they alter a male’s reproductive success against a competitor has seldom been examined . Chinook salmon usually live in the North Pacific Ocean but they swim up rivers in North America and Asia to reproduce . They have also been introduced to several other countries including New Zealand where they are farmed commercially . The fish are highly prized by sport fishermen and are also of cultural significance to certain groups of indigenous people in North America . Barlett et al . studied the semen of chinook salmon , undertaking a series of experiments in which males switched between high and low social status . The experiments show that , as predicted , the sperm of males that changed from high to low social status started to swim faster . These changes in speed were caused by the fluid in the semen and altered the number of eggs that the male’s sperm fertilised when competing against sperm from another male . In their natural range some populations of chinook salmon are declining due to overfishing combined with habitat loss and alteration . The findings of Barlett et al . contribute to a better understanding of how this fish species reproduces , which may lead to the introduction of measures that help natural populations to recover or help to improve commercial farming . Improved knowledge of how the fluid in semen affects sperm activity may also have important consequences for our wider understanding of male fertility in humans and other animals . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology"
] | 2017 | Sperm competition risk drives rapid ejaculate adjustments mediated by seminal fluid |
Human calcium-sensing receptor ( CaSR ) is a G-protein-coupled receptor ( GPCR ) that maintains extracellular Ca2+ homeostasis through the regulation of parathyroid hormone secretion . It functions as a disulfide-tethered homodimer composed of three main domains , the Venus Flytrap module , cysteine-rich domain , and seven-helix transmembrane region . Here , we present the crystal structures of the entire extracellular domain of CaSR in the resting and active conformations . We provide direct evidence that L-amino acids are agonists of the receptor . In the active structure , L-Trp occupies the orthosteric agonist-binding site at the interdomain cleft and is primarily responsible for inducing extracellular domain closure to initiate receptor activation . Our structures reveal multiple binding sites for Ca2+ and PO43- ions . Both ions are crucial for structural integrity of the receptor . While Ca2+ ions stabilize the active state , PO43- ions reinforce the inactive conformation . The activation mechanism of CaSR involves the formation of a novel dimer interface between subunits .
The extracellular calcium-sensing receptor ( CaSR ) is responsible for the maintenance of a stable blood Ca2+ level ( Brown et al . , 1993; Hofer and Brown , 2003 ) . It senses fluctuations in the circulating Ca2+ concentration and controls Ca2+ homeostasis by ( 1 ) modulating the production of parathyroid hormone in parathyroid glands , and ( 2 ) regulating the reabsorption of Ca2+ in kidney and bone ( Brown , 2013 ) . Abnormal function of CaSR is associated with Ca2+ homeostatic disorders ( Brown , 2007; Hendy et al . , 2009; Ward et al . , 2012 ) . Loss-of-function mutations in CaSR lead to potentially fatal neonatal severe primary hyperparathyroidism , while gain-of-function mutations cause autosomal dominant hypocalcemia ( Hendy et al . , 2009; Ward et al . , 2012 ) . CaSR also plays important roles in biological processes unrelated to Ca2+ balance , including fetal development ( Riccardi et al . , 2013 ) , nutrient sensing ( Conigrave and Hampson , 2006 ) , and regulation of neuronal excitability ( Ruat and Traiffort , 2013 ) . The functional diversity of CaSR results from its ability to activate multiple signaling pathways through Gq/11 , Gi/o , G12/13 and Gs proteins ( Conigrave and Ward , 2013; Hofer and Brown , 2003; Magno et al . , 2011 ) , and to respond to a variety of ligands ( Magno et al . , 2011; Saidak et al . , 2009 ) . The general consensus is that the principal agonist of CaSR is extracellular Ca2+ ( Hofer and Brown , 2003 ) . Other orthosteric agonists include various divalent and trivalent cations , polyamines and cationic polypeptides ( Magno et al . , 2011; Saidak et al . , 2009 ) . CaSR function can be regulated by endogenous and synthetic allosteric modulators , extracellular pH and ionic strength ( Jensen and Brauner-Osborne , 2007; Quinn et al . , 2004 , 1998; Saidak et al . , 2009 ) . Aromatic and aliphatic L-amino acids such as L-Phe and L-Trp increase the sensitivity of CaSR toward Ca2+ ( Conigrave et al . , 2000 ) and are considered as positive allosteric modulators of the receptor ( Saidak et al . , 2009 ) . Previous studies have also demonstrated that L-amino acids can activate the receptor provided that Ca2+ concentration is above a threshold ( Conigrave et al . , 2004; Conigrave et al . , 2000; Rey et al . , 2005; Young and Rozengurt , 2002 ) . For this reason , L-amino acids have been called allosteric activators ( Conigrave et al . , 2004 ) . Finally , it has been proposed that Ca2+ ions and amino acids may act as co-agonists of the receptor ( Conigrave et al . , 2000; Young and Rozengurt , 2002 ) . However , this view has not been widely accepted . CaSR is a class C G-protein-coupled receptor ( GPCR ) . Within this family , CaSR and metabotropic glutamate receptors ( mGluRs ) function as disulfide-linked homodimers ( Bai et al . , 1998; Pidasheva et al . , 2006; Ray et al . , 1999; Romano et al . , 1996; Ward et al . , 1998; Zhang et al . , 2001 ) , while GABAB and taste receptors are obligatory heterodimers ( Jones et al . , 1998; Kaupmann et al . , 1998; Kuner et al . , 1999; Nelson et al . , 2002 , 2001; Ng et al . , 1999; White et al . , 1998 ) . Ligand binding to CaSR takes place within a large extracellular Venus Flytrap ( VFT ) module that consists of two domains ( LB1 and LB2 ) ( Brauner-Osborne et al . , 1999; Hendy et al . , 2013; Mun et al . , 2004 ) . In addition , CaSR contains a cysteine-rich ( CR ) domain that connects the VFT module to the transmembrane region ( Hendy et al . , 2013 ) . This CR region is present in all class C GPCRs except GABAB receptor and is required for receptor activation ( Hauache et al . , 2000; Hu et al . , 2000; Huang et al . , 2011 ) . However , the activation mechanism of CaSR remains unknown . Structural information for class C GPCRs is available for the extracellular domains ( ECD ) of mGluRs ( Kunishima et al . , 2000; Muto et al . , 2007; Tsuchiya et al . , 2002 ) and GABAB receptor ( Geng et al . , 2013 , 2012 ) , as well as the transmembrane domains of mGluRs ( Dore et al . , 2014; Wu et al . , 2014 ) . Here , we present the first crystal structures of the entire extracellular domain of human CaSR in two different functional states . These structures reveal novel binding sites for Ca2+ , PO43- and L-Trp , identify L-Trp as an agonist of the receptor , and demonstrate that these ions and amino acids collectively control the function of CaSR .
The ECD of human CaSR was secreted from baculovirus-infected insect cells as a disulfide-tethered homodimer ( Figure 1—figure supplement 1 ) . It contains 11 potential N-linked glycosylation sites . Disruption of three of the glycosylation sites did not alter CaSR signaling ( Figure 1—figure supplement 1 ) . Formation of well-diffracting crystals required partial deglycosylation of the receptor through mutation and enzymatic digestion . We obtained two different forms of CaSR ECD crystals . Form I was crystallized in the absence and presence of 2 mM Ca2+ , and form II in the presence of 10 mM L-Trp and 10 mM Ca2+ ( Table 1 ) . 10 . 7554/eLife . 13662 . 003Table 1 . Data collection and refinement statistics . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 00310 . 7554/eLife . 13662 . 004Table 1—source data 1 . Statistics for anomalous data collection . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 00410 . 7554/eLife . 13662 . 005Table 1—source data 2 . Data collection and refinement statistics for endogenous ligand-bound CaSR ECD . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 005Functional stateInactive ( 2 mM Ca2+ ) Active ( 10 mM Ca2+ , 10 mM L-Trp ) CrystalForm IForm IIData collectionSpace groupF222C2Wavelength ( Å ) 0 . 97921 . 7712Cell dimensions a , b , c ( Å ) 126 . 3 , 150 . 2 , 214 . 6107 . 7 , 127 . 5 , 146 . 8 α , β , γ ( ° ) 90 . 0 , 90 . 0 , 90 . 090 . 0 , 108 . 7 , 90 . 0Resolution ( Å ) 88 . 1 - 3 . 1 ( 3 . 6 - 3 . 1 ) 139 . 0 - 2 . 6 ( 2 . 9 - 2 . 6 ) Rsym or Rmerge0 . 051 ( 0 . 702 ) 0 . 043 ( 0 . 575 ) I / σI21 . 3 ( 2 . 1 ) 22 . 1 ( 2 . 3 ) Completeness ( % ) 99 . 9 ( 100 . 0 ) 98 . 0 ( 97 . 1 ) Redundancy6 . 6 ( 6 . 8 ) 6 . 9 ( 6 . 8 ) CC1/2 ( % ) 100 . 0 ( 93 . 1 ) 99 . 9 ( 96 . 9 ) RefinementResolution ( Å ) 107 . 2 - 3 . 137 . 5 - 2 . 6No . of reflections16 , 74748 , 839Rwork / Rfree ( % ) 22 . 2 / 23 . 921 . 1 / 22 . 2No . of atoms Protein45648454 Ligand ( Trp ) -30 Cation ( Ca2+ ) 18 Anion15 ( SO42- ) 20 ( PO43- ) Sugar9870 Water43331B-factors ( Å2 ) Protein110 . 468 . 3 Ligand-39 . 9 Cation ( Ca2+ ) 105 . 496 . 3 Anion102 . 1 ( SO42- ) 61 . 0 ( PO43- ) Sugar152 . 980 . 4 Water76 . 953 . 8R . m . s . deviations Bond lengths ( Å ) 0 . 0080 . 009 Bond angles ( ° ) 1 . 151 . 14Values in parentheses are for highest-resolution shell . CC1/2 is defined in reference ( Karplus and Diederichs , 2012 ) . In both CaSR ECD structures , the two protomers interact in a side-by-side fashion while facing opposite directions ( Figure 1; Figure 1—figure supplement 2 ) . Each CaSR ECD protomer consists of three domains , LB1 , LB2 and CR . The two lobe-shaped domains LB1 and LB2 form a VFT module similar to that of mGluRs ( Kunishima et al . , 2000; Muto et al . , 2007; Tsuchiya et al . , 2002 ) and GABAB receptor ( Geng et al . , 2013 , 2012 ) . The relative orientation between the LB2 and CR domains is fixed through an interdomain disulfide linkage ( C236-C561 ) , and the CR domain is positioned to amplify and transmit the conformational variations within the VFT module . 10 . 7554/eLife . 13662 . 006Figure 1 . Crystal structures of human CaSR ECD . ( A ) Inactive-state structure of CaSR ECD homodimer in the presence of 2 mM Ca2+ . ( B ) Active-state structure of CaSR ECD homodimer in the presence of 10 mM Ca2+ and 10 mM L-Trp . Each structure is shown in cartoon ( front view ) or surface ( side view ) representations that are related by a 90°-rotation about the vertical axis . Each protomer is colored according to its individual domains ( LB1 , light blue; LB2 , blue; CR , purple ) . The various ligands ( L-Trp , Ca2+ , PO43- , SO42- ) are displayed as space-filling models . Observed carbohydrates are shown as ball-and-stick models in gray . Disulfide bridges are in yellow . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 00610 . 7554/eLife . 13662 . 007Figure 1—figure supplement 1 . Purification of the CaSR ECD homodimer . ( A ) Superdex-200 size exclusion chromatography of secreted wild-type ( wt ) CaSR ECD . ( B ) SDS gel of wt-CaSR ECD before and after Endo H digestion under reducing condition . ( C ) Superdex-200 size exclusion chromatography of secreted CaSR ECD mutant carrying three glycosylation-site mutations ( N386Q , S402N and N468Q ) . ( D ) SDS gel of purified CaSR ECD mutant from ( C ) under reducing ( +DTT ) and non-reducing ( -DTT ) conditions . ( E ) Dose-dependent Ca2+-stimulated IP accumulation in cells expressing wild-type or mutant CaSR . The full-length CaSR contains either two ( N386Q , S402N ) or three ( N386Q , S402N and N468Q ) glycosylation-site mutations . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 00710 . 7554/eLife . 13662 . 008Figure 1—figure supplement 2 . Different conformational states of CaSR ECD . ( A ) Cartoon representation of the inactive ( form I ) CaSR ECD crystal structure . ( B ) Cartoon representation of the active ( form II ) CaSR ECD crystal structure . Each homodimer structure is shown in five views: front ( center ) , top ( top ) , bottom ( bottom ) , and two side ( center left and right ) views . Each protomer is colored according to its individual domains ( LB1 , light blue; LB2 , blue; CR , purple ) . The various ligands ( L-Trp , Ca2+ , PO43- , SO42- ) are displayed as space-filling models . Observed carbohydrates are shown as ball-and-stick models in gray . Disulfide bridges are in yellow . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 008 Form I crystal structure of CaSR ECD represents the inactive configuration since the VFT modules of both protomers are in the open conformation associated with the resting state ( open-open ) , and the interdomain cleft is empty . In addition , each protomer structure contains one Ca2+ ion and three SO42- ions ( Figure 1A ) . In the form II crystal structure , both protomers of CaSR ECD have the closed conformation associated with agonist binding ( closed-closed ) . Surprisingly , the ligand-binding cleft of each protomer is solely occupied by an L-Trp molecule . Ca2+ is bound at four novel sites in the CaSR ECD structure , including one at the homodimer interface . Each CaSR ECD molecule also contains two PO43- ions ( Figure 1B ) . Agonist binding induces large conformational changes within the CaSR ECD homodimer . First , the VFT module of each protomer undergoes domain closure . Alignment based on the LB1 domains showed that the LB2 domains of inactive and active structures are related by a 29° rotation ( Figure 2A ) . Second , the LB2 domains of the two protomers approach each other , resulting in an expansion of the homodimer interactions involving LB2 domains . Third , the CR domains of the two subunits interact to form a large homodimer interface that is unique to the active state . The CR domains are brought into close contact by the motion involving LB2 since the two domains are rigidly associated within each subunit . Finally , the structural reorganization of CR domains reduces the distance between the C-termini of the two subunits from 83 Å to 23 Å ( Figure 2B , C ) . This CR domain movement may cause reorientation of the transmembrane domains during receptor activation . 10 . 7554/eLife . 13662 . 009Figure 2 . Agonist-induced conformational changes . ( A ) Superposition of the inactive ( orange ) and active ( blue ) CaSR ECD structures based on the LB1 domain of one protomer ( front view , left; side view , right ) . Green line is the axis of rotation that relates the LB2 domains of the superimposed protomers ( rotation χ = 29° , screw translation τχ = −2 . 2 Å ) . ( B , C ) Surface representation of inactive ( B ) and active ( C ) structures in front ( top ) and bottom ( bottom ) views . Distance between C-termini of the two subunits ( yellow ) is marked by dashed line for each homodimer . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 009 The inactive structure of CaSR ECD shows that subunit association in the resting state is primarily mediated by the LB1 domains ( Figure 3A; Figure 3—figure supplement 1 ) . This dimer interface is largely conserved in the active structure , indicating that the LB1-LB1 interaction mostly serves to faciliate dimerization between receptor subunits ( Figure 3B; Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 13662 . 010Figure 3 . Homodimer interface . ( A ) Cα trace of inactive structure with elements involved in homodimer formation highlighted by cartoons . The interface is divided into three regions I , II , and IV . Site II is further separated into two symmetrical parts II_a and II_b . Specific contacts at each interface region are shown in surrounding panels . Dashed lines indicate hydrogen bonds . ( B ) Cα trace of active structure showing elements involved in homodimer formation . The interface is divided into six regions , I , II , III , IV , V , and VI . Specific contacts at the interface areas III , IV , V , and VI are shown in surrounding panels . For both structures , the domains involved in dimerization at each interface region are: I: LB1-LB1; II: LB1-LB1; III: LB2-LB1; IV: LB2-LB2; V: LB2-CR; VI: CR-CR . ( C ) Dose-dependent Ca2+-stimulated IP accumulation in cells transiently expressing wild-type ( wt ) or mutant CaSR . Naturally occurring inactivating mutations L159P , R172G , D215G , R227L , R551K , and G557E are located at the homodimer interface . The single mutation W458A was designed based on structure to affect receptor homodimerization . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 01010 . 7554/eLife . 13662 . 011Figure 3—figure supplement 1 . Homodimer interface . ( A , B ) Cα traces of inactive ( A ) and active ( B ) CaSR ECD structures . Each structure is presented in three views: top ( left ) , front ( center ) and side ( right ) . Structural elements involved in homodimer formation are highlighted by cartoons . ( C , D ) A detailed view of the structural elements at the homodimer interface region I of inactive ( C ) and active ( D ) CaSR ECD structures . For each structure , the angle between the D-helices of the two subunits is shown and is used to represent dimer orientation at this interface . ( E , F ) Cα traces of inactive ( E; PDB code: 4MQE ) and active ( F; PDB code: 4MS4 ) GABAB receptor ( GABABR ) ECD structures . ( G , H ) A detailed view of the structural elements at the heterodimer interface of inactive ( G ) and active ( H ) GABABR ECD structures . For each structure , the angle between the C-helices of the two subunits is shown . ( I , J ) Cα traces of inactive ( I; PDB code: 1EWT ) and active ( J; PDB code: 1EWK ) mGluR1 VFT module structures . ( K , L ) A detailed view of the structural elements at the homodimer interface of inactive ( G ) and active ( H ) mGluR1 VFT structures . For each structure , the angle between the C-helices of the two subunits is shown . Panels ( E ) – ( L ) were adapted from Supplementary Figure 13 of ( Geng et al . , 2013 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 01110 . 7554/eLife . 13662 . 012Figure 3—figure supplement 2 . Homodimer interface . Specific contacts at the interface areas I , II_a , and II_b of the active CaSR ECD structure . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 012 The LB1-LB1 dimer interface buries over 3800 Å2 of solvent-accessible surface area , and can be divided into two regions . Site I is located at the center of LB1 domain and is flanked on either side by the two symmetric parts of site II ( Figure 3A , B ) . Site I is formed by two central helices ( B , D ) of each protomer ( Figure 3A , B; Figure 3—figure supplement 1 ) . A dimer interface formed by the same structural elements is also observed in mGluR and GABAB structures ( Geng et al . , 2013; Kunishima et al . , 2000; Muto et al . , 2007; Tsuchiya et al . , 2002 ) . For mGluRs , transition between the resting and active configurations causes a 70°-rotation in the LB1-LB1 homodimer interface ( Kunishima et al . , 2000 ) ( Figure 3—figure supplement 1 ) . For both CaSR and GABAB receptors ( Geng et al . , 2013 ) , however , agonist binding only induces a small 5°-rotation in the orientation of this interface , and the LB1-LB1 dimer interface is largely conserved throughout the activation process ( Figure 3—figure supplement 1 ) . The dimer interactions at site I of CaSR are predominantly hydrophobic and involve tightly packed leucine and phenylalanine residues ( L112 , L156 , L159 , and F160 ) . The disease-causing mutation L159P renders the receptor less sensitive to Ca2+ ( Grant et al . , 2011; Hendy et al . , 2009 ) ( Figure 3C ) . In addition , site I features an inter-subunit disulfide bridge located at the tip of helix C . Site II is unique to the CaSR ECD structures . It involves an arm-like long loop stretched out from one subunit to reach its binding partner ( Figure 3A , B; Figure 3—figure supplement 1 ) . The dimer interactions at site II include hydrogen bonds and hydrophobic contacts . Several disease-causing mutations are located at this interface ( S53P , P55L , and Y161C ) ( Hendy et al . , 2009 ) . Substitution of a deeply buried interfacial residue W458 with alanine also decreased the potency of Ca2+ ( Figure 3C ) . These observations indicate that formation of a stable homodimer is important for CaSR function . Agonist binding causes an expansion of the dimer interactions involving LB2 domain . In the inactive homodimer , only minimal contacts occur between the LB2 domains ( Figure 3A ) . In the active state , LB2 of one protomer interacts with all three domains of a second protomer ( Figure 3B ) . These contacts are predominantly hydrophilic , and bury 1000 Å2 of solvent accessible surface area . LB2 mediates dimer interactions primarily through a central helix ( G ) that transverses the domain ( Figure 3B ) . First , the top of helix G contacts LB1 domain of the opposing subunit through two symmetric salt bridges between D215 and R172 ( site III ) . Second , the LB2-LB2 contacts involve the midsection of helix G , and feature a hydrogen bond between R227 and S240 , and water-mediated contacts by E224 ( site IV ) . Finally , LB2 interacts with the CR domain of a second subunit through a Ca2+ ion ( site V ) . This Ca2+ ion bridges the end of helix G in LB2 with a loop in CR domain . An abundance of disease-related mutations are found at these dimer interaction sites , including ( 1 ) R172G and D215G at site III , ( 2 ) R227L and R227Q at site IV , and ( 3 ) G557E at site V ( Hendy et al . , 2009 ) . All these mutations disrupt agonist-dependent dimer contacts , and decreased agonist efficacy ( Heath et al . , 1996; Hendy et al . , 2009; Wystrychowski et al . , 2005 ) ( Figure 3C ) . Agonist binding also induces the formation of a novel homodimer interface between the CR domains that covers approximately 1200 Å2 of solvent accessible surface area ( site VI ) ( Figure 3B ) . The CR-CR interactions are mediated by two β-strands and their connecting loop from each subunit . Key contacts include two cross-subunit hydrogen bonds ( T560 , E558 ) , hydrophobic contacts ( I554 , P569 ) , and electrostatic interactions ( R551 ) . Among these , R551K is a known disease-causing mutation that reduced the receptor response ( Hendy et al . , 2009; Toke et al . , 2007 ) ( Figure 3C ) . In the active structure , the amino acid L-Trp is bound at the interdomain cleft of the VFT module , in agreement with previous mutational data ( Mun et al . , 2005; Mun et al . , 2004; Zhang et al . , 2014 , 2002 ) ( Figure 4A; Figure 4—figure supplement 1 ) . L-Trp facilitates extracellular domain closure of CaSR by contacting both LB1 and LB2 domains of the VFT module ( Figure 4B , C; Figure 4—figure supplement 1 ) . The interactions between CaSR ECD and L-Trp are primarily mediated by hydrogen bonds . ( 1 ) The carboxylic acid group of L-Trp forms hydrogen bonds through both oxygen atoms with LB1 residues S147 , A168 , and S170 . ( 2 ) The backbone nitrogen of L-Trp is hydrogen-bonded to A168 and S170 of LB1 domain . ( 3 ) The indole nitrogen of L-Trp forms two hydrogen bonds with E297 of LB2 domain . ( 4 ) L-Trp is engaged in hydrophobic contacts with both LB1 and LB2 residues including W70 , T145 , Y218 , and A298 . The extensive contacts between backbone atoms of L-Trp and the receptor suggest that other amino acids may bind to CaSR in a similar fashion to induce domain closure . 10 . 7554/eLife . 13662 . 013Figure 4 . L-Trp recognition by CaSR ECD . ( A ) Molecular surface of a L-Trp-bound CaSR ECD protomer in the active structure . L-Trp is displayed as a space-filling model . ( B ) Specific contacts between CaSR ECD ( gray ) and L-Trp ( yellow ) . Mesh represents the final 2Fo-Fc electron density map contoured at 1σ . Hydrogen bonds are represented by dashed lines . ( C ) Schematic diagram of the interactions between CaSR ECD and bound L-Trp . Selected contacts are highlighted; hydrogen bonds , dashed lines; hydrophobic contacts , wiggled lines . ( D ) Dose-response curves of nonradioactive L-Trp inhibiting [3H]-L-Trp binding to CaSR ECD in the presence of 0 mM or 2 mM Ca2+ . ( E–F ) Dose-dependent L-Trp-induced intracellular Ca2+ mobilization at various extracellular Ca2+ concentrations in cells stably expressing CaSR . ( E ) Effect of increments of L-Trp ( final concentrations: 0 . 1 , 0 . 2 , 0 . 5 , 1 . 0 , 2 . 0 , 5 . 0 , 10 , 20 mM ) in the presence of 1 . 5 mM extracellular Ca2+ . The experiments were followed by an exposure to 0 . 5 mM extracellular Ca2+ to demonstrate reversibility of the L-Trp-induced intracellular Ca2+ oscillation , and a later exposure to 8 mM extracellular Ca2+ to demonstrate the maximal response . ( F ) Integrated response curves of L-Trp at 0 . 5 , 1 . 5 , and 2 . 5 mM Ca2+ . Intracellular Ca2+ responses are presented in integrated response units ( IRUs , F340/F380 . min ) . ( G ) Effect of L-Trp on Ca2+-stimulated intracellular Ca2+ mobilization in cells transiently expressing wt CaSR . ( H–L ) Dose-dependent Ca2+-stimulated IP accumulation in cells transiently expressing wild-type or mutant CaSR . Naturally-occurring inactivating mutations Y218S and E297K are located at the L-Trp binding site . The single mutations T145I , S147A , and S170A were designed based on structure to disrupt L-Trp recognition . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 01310 . 7554/eLife . 13662 . 014Figure 4—figure supplement 1 . L-Trp recognition by CaSR ECD . ( A ) Molecular surface of L-Trp-bound CaSR ECD homodimer . L-Trp is displayed as a space filling model . ( B , C ) Specific contacts between CaSR ECD ( gray ) and L-Trp ( yellow ) within each protomer of the active structure . Mesh represents the final 2Fo-Fc electron density map contoured at 1σ . Hydrogen bonds are represented by black dashed lines . ( D ) Superposition of the L-Trp-binding site in CaSR ( gray ) and Glu-binding site in mGluR1 ( cyan ) . Agonist-binding residues that are located at the same locations of the two receptor structures are highlighted . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 01410 . 7554/eLife . 13662 . 015Figure 4—figure supplement 2 . Endogenous ligand of CaSR . ( A ) Active-state structure of CaSR ECD homodimer showing unexplained electron density at the interdomain crevice of each protomer when crystals of CaSR ECD were grown in the presence of 10 mM Ca2+ and absence of any additional L-amino acids . ( B , C ) Close-up view of the region surrounding the extra density . Mesh represents a 2Fo-Fc electron density map contoured at 1σ . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 015 We measured the direct interaction between L-Trp and CaSR ECD by scintillation proximity assay ( SPA ) ( Quick and Javitch , 2007 ) ( Figure 4D ) . CaSR ECD exhibited binding of [3H]-L-Trp in the absence and presence of Ca2+ . Isotope dilution of [3H]-L-Trp with nonradioactive L-Trp led to a displacement of [3H]-L-Trp in a concentration-dependent manner . The addition of 2 mM Ca2+ increased the amount of L-Trp bound to CaSR ECD at any given concentration , suggesting that extracellular Ca2+ enhances L-Trp binding , possibly by affecting the L-Trp-binding affinity and kinetics of CaSR . The half-maximal inhibitory concentration of L-Trp ( IC50 , concentration at which 50% displacement of [3H]-L-Trp was observed ) was approximately 2 mM regardless of whether the experiment was performed in the absence ( 2 . 11 ± 0 . 72 mM ) or presence ( 2 . 04 ± 0 . 10 mM ) of 2 mM Ca2+ . Nevertheless , the binding affinity of L-Trp to CaSR ECD with and without Ca2+ may still differ as it depends on the concentration and binding affinity of the radiolabeled ligand . Further studies are needed to characterize the effect of Ca2+ on L-amino acid binding at the orthosteric agonist site of CaSR . We found that L-Trp directly stimulated intracellular Ca2+ mobilization through CaSR ( Figure 4E , F ) , in agreement with previous findings ( Conigrave et al . , 2004 , 2000; Rey et al . , 2005; Young and Rozengurt , 2002 ) . L-Trp-induced CaSR activation required the presence of extracellular Ca2+ above a threshold level of 0 . 5 mM . The effect of L-Trp on CaSR was concentration-dependent , with an apparent half-maximal effective concentration ( EC50 ) of 0 . 12 ± 0 . 06 mM when extracellular Ca2+ was present at 2 . 5 mM . The efficacy and potency of L-Trp decreased at lower concentrations of Ca2+ that are within the physiological range ( L-Trp EC50 = 0 . 75 ± 0 . 51 mM at 1 . 5 mM Ca2+ ) , consistent with previous proposal that multiple amino acids need to act in concert to control the function of CaSR ( Conigrave et al . , 2000 , 2004 ) . L-Trp is also important for Ca2+-stimulated CaSR response . First , L-Trp elevated the sensitivity of CaSR toward extracellular Ca2+ ( Conigrave et al . , 2000 ) ( Figure 4G ) . The presence of 10 mM L-Trp lowered the EC50 of extracellular Ca2+ by about 30% . Second , the residues involved in L-Trp binding are crucial for Ca2+-dependent receptor activation . Previous and current studies demonstrate that each of the individual mutations S147A , S170A , Y218A , and E297K abolishes Ca2+-induced receptor response ( Silve et al . , 2005; Zhang et al . , 2002 ) ( Figure 4H–L ) . Furthermore , E297K is a naturally occurring inactivating mutation that can lead to life-threatening hyperparathyroidism ( Bai et al . , 1996; Hendy et al . , 2009; Pollak et al . , 1993 ) . These observations suggest that the binding of an amino acid is required for extracellular Ca2+-sensing by CaSR . We identified four distinct Ca2+-binding sites within each protomer of the active structure using anomalous difference maps , and named these sites 1 through 4 ( or 1′-4’ in the second protomer ) ( Figure 5A , B; Figure 5—figure supplement 1; Table 1—source data 1 ) . The inactive structure revealed electron density at site 2 that is consistent with a bound Ca2+ ion ( Figure 5C , D ) . None of the Ca2+-binding sites observed in the CaSR ECD structures has been reported previously . 10 . 7554/eLife . 13662 . 016Figure 5 . Ca2+-binding sites . ( A ) Active-state structure showing peaks in anomalous difference Fourier map ( magenta mesh; 3σ contour level ) that correspond to bound Ca2+ ions . Sites are labeled 1–4 or 1'-4' for each protomer . ( B ) Specific contacts between CaSR ECD and each bound Ca2+ ion within one protomer of the active structure . Anomalous difference Fourier map ( magenta ) : sites 1–3 , 6σ; site 4 , 4 . 5σ . Fo-Fc difference map ( blue ) : sites 1–3 , 4 . 5σ; site 4 , 2 . 5σ . Distances between Ca2+ and oxygen atoms ( dashed lines ) are within 3 . 0 Å . Dashed lines between water and protein atoms are hydrogen bonds . ( C ) Inactive-state structure showing peaks in anomalous difference Fourier map ( magenta mesh; 3σ ) at Ca2+-binding sites 2 and 2' . ( D ) Specific contacts between CaSR ECD and bound Ca2+ ion within one protomer of the inactive structure . Anomalous difference Fourier map ( magenta ) : 5σ . Fo-Fc difference map ( blue ) : 4 . 5σ . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 01610 . 7554/eLife . 13662 . 017Figure 5—figure supplement 1 . Ca2+-binding sites in the active homodimer . ( A ) Active-state structure of CaSR ECD showing peaks in anomalous difference Fourier map ( magenta mesh; 3σ contour level ) that correspond to bound Ca2+ ions . Sites are labeled 1–4 or 1'-4' for each protomer . ( B , C ) Specific contacts between CaSR ECD and each bound Ca2+ ion within both protomers of the active structure . Anomalous difference Fourier map ( magenta mesh ) : sites 1–3 , 6σ; site 4 , 4 . 5σ . Fo-Fc difference map ( blue mesh ) : sites 1–3 , 4 . 5σ; site 4 , 2 . 5σ . The distances between Ca2+ and oxygen atoms ( dashed lines ) are within 3 . 0 Å . Dashed lines between water and protein atoms are hydrogen bonds . ( D ) Comparison of inactive ( red ) and active ( blue ) structures in the region of each Ca2+-binding site . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 017 Site 1 is located in a loop region at the top of LB1 domain ( Figure 5A , B ) . The bound Ca2+ ion is primarily coordinated by backbone carbonyl oxygen atoms of I81 , S84 , L87 , and L88 ( site 1′ ) . The structural configuration of site 1 is similar in the inactive and active structures even though it is only occupied in the active state ( Figure 5—figure supplement 1 ) . The disease-causing mutation I81M ( Hendy et al . , 2009 ) is located at site 1 , and it abolished Ca2+-dependent receptor response , possibly by disrupting a tightly packed hydrophobic patch adjacent to Ca2+-binding site 1 ( Figure 6A , B ) . This implies that the local conformation of this loop region is important for receptor function , and the Ca2+ ion stabilizes the observed conformation in the active state . 10 . 7554/eLife . 13662 . 018Figure 6 . Mutational analysis of Ca2+-binding sites . ( A , B ) Dose-dependent Ca2+-stimulated IP accumulation ( A ) and intracellular Ca2+ mobilization ( B ) in cells transiently expressing wt or mutant CaSR . Naturally occurring inactivating mutations I81M and T100I are located at various Ca2+-binding sites . The single mutation N102I was designed based on structure to interfere with Ca2+-binding . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 018 The inactive and active structures share a common Ca2+-binding mode at site 2 , suggesting that the bound Ca2+ is an integral part of the CaSR structure ( Figure 5A–D; Figure 5—figure supplement 1 ) . Site 2 is positioned directly above the interdomain crevice in LB1 domain , and it abuts the L-Trp binding site in the cleft . The Ca2+ ion is coordinated by the hydroxyl group of T100 in both states , and by the carboxyl group of N102 through a water molecule in the active structure . In addition , T145 , another residue lining the site , forms part of the L-Trp binding cleft in the active state . Therefore , an intact Ca2+ site 2 provides an essential framework for L-Trp recognition . Indeed , introducing a hydrophobic residue at this site through mutations T100I , N102I , or T145I nearly eliminated Ca2+-induced receptor activity ( Figure 4D; Figure 6A , B ) . Site 3 is positioned at the edge of the interdomain cleft in LB2 domain ( Figure 5A , B ) . The Ca2+ ion is coordinated by the hydroxyl groups of S302 and S303 either directly ( site 3′ ) or indirectly through water molecules . Alignment of the inactive and active structures in this region showed a small conformational change ( Figure 5—figure supplement 1 ) . Ca2+ ion stabilizes a loop conformation that permits an interdomain hydrogen bond between the neighboring LB2 residue S301 and LB1 residue R66 . Such interaction enhances domain closure of CaSR ECD for receptor activation . Among all four Ca2+-binding sites in CaSR ECD structure , site 4 is most closely associated with receptor activation because it directly participates in the formation of an active receptor conformation . Site 4 is part of the homodimer interface formed upon agonist binding , bridging the LB2 domain of one subunit and CR domain of the second subunit ( Figure 5A , B ) . The Ca2+ ion is coordinated by three interfacial residues , including the carboxylate group of D234 and carbonyl oxygen of E231 and G557 ( Figure 5B ) . The natural mutation G557E ( Hendy et al . , 2009 ) reduced the potency of Ca2+ possibly by affecting backbone conformation , thereby weakening the affinity of Ca2+ for this site ( Figure 6A , B ) . Our structural data indicate that the Ca2+ ion at site 4 stabilizes the active conformation of the receptor by facilitating homodimer interactions between the membrane-proximal LB2 and CR domains . The Ca2+ ions have different peak heights in the anomalous difference maps , which are correlated with different Ca2+-occupancies at various sites . The Ca2+ ions at sites 1 and 2 have strong peaks ( 12 . 1–13 . 1 σ ) , indicating that these are high-occupancy sites . The Ca2+ ions at sites 3 ( 7 . 3 – 9 . 0 σ ) and 4 ( 5 . 8 σ ) have weaker anomalous peaks , which suggest low occupancy and possibly low affinity . The Ca2+ ion at site 4 has the weakest peak compared with other sites , which is consistent with the site being occupied only at elevated Ca2+ concentration for receptor activation . We identified a total of four anion-binding sites in the inactive and active CaSR ECD structures based on anomalous difference maps ( 1–4 or 1'-4’ in the second protomer ) ( Figure 7; Figure 7—figure supplement 1; Table 1—source data 1 ) . Sites 1–3 are located above the interdomain cleft in the LB1 domain , and site 4 is part of LB2 domain . 10 . 7554/eLife . 13662 . 019Figure 7 . Anion-binding sites . ( A ) Inactive-state structure showing peaks in anomalous difference Fourier map ( green mesh; 3σ ) that correspond to bound SO42- ions . Sites are labeled 1–3 or 1'-3' for each protomer . ( B ) Specific contacts between CaSR ECD and each bound SO42- ion within one protomer of the inactive structure . Anomalous difference Fourier map ( green ) : 3 . 5σ . Fo-Fc map ( blue ) : 4σ . Dashed lines represent hydrogen bonds . ( C ) Active-state structure showing peaks in anomalous difference Fourier map ( green mesh; 3σ ) that correspond to bound PO43- ions . Sites are labeled 2 and 4 or 2' and 4' for each protomer . ( D ) Specific contacts between CaSR ECD and each bound PO43- ion within one protomer of the active structure . Anomalous difference Fourier map ( green ) : 3 . 5σ . Fo-Fc map ( blue ) : 4σ . ( E ) Active-state structure of CaSR ECD showing the additional hydrogen bonds formed across the interdomain cleft in the absence of any bound anion at sites 1 and 3 ( left ) . Comparison of inactive ( red ) and active ( light blue ) structures in the region of anion binding sites 1 ( center ) and 3 ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 01910 . 7554/eLife . 13662 . 020Figure 7—figure supplement 1 . Anion-binding sites in the active homodimer . ( A ) Active-state structure of CaSR ECD showing the peaks in anomalous difference Fourier map ( 3σ ) that correspond to bound PO43- ions . The sites are labeled 2 and 4 or 2' and 4' for each protomer . ( B , C ) Specific contacts between CaSR ECD and each bound PO43- ion within both protomers of the active structure . Anomalous difference Fourier map ( green ) : 3 . 5σ . Fo-Fc map ( blue ) : 4σ . Dashed lines represent hydrogen bonds . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 020 In the inactive structure , electron densities revealed that anions were bound at sites 1–3 ( Figure 7A , B ) . In the active structure , only sites 2 and 4 are occupied ( Figure 7C , D ) . We modeled the anions as SO42- ions in the inactive state and PO43- ions in the active state given their respective presence in the crystallization reagents . It is also possible that endogenous anions other than SO42- and PO43- are bound at these sites . The anion at site 1 is coordinated by the guanidine group of R62 and backbone nitrogen of Y63 ( Figure 7B ) . In the absence of a bound anion in the active state , the side chain of LB1 residue R62 reaches across the interdomain cleft to form a salt bridge with E277 of LB2 domain ( Figure 7E ) . This contact stabilizes the closed conformation of the CaSR VFT module . The anion bound at site 2 is held in place by multiple hydrogen bonds with the side chains of R66 , R69 , W70 , and S417 , and main chains of R415 , I416 , and S417 ( Figure 7B , D ) . The structural integrity of this site is important for receptor function . Each of the mutations R66H , R69E , and S417L essentially eradicated receptor signaling ( Figure 8A , B ) . Among these , R66H is a disease-associated mutation ( Hendy et al . , 2009; Pidasheva et al . , 2006 ) . 10 . 7554/eLife . 13662 . 021Figure 8 . Mutational analysis of anion-binding sites . ( A , B ) Dose-dependent Ca2+-stimulated IP accumulation ( A ) and intracellular Ca2+ mobilization ( B ) in cells transiently expressing wt or mutant CaSR . Naturally-occurring inactivating mutations R66H , R69E , and S417L are located at anion-binding site 2 ( or 2' ) . ( C ) Effect of SO42- ion on Ca2+-stimulated IP accumulation in cells transiently expressing wild-type CaSR . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 021 Anion-binding site 3 is adjacent to site 2 . The bound anion is also coordinated by R66 and additionally by T412 ( Figure 7B ) . In the active state , the side chain of LB1 residue R66 forms a hydrogen bond with S301 of LB2 domain ( Figure 7E ) . This interaction also serves to maintain the CaSR ECD in a closed conformation . The anion at site 4 is coordinated by H192 , T195 , K225 and R520 in the active structure ( Figure 7D ) . In the absence of a bound anion , the side chains of several binding-site residues are disordered in the inactive structure . This indicates that the anion serves to stabilize the local conformation of the receptor structure . We measured the effect of anion on Ca2+-dependent receptor response and found that the presence of SO42- slightly decreased receptor activity , increasing the EC50 of Ca2+ by approximately 25% ( Figure 8C ) . This finding confirmed that anions have a negative allosteric effect on the receptor .
Our structural analyses of CaSR ECD provide direct evidence that amino acids are agonists of CaSR , and they act concertedly with Ca2+ to achieve full receptor activation . L-Trp , the amino acid used in this study , fits the role of an orthosteric agonist for CaSR . ( 1 ) It binds at the interdomain crevice of the VFT module , the canonical agonist-binding site for class C GPCRs ( Geng et al . , 2013; Kunishima et al . , 2000; Muto et al . , 2007; Tsuchiya et al . , 2002 ) . ( 2 ) L-Trp shares a common receptor-binding mode with the endogenous agonists of mGluRs and GABAB receptor , which are also amino acids or their analogs ( Geng et al . , 2013; Kunishima et al . , 2000; Muto et al . , 2007; Tsuchiya et al . , 2002 ) . The residues involved in agonist recognition are located at the same positions in the structures of these receptors ( Figure 4—figure supplement 1 ) . For example , a conserved serine residue is responsible for securing the carboxylate of L-Trp , glutamate and GABA in CaSR ( S147 ) , mGluRs ( S165 ) ( Kunishima et al . , 2000 ) and GABAB receptor ( S130 ) ( Geng et al . , 2013 ) , respectively . ( 3 ) L-Trp interacts with both LB1 and LB2 domains to facilitate extracellular domain closure , a crucial first step during CaSR activation . In contrast , no Ca2+ ion is found at the putative orthosteric agonist-binding site to induce domain closure . ( 4 ) Mutations of L-Trp-binding residues completely blocked Ca2+-induced IP accumulation and intracellular Ca2+ mobilization ( Silve et al . , 2005; Zhang et al . , 2002 ) , indicating that L-Trp is required for Ca2+-mediated receptor response . ( 6 ) L-Trp directly activates CaSR-mediated intracellular Ca2+ mobilization in the presence of extracellular Ca2+ . On the other hand , Ca2+ ion fulfills at least three functional roles . First , it maintains the structural integrity of CaSR , as manifested by the importance of Ca2+-binding site 2 for receptor function . Second , it is directly involved in receptor activation . Specifically , the Ca2+ ion at site 4 stabilizes the unique homodimer interface between membrane-proximal LB2 and CR domains in the active state . Third , Ca2+ enhances L-Trp binding to CaSR ECD , possibly by reinforcing the L-Trp bound and active conformation of the receptor . Furthermore , the actions of Ca2+ and amino acids on CaSR are inter-dependent . While the presence of extracellular Ca2+ above a threshold level is required for amino-acid-mediated CaSR activation , amino acids increase the sensitivity of the receptor toward Ca2+ . Taken together , we conclude that amino acids and Ca2+ ions are indeed co-agonists of CaSR , acting jointly to trigger receptor activation . This then led us to an intriguing question: How does Ca2+ ion activate the receptor on its own as demonstrated in various cell-based functional assays ? It is possible that amino acids are present in cell culture media and extracellular fluid at sufficient concentration to prime the receptor to respond to increasing concentrations of Ca2+ . We have obtained crystals of CaSR ECD in the absence of any additional amino acids . Nevertheless , the structure showed a stretch of continuous density in the interdomain cleft that may belong to an endogenous ligand ( Figure 4—figure supplement 2; Table 1—source data 2 ) . Despite multiple attempts , we have not been able to identify the structure of this endogenous ligand through mass spectrometry . In light of the L-Trp-bound CaSR ECD structure , we reason that the endogenous ligand may be an amino acid or even a mixture of amino acids . In the inactive structure , the endogenous ligand was likely removed by citrate buffer ( pH 5 . 5 ) during enzymatic deglycosylation . Metabolic balances of Ca2+ and PO43- are linked through hormonal factors such as parathyroid hormone and fibroblast growth factor-23 , which control the homeostasis of both ions ( Brown , 2013; Quinn et al . , 2013; Tyler Miller , 2013 ) . We therefore reason that the physiologically relevant anion bound to CaSR is likely PO43- , which primarily exists in a HPO42-/H2PO4- mixture in biological systems . First , the anion at site 2 is required for structural stability of the receptor . Second , the anions at sites 1 and 3 appear to stabilize the inactive conformation . Binding of anions at these sites prevent favorable interactions across the interdomain cleft that promote VFT closure . Third , the presence of anion decreases CaSR-mediated IP accumulation . Therefore , these anions may exert a negative modulatory effect on CaSR activity . The presence of anion-binding sites in CaSR may also provide a mechanism for CaSR to sense polycations such as polyamines . Increasing concentrations of polyamines could potentially compete with arginine residues at the anion-binding sites to bind to PO43- , thereby prompting the dissociation of PO43- from relatively weak sites , and releasing their inhibitory effect on the receptor . This would drive CaSR toward its active-state conformation . Our hypothesis would predict that polycations with higher number of positive charges will be more effective agonists . Indeed , previous studies have shown that polyamines mediate an increase in intracellular inositol phosphate and Ca2+ accumulation with the rank order seprmine > spermidine > putrescine ( Cheng et al . , 2004; Quinn et al . , 1997 ) . In summary , activation of CaSR involves an intricate interplay of amino acids , Ca2+ , and possibly PO43- ions . Like other GPCRs ( Rosenbaum et al . , 2011 ) , CaSR exists in a conformational equilibrium between inactive and active states ( Figure 9 ) . ( 1 ) CaSR adopts an open conformation in the resting state , and PO43- ions promote the inactive configuration . ( 2 ) An L-amino acid closes the groove in the extracellular VFT module , thereby inducing the formation of a novel homodimer interface between subunits . ( 3 ) Ca2+ ions stabilize the active state by enhancing homodimer interactions between membrane-proximal domains to fully activate the receptor . The combination of agonist-induced VFT closure and specific association of membrane-proximal CR domains in CaSR will likely lead to rearrangement of the transmembrane domains for receptor activation . 10 . 7554/eLife . 13662 . 022Figure 9 . Activation mechanism of CaSR . The schematic diagram shows the equilibrium between the resting and active states of CaSR and the effects of L-amino acid and Ca2+ binding . DOI: http://dx . doi . org/10 . 7554/eLife . 13662 . 022 Structural and functional data suggest a universal activation mechanism for class C GPCRs . First , agonist causes VFT closure in all three receptor systems of CaSR , mGluRs and GABAB receptor . Second , receptor activation requires the association of membrane-proximal domains . For CaSR , this involves the formation of a novel homodimer interface between the LB2 and CR domains . For GABAB receptor , which lacks the CR region , agonist leads to the formation of a large heterodimer interface between the LB2 domains ( Geng et al . , 2013 ) . Similarly for mGluRs , single-molecule fluorescence resonance energy transfer ( FRET ) studies indicate that the LB2 domains of mGluRs come into proximity to stabilize the active state ( Vafabakhsh et al . , 2015 ) . Furthermore , disulfide crosslinking experiments of mGluR demonstrate that a precise association between the CR domains is sufficient for full receptor activation ( Huang et al . , 2011 ) . Third , agonist binding is accompanied by a decrease in separation between the C-terminal ends of extracellular domains . This will likely result in rearrangement of the transmembrane domain dimer for receptor activation . Indeed , FRET and crosslinking studies have detected movement between interacting transmembrane domains of mGluRs and GABAB receptor upon activation ( Matsushita et al . , 2010; Tateyama et al . , 2004; Xue et al . , 2015 ) . In conclusion , agonist-induced VFT closure that leads to the specific association of membrane-proximal domains is a common mechanism shared by all class C GPCRs during ligand-dependent receptor activation .
The extracellular domain of human CaSR ( 1–612 ) was cloned into the pFBDM vector ( Berger et al . , 2004 ) for expression in baculovirus-infected insect cells . Wild-type CaSR was heavily glycosylated , and we took two approaches to remove carbohydrates from CaSR ECD: ( 1 ) elimination of potential glycosylation sites through mutation , and ( 2 ) enzymatic deglycosylation . Mutants of CaSR ECD were generated in which either two or three N-linked glycosylation sites were eliminated . Both mutant constructs contained the mutations N386Q and S402N , while one also had the additional mutation N468Q . A Flag tag was engineered at the C-terminus of each construct to facilitate affinity purification . In agreement with previous studies of full-length CaSR in mammalian cells ( Ray et al . , 1998 ) , we found that expression of CaSR ECD in insect cells was essentially eliminated when more than three glycosylation sites were disrupted . The CaSR ECD glycosylation mutants were constructed to maximize the reduction of carbohydrate content while maintaining a sufficient expression level for structural studies . Wild-type and mutant CaSR ECD were secreted from sf9 insect cells infected with the corresponding recombinant CaSR ECD virus . The CaSR ECD protein was isolated from cell supernatant by anti-Flag antibody ( M2 ) affinity chromatography , and eluted with 100 µg/ml Flag peptide in 50 mM Tris , pH 7 . 5 and 150 mM NaCl . The protein was further purified by gel filtration chromatography ( superdex 200 , GE Healthcare Life Sciences , USA ) in 20 mM Tris , pH 8 . 0 and 150 mM NaCl to remove aggregates . Finally , the CaSR ECD protein was applied to an ion exchange column ( MonoQ , GE ) in 20 mM Tris , pH 8 . 0 and eluted using a linear salt gradient from 0 to 1 M NaCl . All purification procedures were performed in the presence of 10 mM CaCl2 . To isolate the L-Trp-bound receptor , purified CaSR ECD protein was washed extensively with a solution containing 20 mM Tris , pH 8 . 0 , 10 mM L-Trp , and 10 mM CaCl2 prior to crystallization . This would facilitate the complete saturation of L-Trp- and Ca2+-binding sites within the receptor . We applied an enzymatic deglycosylation procedure to wild-type receptor and the CaSR ECD mutant with two glycosylation-site mutations . Each construct was expressed in sf9 cells in the presence of the N-glycosylation processing inhibitor kifunensine ( 1 mg/L ) to produce high mannose glycoproteins that were sensitive to enzymatic cleavage by endoglycosidase H ( Endo H ) . The cell supernatant was applied to an M2 anti-Flag antibody affinity column , and the bound CaSR ECD protein was eluted with 100 µg/ml Flag peptide in 50 mM Tris , pH 7 . 5 and 150 mM NaCl . Well-folded CaSR ECD was separated from aggregates by gel filtration chromatography in 20 mM Tris , pH 8 . 0 , and 150 mM NaCl . The CaSR ECD protein was subsequently digested overnight with Endo H in 50 mM Na Citrate , pH 5 . 5 . The partially deglycosylated CaSR ECD protein was further purified by ion exchange chromatography in 20 mM Tris , pH8 . 0 using a linear salt gradient from 0 to 1 M NaCl . The protein purification and enzymatic deglycosylation procedures were performed either in the absence or presence of 2 mM CaCl2 , which is within the normal range of Ca2+ concentrations in plasma . Wild-type CaSR ECD did not form well-diffracting crystals possibly because the presence of flexible and heterogeneous carbohydrates on the protein surface interfered with crystallization . The CaSR ECD mutant carrying two glycosylation-site mutations formed diffracting crystals after enzymatic deglycosylation . The crystals were obtained at 20°C in 1 . 5 M Li2SO4 , 100 mM Tris pH 8 . 5 in the absence and presence of 2 mM CaCl2 . Protein crystallization was achieved by hanging drop vapor diffusion method using 24-well VDX plates ( Hampton Research , USA ) . The CaSR ECD crystals were flash-cooled with liquid nitrogen in a cryoprotecting solution containing 3 . 0 M Li2SO4 , 100 mM Tris pH 8 . 5 , and the same concentration of CaCl2 as used for crystallization . These CaSR ECD crystals have the space group F222 with one molecule per asymmetric unit ( form I ) . The CaSR ECD mutant carrying three glycosylation-site mutations crystallized under two different conditions . ( 1 ) In the absence of additional amino acids , CaSR ECD formed the best diffracting crystals at 20°C in 0 . 9 M Na Citrate , 100 mM Tris , pH 8 . 0 , and 10 mM CaCl2 . The crystals were obtained using hanging drop vapor diffusion technique . These crystals were flash-cooled with liquid nitrogen in a cryoprotecting solution containing 0 . 9 M Na Citrate , 100 mM Tris , pH 8 . 0 , 10 mM CaCl2 , and 20% glycerol . ( 2 ) Crystals of the L-Trp-bound CaSR ECD mutant were grown at 20°C in 1 . 6 M NaH2PO4 , 0 . 4 M K2HPO4 , 100 mM Na2HPO4/citric acid pH 4 . 2 , 10 mM CaCl2 , and 10 mM L-Trp . The crystals were obtained by sitting drop vapor diffusion method using 24-well Cryschem plates ( Hampton Research ) . These crystals were flashed-cooled with liquid nitrogen in a cryoprotecting solution containing 20% glycerol and all other components of the crystallization solution . The crystals obtained from both conditions belong to space group C2 and have two molecules per asymmetric unit ( form II ) . Diffraction data were collected at the 24ID-C and 24ID-E beamlines of Advanced Photon Source ( APS ) . To minimize radiation damage at low energy , a native dataset was collected at the energy of Se-K edge ( λ=0 . 9792 Å ) for a single form I crystal of CaSR ECD grown in the presence of 2 mM Ca2+ . This native dataset was used to solve the inactive-state CaSR ECD structure . Form I crystals grown in the absence of additional Ca2+ diffracted to much lower resolution than those obtained in the presence of 2 mM Ca2+ and were therefore not used for structural determination . Multiple anomalous data sets were also collected at λ = 1 . 7712 Å for form I crystals of CaSR ECD , each from a single crystal . A total of five anomalous data sets were obtained for CaSR ECD in the absence of Ca2+ . These data sets were integrated individually by XDS ( Kabsch , 2010 ) , and then scaled and merged using CCP4 programs ( Winn et al . , 2011 ) to calculate the anomalous difference Fourier maps in the absence of additional Ca2+ . Similarly , eight anomalous data sets were collected for CaSR ECD in the presence of 2 mM Ca2+ , and combined to determine the anomalous difference Fourier maps in the presence of plasma concentration of Ca2+ . The anomalous difference maps obtained in the absence and presence of Ca2+ revealed peaks that correspond to bound Ca2+ and SO42- ions at the same positions in the CaSR ECD structure . The averaging of multiple data sets enhanced signal to noise in anomalous diffraction ( Liu et al . , 2014 ) . A total of four anomalous data sets were collected at low energy ( λ=1 . 7712 Å ) for L-Trp-bound form II crystals of CaSR ECD , each from a single crystal . The four data sets were processed individually by XDS ( Kabsch , 2010 ) and CCP4 programs ( Winn et al . , 2011 ) . The data set with the highest resolution limit ( 2 . 6 Å ) was used to determine the active-state CaSR ECD structure in the presence of L-Trp and excess Ca2+ . The four data sets were also scaled and merged for the calculation of anomalous difference Fourier maps . The structure of L-Trp-bound CaSR ECD in form II crystal was solved by molecular replacement . A two-fold non-crystallographic symmetry ( NCS ) axis was identified from the self-rotation function . Polyalanine models generated from the individual LB1 and LB2 domains of mGluR3 ECD structure ( Muto et al . , 2007 ) ( PDB code: 2E4U ) were used as search probes to locate the VFT modules of both CaSR ECD molecules in the crystal . After phase improvement by two-fold NCS-averaging , additional density appeared for the CR domain of CaSR ECD . A complete model of the CaSR ECD homodimer was developed through a succession of manual fittings and iterative refinement . The final model contained the CaSR ECD residues 20–119 , 135–359 and 393–598 in one protomer , and residues 22–122 , 136–360 and 392–602 in the other protomer . Each protomer contained eight intrasubunit disulfide bridges . Electron density was visible for carbohydrate residues ( N-acetyl-glucosamine ) attached to Asn90 , Asn287 , Asn488 , Asn541 of one protomer , and Asn541 of its dimer partner . Finally , each protomer was also bound to one L-Trp ligand , four Ca2+ ions , and two PO43- ions . The Ca2+ and PO43- ions were identified by anomalous difference Fourier maps calculated using data collected at a wavelength of 1 . 7712 Å . We modeled PO43- as the anions in the CaSR ECD structure because they were major components of the crystallization solution . The anomalous scattering of Ca2+ at 1 . 75 Å has been used successfully to identify the Ca2+-binding sites in a voltage-gated calcium channel ( Tang et al . , 2014 ) . Ramachandran analysis places 94 . 9% of all residues in favored regions and 0 . 28% in outlier regions . A molecular replacement solution was also obtained for CaSR ECD in form II crystal in the presence of 10 mM Ca2+ but absence of any additional amino acid . The protein structure of L-Trp-bound CaSR ECD without any bound ligand was used as the search model . A stretch of electron density at the interdomain cleft region indicates the presence of an endogenous ligand . This structure was partially refined because we have not been able to identify the structure of the endogenous ligand by mass spectrometry despite multiple attempts . The structure of partially deglycosylated CaSR ECD in form I crystal was solved using the individual LB1 , LB2 , and CR domains of L-Trp-bound CaSR ECD as search models . The asymmetric unit of the form II crystals contained one CaSR ECD protomer , and it forms a disulfide-linked homodimer with a crystallographic symmetry-related molecule . The final model for a single protomer contained CaSR ECD residues 21–130 and 136–603 . In addition to the eight intrasubunit disulfide linkages within each protomer , the inter-subunit disulfide bond formed by C129 was ordered in the form I CaSR ECD crystal structure . Electron density was also visible for carbohydrate residues ( N-acetyl-glucosamine ) attached to Asn261 , Asn287 , Asn446 , Asn468 , Asn488 , Asn541 , and Asn594 . In addition , each protomer had one Ca2+ ion , and three SO42- ions . The Ca2+ and SO42- ions were identified by anomalous difference Fourier maps calculated from data collected at 1 . 7712 Å . We reasoned that SO42- ions were most likely the anions bound to CaSR ECD since they were used as the precipitant for crystallization . Geometric analysis places 92 . 9% of all residues in favored regions and 0 . 35% as outliers . In the form I CaSR ECD structure , the interdomain groove was empty except for a water molecule; it also forms crystal contacts with a symmetry-related molecule . Molecular replacement searches were carried out using PHASER ( McCoy et al . , 2007 ) . Model building was performed with COOT ( Emsley and Cowtan , 2004 ) . Structural refinement was executed using BUSTER ( Roversi et al . , 2000 ) . Anomalous difference Fourier maps were calculated with PHENIX ( Adams et al . , 2010 ) . Ramachandran statistics were obtained for each structure using MolProbity ( Chen et al . , 2010 ) . Pairwise structural comparison was performed using LSQMAN ( Novotny et al . , 2004 ) . Protein contacts were analyzed using the CCP4 program CONTACT ( Collaborative Computational Project , 1994 ) . Software installation support was provided by SBGrid ( Morin et al . , 2013 ) . Full-length human CaSR was cloned into a pcDNA3 . 1 ( + ) vector ( Life Technologies , USA ) for expression in human embryonic kidney ( HEK ) 293 cells . A Flag tag was inserted after the signal peptide of CaSR . Mutants of CaSR were constructed using the QuikChange mutagenesis system ( Agilent Technologies , California , USA ) . HEK293 T/17 cells ( ATCC ) were transfected by Lipofectamine 2000 ( Life Technologies ) with wild-type or mutant CaSR plasmids . Cells permeabilized with 0 . 5% Triton X100 were used to determine the total expression level of CaSR in transfected cells . Untreated cells were used to determine the cell surface expression level of the receptor . The amount of surface protein detected for each construct was normalized to that found in the total cell lysate . The cells were blocked with 5% milk , and then incubated with mouse anti-Flag M1 antibody ( Sigma-Aldrich , USA ) as the primary antibody to measure CaSR expression . Donkey anti-mouse IRDye 800-labeled antibody ( LiCor Biosciences , Nebraska , USA ) was used as the secondary antibody . Fluorescent signals were measured with an Odyssey Infrared Imager ( LiCor ) . The results of three independent experiments were used for statistical analysis . Most of the mutants were expressed on the cell surface at levels comparable to that of the wild-type receptor . The exceptions were R66H , R69E , I81M , T100I , N102I , and S417L , which reduced the surface expression of the mutant receptors to approximately 70–75% of the wild-type level . Measurement of inositol phosphate ( IP ) accumulation was carried out using the homogenous time-resolved fluorescence ( HTRF ) IP-one Tb kit ( Cisbio Bioassays , USA ) . This assay quantifies the accumulation of inositol 1-monophosphate ( IP1 ) , a degradation product of inositol 1 , 4 , 5-triphosphate ( IP3 ) that is stable in the presence of LiCl . Briefly , HEK293 T/17 cells were transiently transfected with wild-type or mutant full-length CaSR plasmids . The cells were stimulated with increasing concentrations of Ca2+ two days post transfection in a buffer containing 10 mM HEPES pH 7 . 4 , 0 . 5 mM MgCl2 , 4 . 2 mM KCl , 146 mM NaCl , 5 . 5 mM glucose , and 50 mM LiCl . The reaction mixture was then incubated with an IP1 analog coupled to a d2 fluorophore ( acceptor ) and an anti-IP1 monoclonal antibody labeled with Eu Cryptate ( donor ) . The IP1 produced by cells upon activation of CaSR competes with IP1 coupled to the dye d2 for binding to the anti-IP1 antibody . The resulting FRET signal is inversely proportional to the concentration of IP1 in the sample . The fluorescence data was acquired at 620 and 665 nm using an EnVision plate reader ( Perkin Elmer , USA ) after laser excitation at 320 nm . The FRET signal was calculated as the fluorescence ratio ( 665 nm/620 nm ) . Basal activity was determined in the absence of Ca2+ stimulation . The percent stimulation of each receptor mutant was calculated based on the wild-type response obtained under the same condition . Data analysis was performed using the non-linear regression algorithms in Prism ( GraphPad Software , USA ) . Data points represent average ± s . e . m . of triplicate measurements . The effect of anion on Ca2+-stimulated IP accumulation was determined using SO42- instead of PO43- because of the modest solubility of CaHPO4 ( 1 . 5 mM ) , the predominant form of calcium phosphate salt at physiological pH ( 7 . 4 ) . Specifically , dose-dependent Ca2+-induced IP accumulation was measured in the absence and presence of 10 mM Li2SO4 in the reaction buffer . Measurement of intracellular Ca2+ mobilization was performed using a FLIPR Fluorescent Imaging Plate Reader ( FLIPR ) Calcium Assay kit ( Molecular Devices ) . Briefly , HEK293 cells were transiently transfected with wild-type or mutant full-length CaSR plasmids and cultured overnight . The cells were incubated in a loading medium containing 50% Opti-MEM , 50% Ca2+-free Hank’s balanced salt solution , 2 . 5% fetal bovine serum , 20 mM HEPES pH 7 . 4 , 2 . 5 mM probenecid and 2 µM fluorescent Ca2+ indicator Fluo-4 AM ( Life Technologies ) for 1 hr , and then placed into the FLIPR . CaCl2 ( prepared in Hank’s balanced salt solution and 20 mM HEPES , pH 7 . 4 ) was added at 10 s , and changes in fluorescence were monitored over a period of 250 s following excitation at a wavelength of 488 nm and detection at 510–560 nm . Data analysis was performed using the non-linear regression algorithms in Prism ( GraphPad Software ) . Data points represent average ± s . e . m . of triplicate measurements . To measure the potentiating effect of L-Trp on the response of CaSR to extracellular Ca2+ , HEK293 cells were transiently transfected with wild-type CaSR plasmid and cultured for 48 hr . The cells were washed three times with an assay buffer ( 20 mM HEPES , pH 7 . 4 , 1 mM CaCl2 , 1 mM MgCl2 , 1 mg/ml BSA , 5 . 5 mM D-glucose , 5 . 3 mM KCl , 138 mM NaCl , 4 . 2 mM NaHCO3 , 0 . 44 mM KH2PO4 , 0 . 34 mM Na2HPO4 ) to remove any endogenously bound ligand , and pre-incubated with 10 mM L-Trp for 20 min before stimulation with extracellular Ca2+ . HEK293 cells that stably expressed the CaSR ( HEK-CaSR cells ) were cultured on coverslips in 24-well plates , and loaded in the dark with fura2-AM ( 5 µM ) in physiological saline solution ( PSS; 125 mM NaCl , 4 mM KCl , 0 . 1% w/v D-glucose , 1 mM MgCl2 , 20 mM HEPES-NaOH , pH 7 . 45 ) that contained 1 mM CaCl2 , 0 . 8 mM NaH2PO4 and 1 mg/ml BSA for 1 . 5 hr at 37°C . The cells were washed and stored in a fura2-AM-free loading solution prior to experiments . Fura2-loaded HEK-CaSR cells were transferred into a perifusion chamber , placed in the light path of a Zeiss Axiovert fluorescence microscope ( Zeiss , USA ) , and perifused with PSS containing various concentrations of Ca2+ and L-Trp . Fura2-loaded HEK-CaSR cells were excited by a Lambda DG-4 150 Watt xenon light source ( Sutter , Novato , USA ) , using alternating wavelengths of 340 and 380 nm at 0 . 5 s intervals , and imaged at 510 nm . For each data set , regions of interest corresponding to the locations of 10 individual cells were selected and digital images were captured using an AxioCam camera controlled by Stallion SB . 4 . 1 . 0 PC software ( Intelligent Imaging Innovations , USA ) . Single-cell intracellular Ca2+ mobilization data consisted of excitation ratios ( F340/F380 ) plotted against time ( min ) . Ratio data were integrated , expressed as integrated response units ( IRUs ) , and corrected for baseline ( PSS containing 0 . 5 mM Ca2+ ) . Concentration-dependent response data were fitted to the equation:R=b+ ( a−b ) Cn/ ( en+Cn ) in which: a = maximum response; b = basal response; C = activator concentration; e = EC50 in mM; and n = Hill co-efficient . Estimates of curve-fitting parameters were obtained using Prism ( GraphPad Software ) . Wild-type CaSR ECD protein was purified by anti-Flag M2 antibody affinity chromatography and gel filtration chromatography in the absence of additional CaCl2 and amino acids . The protein was incubated in 100 mM NaCitrate , pH 5 . 5 overnight to remove any endogenously bound ligand . The CaSR ECD protein was then separated from any free ligand by gel filtration chromatography in 20 mM HEPES , pH 8 . 0 and 150 mM NaCl . Binding of 0 . 1 mM L-[3H]Trp ( 1 Ci/mmol ) to 600 ng of purified CaSR ECD was measured with the scintillation proximity assay ( SPA ) using 1 . 25 mg/mL Yttrium silicate ( YSi ) protein A SPA beads ( PerkinElmer , USA ) in conjunction with 0 . 125 ng/mL of anti-Flag M2 antibody in 20 mM HEPES , pH 8 . 0 and 150 mM NaCl at 4°C . Increasing concentrations of non-radioactive L-Trp were added to compete for receptor binding in the presence of 0 mM or 2 mM CaCl2 . Each reaction was also performed in the presence of 60 mM non-radioactive L-Trp for background correction . The reactions were allowed to proceed for 1 hr to reach equilibrium . The plates were then counted in a Microbeta counter ( PerkinElmer , USA ) . Data were analyzed using the non-linear regression algorithms in Prism ( GraphPad ) . | Calcium ions regulate many processes in the human body . The calcium-sensing receptor , called CaSR , is responsible for maintaining a stable level of calcium ions in the blood . This receptor can detect small changes in the concentration of calcium ions , and activates signalling events within the cell to restore the level of calcium ions back to normal . Abnormal activity of this receptor is associated with severe diseases in humans CaSR is found in the surface membrane of cells and belongs to a family of proteins called G-protein coupled receptors . Much of the protein extends out of the cell and interacts with calcium ions , phosphate ions and certain other molecules such as amino acids . However , it was not well understood how these small molecules bind to CaSR and how this activates the receptor . Geng et al . have now used a technique called X-ray crystallography to view the three-dimensional structure of the exterior domain of CaSR in its resting state and active state . These structures revealed that , contrary to expectations , calcium ions are not the main activator of the receptor . Instead , Geng et al . found that CaSR adopts an inactive state in the absence or presence of calcium ions , while the active state only forms when an amino acid is bound . Furthermore investigation showed that calcium ions are needed to stabilise the active form , while phosphate ions keep the inactive form stable . Geng et al . also identified the shape changes that must occur as CaSR transitions from its inactive to its active state . In particular , an amino acid binding to the exterior domain causes it to close like a venus flytrap , which is a crucial step in activating the receptor . Taken together , the findings show that the amino acids and calcium ions act jointly to fully activate CaSR . The next steps are to determine the structure of the entire receptor with and without its small molecule partners and to use these structures to design drugs that can alter CaSR’s activity in order to treat human diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2016 | Structural mechanism of ligand activation in human calcium-sensing receptor |
The light environment greatly impacts human alertness , mood , and cognition by both acute regulation of physiology and indirect alignment of circadian rhythms . These processes require the melanopsin-expressing intrinsically photosensitive retinal ganglion cells ( ipRGCs ) , but the relevant downstream brain areas involved remain elusive . ipRGCs project widely in the brain , including to the central circadian pacemaker , the suprachiasmatic nucleus ( SCN ) . Here we show that body temperature and sleep responses to acute light exposure are absent after genetic ablation of all ipRGCs except a subpopulation that projects to the SCN . Furthermore , by chemogenetic activation of the ipRGCs that avoid the SCN , we show that these cells are sufficient for acute changes in body temperature . Our results challenge the idea that the SCN is a major relay for the acute effects of light on non-image forming behaviors and identify the sensory cells that initiate light’s profound effects on body temperature and sleep .
Many essential functions are influenced by light both indirectly through alignment of circadian rhythms ( photoentrainment ) and acutely by a direct mechanism ( sometimes referred to as ‘masking’ ) ( Mrosovsky et al . , 1999; Altimus et al . , 2008; Lupi et al . , 2008; Tsai et al . , 2009; LeGates et al . , 2012 ) . Dysregulation of the circadian system by abnormal lighting conditions has many negative consequences , which has motivated decades of work to identify the mechanisms of circadian photoentrainment ( Golombek and Rosenstein , 2010 ) . In contrast , it has only recently become apparent that light exposure can also acutely influence human alertness , cognition , and physiology ( Chellappa et al . , 2011 ) . As a result , there is a developing awareness of light quality in everyday life ( Lucas et al . , 2014 ) . It is therefore essential to human health and society to elucidate the circuitry and coding mechanisms underlying light’s acute effects . Intriguingly , a single population of retinal projection neurons—intrinsically photosensitive retinal ganglion cells ( ipRGCs ) —have been implicated in the circadian and acute effects of light on many functions , including activity , sleep , and mood ( Göz et al . , 2008; Güler et al . , 2008; Hatori et al . , 2008; LeGates et al . , 2012; Fernandez et al . , 2018 ) . ipRGCs integrate light information from rods , cones , and their endogenous melanopsin phototransduction cascade ( Schmidt et al . , 2011 ) , and relay that light information to over a dozen central targets ( Hattar et al . , 2006; Ecker et al . , 2010 ) . However , the circuit mechanisms mediating ipRGC-dependent functions are largely unknown . One notable exception is the control of circadian photoentrainment . It is accepted that ipRGCs mediate photoentrainment by direct innervation of the master circadian pacemaker , the suprachiasmatic nucleus ( SCN ) of the hypothalamus ( Göz et al . , 2008; Güler et al . , 2008; Hatori et al . , 2008; Jones et al . , 2015 ) . This is supported by studies demonstrating that genetic ablation of ipRGCs results in mice with normal circadian rhythms that ‘free-run’ with their endogenous rhythm , independent of the light/dark cycle ( Göz et al . , 2008; Güler et al . , 2008; Hatori et al . , 2008 ) . Further , mice with genetic ablation of all ipRGCs except those that project to the SCN and intergeniculate leaflet ( IGL ) display normal circadian photoentrainment ( Chen et al . , 2011 ) , suggesting that ipRGC projections to the SCN/IGL are sufficient for photoentrainment . In comparison , the mechanisms by which ipRGCs mediate acute light responses remain largely a mystery . Genetic ablation of ipRGCs or their melanopsin phototransduction cascade blocks or attenuates the acute effects of light on sleep ( Altimus et al . , 2008; Lupi et al . , 2008; Tsai et al . , 2009 ) , wheel-running activity ( Mrosovsky and Hattar , 2003; Güler et al . , 2008 ) , and mood ( LeGates et al . , 2012; Fernandez et al . , 2018 ) . This dual role of ipRGCs in circadian and acute light responses suggests they may share a common circuit mechanism . However , whether the circuit basis for ipRGCs in the acute effects of light and circadian functions is through common or divergent pathways has yet to be determined . ipRGCs project broadly in the brain beyond the SCN ( Hattar et al . , 2002; Hattar et al . , 2006; Gooley et al . , 2003; Baver et al . , 2008 ) . Additionally , ipRGCs are comprised of multiple subpopulations with distinct genetic , morphological , and electrophysiological signatures ( Baver et al . , 2008; Schmidt and Kofuji , 2009; Ecker et al . , 2010; Schmidt et al . , 2011 ) and distinct functions ( Chen et al . , 2011; Schmidt et al . , 2014 ) . Though there are rare exceptions ( Chen et al . , 2011; Schmidt et al . , 2014 ) , the unique roles played by each ipRGC subsystem remain largely unknown . It is currently unknown whether distinct ipRGC subpopulations mediate both the acute and circadian effects of light , and two major possibilities exist for how this occurs: ( 1 ) ipRGCs mediate both acute and circadian light responses through their innervation of the SCN or ( 2 ) ipRGCs mediate circadian photoentrainment through the SCN , but send collateral projections elsewhere in the brain to mediate acute light responses . To date , the predominant understanding has centered on a role for the SCN in both acute and circadian responses to light ( Muindi et al . , 2014; Morin , 2015; Bedont et al . , 2017 ) . However , this model has been controversial due to complications associated with SCN lesions ( Redlin and Mrosovsky , 1999 ) and alternative models proposing a role for direct ipRGC input to other central targets ( Redlin and Mrosovsky , 1999; Lupi et al . , 2008; Tsai et al . , 2009; Hubbard et al . , 2013; Muindi et al . , 2014 ) . Here , we sought to address the question of how environmental light information—through ipRGCs—mediates both the circadian and acute regulation of physiology . To do so , we investigated the ipRGC subpopulations and coding mechanisms that mediate body temperature and sleep regulation by light . We find that a molecularly distinct subset of ipRGCs is required for the acute , but not circadian , effects of light on thermoregulation and sleep . These findings suggest that , contrary to expectations , functional input to the SCN is not sufficient to drive the acute effects of light on these behaviors . These findings provide new insight into the circuits through which light regulates behavior and physiology .
To identify mechanisms of acute thermoregulation , we maintained mice on a 12 hr/12 hr light/dark cycle and then presented a 3 hr light pulse two hours into the night ( Zeitgeber time 14 , ZT14 ) while measuring core body temperature ( Figure 1A ) . The nocturnal light pulse paradigm is well-established for studying acute regulation of sleep and wheel-running activity ( Mrosovsky et al . , 1999; Mrosovsky and Hattar , 2003; Altimus et al . , 2008; Lupi et al . , 2008 ) . We focused first on body temperature because of its critical role in cognition and alertness ( Wright et al . , 2002; Darwent et al . , 2010 ) , sleep induction and quality ( Kräuchi et al . , 1999 ) , metabolic control ( Kooijman et al . , 2015 ) , and circadian resetting ( Buhr et al . , 2010 ) . Body temperature photoentrains to the light/dark cycle with peaks during the night and troughs during the day ( Figure 1B ) . Both rodents and humans utilize ocular light detection to acutely adjust body temperature in response to a nocturnal light pulse ( Dijk et al . , 1991; Cajochen et al . , 2005 ) , though how this body temperature change is initiated by the retina and relayed to the brain is unknown . When we presented wildtype mice with a nocturnal light pulse , we observed a decrease in both body temperature and general activity compared to the previous night ( Figure 1C ) . The decrease in body temperature and activity was sustained for the entire 3 hr stimulus , with moderate rundown ( Figure 1C ) . We observed that acute body temperature regulation only occurred at relatively bright light intensities ( >100 lux ) ( Figure 1—figure supplement 1 ) . This , in combination with previous reports that body temperature regulation is most sensitive to short-wavelength light ( Cajochen et al . , 2005 ) , suggested that it might be mediated by the insensitive and blue-shifted melanopsin phototransduction ( Lucas et al . , 2001; Do et al . , 2009 ) . To test this , we measured body temperature in mice lacking either functional rods and cones ( melanopsin-only: Gnat1-/-; Gnat2-/- ) or lacking melanopsin ( melanopsin KO: Opn4-/- ) . Both genotypes photoentrained their body temperature ( Figure 1D , E ) , with an amplitude indistinguishable from wildtype ( Figure 1F ) . However , we found that acute body temperature decrease to a nocturnal light pulse was present in melanopsin-only mice ( Gnat1-/-; Gnat2-/- ) ( Figure 1G , H and Figure 1—figure supplement 2 ) , but absent from melanopsin knockout mice ( Opn4-/- ) ( Figure 1I , J and Figure 1—figure supplement 2 ) . This indicates that melanopsin is critical for light’s ability to drive acute body temperature decreases , as it is for acute sleep induction ( Altimus et al . , 2008; Lupi et al . , 2008; Tsai et al . , 2009 ) . These results suggest that ipRGCs are the only retinal cells that are necessary and sufficient for acute thermoregulation by light . ipRGCs comprise multiple subtypes ( M1-M6 ) with distinct gene expression profiles , light responses , and central projections ( Schmidt et al . , 2011; Quattrochi et al . , 2019 ) , prompting us to ask which subtypes mediate acute thermoregulation . ipRGCs can be molecularly subdivided based on whether they express the transcription factor Brn3b . Brn3b ( + ) ipRGCs project to many structures including the olivary pretectal nucleus ( OPN ) and dorsal lateral geniculate nucleus ( dLGN ) , but largely avoid the SCN ( Chen et al . , 2011; Li and Schmidt , 2018 ) . In contrast , Brn3b ( – ) ipRGCs project extensively to the SCN and intergeniculate leaflet ( IGL ) , while avoiding the OPN and dLGN ( Chen et al . , 2011 ) . Non-M1 ( i . e . M2-M6 ) ipRGC subtypes express Brn3b , along with the majority of M1 ipRGCs . Interestingly , just 200 ( out of 700–800 ) M1 ipRGCs lack any Brn3b expression ( Chen et al . , 2011 ) . Ablation of Brn3b ( + ) ipRGCs using melanopsin-Cre and a Cre-dependent diphtheria toxin knocked into the Brn3b locus ( Brn3b-DTA: Opn4Cre/+;Brn3bzDTA/+ ) removes virtually all ipRGC input to brain areas aside from the SCN and IGL ( Chen et al . , 2011; Li and Schmidt , 2018 ) , and these mice lack a pupillary light reflex and show deficits in contrast sensitivity , but retain circadian photoentrainment of wheel-running activity ( Chen et al . , 2011; Schmidt et al . , 2014 ) . When we measured body temperature in Brn3b-DTA mice , we found that their body temperature was photoentrained with a similar amplitude to controls ( Figure 2A–C ) . However , despite the presence of melanopsin in the Brn3b ( - ) ipRGCs of Brn3b-DTA mice ( Opn4Cre/+;Brn3bzDTA/+ ) , they did not acutely decrease body temperature in response to a nocturnal light pulse ( Figure 2F , G ) . Importantly , melanopsin heterozygous littermate controls ( Opn4Cre/+ ) displayed normal acute thermoregulation by light ( Figure 2D , E ) , indicating that halving melanopsin gene dosage is not the cause of the impaired body temperature decrease in Brn3b-DTA mice . Additionally , when we compared the change in body temperature of Control to Brn3b-DTA mice during that light pulse , we found that Control mice showed a significantly larger decrease in body temperature ( Figure 2—figure supplement 1 ) . These results demonstrate that Brn3b ( + ) ipRGCs are required for acute thermoregulation by light but not photoentrainment of body temperature and reveal that light information to the SCN is sufficient for circadian photoentrainment of body temperature , but not its acute regulation . Our data thus far suggest that there are two functionally distinct populations of ipRGCs that regulate thermoregulation: ( 1 ) Brn3b ( – ) ipRGCs that project to the SCN to mediate circadian photoentrainment of body temperature and ( 2 ) Brn3b ( + ) ipRGCs that project elsewhere in the brain and are necessary to mediate acute thermoregulation . If Brn3b ( + ) ipRGCs are not just necessary , but also sufficient , for acute thermoregulation , then activation of this population at ZT14 should result in a body temperature decrease . To test if Brn3b ( + ) ipRGCs are sufficient for acute thermoregulation , we expressed a chemogenetic activator in Brn3b ( + ) RGCs ( Figure 3A , Brn3bCre/+ with intravitreal AAV2-hSyn-DIO-hM3Dq-mCherry , we refer to these mice as Brn3b-hM3Dq ) . As a control , we also injected this virus into Control ( Brn3b+/+ ) littermates . We then injected both genotypes first with PBS at ZT14 on the first night , and CNO at ZT14 on the second night . This technique allowed for statistical within animal comparisons of body temperature changes in response to PBS versus CNO injection . Importantly , CNO did not cause a significant decrease in body temperature in the absence of hM3Dq ( Figure 3—figure supplement 1 ) . This technique allowed us to acutely activate the Brn3b ( + ) RGCs with the DREADD agonist clozapine N-oxide ( CNO ) ( Armbruster et al . , 2007 ) . We found that after intravitreal viral delivery , many RGCs were infected , including melanopsin-expressing ipRGCs ( Figure 3A and Figure 3—figure supplement 1 ) . The body temperature of Brn3b-hM3Dq mice photoentrained to a normal light/dark cycle ( Figure 3B ) . Following CNO administration to Brn3b-hM3Dq mice at ZT14 to depolarize Brn3b ( + ) RGCs , we observed a robust decrease in body temperature that lasted at least 6 hr ( Figure 3D ) . Importantly , PBS administration in Brn3b-hM3Dq mice ( Figure 3C ) and nocturnal CNO administration in wildtype control mice ( Figure 3—figure supplement 2 ) had no measurable effect on body temperature , while CNO administration significantly decreased body temperature in Brn3b-hM3Dq compared to pre-injection temperature ( Figure 3—figure supplement 2 ) . Together , these results demonstrate that Brn3b ( + ) ipRGCs mediate the acute effects of light on body temperature though extra-SCN projection ( s ) , while Brn3b ( – ) ipRGCs mediate circadian photoentrainment of body temperature by projections to the SCN and/or IGL . We next examined the contribution of Brn3b ( + ) and Brn3b ( - ) ipRGCs to sleep . To do this , we used EEG and EMG recordings to compare the sleep behavior of Control ( Opn4Cre/+ ) and Brn3b-DTA mice . We first analyzed the daily sleep patterns and proportion of rapid eye movement ( REM ) and non-REM ( NREM ) sleep in Control and littermate Brn3b-DTA animals . We found that Brn3b-DTA mice show normal photoentrainment of sleep and similar percent time of sleep across the 24 hour day , with only one 30 min bin at ZT12 ( light offset ) showing a significant difference between Control and Brn3b-DTA animals ( Figure 4A , B ) . This is consistent with previous reports of normal circadian photoentrainment of daily activity rhythms in Brn3b-DTA mice ( Chen et al . , 2011 ) . Control and Brn3b-DTA mice also showed similar total percent time awake or asleep across an entire day ( Figure 4C ) , though Brn3b-DTA mice showed a small , but significant , increase in the proportion of total sleep that was classified as NREM and decrease in the proportion of total sleep that was classified as REM ( Figure 4—figure supplement 1A ) . We hypothesized that this small difference in sleep at lights-off in Brn3b-DTA mice could be due to a defect in their acute response to light for sleep modulation . To test this , we subjected mice to a 3 hr light pulse from ZT14–17 ( Altimus et al . , 2008 ) , when the homeostatic drive for sleep is low and Control and Brn3b-DTA animals display similar amounts of sleep ( Figure 4A , B ) . We found that in Control mice , a light pulse decreased time awake and increased time asleep relative to baseline ( previous day ) ( Figure 4C , D ) , while in Brn3b-DTA mice a light pulse caused no change in total percent time asleep or awake ( Figure 4F , G ) , but moderately increased sleep in the first 30 min bin ( Figure 4F ) . Importantly , when we compared the time spent asleep during the light pulse between control and Brn3b-DTA animals , the control mice slept significantly more ( Figure 4—figure supplement 2 ) . Neither Control nor Brn3b-DTA animals showed any change in proportion of non-REM or REM sleep in response to the light pulse ( Figure 4—figure supplement 1B , C ) . These data show that Brn3b ( + ) ipRGCs are necessary for the acute light induction of sleep . Consistent with our body temperature data , although Brn3b-DTA mice have apparently normal input to the SCN and show normal circadian photoentrainment of wheel-running activity ( Chen et al . , 2011 ) , body temperature ( Figure 2 ) , and sleep ( Figure 4 ) , this ipRGC innervation of the SCN is not sufficient to drive the normal light induction of sleep . These disruptions in light’s acute effects on thermoregulation and sleep are circuit specific effects because Brn3b-DTA mice showed robust inhibition of wheel running behavior to a 3 hr light pulse delivered from ZT14-17 ( Figure 4—figure supplement 3 ) .
We show here that for the same physiological outcome , the acute effects of light are relayed through distinct circuitry from that of circadian photoentrainment , despite both processes requiring ipRGCs . Our results suggest that for thermoregulation and sleep , ipRGCs can be genetically and functionally segregated into Brn3b ( + ) ‘acute’ cells , and Brn3b ( – ) ‘circadian’ cells . Because Brn3b ( + ) cells largely avoid the SCN , and Brn3b ( – ) cells preferentially target the SCN , our results discount a critical role for the SCN in acute light responses in these two behaviors , and instead implicate direct ipRGC projections to other brain areas ( Gooley et al . , 2003; Hattar et al . , 2006 ) . Surprisingly , Brn3b ( - ) cells are sufficient to drive the acute and circadian effects of light on wheel running activity , demonstrating further divergence in the circuits mediating the acute effects of light on behavior , and suggesting that , unlike for thermoregulation and sleep , acute and circadian regulation of activity is driven via the SCN . Our results indicate that activation of Brn3b ( + ) RGCs at ZT14 using the Brn3bCre line in combination with Gq-DREADDs is sufficient to induce a body temperature decrease . Because other ( non-ipRGC ) RGC types express Brn3b ( Badea et al . , 2009 ) , this manipulation likely also activated multiple non-ipRGCs in addition to Brn3b ( + ) ipRGCs . However , our data indicate that melanopsin signaling ( Figure 1 ) , and therefore ipRGCs , are required for the acute effects of light on thermoregulation . Moreover , when we ablate Brn3b ( + ) ipRGCs , this acute effect of light on thermoregulation is also lost ( Figure 2 ) , again arguing for a necessity of ipRGCs for this behavior . Therefore , though we are unable to specifically activate only Brn3b ( + ) ipRGCs using available genetic tools , we think it highly likely that the temperature changes driven by the activation of all Brn3b ( + ) RGCs is occurring through ipRGCs . The specific Brnb ( + ) ipRGC subtypes that mediate the light’s acute effects on body temperature and sleep remain a mystery . A majority of all known ipRGC subtypes ( M1–M6 ) are lost in Brn3b-DTA mice ( Chen et al . , 2011 ) , with the exception of a subset of ~200 M1 ipRGCs . In agreement with this , ipRGC projections to all minor hypothalamic targets are lost in Brn3b-DTA mice , while innervation of the SCN and part of the IGL remains intact ( Chen et al . , 2011; Li and Schmidt , 2018 ) . This suggests that all non-M1 subtypes and a portion of M1 ipRGCs are Brn3b ( + ) . Each subtype has a distinct reliance on melanopsin versus rod/cone phototransduction for light detection ( Schmidt and Kofuji , 2009 ) . The necessity and sufficiency of melanopsin in mediating acute effects of light on body temperature ( Figure 1 ) and sleep ( Altimus et al . , 2008; Lupi et al . , 2008; Tsai et al . , 2009 ) suggests that a subtype with strong melanopsin , but weak rod/cone photodetection is responsible – possibly either M1 or M2 cells . However , experiments to tease this apart will require novel methods to specifically manipulate ipRGC subtypes that are currently unavailable . The brain areas that mediate the acute effects of light on physiology are essentially unknown . There are many candidate areas that both receive direct ipRGC innervation and have been documented to be involved in light’s acute effects on physiology , including the preoptic areas ( Muindi et al . , 2014 ) , the ventral subparaventricular zone ( Kramer et al . , 2001 ) , and the pretectum/superior colliculus ( Miller et al . , 1998 ) . Aside from the SCN , ipRGC projections to the median ( MPO ) and ventrolateral preoptic ( VLPO ) areas have been the most widely supported . The preoptic areas are involved in sleep and body temperature regulation ( Szymusiak and McGinty , 2008; Nakamura , 2011 ) and are activated by an acute light pulse ( Lupi et al . , 2008; Tsai et al . , 2009 ) . In support of our behavioral findings , ipRGC projections to each of these areas is lost in Brn3b-DTA animals ( Li and Schmidt , 2018 ) . However , ipRGC projections to these areas are sparse ( Gooley et al . , 2003; Hattar et al . , 2006 ) , suggesting their activation by light may be indirect . In contrast , the superior colliculus ( SC ) and pretectum receive robust innervation from ipRGCs ( Hattar et al . , 2002; Hattar et al . , 2006; Gooley et al . , 2003; Ecker et al . , 2010 ) , their lesioning blocks light’s acute effects on sleep ( Miller et al . , 1998 ) , and melanopsin knockout mice lose light-induced cFOS expression in the SC ( Lupi et al . , 2008 ) . However , the SC and pretectum receive robust innervation from many conventional RGCs , making the requirement for melanopsin and ipRGCs in acute sleep and body temperature regulation difficult to reconcile . It is also possible ( and perhaps probable ) , that multiple ipRGC target regions are involved , with distinct areas mediating distinct physiological responses . Future studies silencing each retinorecipient target while activating Brn3b ( + ) ipRGCs will be necessary to tease apart the downstream circuits mediating light’s acute effects on physiology . Alternatively , it remains possible that direct ipRGC control of body temperature is the primary and critical step for many acute responses to light that are mediated by ipRGCs . In support of this possibility , changes in body temperature and heat loss can directly influence sleep induction ( Kräuchi et al . , 1999 ) . This change in sleep is in turn presumably causative of at least some of light’s effects on wheel-running and general activity ( Mrosovsky et al . , 1999 ) . Further , core body temperature can acutely regulate cognition and alertness ( Wright et al . , 2002; Darwent et al . , 2010 ) . It is therefore possible that ipRGCs can have widespread influence on an animal’s basic physiology and cognitive function simply by regulating body temperature . Together , our identification of the photopigment and the retinal circuits mediating acute body temperature and sleep induction will facilitate better methods to promote or avoid human alertness and cognition at appropriate times of day ( Chellappa et al . , 2011 ) . Our results support many recent efforts to capitalize on the specific light-detection properties of melanopsin ( Lucas et al . , 2014 ) , such as its insensitivity and short-wavelength preference , to promote or avoid its activation at different times of day . However , this approach is problematic because acute activation of melanopsin to promote alertness has the unintended effect of shifting the circadian clock ( Provencio et al . , 1994 ) , thereby making subsequent sleep difficult . Our identification that the Brn3b ( + ) ipRGCs specifically mediate light’s acute effects on body temperature provides a cellular basis for developing targeted methods for promoting acute alertness , while minimizing circadian misalignment .
All procedures were conducted in accordance with NIH guidelines and approved by the Institutional Animal Care and Use Committee of Johns Hopkins University . All mice were maintained on a mixed C57Bl/6J; 129Sv/J background and kept on ad libitum food and water under a 12 hr/12 hr light/dark cycle in group housing until experimentation , with temperature and humidity control . Male and female mice between the ages of 2 and 6 months were used for analysis . Each mouse was single-housed at the time of experiment . Surgery was conducted under tribromoethanol ( Avertin ) anestheshia and a telemetric probe ( Starr G2 E-Mitter ) was implanted in the peritoneal cavity to monitor core body temperature and general activity . Data were collected in continuous 1- or 2 min bins using VitalVIEW software and analyzed in Microsoft Excel . All experiments were conducted at least 10 days after surgery . Lights were controlled by a programmable timer and all lights were 6500K CFL bulbs illuminated each cage at ~500 lux . Light intensity ( Figure 1—figure supplement 1 ) was modulated using neutral density filters ( Roscolux ) . Brn3bCre/+ or Control littermate mice were anesthetized with tribromoethanol ( Avertin ) and 0 . 5–1 µl AAV2-hSyn-DIO-hM3Dq-mCherry ( UNC Vector Core ) was injected intravitreally in one eye using a picospritzer and pulled pipet . At least one week later , animals underwent surgery for implantation of telemetric probes ( as above ) . All experiments were conducted at least 10 days after telemetric probe implantation and at least three weeks after viral injection . After behavior , the eyes of each animal were inspected to ensure that >50% infection had been achieved ( assessed by fluorescence detectable across more than half of the retina ) . We routinely saw >80% of the retinas were infected as we have described previously ( Keenan et al . , 2016 ) . Diurnal amplitude was measured by subtracting the mean body temperature for the light cycle ( ZT0-12 ) from the mean body temperature for the dark cycle ( ZT12-24 ) . Mean body temperature during testing used all data from ZT14-17 . Comparisons were performed in one of two ways . First , we compared the mean body temperature during this period on the control ( dark ) night to that on the night where the light pulse was given . Additionally , we compared the change in body temperature between ZT14 ( which served as a baseline ) and the mean body temperature from ZT14-17 between the control night and the night where the light pulse was given . For CNO experiments , injections were carried out near ZT14 , but specific times were recorded for each mouse to align the data to the time of injection . Comparisons of mean body temperature after PBS or CNO utilized the 6 hr following injection . Clozapine-N-oxide ( Sigma ) was prepared as a 0 . 1 mg/ml solution in PBS and injected at 1 mg/kg intraperitoneally at ZT14 . All procedures were conducted in accordance with NIH guidelines and approved by the Institutional Animal Care and Use Committee of Northwestern University . Opn4Cre and Brn3bz-dta were maintained on a mixed C57Bl/6J; 129Sv/J background ( Hattar et al . , 2002; Hattar et al . , 2006; Mu et al . , 2005 ) . Male and female littermate Opn4Cre/+ and Opn4Cre/+; Brn3bz-dta/+ animals between the ages of 2 and 3 months were used for sleep analysis . Male and female littermate Opn4Cre/+ and Opn4Cre/+; Brn3bz-dta/+ mice were used for sleep recordings . Electroencephalogram ( EEG ) and electromyogram ( EMG ) electrode implantation was performed simultaneously at 8 weeks of age . Mice were anesthetized with a ketamine/xylazine ( 98 and10 mg/kg respectively ) and a 2-channel EEG and 1-channel EMG implant ( Pinnacle Technology ) was affixed to the skull . Mice were transferred to the sleep-recording cage 6 days after surgery , tethered with a preamplifier , and allowed 3 days to acclimate to the new cage and tether . Mice were housed in 12:12 light/dark conditions before and after EEG implantation . EEG and EMG recording began simultaneously at the end of the habituation period , which were displayed on a monitor and stored in a computer for analysis of sleep states . The high pass filter setting for both EEG channels was set at 0 . 5 Hz and low pass filtering was set at 100 Hz . EMG signals were high pass filtered at 10 Hz and subjected to a 100 Hz low pass cutoff . EEG and EMG recordings were collected in PAL 8200 sleep recording software ( Pinnacle Technology ) and scored , using a previously described , multiple classifier , automatic sleep scoring system , into 10 s epochs as wakefulness , NREM sleep , or REM sleep on the basis of rodent sleep criteria ( Gao et al . , 2016 ) . Light source for all sleep experiments was a 3000 Kelvin light source at 500 lux . Mice were placed in cages with a 4 . 5-inch running wheel , and their activity was monitored with VitalView software ( MiniMitter ) . Analyses of wheel running activity were calculated with ClockLab ( Actimetrics ) . We used 500 lux light intensity . Mice were initially placed under 12:12 LD masking experiments . Mice were exposed , in their home cage , to a timer-controlled 3 hr light pulse at ZT14-ZT17 . Percent activity for each mouse was normalized to its own activity at ZT13 ( 1 hr before light pulse ) , and analyzed in 30 min bins . Animals were anesthetized with Avertin and euthanized prior to fresh dissection of retinas in PBS . Retinas were fixed in 4% paraformaldehyde ( Sigma ) for at least 1 hr on ice . Retinas were then washed in PBS before staining overnight in anti-OPN4 antibody ( 1:1000 , Advanced Targeting Systems ) and then washed prior to 2 hr in secondary antibody ( 1:1000 goat anti-rabbit AlexaFluor 488 , Life Technologies ) . Retinas were then flat-mounted on slides and imaged on a Zeiss LSM 710 confocal microscope . All statistical tests were performed in Graphpad Prism or R 3 . 4 . 4 . Specific tests are listed in the text and figure legends . Linear mixed models were performed with the R packages lme4 1 . 1–21 and emmeans 1 . 3 . 4 . All raw data are linked to this manuscript and available online . | Light , whether natural or artificial , affects our everyday lives in several ways . Exposure to light impacts on our health and well-being . It plays a crucial but indirect role in helping to align our internal body clock with the 24-hour cycle of day and night , and a burst of bright light in the middle of the night can wake us up from sleep . Decades of research have revealed the circuitry that controls the indirect effects of light on the body's internal clock . A tiny set of cells in the base of the brain called the suprachiasmatic nucleus ( SCN for short ) generates the body’s daily or “circadian” rhythm . A small group of nerve cells in the retina of the eye called intrinsically photosensitive retinal ganglion cells ( ipRGCs ) connect with the SCN . These ipRGCs relay information about light to the SCN to ensure that daily rhythms happen at the appropriate times of day . But scientists do not yet know if the same brain circuits regulate the direct effects of light on alertness . Mice are often used in studies of circadian rhythms but , unlike humans , mice are normally active at night and sleep throughout the day . This means that a burst of bright light in the middle of the night causes mice to become less alert . Now , in experiments with mice , Rupp et al . show there are two separate circuits from the retina to the brain that influence wakefulness . In the experiments , some mice were genetically engineered to only have ipRGCs that connect with the SCN and to lack those that connect with other brain areas . These mice lived in cages with a normal day/night cycle and their body temperature and sleep-related brain activity were monitored as Rupp et al . sporadically exposed them to bright light at night . These mice continued their normal routines and were unaffected by the bursts of light . In a second set of experiments , ipRGCs that do not connect with the SCN were activated in other mice . This caused an immediate and sustained drop in the body temperature of the mice , which is linked to them becoming less alert . The experiments suggest that the circuit that connects ipRGCs to the SCN to align the body’s circadian rhythm with light does not control the direct effect of light on wakefulness . Instead , a separate circuit that extends from ipRGCs to an unknown part of the brain area influences wakefulness . Better understanding this second circuit could allow scientists to develop ways to keep people like emergency personnel or overnight shift workers awake and alert at night while avoiding harmful disruptions to their circadian rhythms . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2019 | Distinct ipRGC subpopulations mediate light’s acute and circadian effects on body temperature and sleep |
Endoplasmic reticulum ( ER ) stress is implicated in many chronic diseases , but very little is known about how the unfolded protein response ( UPR ) responds to persistent ER stress in vivo . Here , we experimentally reconstituted chronic ER stress in the mouse liver , using repeated injection of a low dose of the ER stressor tunicamycin . Paradoxically , this treatment led to feedback-mediated suppression of a select group of mRNAs , including those encoding the ER chaperones BiP and GRP94 . This suppression was due to both silencing of the ATF6α pathway of UPR-dependent transcription and enhancement of mRNA degradation , possibly via regulated IRE1-dependent decay ( RIDD ) . The suppression of mRNA encoding BiP was phenocopied by ectopic overexpression of BiP protein , and was also observed in obese mice . Our findings suggest that persistent cycles of UPR activation and deactivation create an altered , quasi-stable setpoint for UPR-dependent transcriptional regulation—an outcome that could be relevant to conditions such as metabolic syndrome .
The western obesity epidemic has exposed one of the consequences of modernity: the increasing prevalence of non-infectious chronic diseases that progress down increasingly irreversible paths over the course of years to decades . Diabetes , atherosclerosis , hypertension , steatohepatitis , and the like join neurodegenerative disorders , cancers , lung disease and other chronic diseases in driving morbidity and mortality in the western world ( Ezzati et al . , 2002 ) . As a class , these diseases necessarily entail a gradual deterioration of cellular and organ function , rather than an acute collapse . Thus , treating or reversing them requires understanding how persistent but otherwise modest stimuli alter the activity of key cellular pathways . One cellular pathway increasingly implicated in the progression of a number of chronic diseases is the Unfolded Protein Response ( UPR ) . The UPR is activated by disruption of the protein folding capacity of the endoplasmic reticulum , otherwise known as ‘ER stress’ ( Walter and Ron , 2011 ) . Unremitted ER stress and/or a dysregulated UPR appear to contribute to hepatic steatosis and steatohepatitis ( Malhi and Kaufman , 2011 ) , atherosclerosis ( Zhou and Tabas , 2013 ) , colitis and inflammatory bowel disease ( Kaser et al . , 2013 ) , hypertension ( Young and Davisson , 2015 ) , and many others . As an organelle that carries out several essential cellular processes ( protein processing , calcium storage , lipogenesis , certain metabolic steps , etc . ) and that is physically and functionally intertwined with many other critical cellular pathways , the ER is sensitive to a range of diverse stimuli . For example , ER stress is observed in the livers of obese mice ( Ozcan et al . , 2004 ) , and this response has been attributed to an excess load of nascent client proteins in the organelle due to nutrient-mediated stimulation of mTOR activity ( Ozcan et al . , 2008 ) . Other physiological stimuli that elicit ER stress include nutritional status and the activity of metabolic pathways ( Oyadomari et al . , 2008; Tyra et al . , 2012; Shao et al . , 2014 ) , differentiation cues ( Iwakoshi et al . , 2003; van Anken et al . , 2003; Lee et al . , 2005 ) , inflammatory signals ( Zhang et al . , 2006; Hotamisligil , 2010 ) , and many others . While the UPR is capable of responding to excessive ER stress by initiating cell death cascades ( Sano and Reed , 2013 ) , chronic stress is instead likely to result not in cell death ( at least not for most cells , most of the time ) , but instead in a persistent burden on ER function that must be accommodated by the UPR . And yet , for many chronic diseases in which ER stress is implicated , it is not clear whether disease results from a UPR that becomes progressively dysregulated and thus responds inappropriately; or whether the UPR becomes progressively neutered and simply becomes increasingly unresponsive . Experimental manipulation of the UPR has sketched the framework of a canonical UPR that is initiated by three ER-resident proteins and that culminates in transcriptional augmentation of the ER protein processing capacity and other non-transcriptional mechanisms to alleviate ER load . The inositol-requiring enzyme 1α ( IRE1α ) pathway results in both the production of the XBP1 transcription factor ( Yoshida et al . , 2001; Calfon et al . , 2002; Lee et al . , 2002 ) and the degradation of ER-associated mRNAs in a process known as Regulated IRE1-Dependent Decay via the IRE1 endonuclease activity ( Hollien and Weissman , 2006; Hollien et al . , 2009 ) . The PKR-like endoplasmic reticulum kinase ( PERK ) stimulates production of the ATF4 transcription factor ( Harding et al . , 2000 ) and also transiently reduces ER protein load by phosphorylation of the translation initiation factor eIF2α ( Shi et al . , 1998; Harding et al . , 1999 ) . ATF6 ( comprising both α and β paralogs ) initiates the third pathway when ATF6 is itself cleaved by regulated intramembrane proteolysis in the Golgi , liberating a transcriptionally active N-terminal fragment ( Ye et al . , 2000 ) . These pathways are all robustly activated by acute exposure of cells to high doses of ER stress-inducing drugs , which is an effective tool for understanding what the system is capable of but not as effective for understanding how the cellular response to chronic stress differs from these canonical pathways . Evidence for the involvement of ER stress in chronic diseases in general comes from two lines of investigation: detecting markers of ER stress in the affected tissues of human patients or mouse models thereof , and observing disease-like phenotypes in mice with genetic ablations of UPR signaling pathways . Yet while studies of this sort provide prima facie evidence for UPR involvement , they do not address how the UPR becomes dysregulated during chronic disease , if in fact it even does . Further , animal models of chronic diseases such as those associated with obesity tend to influence innumerable cellular pathways and make it difficult to isolate the effects of ER stress per se from the many other confounding factors . With these considerations in mind , we set out to address how the UPR responds to chronic stress in vivo . Following the example of our previous work in vitro ( Rutkowski et al . , 2006 ) , here we describe how UPR signaling in the liver becomes altered during repeated exposure to stress , the mechanisms by which these alterations occur , and their potential relevance to liver dysfunction during obesity .
Our first step in experimentally reconstituting chronic ER stress was to identify conditions that were as specific as possible in perturbing ER function with minimal pleiotropic effects , and were sufficient to activate the UPR without causing overt toxicity . In doing so , we hoped to be able to understand how the UPR responds to chronic stress , divorced from the confounding influences of chronic diseases on other cellular pathways . Toward this end , we used tunicamycin ( TM ) to induce ER stress in vivo; TM blocks N-linked glycosylation , an ER-specific protein modification . It has few known off-target effects , most prominently targets the kidneys and the liver , and is not lethal in wild-type animals at even relatively high doses ( Foufelle and Fromenty , 2016 ) . Its effects on the liver can also be mimicked , albeit less readily , by other agents that elicit ER stress in that organ ( Rutkowski et al . , 2008; Chikka et al . , 2013 ) , further justifying its use here . Following an approach analogous to one we described previously in cultured cells ( Rutkowski et al . , 2006 ) , we looked for a dose of TM that was capable of inducing the UPR , but that was significantly less toxic than the common experimental dose of 1 mg/kg—which is known to elicit hepatocellular death and inflammation ( Foufelle and Fromenty , 2016 ) . Using qRT-PCR to detect changes in expression of the UPR sentinel genes Hspa5 ( encoding BiP ) and Ddit3 ( encoding CHOP ) , we found that doses of TM below 0 . 1 mg/kg were capable of eliciting attenuated UPR activation ( Figure 1A ) . 10 . 7554/eLife . 20390 . 003Figure 1 . Mice adapt to repeated exposure to TM despite persistent stress . ( A ) Livers from wild-type mice collected 8 hr after injection with the indicated concentrations of TM ( in mg/kg ) were probed by qRT-PCR for expression of Hspa5 ( encoding BiP ) and Ddit3 ( encoding CHOP ) . n = 2–3 animals/group . Significance was determined by one-way ANOVA . Here and elsewhere , *** , p<0 . 001; ** , p<0 . 01; * , p<0 . 05 . ; NS = not significant ( B ) Livers were harvested after treatment for 8 hr with vehicle or 0 . 025 mg/kg TM and protein lysates were probed by immunoblot to detect BiP or CHOP . Calnexin was used as a loading control . For all blots and gels in this paper , each lane represents a separate animal . Hairlines are used for visual clarity only . ( C ) Chronic stress treatment schematic . Mice were weighed and then injected daily with 0 . 025 mg/kg TM ( red arrows ) ; liver samples were collected at the indicated times ( blue arrows ) for downstream analysis . The naming convention is as follows: ‘D’ indicates the number of daily injections received while ‘h’ indicates the time of tissue collection after the last injection . Here and elsewhere , not treated animals ( NT ) received injections of vehicle at the same times that D5-24h mice received TM . ( D ) Formalin-fixed liver sections from animals treated as in ( C ) were collected , fixed , and stained by H and E at the indicated times . For each condition , representative images from two separate mice are shown . Note the extensive cytoplasmic vacuolization in the D1-24h animals . ( E , F ) Protein lysates from animals treated as in ( C ) were isolated and probed for expression of the ER-resident glycoprotein TRAPα ( E ) and BiP , CHOP , ADRP , and α-actin ( loading control ) ( F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20390 . 00310 . 7554/eLife . 20390 . 004Figure 1—source data 1 . Contains raw and transformed Ct values for qRT-PCR experiment in Figure 1A . DOI: http://dx . doi . org/10 . 7554/eLife . 20390 . 004 Based on this finding , we chose 0 . 025 mg/kg as the dose most likely to be tolerated long-term by mice without leading to significant toxicity , and we used it as the basis for induction of chronic stress . As expected , acute treatment ( 8 hr ) with this dose of TM was sufficient to lead to upregulation of both BiP and CHOP at the protein level ( Figure 1B ) , although for CHOP this induction was much more modest than commonly elicited by higher doses of TM . To test the effects of chronic treatment with 0 . 025 mg/kg TM , we injected animals with this dose ( or with vehicle ) every day for five days ( or longer ) , as shown in Figure 1C . Animals were sacrificed approximately 24 hr after the most recent dose ( i . e . , immediately before the next injection would have taken place ) . After sacrifice , livers were harvested and analyzed for molecular and histological markers of ER stress . The predominant acute effect of TM on the liver , as revealed by hematoxylin and eosin staining , was the formation of intracellular vacuoles in hepatocytes ( Figure 1D ) that correspond to accumulated lipid droplets resulting from impaired lipid metabolism ( Rutkowski et al . , 2008 ) . These vacuoles were abundant in animals administered a single dose of 0 . 025 mg/kg TM , but , remarkably , were absent by the end of the fifth day of challenge . Other than these changes , there were few signs of gross liver damage such as necrosis or fibrosis ( data not shown ) . Thus , at least by these histological criteria , animals are capable of maintaining liver function in the face of ongoing stress . The apparent normalization of liver histology could be explained if animals became resistant to TM over time . However , persistent underglycosylation of the ER-resident glycoprotrein TRAPα at day five confirmed that the drug remained active upon repeated dosing ( Figure 1E ) . In order to test whether ER stress persisted in the chronic condition despite the apparent normalization of liver histology , we examined expression of BiP and CHOP proteins . Immunoblotting showed that BiP upregulation persisted throughout the time-course ( Figure 1F ) , as did upregulation of the ER chaperone GRP94 ( not shown ) . In contrast , and consistent with the effects of chronic stress on cultured cells ( Rutkowski et al . , 2006 ) , upregulation of the highly labile CHOP protein was lost even by D1-24h , and remained mostly undetectable in animals treated for 5 days ( Figure 1F ) . We also monitored the lipid droplet marker protein ADRP , the expression of which correlates with lipid droplet content . Consistent with histological data ( Figure 1D ) , ADRP expression—and therefore intracellular lipid accumulation—was elevated in animals treated for one day , but returned to basal levels in animals treated for five days ( Figure 1F ) . Of note , we also performed a similar experiment in which the chronic treatment was extended to 15 days . Results were similar to the 5 day treatment ( not shown; see also Figure 2A ) . 10 . 7554/eLife . 20390 . 005Figure 2 . Chronic stress suppresses a subset of UPR target genes . ( A ) Mice were injected with 0 . 025 mg/kg TM daily for up to 18 days , and livers were collected 24 hr after the last injection as indicated . Expression of Hspa5 and Ddit3 was measured by qRT-PCR . Expression is given relative to vehicle-treated control mice . Data for all qRT-PCR experiments is shown on a log scale . n = 2–3 animals/group ( B ) Expression of Hspa5 using an alternate primer pair spanning a different exon/intron junction confirms results from ( A ) . ( C ) qRT-PCR of a group of UPR target genes from animals treated with vehicle or TM for 5d . n = 8 animals/group from two experiments . ( D ) Expression of the indicated metabolic genes was determined by qRT-PCR from an experiment similar to ( C ) , except injections were performed for 7d . n = 4 animals/group . DOI: http://dx . doi . org/10 . 7554/eLife . 20390 . 00510 . 7554/eLife . 20390 . 006Figure 2—source data 1 . Contains raw and transformed Ct values for qRT-PCR experiments in Figure 2A–D . DOI: http://dx . doi . org/10 . 7554/eLife . 20390 . 006 Taken together , these data indicate that chronic TM treatment caused persistent ER stress in the liver without permanently perturbing liver anatomy . They suggest that , as with cultured cells , hepatocytes in vivo are capable of adapting to long-term disruption of ER function if the stress is sufficiently mild . Thus , we used these conditions to investigate how the UPR is regulated during such adaptation . As the UPR is fundamentally a gene regulatory program , we examined the effects of chronic stress on UPR-dependent transcription . Despite Ddit3 mRNA being upregulated by acute stress ( Figure 1A ) , its expression returned to basal levels by 24 hr ( D1-24h ) and remained at similar levels at every subsequent time point , even after 18d of repeated treatment ( Figure 2A ) . This finding mirrored the behavior of CHOP protein ( Figure 1F ) . Surprisingly , however , Hspa5 mRNA levels—while elevated by acute stress ( Figure 1A ) —were significantly diminished below basal levels by 24 hr and continued to diminish further through the first five days , remaining suppressed thereafter ( Figure 2A ) . Similar results were obtained using qRT-PCR primers bridging a separate pair of exons in Hspa5 mRNA , demonstrating that this suppression reflected total Hspa5 mRNA levels rather than alternate splicing ( Figure 2B ) . Using the D5-24h point as representative of chronic stress conditions , we next surveyed other UPR-regulated gene targets . While the expression of most mRNAs , including Ddit3 , was not significantly different in the chronic condition , a subset displayed a similar downregulation to Hspa5 . These included mRNAs encoding the ER-associated degradation factor Herpud1 and Hsp90b1 ( encoding the ER chaperone GRP94 ) . ( Figure 2C ) . We also assessed the expression of metabolic genes that we and others have previously shown to be suppressed by ER stress in the liver ( Rutkowski et al . , 2008; Yamamoto et al . , 2010; Arensdorf et al . , 2013a ) . While each of these was downregulated in the acute condition as expected ( not shown ) , none was significantly suppressed in the chronic condition ( Figure 2D ) . This finding provides evidence that the suppression of Hspa5 and other mRNAs during chronic ER stress is mechanistically distinct from the pathways that suppress metabolic gene expression during acute stress . The observation that all of the assessed UPR targets either were unregulated or were suppressed in the chronic condition suggested that , despite persistent upregulation of at least some proteins such as BiP , transcriptional regulation was a more dynamic process and might be suppressed during chronic stress , at least by the 24 hr that elapsed between each chronic stress time point and its most proximate dose of TM . One of the commonalities among the genes that were suppressed below basal levels ( Hspa5 , Herpud1 , Hsp90b1 ) is that each has been identified as a target of the ATF6α pathway of the UPR ( Yoshida et al . , 2001; Yamamoto et al . , 2004; Wu et al . , 2007 ) . Thus , we hypothesized that UPR-dependent transcription must not only be deactivated to basal ( i . e . , unstressed ) levels in the chronic condition , but also that at least ATF6α activity would be even further suppressed . To test this hypothesis , we first assessed expression of the key UPR transcriptional regulators by immunoblot from liver nuclei . As a positive control , acute stress ( 8 hr treatment with 0 . 025 mg/kg TM as in Figure 1A ) led to elevated expression of ATF4 , XBP1spl and ATF6αcl; however , all three were undetectable in the chronic condition ( Figure 3A ) . This result was mirrored in a chromatin-immunoprecipitation ( ChIP ) experiment . For this experiment , the Hspa5 promoter was recovered by an ATF6α antibody following 8 hr of treatment served as expected ( Yoshida et al . , 1998 ) . In contrast , no ATF6α binding above background was detected in the chronic condition ( Figure 3B ) . These ChIP data suggested that transcription of Hspa5 was completely lost in the chronic condition , which was confirmed by ChIP directed against RNA Polymerase 2 ( Pol2 ) at the Hspa5 locus: As expected , Pol2 was recovered near the Hspa5 transcriptional start site ( −215 to +0 ) , and this binding was unaffected by the presence or absence of ER stress ( Figure 3C ) . This binding reflects Pol2 that is poised at the Hspa5 promoter waiting to be engaged during the transcriptional activation process ( Muse et al . , 2007 ) . In contrast , while Pol2 was abundantly recovered from the body of the Hspa5 locus during acute stress ( Intron 7 , or +2727 to +2906 , which reflects elongating polymerase ) , this binding was completely lost in the chronic condition ( Figure 3C ) . 10 . 7554/eLife . 20390 . 007Figure 3 . Chronic stress silences ATF6α-dependent transcription . ( A ) After treatment of mice with 0 . 025 mg/kg TM for either 8 hr or 5d , nuclei were isolated from liver lysates and probed with antibodies against ATF4 , XBP1 , or ATF6α . Specificity of all antibodies was confirmed by immunoblot from knockout or overexpression cells or liver lysates ( not shown ) . The loading control for ATF4 and XBP1 blots was PARP . The loading control for the ATF6α immunblot is a nonspecific background band ( * ) . ( B ) After TM treatment as in ( A ) , control Ig or antibodies against ATF6α were used to immunopurify the Hspa5 promoter region , which was then amplified by qPCR using primers directed against two overlapping regions in the promoter containing the ER stress elements ( ERSEs ) or a control sequence . Number is given relative to the transcriptional start site . n = 3–4 animals/group ( C ) Same as ( B ) , except ChIP was performed using antibodies against RNA Polymerase 2 . The −215 to + 8 regions detects poised polymerase and so Pol2 binding does not change , while the Intron seven region ( +2727 to+2906 ) detects elongating Pol2 . ( D ) Wild-type or Atf6-/- mice were treated with 0 . 025 mg/kg TM or vehicle for 5d . qRT-PCR expression of Hspa5 in the liver was assessed using both primer sets . Hspa5 expression was suppressed to comparable levels by either chronic stress or deletion of ATF6α , but chronic stress had no effect in Atf6-/- animals . n = 3–5 animals/group . DOI: http://dx . doi . org/10 . 7554/eLife . 20390 . 00710 . 7554/eLife . 20390 . 008Figure 3—source data 1 . Contains raw and transformed Ct values for qPCR and qRT-PCR experiments in Figure 3B–D . DOI: http://dx . doi . org/10 . 7554/eLife . 20390 . 008 Taken together , these results demonstrate that ATF6α-dependent transcription ( and possibly XBP1- and ATF4-dependent transcription as well ) is silenced in the chronic condition . However , for this inhibition of transcription to result in suppression of Hspa5 mRNA expression compared to untreated animals , ATF6α must contribute not only to stress-dependent transcription of Hspa5 but also to the basal regulation of Hspa5 . Confirming this prediction , we found that the basal expression of Hspa5 in animals lacking ATF6α was reduced by approximately 50 percent relative to wild-type animals , and that chronic stress then had no further effect on Hspa5 expression ( Figure 3D ) . Thus , loss of ATF6α alone was sufficient to mimic chronic stress treatment . From these data , we conclude that at least the ATF6α pathway of the UPR is engaged in the liver absent exogenous stress , and that even this basal component can be inhibited upon chronic stress . To this point , our data establish that under conditions of chronic stress , UPR-dependent transcription is attenuated , including , apparently , the contribution of ATF6α to basal Hspa5 expression . This attenuation could come about in either of two ways . One possibility is that some fundamental feature of chronic stress renders the UPR increasingly refractory . This means that after 5 days of activation by successive bouts of stress the UPR would no longer be responsive; this is a ‘preconditioning’ model . The other possibility is an ‘augmented deactivation’ model in which the UPR is activated in full or nearly so upon each exposure to stress , but the temporal dynamics of its activation and resolution become progressively altered . In other words , the question is whether chronic stress mutes the activation of the UPR ( preconditioning ) or enhances its deactivation ( augmented deactivation ) . To distinguish between these possibilities , we examined the temporal response of the UPR in animals exposed to either the first dose of TM or the fifth dose ( Figure 4A ) . Rather than following the production of UPR-regulated transcription factors as in Figure 3 , we instead monitored activation of the UPR sensor IRE1α , which serves as a direct , robust , and sensitive readout for the capacity of the UPR to be activated . IRE1α was monitored using Xbp1 mRNA splicing ( Figure 4B ) . This experiment demonstrated that the UPR remained essentially as stress-responsive on the fifth day as on the first; splicing of Xbp1 8 hr after the first injection and 8 hr after the fifth were induced to comparable levels . Supporting the idea that the UPR remained responsive , repeated TM treatment continued to induce short-term lipid dysregulation , as evidenced by abundant cytoplasmic vacuoles in D5-8h livers ( Figure 4C ) . Yet the deactivation of the UPR , or at least the Xbp1 splicing activity of IRE1α , proceeded more rapidly on day five than on day 1; substantial Xbp1 splicing was seen 14 hr after the first injection but not 14 hr after the fifth ( Figure 4B ) . Likewise , as shown in Figure 1D , lipid dysregulation was observed 24 hr after the first injection but not 24 hr after the fifth . We also monitored Hspa5 mRNA expression over a similar time course , and found that its expression mirrored this pattern—namely , that Hspa5 expression was well-induced after both treatments , but that its attenuation proceeded more rapidly in the chronic condition ( Figure 4D ) . Similar results were also obtained for Hsp90b1 ( encoding GRP94 ) mRNA ( not shown ) . Collectively , our data support the augmented deactivation model described above . Thus , the chronic treatment opens a window into the mechanisms of UPR deactivation , about which relatively little is known since ER stress is usually studied in the context of stresses sufficiently severe as to kill cells . 10 . 7554/eLife . 20390 . 009Figure 4 . Accelerated degradation diminishes Hspa5 mRNA levels during chronic stress . ( A ) Schematic showing treatment protocol; livers were harvested 8 , 14 , or 20–24 hr after either the first TM injection or the fifth . ( B ) Conventional RT-PCR was used to distinguished spliced ( spl ) from unspliced ( us ) Xbp1 mRNA in samples treated as in ( A ) . Image is inverted black-to-white for greater visual clarity . Each lane represents a separate animal . ( C ) H and E staining of liver sections harvested at the D1-8h or D5-8h timepoints . ( D ) Hspa5 mRNA expression was assessed by qRT-PCR from samples treated as in ( B ) . n = 3–4 animals per group ( E ) Primary hepatocytes were isolated from a wild-type mouse , and treated with TM in the presence or absence of actinomycin D ( ActD ) to inhibit transcription as described in Materials and methods . Hspa5 half-life was calculated from these data . n = 3 plates/group ( F ) Wild-type or Ern1-/- ( lacking IRE1α ) mouse embryonic fibroblasts ( MEFs ) were treated with TM for 8 hr and expression of the RIDD target Bloc1s1 was determined by qRT-PCR . n = 3 plates/group ( G ) Animals were treated as in ( A ) and Bloc1s1 expression was detected by qRT-PCR . ( H ) Animals were treated for 8 hr or 5d with 0 . 025 mg/kg TM . Expression of Fn1 was determined by qRT-PCR . n = 4 animals/group . DOI: http://dx . doi . org/10 . 7554/eLife . 20390 . 00910 . 7554/eLife . 20390 . 010Figure 4—source data 1 . Contains raw and transformed Ct values for qRT-PCR experiments in Figure 4D–H . DOI: http://dx . doi . org/10 . 7554/eLife . 20390 . 010 These experiments also yielded an additional insight , which was that the rapid loss of Hspa5 mRNA by the fifth day might not be accounted for solely by the loss of Hspa5 transcription . Hspa5 mRNA levels decreased more than 30-fold between 8 hr and 24 hr on the fifth day , for a half-life of approximately 3 hr ( Figure 4D ) . However , by monitoring Hspa5 expression after treatment of primary hepatocytes with the transcription inhibitor Actinomycin D , we measured Hspa5 mRNA half-life as 8 . 2 hr after treatment with TM ( Figure 4E ) , consistent with our previous measurement in MEFs ( Rutkowski et al . , 2006 ) . The simplest explanation for these results is that deactivation of the UPR is accompanied by stimulated degradation of Hspa5 mRNA , for even the loss of Hspa5 that occurs on the first day of treatment ( Figure 4D ) is too rapid to be accounted for by inhibition of transcription alone . A candidate pathway for stimulated degradation of Hspa5 mRNA is Regulated IRE1-Dependent Decay ( RIDD ) , which is the process by which activated IRE1 directly degrades ER-associated mRNAs ( Hollien et al . , 2009 ) . This process is thought to be counter-adaptive , facilitating cell death during severe ER stress ( Han et al . , 2009; Upton et al . , 2012 ) . However , it has also been implicated in metabolic regulation in the liver ( Cretenet et al . , 2010; So et al . , 2012 ) , implying that perhaps its function is more nuanced . Hspa5 , Hsp90b1 , and Herpud1 mRNAs would be expected to localize to the ER by virtue of encoding proteins bearing ER targeting signals . Hspa5 at least has also been proposed to be a RIDD target , though with unclear functional significance ( Han et al . , 2009 ) . To test whether RIDD was active in the chronic stress condition , we assessed the expression of the well-validated RIDD target Bloc1s1 ( aka Blos1 ) . Among putative RIDD targets , only Bloc1s1 has been documented in multiple reports , and in highly diverse cell types ( Bright et al . , 2015 ) . We confirmed that Bloc1s1 is a bona fide RIDD target in MEFs lacking IRE1α , in which its downregulation by ER stress was completely lost ( Figure 4F ) . As expected , Bloc1s1 expression was reduced in the liver by acute ( 8 hr ) TM treatment ( Figure 4G ) . Furthermore , while Bloc1s1 expression returned to normal levels 24 hr after the first TM treatment , it remained suppressed 24 hr after the fifth treatment ( Figure 4G ) . Suppression of a RIDD target in the chronic condition was also seen for Fn1 , which we found to be a potential RIDD target in the liver based on microarray data from wild-type and liver-specific Ern1-/- ( encoding IRE1α ) mice treated with TM ( Zhang et al . , 2011 ) ( Figure 4H ) . The behavior of Bloc1s1 and Fn1 is in contrast to genes such as Ddit3 , Wars , etc . , that were transcriptionally induced by acute ER stress , but which did not remain induced in the chronic condition ( Figure 2C ) , as well as to metabolic genes that are transcriptionally repressed by acute ER stress , which did not remain repressed in the chronic condition ( Figure 2D ) . Computational modeling suggests that mRNAs whose expression is regulated by stimulated degradation should return to basal expression levels much more rapidly than those whose regulation is transcriptional , when both mechanisms are deactivated ( Arensdorf et al . , 2013b ) . Thus , this persistence of Bloc1s1 and Fn1 suppression might suggest that , in contrast to the pathways of transcriptional regulation , RIDD activity persists through the cycles of UPR activation and deactivation that characterize the chronic stress state , even while the Xbp1 splicing activity of IRE1α is attenuated . Conclusively demonstrating that RIDD is active under these conditions will ultimately require replacement of endogenous IRE1α by a form in which the Xbp1 splicing and RIDD activities can be separated and modulated ( Han et al . , 2009 ) . Thus , while the hypothesis that RIDD activity reduces the expression of Hspa5 and other genes is an appealing possible mechanism , other degradative pathways are possible . Our results to this point suggest that with each successive exposure to stress , the UPR becomes more efficient at deactivation , even to the point of overshooting , while still retaining its activation capacity . The mechanisms of deactivation , however , are poorly understood . One pathway by which the UPR is deactivated is upregulation of BiP , which turns off the response in two ways: by improving ER protein folding ( in concert with the many other factors regulated by the UPR ) ; and by directly binding to the UPR stress sensors ( IRE1α , ATF6α , and PERK ) and repressing their activation ( Bertolotti et al . , 2000; Shen et al . , 2002 ) . In the chronic condition , Hspa5 mRNA was suppressed despite ongoing upregulation of BiP protein ( Figure 1F ) ; the same was true for GRP94 ( not shown ) . Given the repeated cycles of activation that characterize the chronic response ( Figure 4 ) , it is perhaps unsurprising that BiP accumulates to the extent that it does; we have previously measured the half-life of BiP protein in MEFs at approximately two days ( Rutkowski et al . , 2006 ) , and it appears to also be long-lived in hepatocytes ( not shown ) . This persistent BiP protein expression thus provides a potential means by which UPR deactivation could be accelerated . To test this hypothesis , we ectopically overexpressed BiP protein in the liver using recombinant adenovirus ( ad-BiP , aka AdGRP78 ) ( Young et al . , 2012 ) . After injection of ad-BiP , we observed a significant increase in BiP protein expression compared to ad-GFP expressing animals ( Figure 5A ) . qRT-PCR confirmed the efficacy of exogenous BiP overexpression ( Figure 5B ) . Using primers that detect the 5’ UTR of Hspa5 mRNA ( and thus amplify only endogenous Hspa5 and not exogenous ) , we found that , like chronic stress , BiP protein overexpression suppressed Hspa5 mRNA expression ( Figure 5C ) . Similar results were obtained for Hsp90b1 , while overexpression had no effect on Ddit3 expression ( Figure 5C ) . Surprisingly , Bloc1s1 expression was also suppressed under these conditions , again mirroring the chronic condition ( Figure 5D ) . This finding hints that the apparent persistence of RIDD in the chronic condition ( Figure 4G , H ) might reflect a unique sustainment of that activity of IRE1α by BiP during recovery from stress when the transcriptional limbs of the UPR are suppressed . 10 . 7554/eLife . 20390 . 011Figure 5 . Overexpression of exogenous BiP is sufficient to repress endogenous Hspa5 expression . ( A ) Animals were injected with recombinant adenovirus expressing either GFP or BiP . 5d after injection , mice were sacrificed and BiP was probed by immunoblot . Loading control was calnexin . ( B ) Both primer sets detected elevated Hspa5 expression in ad-BiP mice , which is attributable to the contribution of the exogenous Hspa5 . n = 8 animals/group for regular Hspa5 primers and 3–4 animals/group for Hspa5 alternate primers . ( C , D ) Expression of endogenous Hspa5 from the genomic locus ( gen-Hspa5 ) , Hsp90b1 , or Ddit3 ( C ) or Bloc1s1 ( D ) was assessed in ad-GFP and ad-BiP animals by qRT-PCR . n = 8 animals/group from two experiments . ( E ) Wild-type animals were treated for 10d with 500 mg/kg TUDCA , and then for 8 hr with 0 . 025 mg/kg TM , and expression of the indicated genes was detected by qRT-PCR . TUDCA was sufficient to reduce stress-induced expression of these genes approximately two-fold or more , but not to suppress basal Hspa5 expression . n = 3–6 animals/group . DOI: http://dx . doi . org/10 . 7554/eLife . 20390 . 01110 . 7554/eLife . 20390 . 012Figure 5—source data 1 . Contains raw and transformed Ct values for qRT-PCR experiments in Figure 5B–E . DOI: http://dx . doi . org/10 . 7554/eLife . 20390 . 012 The effects of BiP overexpression on Hspa5 and Hsp90b1 mRNA could be due either to enhanced protein folding in the ER or to direct effects of BiP on the UPR sensors independent of the protein folding capacity . We found that treating mice with the pharmacological chaperone TUDCA ( Ozcan et al . , 2006 ) —which reduced upregulation of ER stress markers by TM approximately 2-fold—had no effect on its own on any UPR gene tested ( Figure 5E ) . These data show that improved ER protein folding capacity was not sufficient to suppress Hspa5 expression . While they do not completely rule out the possibility that improved ER protein folding ultimately drives Hspa5 suppression in the chronic condition , they nonetheless lead us to favor a model whereby BiP directly modulates the activities of at least ATF6α and IRE1α . To what extent are the regulatory events described here relevant to physiological chronic stresses ? Repeated dosing with TM , while effective as a tool for inducing ER stress , is nonetheless doubtless much more robust and focal a stress than anything commonly encountered physiologically . However , the cyclic nature of the stimulus—bouts of stress followed by periods of recovery—might to some extent mirror the stress caused by metabolic flux , which is tied to feeding and fasting cycles and is thus by its nature cyclic . The notion that overnutrition—i . e . , consumption of too much food per meal and/or too many meals—might represent a chronic ER stress was first suggested in 2004 , in seminal work demonstrating that ER stress was observed in the livers of obese mice , and was tied to hepatic insulin resistance ( Ozcan et al . , 2004 ) . Thus , we considered the possibility that feeding is capable of eliciting the same changes in gene regulation as seen in the TM-induced chronic condition . Consistent with previous reports ( Oyadomari et al . , 2008; Pfaffenbach et al . , 2010 ) , we found that feeding elicited ER stress in mouse livers ( Figure 6A ) . Interestingly , at least at this timepoint the effect was much more robust for some targets ( Hspa5 , Hsp90b1 , and Xbp1 ) than others; in fact , targets of the PERK pathway ( Ddit3 , Ppp1r15a [encoding GADD34] , and Wars ) were not upregulated to a statistically significant extent under these conditions . Feeding also elicited splicing of Xbp1 mRNA as measured by qRT-PCR using primers that preferentially amplify only the spliced form ( Xbp1 ( s ) , Figure 6A ) , albeit to a very modest extent ( once the change in total Xbp1 levels is accounted for ) . While it has been proposed that a high fat diet is intrinsically stress-inducing ( Ozcan et al . , 2004 ) , at least in this instance we found no difference between the level of stress induced by diets containing either high fat or no fat ( Figure 6A ) . This result indicates that nutrient intake itself elicits ER stress , and is consistent with the idea that the stress is driven by increased anabolism ( Ozcan et al . , 2008 ) rather than fat per se . As we speculated , this stress is modest compared to that induced by TM , for which most UPR targets were induced to an extent that was at least an order of magnitude greater than feeding ( Figure 6B ) . 10 . 7554/eLife . 20390 . 013Figure 6 . Genetically-induced obesity phenocopies chronic stress . ( A ) Wild-type animals were fasted overnight , and then provided food containing either 45% fat or no fat for 4 hr . Expression of the indicated genes was assessed by qRT-PCR . n = 3 animals/group ( B ) Wild-type animals were treated with 1 mg/kg TM for 4 hr , and the same genes as in ( A ) were detected by qRT-PCR . #; p<0 . 1 . n = 3 animals/group ( C–E ) Livers from five month-old female Lepob/ob or Leprdb/db mice or age-matched wild-type mice were probed for expression of genes that were suppressed by chronic stress ( C ) or not-suppressed ( D ) , or of the RIDD target Bloc1s1 and the putative RIDD target Fn1 ( E ) . n = 4 animals/group . DOI: http://dx . doi . org/10 . 7554/eLife . 20390 . 01310 . 7554/eLife . 20390 . 014Figure 6—source data 1 . Contains raw and transformed Ct values for qRT-PCR experiments in Figure 6A–E . DOI: http://dx . doi . org/10 . 7554/eLife . 20390 . 014 If feeding represents an ER stress , then repeated feeding should represent a chronic stress , particularly if it outpaces the ability of the organelle to compensate . Thus , we examined gene expression in Lepob/ob or Leprdb/db mice , which are obese due to mutations in leptin or the leptin receptor , respectively , that compromise appetite control and lead to overfeeding ( Campfield et al . , 1995; Halaas et al . , 1995; Pelleymounter et al . , 1995; Chen et al . , 1996; Lee et al . , 1996 ) . We observed a similar pattern of gene regulation in these animals as in animals treated with chronic TM; Hspa5 and Hsp90b1 were both significantly suppressed ( Figure 6C ) , while other UPR targets were not suppressed ( and , indeed , appeared to be upregulated; Figure 6D ) . In addition , again mirroring the results seen in TM-treated animals , expression of Bloc1s1 was reduced in Leprdb/db mice and potentially in Lepob/ob mice as well , and Fn1 was strongly suppressed in both ( Figure 6E ) . These results suggest that suppression of the mRNA expression of Hspa5 and Hsp90b1 characterizes both pharmacological and dietary models of chronic stress . Though they do not yet establish that suppression proceeds by the same mechanism in both , they highlight the potential for the experimental induction of chronic ER stress to be a useful tool for understanding how gene regulatory patterns are altered by chronic physiological stresses .
In this paper , we have described the first , to our knowledge , experimental reconstitution of chronic , tractable ER stress in a mammal in vivo . This work builds upon previous efforts to recreate chronic ER stress in fish larvae ( Cinaroglu et al . , 2011 ) , and to elicit chronic ER stress in mammals by overexpression of misfolded ER client proteins or deletion of ER chaperones ( e . g . , [Ji et al . , 2011; Zode et al . , 2011; Kawasaki et al . , 2015] ) . We have used this system as a tool to probe how UPR signaling becomes altered during persistent and/or repeated activation . Our investigation has several novel implications for the mechanisms of UPR signaling . The data presented here lead us to propose a working model to account for the altered expression of Hspa5 ( Bip ) and other UPR transcriptional targets during chronic stress ( Figure 7 ) . In the basal ( i . e . , nominally ‘unstressed’ ) state , activation of ATF6α contributes to the expression of Hspa5 mRNA . We arrive at this conclusion because expression of Hspa5 was diminished in the absence of ATF6α even without an exogenous stress ( Figure 3D ) . This finding implies that normal physiological function in the liver entails periods of UPR activation—perhaps not surprisingly , given that feeding itself activates the UPR ( Figure 6A ) . BiP is one of the most abundant proteins in the ER lumen ( Gething , 1999 ) , and other non-stress-dependent mechanisms contribute to its basal regulation as well ( Resendez et al . , 1988 ) . Upon exposure to a conventional ER stress , all three limbs of the UPR are activated ( Rutkowski et al . , 2006 ) , leading to robust production of stress-specific target transcripts such as Ddit3 ( Chop ) , as well as augmentation of transcripts encoding ER chaperones , such as Hspa5 and Hsp90b1 ( Grp94 ) . The attendant improvement of ER protein folding then presents a problem for the cell: how to avoid overproduction of UPR targets . As the UPR stress-sensing molecules become deactivated , naturally labile factors such as CHOP ( both mRNA and protein ) are readily lost ( Rutkowski et al . , 2006 ) . However , ER chaperones such as BiP and GRP94 are long-lived at both the protein and mRNA levels ( Rutkowski et al . , 2006 ) , and continued overproduction of these proteins would presumably greatly tax cellular resources and potentially constitute their own burden on the organelle . Thus , BiP overexpression appears to initiate a negative feedback loop , both suppressing UPR activity ( including the basal activity of ATF6α ) and also indirectly stimulating the degradation of its own mRNA . In the context of incidental exposure to stress , or of the low-level stresses that occur during normal physiological function , we presume that this depression of Hspa5 mRNA levels is transient . However , during repeated stress , we propose that it constitutes a resetting to a new quasi-stable setpoint for UPR transcriptional regulation . 10 . 7554/eLife . 20390 . 015Figure 7 . Model for UPR dynamics during chronic stress . See Discussion for details . DOI: http://dx . doi . org/10 . 7554/eLife . 20390 . 015 One of the founding observations of the UPR field was that activation of the ER stress sensing molecules required their dissociation from BiP in both yeast and mammals ( Bertolotti et al . , 2000; Shen et al . , 2002 ) . Originally , the purpose attributed to BiP binding was to hold the stress sensors in a quiescent state and so to play a role in UPR activation . However , the suggestion that at least yeast Ire1p can be activated by direct binding to unfolded proteins ( Kimata et al . , 2004; Gardner and Walter , 2011; Gardner et al . , 2013 ) has cast doubt on this model . Our data raise the possibility that perhaps the role of BiP binding is to instead modulate UPR deactivation . The objective of this work from the outset was to develop an experimental system for eliciting chronic ER stress specifically , in order to be able to determine how the regulation and output of the response changed under such conditions and so identify fingerprints of chronic ER stress associated with disease states . An important caveat is that most agents other than TM that specifically induce ER stress in vitro fail to do so easily or effectively in vivo . While relatively specific in inducing ER stress due to its blockage of N-linked glycosylation , TM also is capable of disrupting carbohydrate metabolism ( Olden et al . , 1979 ) and protein palmitoylation ( Patterson and Skene , 1995 ) , although whether it does so at the doses used here is unclear . In addition , it is conceivable that the chronic stress phenotype is caused not by the ER stress-inducing effect of TM , but by underglycosylation of some specific key protein . However , we do not favor this alternative explanation in part because the degree of protein underglycosylation under these conditions is relatively modest ( e . g . , Figure 1E ) , and also because of the ability of BiP overexpression to phenocopy Hspa5 downregulation ( Figure 5 ) . In addition , the strikingly similar phenotype in Lepob/ob and Leprdb/db mice vouches for this approach . One difference between the chronic condition on one hand and the steady-state in obese mice on the other is that the latter do not appear to express elevated BiP protein levels ( not shown ) , which at face value seems at odds with the idea that elevated BiP protein is needed to both suppress ATF6α and activate or perpetuate RIDD . However , as we have demonstrated , the magnitude of ER stress elicited by feeding is considerably lower than that elicited by tunicamycin ( Figure 6; also compare Figure 6A with Figure 4D ) , yet also much more repetitive than the once-daily TM challenge . It is plausible that the less robust but more constant stimulus of overfeeding results in modest alterations in expression of BiP protein that have little effect on Hspa5 mRNA over the course of meals or days but great effect over the course of months—essentially that small increases in BiP protein levels could produce small decreases in BiP mRNA expression that accumulate over time . It is possible that further moderating the dose of TM and lengthening its extent ( from days to weeks/months ) would more closely mimic the phenotype observed in obese animals , although this would be experimentally cumbersome . It is also worth noting that the suppression of Hspa5 and Hsp90b1 mRNAs might be achieved by other mechanisms that suppress ATF6α activation while perpetuating RIDD . For instance , it has been reported that ATF6α expression is lost in obese animals ( Wang et al . , 2009 ) . Along these lines , we observed that Atf6 mRNA is also significantly downregulated by chronic stress , albeit not as robustly as Hspa5 ( not shown ) , raising the possibility of a positive feedback loop for suppressing Hspa5 expression even as BiP protein levels subside . Alternatively , posttranslational modifications that alter BiP’s interactions with its binding partners ( Chambers et al . , 2012; Wang et al . , 2014; Preissler et al . , 2015 ) might also play a role . Ultimately , the clearest way to test whether the mechanism we describe here is at work in obese animals might be to intercross these animals with Atf6-/- animals; as when given repeated TM injections , these animals should have basally suppressed Hspa5 mRNA expression even at an early age , but then there should be no further suppression as the animals age and become obese . Even if the mechanism turns out to be entirely distinct , the description here of suppressed expression of Hspa5 , and Hsp90b1 and potentially elevated RIDD activity in these animals has potential implications for their ability to maintain ongoing ER homeostasis . It is conceivable that an altered UPR ‘setpoint’ with reduced expression of Hspa5 and other UPR mRNA targets could render the animals more sensitive to ER stress-induced cell death and exacerbate the hepatic inflammation associated with obesity ( Gregor and Hotamisligil , 2011 ) . A surprising conclusion from this work is that RIDD might remain active in the chronic condition , even as UPR-dependent transcriptional signaling is shut off . Admittedly , conclusively demonstrating RIDD activity in vivo is challenging; here we have relied on repression of the well-documented RIDD target Bloc1s1 as the primary readout for RIDD activity . Bloc1s1 expression was repressed in the chronic condition ( Figure 4G ) , upon BiP overexpression ( Figure 5D ) , and in Leprdb/db animals ( Figure 6E ) . Phos-tag immunoblotting showed substantial phosphorylation of IRE1α only in animals treated with a high acute dose of TM , and antibodies purported to be specific for phospho-IRE1 did not produce bands that were absent in cells lacking IRE1α in control experiments ( not shown ) . These limitations made it impossible to directly test whether IRE1α remained phosphorylated in the chronic condition . In addition , there is as yet no functional assay for confirming that RIDD is active in vivo . Rather , substrates are generally identified as putative RIDD targets based on an in vitro assay using purified IRE1α ( Hollien and Weissman , 2006; Hollien et al . , 2009 ) . Then , their RIDD dependence is confirmed by their suppression during ER stress in an IRE1α-dependent XBP1-independent manner . However , Hspa5 and Hsp90b1 are both also transcriptionally upregulated by XBP1 ( Lee et al . , 2003 ) , which would likely confound analysis of their regulation in animals lacking IRE1α . In any case , our data hint at a previously unappreciated role for RIDD , which has generally been considered an anti-adaptive signaling mechanism to this point ( Han et al . , 2009; Lerner et al . , 2012; Upton et al . , 2012; Maurel et al . , 2014 ) . The purpose of the clearance of ER-resident proteins might not so much be to neuter the organelle and hasten its dysfunction during severe stress ( Maurel et al . , 2014 ) as to help turn the response off once stress has been overcome . If RIDD is indeed active under these conditions , such a role could help explain the previously puzzling finding that Hspa5 is a potential RIDD target ( Han et al . , 2009 ) . Assuming RIDD is active during chronic stress , our data provide further evidence that the Xbp1 splicing and RIDD activities of IRE1α can be dissociated ( Lin et al . , 2007; Han et al . , 2008 ) . They raise the possibility that this dissociation occurs not necessarily to separate adaptive from apoptotic outputs so much as to separate activation from deactivation . Given that most stresses encountered by most cells most of the time are likely to be mild and transient—such as that encountered by feeding—the death-accelerating function of RIDD might simply be a byproduct of a normal physiological role in restoring the UPR rheostat . Indeed , a parallel can be drawn with the UPR-regulated transcription factor CHOP which , though it clearly promotes cell death during severe stress ( Zinszner et al . , 1998 ) also might promote normal cellular function during milder stress both by potentiating the dephosphorylation of eIF2α ( Marciniak et al . , 2004 ) and by regulating lipid metabolism ( Tyra et al . , 2012; Chikka et al . , 2013 ) toward the ultimate improvement of ER function . The approach we took to experimentally reconstituting chronic ER stress in vivo was conceptually analogous to our earlier reconstitution of chronic stress in vitro in MEFs ( Rutkowski et al . , 2006 ) . However , neither MEFs nor even cultured hepatocytes showed the same suppression of Hspa5 expression as seen in vivo . This discrepancy points to the existence of other as yet unappreciated factors that influence UPR output in the liver . Indeed , it is now becoming clear that , overlaid atop the structural framework of the canonical UPR lie modulatory pathways for altering UPR sensitivity and output , with potential implications for signaling during chronic stress . These include posttranslational modifications either of the stress sensors—such as S-nitrosylation of IRE1α ( Yang et al . , 2015 ) —or of ER chaperones , such as ADP ribosylation ( Chambers et al . , 2012 ) or AMPylation ( Preissler et al . , 2015 ) of BiP . It will be interesting to test whether any of these modifications is active during the chronic condition . Finally , our results underscore the remarkable capacity of the UPR to adapt to chronic stress . While the severe experimental stresses that lead to massive cell death and organ dysfunction have proven useful in elucidating the signaling capabilities of the UPR when activated to its maximum , they yield little about the ebb and flow of the response at it is most likely to be activated in physiological scenarios . We hope that our approach will stimulate further efforts to more closely mimic physiological stresses in vivo , as these will be essential in linking chronic ER stress to human disease .
All protocols for animal use were reviewed and approved by the University Committee on Use and Care of Animals at the University of Iowa . Animals were fed standard rodent chow and housed in a controlled environment with 12 hr light and dark cycles . Animals used were of both genders unless otherwise noted , with control and experimental groups having similar composition . Animal numbers needed for each experiment were determined based on previous experiments . Animals were fasted for 4 hr prior to sacrifice , which was carried out during the lights-on period . Atf6-/- ( RRID:MGI:3723589 ) ( Wu et al . , 2007 ) animals have been backcrossed into the C57BL/6J strain ( RRID:IMSR_JAX:000664 ) for >10 generations . For chronic TM ( EMD Millipore , Billerica , MA ) treatment , 8 to 12 week old C57BL/6J or Atf6-/- mice were injected once per day ( ~1 hr after lights-on ) intraperitoneally with vehicle or the indicated dose of Tunicamycin dissolved in 150 mM dextrose or PBS and were sacrificed at various timepoints after last injection . For TUDCA ( EMD Millipore ) treatment , animals were injected IP with vehicle or 500 mg/kg TUDCA in PBS once daily , 1 hr after lights-on for 10 days prior to sacrifice . 8 hr prior to sacrifice ( 3 hr after lights-on ) , animals were injected with TM or vehicle . For high fat/no fat feeding , animals were habituated for 5d to a standard defined diet ( D12450B , Research Diets , New Brunswick , NJ ) , fasted 20 hr , and then either injected with 1 mg/kg TM for 4 hr or provided access to a high fat diet ( D12451 , 45% fat ) or a no-fat diet ( D10062804 , 0% fat ) for 4 hr . Lepob/ob and Leprdb/db samples were taken from five month-old female mice . Harvested liver tissue pieces were either frozen immediately for RNA or protein analysis , fixed in formalin ( for histology ) , processed for nuclear isolation , or minced and fixed in formaldehyde ( for ChIP ) . Nuclei were isolated as described ( Rutkowski et al . , 2008 ) . For histological analysis , formalin-fixed tissues were embedded in paraffin , sectioned , and stained with hematoxylin and eosin . Antibodies used for immunoblot were as in ( Chikka et al . , 2013 ) , and additionally as follows: ATF4 ( sc200 , Santa Cruz , Dallas , TX; RRID:AB_2058752 ) ; XBP1 ( sc7160 , Santa Cruz; RRID:AB_794171 ) ; ATF6α ( courtesy of R . Kaufman ) . Antibodies were generally incubated on blots in TBS-Tween + 5% milk at room temperature for 1 hr except for ATF6α antibody , which was incubated in PBS + milk ( no Tween-20 ) at 4°C overnight . MEFs were harvested as described previously ( Lee et al . , 2002 ) from Ern1-/- or Ern1+/+ mouse embryos ( RRID:MGI:3723591 ) . Genotype was confirmed by PCR . Isolation of primary hepatocytes was as in ( Fan et al . , 2004 ) with minor modifications . Hepatocytes were isolated from C57BL/6J mice . Mice were anesthetized with isofluorane . The liver was perfused through the portal vein first with Perfusion Buffer Solution and then with Liver Digest Medium . Formulations were as follows: Liver Perfusion Media: HBSS , no calcium , no magnesium , no phenol red ( Life Technologies , Carlsbad , CA ) , 0 . 5 mM EGTA , 0 . 5 mM EDTA , 25 mM HEPES , Penicillin-Streptomycin ( 10 , 000 U/mL ) ( Life Technologies ) and 0 . 2% BSA ( Research Products International , Mt . Prospect , IL ) ; Liver Digest Media ( 50 mL for 25g mouse ) : HBSS , calcium , magnesium , no phenol red ( Life Technologies ) , 0 . 25 mM HEPES , Penicillin-Streptomycin ( 10 , 000 U/mL ) ( Life Technologies ) , 3 . 6 mg Trypsin Inhibitor ( Sigma , St . Louis , MO ) , 28 mg Collagenase Type IV ( Life Technologies ) added fresh . Digest flow rates were 5 mL/min for 5 min for perfusion and 10 min for digest . The liver was then quickly dispersed and filtered through a sterile 100 µm mesh . Hepatocyte suspensions were then centrifuged at 50x g for 3 min and resuspended to a density of 5×106 cells/ml in DMEM . Viable hepatocytes in the pellet were washed three times and then plated on collagen-coated tissue culture plates in DMEM with 10% calf serum and 100 µg/ml penicillin and streptomycin . After overnight culture , the medium was replaced with F-12 medium containing insulin ( 10 µg/ml ) , dexamethasone ( 67 ng/ml ) , triiodothyronine ( 67 . 3 ng/ml ) , penicillin ( 100 units/ml ) , and streptomycin ( 0 . 1 mg/ml ) . Primary hepatocytes were treated with 5 µg/mL of TM for 4 hr before addition of ActinomycinD ( ActD; Sigma ) to 5 µg/mL . Cells were collected in triplicate 15 min after addition ( considered time 0 ) , 2 hr 15m , and 4 hr 15m after treatment . As all in vitro experiments involved primary cells , cells were not mycoplasma tested . Ad-BiP was created as described ( Young et al . , 2012 ) and amplified by the University of Iowa Gene Transfer Vector Core . Control virus expressing GFP only ( ad-GFP ) was also obtained from the Vector Core . 3×109 pfu/mouse were administered through tail vein injections for hepatic expression . Experiments were performed five days later . RNA and protein analyses were performed as described ( Rutkowski et al . , 2006 ) . Immunoblots were imaged using the ChemiDoc-It imaging system ( UVP , LLC , Upland , CA ) with on-chip integration and empirically derived exposure settings . Membranes were sliced into appropriate molecular weight ranges for blotting; membranes were never stripped and reprobed . Images were processed using Adobe Photoshop . Black hairlines are solely to aid in visual assessment . All contrast adjustments were performed uniformly . Primer sequences and methods utilized for real-time PCR analysis have been published previously ( Rutkowski et al . , 2006 , 2008; Arensdorf et al . , 2013a ) . Xbp1 RT-PCR was as described ( Tyra et al . , 2012 ) . Additional primers are described here: Xbp1 ( s ) : Fwd GAGTCCGCAGCAGGTG , Rev GTGTCAGAGTCCATGGGA; genomic Hspa5 Primer ( 5’ UTR of Hspa5 gene ) Fwd TAAGACTCCTGCCTGACTGC , Rev GGAATAGGTGGTCCCCAAG; ChIP primers: Hspa5 Promoter ( −215 to +8 bp ) Fwd CATTGGTGGCCGTTAAGAATGACCAG , Rev AGTATCGAGCGCGCCGTCGC; Hspa5 Intron 7 ( +2727 to +2906 ) Fwd GGGAGGACTGTTGCTTTAGG , Rev TGAATGAACTCTTGCCATCTTC . ChIP was performed as in ( Arensdorf and Rutkowski , 2013; Chikka et al . , 2013 ) with minor modifications . Formaldehyde-fixed liver tissues were weighed out after being quenched with 1 . 5M Tris-HCl and pulverized in a metal mortar and pestle . Samples were sonicated using a Covaris E220 sonicator . Chromatin was immunoprecipitated using ATF6α ( sc-22799x , Santa Cruz; RRID:AB_2242950 ) or Pol II ( sc-899x , Santa Cruz; RRID:AB_632359 ) antibodies or non-specific IgG ( 12–370 , Millipore; RRID:AB_145841 ) overnight at 4°C . DNA was purified using a standard phenol/chloroform extraction protocol . Samples were then analyzed by quantitative real-time PCR with an annealing temperature of 58°C . Groups were compared by one-way ANOVA . All quantitative data are presented as means ± S . E . M . Tukey’s Post-Hoc analysis was used when comparing multiple conditions for the same readout . Statistical comparisons for qRT-PCR and qPCR data were carried out prior to transforming data out of the log scale . All replicates were biological rather than technical , and ‘n’ numbers are given in figure legends; for animal experiments , this refers to number of animals . For cell culture experiments , this refers to numbers of independently-treated plates . Each experiment was performed at least twice , and some experiments ( such as chronic stress treatments ) were performed many times . There was no removal of outliers; no data were discarded unless there was an unambiguous indication that the experiment failed for technical reasons . | Toxic chemicals , extreme temperatures and other abnormal environmental conditions can cause the cells in our bodies to become stressed . Several kinds of stresses overwhelm a compartment in the cell called the endoplasmic reticulum , which is critical for processing new proteins so that they can work correctly . Endoplasmic reticulum stress has been linked to long-term diseases such as diabetes , cancer and neurodegenerative diseases . Most of what is known about how cells sense and respond to endoplasmic reticulum stress comes from studies on isolated cells that were subjected to harsh conditions that cells cannot tolerate for longer than a day or two . By contrast , little is known about how cells within whole organisms respond to milder but longer-lasting endoplasmic reticulum stress , which is closer to what occurs during disease . To investigate this issue , Gomez and Rutkowski treated mice repeatedly with a chemical that causes mild endoplasmic reticulum stress in the liver . The cells exposed to this persistent stress responded differently to those exposed to severe short-term stress . Whereas short-term stress causes liver cells to turn on genes that help the endoplasmic reticulum to process proteins more efficiently , long-term stress causes cells to turn off some of those genes . Further investigation revealed that cells in the livers of obese mice show similar patterns of gene activity as cells exposed to long-term endoplasmic reticulum stress . The findings presented by Gomez and Rutkowski could therefore also help us to understand more about the liver problems that often occur during obesity and diabetes . Further studies are now needed to examine exactly how long-lasting stress can shut off the cells’ protective mechanisms . Future experiments could also investigate whether other types of cells and organs respond to long-term endoplasmic reticulum stress in the same way as cells in the liver . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2016 | Experimental reconstitution of chronic ER stress in the liver reveals feedback suppression of BiP mRNA expression |
To understand the function of eukaryotic cells , it is critical to understand the role of protein-protein interactions and protein localization . Currently , we do not know the importance of global protein localization nor do we understand to what extent the cell is permissive for new protein associations – a key requirement for the evolution of new protein functions . To answer this question , we fused every protein in the yeast Saccharomyces cerevisiae with a partner from each of the major cellular compartments and quantitatively assessed the effects upon growth . This analysis reveals that cells have a remarkable and unanticipated tolerance for forced protein associations , even if these associations lead to a proportion of the protein moving compartments within the cell . Furthermore , the interactions that do perturb growth provide a functional map of spatial protein regulation , identifying key regulatory complexes for the normal homeostasis of eukaryotic cells .
Post-translational protein modifications such as phosphorylation or ubiquitylation often alter the affinity of one protein for other proteins or cellular components , which drive their movement within the cell ( Scott and Pawson , 2009 ) . Protein relocalization is critical for many cellular processes , including the asymmetric division of adult stem cells , which underlies metazoan development . The importance of protein localization is also highlighted by diseases ranging from cystic fibrosis to cancer that result , in part , from protein mislocalization ( Hung and Link , 2011 ) . The evolution of new modes of protein regulation requires new associations to form , but currently we do not know how tolerant the cell is of novel protein interactions . For example , can a nuclear kinase relocate to the cytoplasm without consequence ? Various methodologies have been developed to allow specific affinity-based relocation of proteins in vivo . For example , some systems are designed to disable a location-specific function by sequestering proteins to a specific compartment ( Haruki et al . , 2008; Robinson et al . , 2010 ) . Alternatively , a leucine zipper-based system was developed to screen for pairwise protein associations , provided that selection for a phenotype is possible ( Devit et al . , 2005 ) . However , none of these approaches have systematically assessed the effects of creating pairwise protein associations , one at a time , across the entire proteome . To address this , we made use of the Synthetic Physical Interaction ( SPI ) system ( Olafsson and Thorpe , 2015 ) to create high-affinity interactions between each of the ~six-thousand members of the eukaryotic yeast proteome and target proteins in each of the major cellular compartments . This has allowed us to assay the effect of each of these in vivo binary protein interactions individually upon the normal growth of cells . We find that most protein-protein interactions are benign to the normal growth of cells , but that specific interactions do perturb growth - these interactions are termed Synthetic Physical Interactions or SPIs ( Olafsson and Thorpe , 2015 ) . The SPIs are enriched for functional regulators , indicating that constitutive colocalization of a regulator with its target causes a growth defect . We are able to use SPIs to identify novel regulatory proteins; for example , we examine SPIs between the kinetochore protein Nuf2 and both Hmo1 and Sgf29 and find that these two proteins are required to regulate the levels of outer kinetochore proteins . Furthermore , the SPIs correlate with the quaternary structure of large protein complexes such as the kinetochore or nuclear pore . As such , the SPIs provide a powerful tool to complement existing physical and genetic interactions .
The SPI system uses a GFP-binding protein ( GBP ) derived from an alpaca antibody ( Rothbauer et al . , 2006 ) , which when fused to a target protein of interest creates binary associations in vivo with GFP-tagged proteins ( Rothbauer et al . , 2006; Rothbauer et al . , 2008; Grallert et al . , 2013 ) . We define a target protein as one fused with the GBP and a query protein as one tagged with GFP . By introducing GBP-target proteins into strains encoding GFP-query proteins , we induce an affinity between the target and query proteins via the strong binding of GBP to GFP . We used the Selective Ploidy Ablation technique ( Reid et al . , 2011 ) to introduce a plasmid encoding the GBP-target protein into the collection of ~6000 GFP strains , each of which has a chromosomally integrated GFP introduced at the 3’ end of a specific open-reading frame ( Huh et al . , 2003 ) . In each resulting haploid strain , the GBP-target protein is plasmid-encoded and the GFP-query protein is endogenously-encoded; we are therefore able to create a binary protein-protein interaction and assess the effects of this interaction upon growth . We used two independent controls , which were separately transferred into the GFP collection . The first control encodes the GBP alone , and the second encodes the target protein . These two constructs control both for the effects of binding a protein to the GFP tag and also for the ectopic expression of the target gene in each GFP strain . We chose 23 different target proteins that represent 18 of the major cellular compartments ( Figure 1A and Figure 1—source data 1 ) , such as the nucleus ( Pus1 and Rad52 ) , the cell membrane ( Psr1 ) , and the endoplasmic reticulum ( Sec63 ) . The genes encoding these target proteins were fused with GBP and transferred into every strain of the GFP collection ( Figure 1—source data 1 ) . Thus , for each target protein , we create ~6000 strains each of which contains the target GBP-tagged protein together with a specific GFP-query protein . The effect on growth was assayed by comparing the colony sizes of strains containing the GBP-GFP interaction with the two controls ( Figure 1B , C ) ( Dittmar et al . , 2010 ) . The two controls gave equivalent results ( Figure 1—figure supplement 1 and as previously reported Olafsson and Thorpe , 2015 ) and consequently an average growth score was used . 10 . 7554/eLife . 13053 . 003Figure 1 . Quantitative analysis of the effects of binding proteins throughout the cell . ( A ) A schematic diagram of S . cerevisiae indicating the cellular compartments and target proteins within the cell that were associated with each member of the proteome . ( B ) A 1536 colony plate from the Sec63 screen . The inset below shows the highlighted row from the Sec63-GBP plate , the Sec63 control plate and the GBP-only control plates respectively . Growth defects are indicated with a black line . ( C ) The z-scores of all 5734 proteins in each of the 23 screens . For each screen , the strains are ranked according to order of z-scores , positive z-scores indicate a growth defect relative to controls . The inset highlights the strains with the largest growth defects in each screen . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 00310 . 7554/eLife . 13053 . 004Figure 1—source data 1 . Z-scores of growth defects caused by protein associations . The z-scores for each binary protein interaction are listed for each control . The mean smoothed z-scores , plotted in panel C , are listed separately . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 00410 . 7554/eLife . 13053 . 005Figure 1—source data 2 . Plasmids used in this study . A list of the plasmids used to generate the SPI data are listed . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 00510 . 7554/eLife . 13053 . 006Figure 1—figure supplement 1 . The correlation between the two controls . ( A ) x-axis: z-scores quantifying the difference in colony sizes between strains expressing the target protein fused with GBP and strains expressing the target protein alone , to control for the effects of ectopic expression . y-axis: z-scores quantifying the difference in colony sizes between strains expressing the target protein fused with GBP and strains expressing the GBP alone , to control for the effects of binding a protein to the GFP tag . A positive z-score indicates a growth defect of the GBP-tagged target protein relative to its control . ( B ) The relative growth ( Log Growth Ratios , LGRs ) from the high-density re-testing of Sec63-GBP SPIs relative to the two controls ( the target protein alone – x-axis and the GBP alone – y-axis ) . The relative growth of the negative controls ( no GFP tagged protein ) are shown in green , whereas the retested Sec63 SPIs are shown in blue . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 006 We expected that many of the forced associations would disrupt cellular homeostasis , but unexpectedly , we found that 98% of GBP–GFP combinations ( 129 , 098 out of 131 , 882 ) do not affect the growth of cells ( Figure 1C ) . These data imply that cells are surprisingly permissive for most protein-protein interactions and as a corollary that cells are broadly tolerant of proteins being relocated within the cell . In cases where fluorescent imaging was able to detect protein relocalization , we confirmed that ~72% of interactions do occur . Since the GBP tag is linked to red fluorescent protein ( RFP ) , we were able to assay colocalization with GFP . We examined 552 GBP-GFP combinations - each of the 23 GBP-tagged target proteins separately combined with a random selection of 24 GFP-tagged query proteins - using live cell imaging ( Figure 2—source data 1 , for examples see Figure 2 and Figure 2—figure supplement 1 ) . Of the 524 GBP-GFP combinations that we could score , 210 ( 40% ) are already in the same compartment and so we cannot determine whether GFP and GBP associate , of the remaining 314 , 225 were detectably colocalized ( Figure 2C ) , indicating that in the majority of cases the protein-protein associations do occur ( Figure 2 , Figure 2—figure supplement 1 and Figure 2—source data 1 ) . These observations are therefore consistent with the notion that most synthetic protein-protein interactions do not cause a growth defect . 10 . 7554/eLife . 13053 . 007Figure 2 . Colocalization of target GBP protein and query GFP proteins . ( A ) Cdc11-GBP relocalizes to the Golgi when bound to Sec26-GFP . ( B ) Cdc55-GFP relocalizes to the mitochondria when bound to Om45-GBP . ( C ) Bar chart of the proportion of colocalization ( n=552 ) , note that the colocalized category includes 210 combinations where the target and query proteins are within the same compartment and so protein-protein association will not be apparent from this microscopy analysis . ( D ) Bar chart of the direction of movement of GFP and GBP ( n=225 ) . ‘To query protein - GFP’ indicates relocation of the majority of the GBP target protein to GFP ( see A ) ; ‘To target protein - RFP’ denotes relocation of the majority of the GFP query proteins to the GBP-RFP target ( see B ) . ‘Both locations’ indicates that GBP and GFP proteins are in both their normal location and those of the other protein ( e . g . Figure 2—figure supplement 1B ) . ‘Neither location’ denotes both GFP and GBP proteins are colocalized , but not to either of their normal locations ( e . g . Figure 2—figure supplement 1C ) , whereas ‘Regionally colocalized’ indicates one protein is in the same region of the cell as the second protein , but not completely colocalized ( see E ) . ‘Foci only’ designates that the proteins relocalized to discrete foci ( see Figure 2—figure supplement 1E ) . Two categories are omitted from this analysis , first those cells which were uncharacterized , typically because the cells were dead . Second , cells in which the target and query protein reside in the same cellular location , such that microscopy is not informative on whether or not they are associated , this latter category make up ~40% of our combinations . ( E ) Hta2-GBP is displaced from the nucleus when bound to Spt6 . The scale bars are 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 00710 . 7554/eLife . 13053 . 008Figure 2—source data 1 . Localization summary . These data provide a detailed list of the localization type for each GBP-GFP protein pair . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 00810 . 7554/eLife . 13053 . 009Figure 2—figure supplement 1 . Fluorescent images of colocalization between GBP and GFP . ( A ) GBP cololcalizes with Rpa49-GFP . ( B ) Both GFP and RFP can be detected in the nucleus and peroxisomes when Ski6-GFP and SKL-GBP bind . ( C ) Pan1-GFP and Om45-GBP colocalize in neither of their expected locations . ( D ) Med7-GFP and Pus1-GBP both localize to the nucleus , so displacement due to colocalization is impossible to detect . ( E ) Pan1-GFP and Nup53-GBP colocalize to form foci in the cell . ( F ) Okp1-GFP and Cdc11-GBP do not colocalize . The GFP kinetochore signal and RFP bud neck signal are still clearly visible in the combined strain . Scale bars are 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 00910 . 7554/eLife . 13053 . 010Figure 2—figure supplement 2 . Colocalization of Pil1-GBP with histone subunits . Localization of PIL1-GBP-RFP is shown , along with Htb1-GFP , Htb2-GFP , Hta2-GFP , and Hhf2-GFP . DIC , GFP , RFP and merged images then show the colocalization of each histone subunit ( GFP ) with Pil1-GBP-RFP . Scale bars are 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 010 The microscopy analysis also allows us to examine whether the GBP-tagged target protein recruits the GFP protein to its location or vice versa . We anticipated that each binary protein association would create a ‘tug-of-war’ between the target protein and the query protein . The image data support this notion; where it is possible to distinguish the location of two proteins in the cell , we observed that there are roughly equal instances of the GBP protein recruiting the GFP protein as the reverse ( Figure 3 and Figure 3—figure supplement 1 ) . However , this generalization is not true for some classes/types of proteins . When we look at individual GBP or GFP proteins , we find that structural components more often recruit proteins to their location than enzymes that are not anchored to a specific location ( Figure 3 , Figure 3—figure supplement 1 and Figure 2—source data 1 ) . For example , GFP-tagged cytosolic query proteins such as Cdc55 and Snf1 mostly relocalize to their target proteins ( Figure 2B and Figure 3—figure supplement 1 ) , whereas the GFP-tagged nucleolar proteins Rpa49 and Pwp2 more often recruit GBP-tagged target proteins to their location ( Figure 3—figure supplement 1 ) . There are some rare cases where the two proteins localize to both locations and also where one or both proteins mislocalize to a new location that is foreign to both ( Figure 2D ) . An example of the latter is the recruitment of the nucleosome remodeling protein , Spt6 , to the histone subunit Hta2 . Constitutive recruitment of a nucleosome remodeler to the chromatin might be expected to give a phenotype and indeed we find that the histone subunit Hta2-GBP is strikingly no longer restricted to the nucleus ( Figure 2E ) concomitant with a strong growth defect . It is possible that we are overestimating the extent of relocalization caused by the GFP-GBP interaction . First , since the target and query proteins are not stoichiometrically matched , some of the GFP or GBP protein will likely remain at its native location . Second , it is possible that in some cases either the GFP tag or the GBP tag is cleaved from its query or target protein respectively , thus giving a false indication of colocalization . It is also possible that imaging underestimates the proportion of relocalization , since we could not score the 210 combinations where proteins are already in the same compartment , these are perhaps more likely to associate via the GFP-GBP interaction . Furthermore , it should be noted that in some cases where we could not detect that the GFP and GBP proteins were colocalized , there was nevertheless either a growth phenotype or a change in the location of one of the proteins . For example , of the 15 Iqg1 associations that failed to show protein colocalization ( Figure 3 ) , 14 show mislocalization of either the Iqg1 target protein or the GFP query protein . 10 . 7554/eLife . 13053 . 011Figure 3 . Direction of colocalization . ( A ) The proportion of the 24 query proteins that colocalized in the direction indicated . Categories used to characterize the direction of colocalization are described in Figure 2 . The ‘Uncharacterized’ category includes strains where there were no cells to image , which is often the case if the interaction perturbs growth . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 01110 . 7554/eLife . 13053 . 012Figure 3—figure supplement 1 . Direction of colocalization . A pie chart for each of the 24 randomly selected query proteins shows the proportion of the 23 target proteins observed to have colocalized in the direction indicated . Categories used to characterize the direction of colocalization are described in Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 012 Around 2% of the forced interactions restrict growth ( Figure 1C , 4A and Figure 1—source data 1 ) . However , we note that of the 6000 GFP-tagged proteins used in this study , only ~4000 have been validated and are clearly observable ( Huh et al . , 2003 ) . We therefore reanalyzed the proteome-wide data using only 3905 GFP strains with unambiguous fluorescence signal ( Tkach et al . , 2012 ) and find that ~3% restrict growth ( Figure 4—source data 1 ) , consistent with the notion that most protein-protein associations do not restrict growth . We did not use a specific threshold cutoff to define a SPI , rather we confirmed the SPIs with the greatest impact on cell growth for each GBP by repeating the assay starting with the strongest interaction and proceeding sequentially through the SPIs until the false discovery rate ( FDR ) reached 40% ( Figure 4—figure supplement 1 ) . Associations that produced a growth defect relative to controls with 16 replicates in the confirmation experiments are considered SPIs . Thus , some SPIs result from relatively mild growth defects , as outlined in Figure 4—source data 1 . We note that the false negative rate may be significant , since we did not test further than the 40% FDR and due to the limitations of measuring growth by colony size . Using this approach , we confirmed 2784 SPIs in total produced by 727 GFP-tagged query proteins with one or more of the 23 target proteins ( Figure 4—figure supplement 2 and Figure 4—source data 1 ) . 10 . 7554/eLife . 13053 . 013Figure 4 . Comparisons of synthetic physical interaction screens . ( A ) Cluster analysis of the SPI data . The 23 screens are arranged horizontally and the 727 GFP strains clustered vertically . High z-scores ( positive; >2 ) in yellow and low ( negative; < -1 ) scores in blue . Three distinct clusters are highlighted ( a , b , and c ) and described in Figure 4—figure supplement 6 . ( B ) Spearman’s Rank Correlation Coefficients for the different SPI screens shows similar compartments give similar SPIs , for example , Sec63 and Loa1 cluster together as do two kinetochore proteins Nuf2 and Dad2 . ( C ) The notched box-and-whisker plot indicates the distributions of the retest log growth ratios and indicates that SPIs produced by a query protein and a target protein from different compartments produce stronger growth defects than those from the same compartments ( ***indicates a p-value = 1 . 8x10-5 , Wilcoxon's rank-sum ) . The plot shows the median value ( bar ) and quartiles ( box ) , the whiskers show the minimum of the range or 1 . 5 interquartile ranges , outlying data points are indicated as circles and the notches indicate the 95% confidence intervals of the medians . ( D ) The GFP proteins with SPIs have , on average , more protein-protein interactions than non-SPI query proteins , the notched box-and-whisker plot is in the same format as panel B ( ***indicates a p-value < . 2x10-16 , Wilcoxon’s rank-sum ) . The 727 SPI query proteins ( red ) are superimposed upon the yeast interactome with proteins with ≥10 interactions shown as larger squares . ( E ) The CLIK interaction density plot for Sec63 is shown ( see Figure 4—figure supplement 5 for the other CLIK plots ) . The ~500 Sec63 associations that show the strongest growth restriction have a high interaction density ( inset ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 01310 . 7554/eLife . 13053 . 014Figure 4—source data 1 . High-density retesting of the SPIs . All the high-density SPI data to retest the strongest interactions are listed together with a list of those GFP strains that produce a reproducible SPI with each of the 23 GBP proteins . We also include a list of the growth data for a subset of GFP proteins whose expression and location is well characterized . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 01410 . 7554/eLife . 13053 . 015Figure 4—figure supplement 1 . False discovery rates ( FDR ) . Strains were ordered from highest to lowest z-score and the strongest 80 growth defects were retested with 16 replicates ( 80 strains with 16 controls per plate ) . The FDR was calculated for each batch of 20 strains working down the list of z-scores , shown here in blue . Black lines indicate three-point moving averages . Vertical dashed lines group points from each retest plate together . No more retests were performed once a screen had reached 40% FDR , indicated by a horizontal green dashed line . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 01510 . 7554/eLife . 13053 . 016Figure 4—figure supplement 2 . Frequent SPIs are with lower abundance query proteins . ( A ) The distribution of SPIs is shown as a bar chart illustrating that most SPI query proteins give a SPI with only one or a few target proteins . Nevertheless , 75 query proteins have SPIs with at least 10 target proteins , the ‘Frequent SPIs’ . ( B ) Frequent SPI query proteins have a lower mean abundance than non-frequent SPI query proteins ( t-test , p>0 . 006 ) . Error bars indicate standard error of the mean . ( C ) Frequency data shown in ( A ) plotted in terms of protein abundance . The mean abundance values in each category are indicated as red lines a linear trendline is shown in black . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 01610 . 7554/eLife . 13053 . 017Figure 4—figure supplement 3 . The effect of protein abundance on the number of verified SPIs . ( A ) The number of verified SPIs in each abundance category is shown for each screen , where abundance categories are bins containing an equal number ( 421 ) of proteins . The dashed line indicates the number of SPIs expected per category if distribution was entirely unbiased . ( B ) GBP-tagged protein levels were assessed via RFP fluorescence . Addition of copper to the medium increased the total protein concentration by as much as twofold . ( C ) 400 GFP strains were chosen to provide representatives in each protein abundance category , with a bias toward those in the highest abundance category . These GFP strains were retested with the four GBP-tagged proteins at different copper concentrations ( 0 , 20 , and 80 µM ) and the proportion of SPIs within each category are indicated . Addition of copper did not increase the proportion of SPIs specifically with the high-abundance protein categories . It is of note that increasing the amount of Hta2-GBP produced more SPIs in all abundance categories , whereas increasing the amount of Nop10-GBP reduced the number of SPIs in all categories . Hence , although the amount of GBP protein can affect the SPIs , it does not correlate with GFP protein levels . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 01710 . 7554/eLife . 13053 . 018Figure 4—figure supplement 4 . Frequent SPIs are rarely dominant . ( A ) From the Nuf2 SPI screen 41 frequent SPI query proteins ( left panel ) and 40 non-frequent SPI query proteins ( right panel ) were retested both as haploids and diploids . The numbers in the box to the left of the ORF name indicate the number of screens , out of 23 , that the GFP-query protein was detected as a SPI . All the frequent SPI query proteins were not dominantly restricted for growth when associated Nuf2 , compared with 15% ( 6/40 ) of the non-frequent SPI query proteins . ( B ) An example of the raw data with one frequent SPI ( Nuf2-Tbf1 ) and one non-frequent SPI ( Nuf2-Pan1 ) illustrate the suppression of the former SPI in diploid cells . ( C ) The mean log growth ratio of the combined 41 frequent and 40 non-frequent SPI query proteins from the Nuf2 SPI screen shows that there is no difference between frequent and non-frequent haploid SPI query proteins , in contrast the frequent diploid SPI query proteins are suppressed compared to non-frequent diploid SPI query proteins . Error bars indicate standard error of the mean , and **indicates a t-test p-value of 0 . 006; n . s . indicates not statistically significant . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 01810 . 7554/eLife . 13053 . 019Figure 4—figure supplement 5 . CLIK ( Cutoff Linked to Interaction Knowledge ) outputs for each of the 23 screens . Strains are ranked according to z-score and plotted in this order along the x- and y- axes ( at position 0 is the strain with the strongest growth defect ) . Points are plotted where a genetic or physical interaction exists between two strains , and colors indicate high density of interactions . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 01910 . 7554/eLife . 13053 . 020Figure 4—figure supplement 6 . Analysis of three sub-clusters from the 727 SPI heat-map from Figure 3A . ( A ) Many components of the mediator complex , transcription factor complex , and mRNA cleavage and polyadenylation specificity factor complex cluster together the SPI data ( cluster a in Figure 4A ) . ( B ) Almost all components of the COP1 vesicle coat ( or coatomer ) , nuclear pore outer ring ( specifically the NUP84 complex ) , and signal recognition particle ( SRP ) subunits cluster together ( cluster b in Figure 4A ) . ( C ) Members of the TRAMP complex cluster together and are specifically SPIs with the target protein Pus1 ( cluster c in Figure 4A ) . p-Values for gene ontology ( GO ) term enrichments are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 020 One possible cause of the SPIs is that the target protein would sequester the GFP-tagged query protein away from its normal location . Should this be the case , we would expect low-abundance proteins to be more susceptible to growth defects . However , this is not generally the case for most SPIs , consistent with our earlier findings ( Olafsson and Thorpe , 2015 ) , since we found there was no correlation between protein abundance and the z-scores ( a relative measure of growth ) from the 23 GBP screens ( R2 values ≤0 . 004 ) . To address the issue further we grouped all GFP strains into eight categories based upon the abundance of their GFP proteins , each group has 421 proteins . We then plotted the proportion of GFP strains within each group that produced SPIs with a given GBP target ( Figure 4—figure supplement 3A ) . Broadly , there are no abundance categories that are consistently enriched for SPIs with all GBP associations . However , we did note that in some cases the group of most abundant proteins had fewer SPIs than the other groups ( for example Hta2 and Sec63 , see Figure 4—figure supplement 3A ) . To assess whether the levels of the GBP-tagged protein would influence the SPIs , we altered the GBP-tagged protein levels by virtue of their constitutive copper promoter . The CUP1 promoter functions in the absence of copper and its expression can be gradually increased by adding copper to the growth media . We confirmed that upon addition of increasing amounts of copper , the levels of the GBP target proteins increased , as assayed by quantitative fluorescence imaging of the RFP tag attached to GBP ( Figure 4—figure supplement 3B ) . We then retested 400 GFP strains , representing both high and low abundance proteins , with four different GBP target proteins , two of which had less SPIs with high-abundance proteins than expected ( Hta2 and Sec63 ) . The results indicate that increasing the expression of the GBP proteins does not specifically increase the number of SPIs within high abundance categories ( Figure 4—figure supplement 3C ) . Nevertheless , we expected that a subset of proteins would be particularly sensitive to the effects of forced association and relocalization and this proved true . When we examine all the 727 SPI query proteins collectively ( Figure 4A ) , we find that 75 GFP query proteins produce SPIs with at least 10 of our 23 GBP-tagged target proteins ( Figure 4—figure supplement 2A ) . These ‘frequent SPI query proteins’ are on average of lower abundance than less frequent SPI query proteins ( Figure 4—figure supplement 2B , C ) , also they are enriched for essential genes ( ≈83% ) and for proteins whose gene ontology ( GO ) terms include RNA metabolism ( p-value = 9 . 26x10-5 ) , mRNA polyadenylation ( p-value = 1 . 63x10-9 ) , cytoplasmic and nuclear transport ( p-values = 1 . 14x10-8 and 1 . 69x10-7 , respectively ) , microtubule nucleation ( p-value = 5 . 09x10-8 ) , and spindle pole body ( p-value = 3 . 22x10-8 ) . We have previously shown that these interactions are mostly suppressed by having an untagged copy of the query protein present in the cell ( Olafsson and Thorpe , 2015 ) . In heterozygous diploid strains , the untagged version of the SPI query protein is able to complement for the tagged version of the protein that is mislocalized via its association with the target protein . To confirm that the frequent SPI query proteins fall into this category we retested 41 SPIs from the Nuf2 screen that fall into the frequent SPI query proteins group and 40 from the non-frequent SPI query proteins group . Consistent with our expectation all 41 frequent SPIs are suppressed in heterozygous diploid cells , whereas 15% ( 6 out of 40 ) SPIs in the non-frequent group were reproduced in diploid cells ( Figure 4—figure supplement 4 ) . Thus , we conclude that these frequent SPI query proteins are predominantly those whose essential function is location-dependent and whose sequestration to another compartment results in a growth defect ( as is routinely achieved using other systems Haruki et al . , 2008 ) . To understand whether associations to similar areas of the cell create growth defects from common sets of query proteins , we compared the SPIs generated from each target protein . Spearman’s correlation coefficient analysis ( Lubbock et al . , 2013 ) indicates that , in specific cases , SPI screens using target proteins from the same cellular compartment give similar SPIs ( Figure 4B ) . For example , the Pus1 and Rad52 target proteins , which are both in the nucleus , produce SPIs with a similar set of query GFP proteins . However , it is interesting to note that some target proteins from the same cellular compartment give quite distinct sets of SPIs . For example , the SPI data for nuclear proteins Nop10 ( nucleolus ) , Heh2 ( nuclear membrane ) , and Hta2 ( histone ) cluster together but are distinct from both Pus1 and Rad52 ( two nuclear enzymes ) . We suggest that these SPIs segregate into two different classes because Pus1 and Rad52 are non-essential nuclear enzymes , whereas Nop10 , Heh2 , and Hta2 are structural components , which may be more sensitive to movement . We next asked whether SPI query proteins would be located in the same cellular compartment as their target protein . SPIs between query and target proteins that normally localize to the same cellular compartment are enriched ( 10 . 4% of our confirmed SPIs are with target and query proteins from the same compartment , versus an expected value of 7 . 1% for the full dataset , p-value = 1 . 8x10-9 , Fisher's Exact test ) . Also , this notion is true in specific cases , particularly for nuclear proteins . For example , SPIs with a nucleolar protein , Nop10 , are enriched for nucleolar components ( 21 out of 115 , p-value = 1 . 8x10-8 , Fisher’s exact test ) or SPIs with the microtubule-associated kinetochore component Nuf2 , which are enriched for microtubule components ( described below ) . This pattern was typical of nuclear proteins , but not evident for other proteins: for example , the SPIs with the mitochondrial protein Om45 did not include any mitochondrial proteins . However , it should be noted that although there are more SPIs between proteins in the same compartment , SPIs produced by proteins in different compartments tend to give a greater growth defect ( Figure 4C ) . Unexpectedly , we find that SPI query proteins are enriched for characterized physical interactions , compared with non-SPIs ( p<2 . 2 x 10–16 , Wilcoxon’s rank-sum ) . This is visualized by overlaying all the confirmed SPI query proteins onto a graph of the yeast physical interaction dataset ( HINT database ( Das and Yu , 2012 ) , Figure 4D ) . We also asked the same question for each SPI screen using the Cutoff Linked to Interaction Knowledge tool ( CLIK ) , which examines quantitative data for interaction density ( Dittmar et al . , 2013 ) . The CLIK tool ranks all genes/proteins by their z-score ( high scores bottom left , low scores top right ) and then plots the interaction density between all proteins ( using data from the Biogrid database Stark et al . , 2006 ) . If , from a specific target screen , the most growth restricted query proteins are collectively enriched for genetic or physical interactions then a cluster of high density will be visible in the bottom left of the density plot . Most SPI screens have a strong enrichment for genetic and physical interactions indicating that the strongest SPIs share interactions ( Figure 4E and Figure 4—figure supplement 5 ) , which is a predictor of common function . The overlap with physical interactions is particularly surprising; indicating that proteins that normally interact together can induce a growth defect when constitutively bound . Collectively , these observations are consistent with the idea that proteins and their regulators are often located within the same compartment , but their temporal or spatial physical association is tightly regulated . The SPIs for each target protein are also enriched for proteins involved in regulating their function . Gene ontology enrichment analysis for the SPIs demonstrates that specific functional classes of proteins are enriched for each cellular compartment . For example , SPIs for the DNA repair protein Rad52 are enriched for components of the nuclear pore ( Ndc1 , Nic96 , Nup1 , Nup85 , Nup49 , Nup57 , Nup84 , Nup145 , and Nup192; p-value = 7x10-9 ) , specifically the Nup84 complex , which functions in specialized types of DNA repair ( Nagai et al . , 2008 ) . Another example is the kinetochore protein Nuf2 , whose SPIs are enriched for proteins involved in microtubule organization ( Ark35 , Bir1 , Cbf2 , Cdc14 , Ctf19 , Dad2 , Dad4 , Dsn1 , Ipl1 , Kip1 , Kip3 , Okp1 , Spc24 , Spc29 , Spc42 , Spc105 , Spc110 , Stu1 , and Tub4; p-value = 8x10-13 ) . Nuf2 is an outer kinetochore protein whose calponin-homology domain directs microtubule binding ( Wei et al . , 2007; Ciferri et al . , 2008 ) . As such , the SPIs may include numerous novel regulators of their target proteins ( Olafsson and Thorpe , 2015 ) . To test this , we examined three Nuf2 SPIs in more detail . Hmo1 , Sgf29 ( both chromatin-associated proteins ) and Sst2 ( a GTPase activating protein ) all gave a strong SPI phenotype with the kinetochore protein Nuf2 . Only one of these mutants , hmo1∆ , gives a chromosomal instability phenotype ( Stirling et al . , 2011 ) and none have a reported role in kinetochore function . The SPI data ( Figure 4A ) cluster Hmo1 adjacent to Dad4 , an outer kinetochore protein and with other kinetochore proteins ( Mcm21 , Okp1 , Nkp2 , Ctf19 , and Spc24 ) . To test whether the Hmo1-Nuf2 SPI was unique in the kinetochore , we tested various other kinetochore target proteins fused with GBP in an Hmo1-GFP strain . We find that in addition to Nuf2 , Hmo1 has SPIs with Mif2 and Ctf19 , but not Kre28 , Mtw1 , Dad2 , Ctf3 , Chl4 , Skp1 , Cnn1 , or Cbf1 ( Figure 5A ) . These data suggest that the Hmo1 SPI is specific to central/outer kinetochore components . We examined fluorescently tagged kinetochore proteins in hmo1∆ , sgf29∆ , and sst2∆ cells . We chose two kinetochore proteins , Mtw1 and Dad4 , both of which are at the central and outer kinetochore , respectively , and adjacent to Nuf2 , Ctf19 and Mif2 and also have been used in quantitative studies ( Joglekar et al . , 2006; Ledesma-Fernández and Thorpe , 2015 ) . Strikingly , we find that hmo1∆ and , to a lesser extent , sgf29∆ cells both have elevated levels of Dad4 outer kinetochore protein associated with their centromeres , although the levels of Mtw1 were unaffected ( Figure 5B , C and 5D ) . However , Hmo1 stimulates the activity of the SWI/SNF chromatin remodeling complex ( Hepp et al . , 2014 ) and therefore may affect expression of the DAD4 gene . To test whether the hmo1∆ mutant was affecting Dad4 protein levels we quantified total cellular Dad4-YFP fluorescence in wild-type and mutant cells and find approximately one third of hmo1∆ cells have higher levels of Dad4 than those found in wild-type cells ( Figure 5—figure supplement 1 ) . Nearly half of the hmo1∆ cells have Dad4 levels in the wild-type range ( +/- one standard deviation of the wild-type mean ) ; hence cellular Dad4 protein levels are not sufficient to explain the aberrant Dad4 foci seen in most hmo1∆ cells ( Figure 5B ) . Furthermore , it has previously been shown that Hmo1 is associated with purified kinetochores ( Akiyoshi et al . , 2010 ) , consistent with a direct role at the kinetochore . These data support the notion that in specific cases SPIs define functional regulators . 10 . 7554/eLife . 13053 . 021Figure 5 . Nuf2 SPIs affect kinetochores . ( A ) The Hmo1-GFP query protein encoding strain was transformed separately with 13 plasmids encoding different kinetochore proteins target proteins tagged with GBP ( 4 replicates each ) . The growth relative to controls ( GBP alone and target protein alone ) was assessed as in Figure 1 . ( B ) Deletion of HMO1 , SGF29 , and SST2 were separately introduced into strains encoding Dad4-YFP and Mtw1-YFP at their endogenous loci . Fluorescence imaging of these strains reveals that hmo1∆ mutants have large-bright Dad4-YFP kinetochore foci ( red arrows ) and some weak foci ( green arrows ) . sgf29∆ mutants contain bright Dad4-YFP foci ( red arrows ) . In all cases , there are no effects upon Mtw1-YFP foci ( right panels ) . Scale bars in all images are 5 µm . ( C ) Quantitation of the Dad4-YFP kinetochore foci fluorescence levels from these cells indicates that the levels of Dad4-YFP at kinetochores are affected by deletion of either HMO1 or SGF29 . The left notched box and whiskers plot indicates the median ( background subtracted ) fluorescence values of kinetochore foci in relative units . The plot shows the median value ( bar ) and quartiles ( box ) , the whiskers show the minimum of the range or 1 . 5 interquartile ranges , outlying data points are indicated as circles ( note that several outlying data points are not shown as they are beyond the scale of the plot ) . The notches indicate the 95% confidence intervals of the medians ( ***indicates p-values <10–10 from a Wilcoxen’s rank-sum test ) . It should be noted that the distribution of kinetochore intensities do not conform to a normal distribution , particularly for the hmo1∆ mutant . The right panels show the distribution of fluorescent intensities of kinetochore foci of the same data plotted to the left ( note that several outlying data points are beyond the scale of the plot ) . These data indicate the abundance of both the low and high intensity Dad4 foci of the hmo1∆ mutant ( green and red arrowheads in Figure 5A , respectively ) ( D ) Mtw1 kinetochore foci fluorescence levels are plotted as in panel C , we could not detect a difference from wild type cells in all three mutants . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 02110 . 7554/eLife . 13053 . 022Figure 5—figure supplement 1 . Cellular levels of Dad4-YFP in wild-type and hmo1∆ cells . The histogram indicates the levels of Dad4-YFP fluorescence ( relative units , r . u . ) in both wild-type ( blue ) and hmo1∆ ( red ) cells . The plots below show the mean cellular fluorescence ( wild type=972 r . u . and hmo1∆=1591 r . u . and the error bars indicate the standard deviation of the mean ) . The dashed lines indicate + and – one standard deviation of the mean of wild-type cells . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 022 For each cellular compartment there are relatively few GFP proteins that produce SPIs with just one target protein . The query GFP proteins that produce SPIs have on average 3 . 8 SPIs with the 23 target proteins . However , those GFP proteins with just one SPI may be informative . For example , the histone subunits Hta2 , Htb1 , Htb2 , and Hhf2 as well as the chromosomal proteins Bub1 and Mft1 have unique SPIs with the eisosome component Pil1 . These interactions may indicate a nuclear role for Pil1 , which relocalizes from the plasma membrane in response to DNA damage ( Tkach et al . , 2012 ) and associates with histones and chromosomal proteins ( Lambert et al . , 2009; Akiyoshi et al . , 2010 ) . Indeed , the Pil1-histone SPIs result from Pil1 recruitment into the nucleus ( Figure 2—figure supplement 2 ) . Since selected SPI query proteins are enriched for physical and genetic interactions and contain proteins involved in regulating the biology of their target , we next performed hierarchical clustering analysis in order to test whether SPI data can be used to assess functional associations between proteins ( Figure 4A ) . We find that query proteins from specific large functional complexes cluster together , for example , the mediator complex , which is involved in activating transcription , clusters together as do members of the COP1 coatomer , the outer ring of the nuclear pore , the signal recognition particle and TRAMP complex ( Figure 4—figure supplement 6 ) . It is important to note that SPIs are not a substitute for physical interaction data , but rather represent a common phenotype in response to forced association . Collectively , the clustering of protein complexes , gene ontology enrichment and physical and genetic enrichment indicate that specific target proteins show SPIs with sets of query proteins that share a common location , potentially common components of larger protein complexes . Thus , although the proteome-wide SPI data themselves do not directly give structural information , the SPI data groups query proteins within these known protein complexes . We next asked whether the SPI data would correlate with the quaternary structure of multi-protein complexes , since protein associations with one part of a complex may give a similar growth phenotype that contrasts with a different part of that same complex . We chose the kinetochore as an example , since this is a large array of between 60 and 100 proteins that are arranged into defined sub-complexes ( Biggins , 2013 ) . We selected these proteins ( and some kinetochore-associated proteins ) and clustered them based upon their SPI scores from the 23 screens . We find that key sub-complexes within the kinetochore are clustered together purely based upon their 23 SPI scores ( Figure 6 ) . For example , three of the four members of the COMA complex cluster together ( Ctf19 , Okp1 , and Mcm21 ) with two members of the Ctf3 complex ( Mcm22 and Nkp2 ) , and Cse4 and Chl4 , which are all part of the constitutive centromere associated network ( CCAN ) of inner kinetochore proteins that bind to centromeric DNA . Three of the four MIND complex members ( Dsn1 , Nnf1 , and Nsl1 ) also cluster with Spc24 , Kre28 and Nuf2 , which are all part of the KMN network of outer kinetochore proteins . In contrast , the DAM/DASH complex , which is composed of 10 different proteins , segregates into distinct clusters ( with Dad2 , 3 , and 4 distinct from Dam1 , Ask1 , Dad1 , Spc34 , and Duo1 ) . Dad2 , 3 and 4 are small central domain subunits of the DAM/DASH complex that are important for structural integrity of the complex and therefore potentially sensitive to association with other proteins ( consequently they have many SPIs ) . In contrast Dam1 , Duo1 , and Spc34 are key interaction hubs for the decameric complex ( Shang et al . , 2003 ) and Ask1’s C-terminus plays an important role in intercomplex interactions ( Ramey et al . , 2011 ) . Thus these proteins form external surfaces on the complex , which may be more tolerant of protein association . A similar correlation with the quaternary structure can be made for another large protein assembly , the nuclear pore complex ( Figure 6—figure supplement 1 ) . Hence , although SPIs do not substitute for physical interaction data they indicate a common phenotype produced by specific protein-protein associations . 10 . 7554/eLife . 13053 . 023Figure 6 . Cluster analysis of kinetochore and associated proteins using the SPI data are plotted as a heat-map . High z-scores ( positive; >2 ) are shown in yellow and low ( negative; < -1 ) scores in blue ( as in Figure 4A ) . The different protein complexes within the kinetochore are color-coded as indicated in the legend . Based on the SPI data alone , key complexes within the kinetochore cluster together as indicated by the colored boxed regions of the plot . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 02310 . 7554/eLife . 13053 . 024Figure 6—figure supplement 1 . Clustering analysis of nuclear pore complex ( NPC ) . The SPI data are sufficient to cluster NPC subunits and karyopherins into some of the key functional complexes , such as the Nup84 complex of the outer ring of the nuclear pore . The data also cluster Gle1 with Dbp5 . Gle1 normally regulates the activity of Dbp5 . z-scores of NPC and NPC-associated GFP strains in all of the SPI screens are plotted as a heat-map . High z-scores ( positive; >2 ) in yellow and low ( negative; < -1 ) z-scores in blue . Both the screens and the GFP-tagged genes are clustered as in Figure 4A . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 024
The SPI technology has allowed us to create binary protein associations throughout the cell and in many cases these interactions result in protein relocalization . However , only a small fraction of these interactions lead to a measurable growth phenotype , suggesting that cells are highly tolerant of both protein mislocalization and protein-protein associations . There are exceptions , proteins that do affect growth in almost any location . For example the ubiquitin hydrolase , Doa4 and numerous proteins involved in transport ( Figure 4—source data 1 ) . Furthermore , there are proteins whose association with specific proteins causes a growth defect . We find that these SPIs are enriched for proteins that physically interact ( Figure 4 ) . Collectively the SPI data allow us to both identify regulatory proteins ( Olafsson and Thorpe , 2015 and Figure 5 ) and provide information on quaternary structure of specific large complexes within the cell ( Figure 6 ) . These data illustrate that SPIs can be used , like physical interactions , to reveal the functional organization of the cell . However , since the readout of SPIs is phenotypic , in this case cell growth , the SPIs indicate functional interactions rather than physical interactions per se . Thus , the SPI methodology provides a powerful in vivo proteomics tool to map the mechanisms underlying spatial regulation within cells . The SPI technology may be particularly informative to define interactions that are detrimental under conditions of stress , drug treatment or other specific cellular perturbations . Many disease pathologies result , at least in part , from the mislocalization of proteins in cells ( Hung and Link , 2011 ) . Recent studies are discovering the extent to which specific drugs induce global changes in protein location ( Tkach et al . , 2012; Breker et al . , 2013; Chong et al . , 2015 ) . Combining this cellular pharmacodynamics knowledge with SPI data opens the possibility of using drugs to induce therapeutic changes in protein localization; of the 727 SPI query proteins identified here , ~76% ( 549 ) have human homologs compared to 56% ( 3766 ) of the whole yeast genome ( 6604 ORFs ) ( YeastMine , Balakrishnan et al . , 2012 ) . This study provides the first comprehensive map of the effects of forced protein associations within cells .
All yeast strains used in this study are listed in Table 1 . W303 strains are ADE2+RAD5+ derivatives of W303 ( can1-100 his3-11 , 15 leu2-3 , 112 ura3-1 unless otherwise indicated Thomas and Rothstein , 1989; Zou and Rothstein , 1997 ) . GFP strains are all based upon BY4741 ( his3∆1 leu2∆0 met15∆0 ura3∆0 Brachmann et al . , 1998; Huh et al . , 2003 ) . Yeast were grown in standard growth medium including 2% ( weight/volume ) of the indicated carbon source ( Sherman , 2002 ) . Yeast plasmids were created using the gap-repair cloning technique , which combines a linearized plasmid with PCR products using in vivo recombination . All PCR products were generated using primers from Sigma Life Science and PfuII Ultra proof reading polymerase ( Agilent Technologies , UK ) or Q5 polymerase ( New England Biolabs , USA ) . All plasmid constructs ( listed in Figure 1—source data 2 ) were validated using Sanger sequencing ( Beckman Coulter Genomics , UK ) . 10 . 7554/eLife . 13053 . 025Table 1 . Yeast strains used in this study . DOI: http://dx . doi . org/10 . 7554/eLife . 13053 . 025Strain nameGenetic backgroundRelevant genotypeReferenceW8164-2BW303MATα CEN1-16::Gal-Kl-URA3 ( Zou and Rothstein , 1997 ) GFP strainsBY4741MATa his3∆1 leu2∆0 met15∆0 ura3∆0 XXX-GFP::HIS3 ( Huh et al . , 2003 ) PT147-7CW303MATa TRP1 lys2∆ DAD4-YFP::NAT SPC42-RFP::This studyPT12-13DW303MATa TRP1 MTW1-YFP hmo1∆::KANThis studyT403W303MATa TRP1 lys2∆ DAD4-YFP::NAT SPC42-RFP::HYG hmo1∆::KANThis studyT404W303MATa TRP1 lys2∆ DAD4-YFP::NAT SPC42-RFP::HYG sgf29∆::KANThis studyT402W303MATa TRP1 lys2∆ DAD4-YFP::NAT SPC42-RFP::HYG sst2∆::KANThis studyT406W303MATa TRP1 MTW1-YFP hmo1∆::KANThis studyT407W303MATa TRP1 MTW1-YFP sgf29∆::KANThis studyT405W303MATa TRP1 MTW1-YFP sst2∆::KANThis study The SPA screening method is a mating-based approach for yeast transformation , and we followed the established protocol ( Reid et al . , 2011 ) . The SPA method relies upon a universal donor strain ( UDS , W8164-2B ) that includes conditional centromeres on each and every chromosome . This strain is transformed with a plasmid encoding the GBP-tagged target protein ( or controls ) and then mated en masse with the collection of GFP strains . The resulting diploids are converted back to haploids by first destabilizing and then counter-selecting against all of the chromosomes from the UDS . The resulting colonies are then assessed for growth by measuring colony size as described below . In the first step , plasmid constructs ( encoding GBP alone , target protein alone or target-GBP ) were transferred into the UDS by transformation . The three resulting strains were separately mated with arrays of MATa GFP strains ( Huh et al . , 2003 ) on YPD agar plates for 24 hr . The resulting colonies were then copied to synthetic galactose medium lacking leucine to destabilize the donor chromosomes for 24 hr . Finally , colonies were copied onto galactose medium lacking leucine , including the drug 5-Fluoroorotic acid ( 5-FOA ) to counter-select against the UDS chromosomes . Plates were then grown at 30˚C for 48–72 hr prior to imaging . All mating and copying of yeast colonies utilized a RoToR pinning robot ( Singer Instruments , UK ) with a minimum of four replicates per strain . After SPA screening , the resulting agar plates were scanned using a desktop flatbed scanner ( Epson V750 Pro , Seiko Epson Corporation , Japan ) at 300 dpi resolution in transmission mode . These images were processed and analyzed using the ScreenMill suite of software ( Dittmar et al . , 2010 ) , which assesses growth based upon the two-dimensional size of the colonies . The software was run in default mode , both for the kinetochore-specific screen and for the proteome-wide screen . For retesting strains for growth defects , plate images were normalized using specific controls on the plate as a reference , rather than the default plate median . This is necessary when the majority of the strains on a plate are affected since this will influence the plate median . Colonies arrayed on agar plates often grow faster on one side of the plate than the other . This growth effect can be caused by temperature or humidity gradients within incubators , variable thickness of agar ( and hence concentration of nutrients ) , or uneven pinning pressure during plate copying . These anomalies can result in one side of the plate producing an overall higher z-score than the other . To correct for these type of biases , algorithms adjust colony size data to reflect overall even growth across a plate ( Collins et al . , 2006; Baryshnikova et al . , 2010 ) . The ScreenMill suite of software used for our analysis does not contain such corrections and so we employed a simple algorithm to correct z-scores for spatial anomalies ( Olafsson and Thorpe , 2015 ) . To examine the levels and location of tagged proteins within the cells , we used epifluorescence microscopy . Log phase cells were embedded in 0 . 7% low melting point agarose dissolved in the appropriate growth medium . The depth of agarose between the slide and coverslip is fixed at 6–8µm , slightly larger than the diameter of the average yeast cell , which maintains a consistent distance from the coverslip to the cell nucleus . Cells were imaged with a Zeiss Axioimager Z2 microscope ( Carl Zeiss AG , Germany ) , using a 63x 1 . 4NA oil immersion lens , illuminated using a Zeiss Colibri LED illumination system ( GFP=470 nm , YFP=505 nm , and RFP=590 nm ) . Bright field contrast was enhanced with differential interference contrast ( DIC ) prisms . The resulting light was captured using either a Hamamatsu ORCA ERII CCD camera containing an ER-150 interline CCD sensor with 6 . 45 µm pixels , binned 2x2 ( Hamamatsu Photonics , Japan ) or a Hamamatsu Flash 4 Lte . CMOS camera containing a FL-400 sensor with 6 . 5 µm pixels , binned 2x2 . The exposure times were set to ensure that pixels were not saturated and were identical between control and experimental images . All images were acquired using either Axiovision or Zen software from Zeiss . Images shown in the figures were prepared using Volocity imaging software ( Perkin Elmer Inc . , USA ) and control and experimental images have identical linear contrast adjustments unless otherwise stated . To quantify the relative amount of RFP in cells containing GBP-RFP tags we used custom scripts for the Volocity image analysis software ( Perkin Elmer Inc . USA ) . Briefly , red fluorescence regions were identified within the three-dimensional images based upon an intensity threshold . These regions were then dilated by a fixed amount ( ~600 nm ) in each direction to ensure that we assay all of the red fluorescence signal . The regions were further dilated ( 2 . 4 µm ) to create an outer background region , which was subtracted from each fluorescence measurement ( the script is available online https://sourceforge . net/projects/berry-et-al/files/RFP_quantitation . assf/download ) . To quantify the relative levels of Dad4-YFP and Mtw1-YFP kinetochore proteins within kinetochore foci , we employed a custom ImageJ script ( Ledesma-Fernández and Thorpe , 2015 ) . To quantify the total cellular levels of Dad4-YFP we measured the YFP fluorescence signal from maximum projection images ( from a stack of vertically separated z planes ) for each cell and subtracted a mean background signal specific to each image ( this script is available at https://sourceforge . net/projects/berry-et-al/files/general_cell_quan . ijm/download ) . Michael Eisen’s cluster program ( version 3 . 0 ) was used to cluster the SPI data ( Eisen et al . , 1998 ) . We used hierarchical centroid linkage clustering of both the GBP screens and the GFP-tagged genes . For the quaternary structure examples ( Figure 6 , Figure 4—figure supplement 6 and Figure 6—figure supplement 1 ) only a selected subset of the GFP strains were used for the cluster analysis . Cluster diagrams were visualized using Java Treeview ( Saldanha , 2004 ) . Gene ontology enrichment analysis was performed using the GOrilla algorithm ( cbl-gorilla . cs . technicon . ac . il [Eden et al . , 2009] ) . | Our actions often depend on who we interact with: parents , teachers , friends , colleagues . So it is for proteins in the cell: their function depends on which other proteins they work with . If a protein interacts with new partners or ends up in a new neighborhood of the cell , it can perform an entirely unexpected role , rewiring how that cell works . There are millions of possible protein-protein interactions , but it is not known how cells behave if their proteins are forced into new associations . For example , how many of these associations affect how well the cell can grow ? Using budding yeast , Berry et al . were able to associate every protein in the cell with proteins from each of the major areas of the cell such as the nucleus , cell membrane or mitochondria . These new associations and relocations were then examined to see how many of them caused problems , slowing the cell’s growth or killing it . Unexpectedly , most forced associations had no detectable effect , indicating that the cell is remarkably tolerant of new protein-protein interactions . This contradicts a common idea that proteins are very fussy about their partner proteins , and will not work properly if they are forced into new interactions . The associations that do cause a growth defect are often between proteins that normally work together , indicating that their association is normally carefully controlled during the normal growth of cells . In some cases these forced associations identified previously unknown regulators of cell behavior . Proteins that interact with the wrong partners or are in the wrong place within cells cause a number of diseases . Future forced association experiments will allow us to examine such interactions and possibly search for drugs that will correct the problem . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"tools",
"and",
"resources"
] | 2016 | Synthetic protein interactions reveal a functional map of the cell |
Mobile integrons are widespread genetic platforms that allow bacteria to modulate the expression of antibiotic resistance cassettes by shuffling their position from a common promoter . Antibiotic stress induces the expression of an integrase that excises and integrates cassettes , and this unique recombination and expression system is thought to allow bacteria to ‘evolve on demand’ in response to antibiotic pressure . To test this hypothesis , we inserted a custom three-cassette integron into Pseudomonas aeruginosa and used experimental evolution to measure the impact of integrase activity on adaptation to gentamicin . Crucially , integrase activity accelerated evolution by increasing the expression of a gentamicin resistance cassette through duplications and by eliminating redundant cassettes . Importantly , we found no evidence of deleterious off-target effects of integrase activity . In summary , integrons accelerate resistance evolution by rapidly generating combinatorial variation in cassette composition while maintaining genomic integrity .
Given the mounting threat posed by antibiotic resistance , we need a better understanding of the mechanisms used by bacteria to evolve resistance to antibiotics . Mobile integrons ( MIs ) are widespread elements providing a platform for the acquisition , shuffling , and expression of gene cassettes , many of which are antibiotic resistance genes ( Recchia and Hall , 1995; Escudero et al . , 2015 ) . These elements are typically associated with transposons and conjugative plasmids and have played an important role in the evolution of resistance in pathogenic bacteria ( Partridge et al . , 2018 ) . Five classes of MIs , based on the sequence of their integrase , have arisen independently through association with diverse mobile elements , but the class 1 integron is , by far , the most prevalent and clinically relevant . The first multidrug resistance plasmids that were isolated in the 1950s carried class 1 MIs ( Liebert et al . , 1999; Mitsuhashi et al . , 1961; Rownd et al . , 1966; Stokes and Hall , 1989 ) , and recent surveys have shown that class 1 integrons are found in a substantial fraction of isolates of Escherichia coli ( Halaji et al . , 2020; Rao et al . , 2006; Yu et al . , 2003 ) , Klebsiella pneumoniae ( Firoozeh et al . , 2019; Li et al . , 2013 ) , Pseudomonas aeruginosa ( Oliver et al . , 2015; Ruiz-Martínez et al . , 2011 ) , and Acinetobacter baumanii ( Chen et al . , 2015; Turton et al . , 2005 ) . Mobile integrons consist of an integrase encoding gene named intI followed by a recombination site , attI ( Hall et al . , 1991; Partridge et al . , 2000 ) and a variable array of mobile gene cassettes ( typically 2–5 in mobile integrons ) ending each in a characteristic hairpin recombination site called the attC site ( Hall et al . , 1991 ) . Integron cassettes usually lack a promoter , and their expression is driven by the Pc promoter located upstream of the array ( Collis and Hall , 1995 ) , such that expression levels are highest for cassettes closest to the promoter ( Collis and Hall , 1995 ) . The SOS response ( named after the Morse code . . . --- . . . ) induces the expression of integrases ( Guerin et al . , 2009 ) , which then allows for the efficient integration and excision of cassettes in the array through attC × attI and attC × attC reactions , respectively ( Collis and Hall , 1992 ) . A peculiarity of this system is that integron recombination is semi-conservative , as only the bottom strand of the cassette is excised from the array following a recombination pathway that includes a replication step ( Loot et al . , 2012 ) . The implication of this is that cassette excision and re-integration can lead to two different results: either a ‘cut and paste’ outcome , resulting in the movement of a cassette within an array , or a ‘copy and paste’ outcome , leading to the insertion of a duplicate copy of a cassette in the conserved array ( Escudero et al . , 2015 ) . An overview of the mechanisms of integron activity is presented in Figure 1a . Due to the stress-inducible regulation of integrase activity , integrons have been proposed to accelerate bacterial evolution by providing ‘adaptation on demand’ ( Escudero et al . , 2015 ) . According to this hypothesis , integrase-mediated cassette re-shuffling in stressful environments allows bacteria to optimize cassette expression and maximize fitness: useful cassettes can be brought forward to ensure maximal expression , while unnecessary cassettes can be kept at the end of the array as a low cost memory of once-adaptive functions , ready to be brought forward when needed ( Escudero et al . , 2015 ) . Stress-inducible regulation also helps to minimize the costs associated with integrase expression ( Lacotte et al . , 2017; Starikova et al . , 2012 ) , which are thought to come from increased genomic instability created by off-target integrase activity ( Harms et al . , 2013 ) . Although antibiotics have diverse modes of action , many of the most common classes of antibiotic cause DNA damage that induces the SOS response ( Kohanski et al . , 2010 ) . This link between antibiotic exposure and integrase activity suggests that cassette re-shuffling may allow pathogenic bacteria to rapidly adapt to novel antibiotic challenges . While the molecular mechanisms of integron shuffling are known in detail , our ability to understand the evolutionary benefits provided by this fascinating genetic platform is limited by our understanding of the population biology of integron-mediated antibiotic resistance . For example , constitutive over-expression of the integrase enzyme has been shown to accelerate the evolution of chloramphenicol resistance through the loss of cassettes between Pc and the resistance cassette as well as the formation of co-integrates between integron copies ( Barraud and Ploy , 2015 ) . However , to the best of our knowledge , the benefits of cassette shuffling under the integrase natural promoter and its associated LexA regulation have never been investigated . This is an important limitation , as parameters such as the re-insertion rate of excised cassettes and fitness costs of integrase expression are predicted to have a strong impact on the evolutionary benefits of the integrase ( Engelstädter et al . , 2016 ) . Moreover , integron cassette shuffling has rarely been studied in the large , natural plasmids where class 1 integrons often occur . Here we directly test the ‘adaptation on demand’ hypothesis using experimental evolution in populations of P . aeruginosa carrying a variant of the broad host range plasmid R388 . We replaced the naturally occurring class 1 mobile integron in R388 with a customized integron containing three antibiotic resistance cassettes: dfrA5 ( a trimethoprim resistant dihydrofolate reductase [Sundström et al . , 1988] ) , blaVEB-1 ( an extended-spectrum-β-lactamase [Poirel et al . , 1999] ) , and aadB ( an aminoglycoside-2′-adenylyltransferase [Cameron et al . , 1986] ) . To directly investigate the role of integrase activity in this system , we also generated a truncated integrase mutant that allows normal levels of cassette expression , but not recombination . Using this system , we found that integrase activity leads to rapid and extensive cassette re-shuffling in response to strong selection for increased gentamicin resistance . Specifically , integrase activity caused the insertion of duplicate copies of aadB cassettes in the first position of the integron , followed by the loss of redundant cassettes . Crucially , this accelerated the ability of populations to adapt to antibiotic stress , providing good support for the ‘adaptation on demand’ hypothesis . Finally , we show that rapid duplications can also occur with blaVIM-1 cassettes , which confer resistance to ‘last line of defense’ carbapenem antibiotics , in a recently isolated clinical plasmid under meropenem selection .
We replaced the naturally occurring class 1 integron of plasmid R388 with all six possible configurations of a class 1 integron containing three antibiotic resistance cassettes , including dfrA5 , aadB , and blaVEB-1 and transformed our integron variants into P . aeruginosa PA01 ( Figure 1b ) . Integrons have played an important role in the evolution of antibiotic resistance in the opportunistic pathogen P . aeruginosa and are highly prevalent in P . aeruginosa high-risk clones ( Oliver et al . , 2015 ) . As expected , resistance levels to gentamicin declined as the aadB cassette moved further away from the integrase ( Figure 1c ) . Interestingly , the relationship between aadB position and resistance was not linear: we observed a 6- to 24-fold difference in minimum inhibitory concentration ( MIC ) between arrays containing aadB in first and second positions , but a less than twofold difference between arrays with aadB in second and third places . In order to investigate the mechanisms behind this trend , we measured the aadB cassette transcription levels of the different arrays . Instead of a sharp drop , we observed a steady decrease depending on the cassette position ( Figure 1—figure supplement 1a ) . Previous work has shown that two short open-reading frames contained within the attI site can substantially enhance the translation of a cassette lacking a Shine–Dalgarno ( SD ) sequence when the cassette is located in first position ( Hanau-Berçot et al . , 2002; Papagiannitsis et al . , 2017 ) , showing that cassette position can also modulate translation levels ( Hanau-Berçot et al . , 2002; Jacquier et al . , 2009 ) . Interestingly , the aadB cassette contains a reduced SD box ( Figure 1—figure supplement 1b ) , suggesting that the steep gradient in gentamicin resistance between first and second positions was mostly due to decreased translation . Given the strong effect of aadB cassette position on gentamicin resistance , we decided to use this combination of cassette and antibiotic to experimentally test the hypothesis that integrase activity accelerates resistance evolution . To properly measure the effect of integrase activity on evolvability , we constructed a ΔintI1 mutant of the A3 array lacking 818 bp of intI1 ( total length is 1014 bp ) but conserving the Pc and Pint promoters . We challenged independent populations of WTA3 and ΔintI1A3 with increasing doses of gentamicin using an ‘evolutionary ramp’ design ( Gifford et al . , 2018; San Millan et al . , 2016 ) . Importantly , we did not detect any difference in initial gentamicin resistance ( MICexp = 24 mg/L ) between strains with array A3 or the ΔintI1A3 mutant in the conditions of the evolution experiment ( see Materials and Methods ) . We passaged 65 populations of each strain , starting at 1/8 MIC ( i . e . 3 mg/L ) and doubling the concentration of gentamicin each day until reaching 1024 times ( 24 . 5 g/L ) the initial MIC ( Figure 2a ) . As controls , 15 populations of each strain were passaged without antibiotic ( no selection for gentamicin resistance ) , while 15 populations were passaged at a constant dose of 1/8 MIC ( 3 mg/L ) to generate weak selection for gentamicin resistance and plasmid maintenance . The rapid increase in antibiotic concentration in the ‘evolutionary ramp’ treatment ensures that populations must either evolve increased resistance or face extinction ( once concentrations exceed the MIC of the parental strains ) . Given this , measuring the rate at which populations go extinct provides a way to measure the evolvability of a strain . Crucially , populations of WTA3 populations with a functional integrase had a higher survival rate than those of the ΔintI1A3 mutant , showing that the integrase can increase evolvability for antibiotic resistance ( Figure 2b; log-rank test: Chisq = 17 . 7 , df = 1 , p=3e-05 ) . We did not detect any extinctions in either of the controls , showing that the higher extinction rate of ΔintI1A3 populations was driven by exposure to high doses of gentamicin . We confirmed the evolution of high level of gentamicin resistance by measuring the MICs of a subset of populations from the final time point ( Figure 2c ) . We observed a similar level of resistance between ΔintI1A3 ( mean MIC = 24 , 200 mg/L; s . d . = 1864; n = 5 ) and WTA3 ( mean MIC = 27 , 800 mg/L; s . d . = 6031; n = 15 ) populations ( t = 2 . 044 , p=0 . 056 ) , showing that integrase activity increased the likelihood of resistance evolution , but did not impact the final resistance levels of surviving populations . To understand the mechanisms by which integrase activity accelerates evolution , we sequenced DNA extracted from randomly chosen populations at a mid-point of the experiment ( ×4 MIC; n = 24 WTA3 and 22 ΔintI1A3 ) and all populations that survived until the end of the experiment ( ×1024 MIC; n = 21 WTA3 and 6 ΔintI1A3 populations ) . We found evidence for widespread cassette re-arrangement in WTA3 populations and identified four novel integron structures that were formed by insertion of the aadB cassette and/or deletions of blaVEB-1-dfrA5 ( Figure 3a , b and Supplementary files 2 and 3 ) . The junction sites for cassette insertions and deletions were consistent with integrase activity: recombination happened at the 5′-GTT-3′ triplet of the attI1 , attCaadB , and attCdfrA5 sites . We did not find any evidence for cassette re-arrangements in ΔintI1A3 populations or in control WTA3 populations that we selected at a low dose of gentamicin ( ×1/8 MIC ) , while the entire R388 plasmid was lost in all ΔintI1A3 and WTA3 populations passaged without antibiotic . Cassette re-arrangements were found in most populations at the ×4 MIC time point , and approximately 90% of populations ( 19/21 ) contained cassette re-arrangements by the end of the experiment , highlighting the importance of integrase activity in resistance evolution . Integron structural polymorphisms were found in 50% of populations ( 12/24 ) at ×4 MIC , but this within-population diversity was transient and almost all populations contained a single dominant integron structure by the final time point . The most common novel integron structure contained a ‘copy and paste’ insertion of aadB in first position via attI × attCaadB recombination ( ie aadB-blaVEB-1-dfrA5aadB ) . We measured the impact of this novel array on gentamicin resistance level by transferring the evolved plasmids into the ancestral chromosomal background ( Figure 3e ) . This novel integron is associated with a large increase ( 64-fold ) in gentamicin resistance due to the dominant effect of first position on aadB , with similar levels to our constructed arrays with aadB in first position ( Figure 3e ) . Interestingly , we did not identify any aadBblaVEB-1dfrA5 arrays , which would be the result of an aadB excision followed by reintegration of aadB in first position within the same array , highlighting the prevalence of ‘copy and paste’ cassette insertions . Degenerate integrons that lack the blaVEB-1 and dfrA5 cassettes ( i . e . either aadB or aadBaadB arrays ) were also present at relatively high frequency at both the ×4 and ×1024 MIC time points . Interestingly , in mixed arrays populations , these reduced arrays were always observed in conjunction with the aadB-blaVEB-1dfrA5aadB array and never with the ancestral array . This repeated association provides good evidence that degenerate arrays evolved via aadB insertion in first position , to form the common aadB-blaVEB-1dfrA5aadB array , followed by the en bloc deletion of the other cassettes ( blaVEB-1dfrA5aadB or blaVEB-1dfrA5 ) to form aadB and aadBaadB arrays . Recombination happened at the 5′-GTT-3′ triplet of the R box of attCaadB and attCdfrA5 sites , suggesting that these deletions were driven by integrase activity , although we cannot rule out the possibility that the blaVEB-1dfrA5aadB cassette deletion was driven by homologous recombination between aadB cassettes . The relative prevalence of these two degenerate arrays did not change between the ×4 and ×1024 MIC time points ( four aadB against three aadBaadB arrays at ×4 MIC and three aadB against three aadBaadB arrays at ×1024 MIC ) , which suggests that the second aadB cassette in the aadBaadB array is redundant . In line with this argument , we found a marginal difference ( ×2 ) in MIC between evolved plasmids carrying aadB ( mean = 512 , s . d . = 0 ) and aadBaadB ( mean = 939 , s . d . = 121 ) . Finally , we found arrays containing a duplicate copy of aadB at the end of the array ( blaVEB-1dfrA5aadBaadB ) , which are likely to have been formed by the insertion of an aadB cassette in the middle or at the end of the array through the less frequent attC × attC integration ( intermolecular ) reaction . These arrays were only found at the ×4 MIC time point , strongly suggesting that they conferred a small increase in gentamicin resistance that was ultimately an evolutionary dead end under strong selection for elevated resistance . In addition to changes in integron structure , we found widespread integron evolution by mutations in both the WTA3 and ΔintI1A3 populations . Mutations in blaVEB-1 were found in more than 80% of WTA3 and ΔintI1A3 populations from the ×4 MIC time point , and in almost all populations where the blaVEB-1 cassette was maintained at the ×1024 MIC time point , including 5/6 ΔintI1A3 and 16/16 WTA3 populations . All mutations in blaVEB-1 were amino acid substitutions ( n = 9 ) or indels ( n = 16 ) and the 23 amino acid signal peptide was a hotspot for mutations ( 12 of 25 blaVEB-1 mutations ) , suggesting strong selection to eliminate the secretion of this redundant resistance protein ( Supplementary files 2 and 3 ) . Furthermore , similar blaVEB-1 mutations were also found in the ⅛ MIC controls , demonstrating that these mutations were beneficial under low doses of gentamicin , as we would expect if this cassette imposed an important fitness cost . It is unclear if this cost of blaVEB-1 was driven by the presence of gentamicin ( i . e . collateral sensitivity ) because the entire R388 plasmid was lost in every control population that was passaged in antibiotic-free medium . Parallel evolution also occurred close to the putative translation initiation site of the aadB cassette . These mutations were very rare at the ×4 MIC time points ( 2/46 populations ) , but were present at a high frequency in ΔintI1A3 populations from the final time point ( 4/6 populations ) and are linked with high level of gentamicin resistance ( Figure 3e ) . We speculate that these mutations were favored in ΔintI1A3 populations as they increased the translation rate of the weakly expressed third position aadB cassette and offer an alternative mechanism to increase aadB expression in the absence of re-arrangements . Similarly , we identified one 41 bp deletion within the dfrA5 attC site of a WTA3 population , which may increase translational coupling with the previous dfrA5 cassette ( as in Jacquier et al . , 2009 ) or lead to the creation of a fused protein with part of the previous cassette . Finally , we observed similar extended deletions in one WTA3 population and in the 1/8 MIC WTA3 pooled control . These deletions occur between attCaadB and different positions within the plasmid trwF gene , effectively deleting most of the genes involved in mating pore formation ( Figure 3—figure supplement 1 ) , with the sequence of the junction sites pointing toward potential off-target activity of the integrase ( abundance of 5′-GNT-3′ secondary sites ) ( Figure 3—figure supplement 3 ) . The integron integrase is known to have off-target effects , suggesting that integrase activity may also have an important effect on chromosomal evolution by reducing genomic stability through recombination between chromosomal pseudo-attC or pseudo-attI sites , leading to an increase in deletions and re-arrangements ( Harms et al . , 2013 ) . Chromosomal evolution occurred more rapidly in the ΔintI1A3 populations than in the WTA3 populations , as demonstrated by the high cumulative frequency of mutations in ΔintI1A3 populations ( mean = 1 . 01; s . e . = 0 . 17 , see Figure 4—figure supplement 2 ) at ×4 MIC compared to WTA3 populations ( mean = 0 . 48; s . e . = 0 . 11; t = −2 . 69 , df = 35 . 83 , p=0 . 01 , Welch two-sample t-test ) . However , accelerated chromosomal evolution in the absence of integrase activity was short-lived , and cumulative frequency of mutations in the WTA3 and ΔintI1A3 populations was almost identical at the end of the experiment ( mean WTA3 = 2 . 45 SNPs; mean ΔintI1A3 = 2 . 65 SNPs , see Figure 4—figure supplement 1 ) . Crucially , we found only one case of chromosomal recombination , with a 3 . 5 kb deletion between the two highly homologous ccoN1 and ccoN2 cytochrome C subunits in one WTA3 population ( Supplementary file 2 ) , showing that off-target effects of the integrase were undetectable in our experiment , in spite of our extensive genomic sequencing . In total , we identified 41 different SNPs and 58 short indels in eight intergenic regions and 41 genes , with a similar spectrum of mutations in the ramping WTA3 and ΔintI1A3 populations ( Figure 4; Supplementary files 2 and 3 ) . Several lines of evidence indicate that the overwhelming majority of mutations were beneficial mutations that reached high frequency as a result of selection . First , many of the mutated genes have known roles in antibiotic resistance; for example , 11/41 mutated genes have also been identified in an aminoglycoside resistance screen in P . aeruginosa ( Schurek et al . , 2008 ) . Second , parallel evolution was very common . Repeated evolution occurred in 11 of 42 ( 26% ) genes and 3 of 8 ( 38% ) intergenic regions with 68% of mutations occurring in these genes . Only 1 of 41 mutations in coding regions was synonymous , providing evidence that the rapid evolution of proteins was driven by positive selection , and not simply by an elevated mutation rate . Finally , we found almost no overlap between the genes that were mutated in the ramping populations and the controls , implying that the evolutionary response of the ramping populations was dominated by selection for high levels of gentamicin resistance . For both control regimens , the chromosomal genes with the most mutations were cdrA ( PA4625 ) , involved in biofilm formation ( Reichhardt et al . , 2018 ) , and PA1874 , a hypothetical protein . Similarly to the ramping populations , no wide-scale chromosomal re-arrangements were observed , and no major differences could be found between the distribution of mutations between WTA3 and ΔintI1A3 populations in both control regimens at x1024 MIC ( Figure 4—figure supplement 1 ) . In the ramping populations , the initial stages of adaptation to gentamicin were driven by mutations in a very diverse set of genes , with a strong bias toward genes that are involved in translation , such as rluD and PA0668 . 4 , which encodes for the 23S ribosomal RNA ( Figure 4 ) . Interestingly , we observed divergent mutational trajectories of evolution in the WTA3 and ΔintI1A3 backgrounds: the number of genes that were mutated in both backgrounds was small ( n = 7 ) relative to the total number of mutated genes in either background ( n = 26 ) and the correlation in mutation frequencies across backgrounds was weak ( rho = 0 . 029; p=0 . 89 Spearman test ) . Continued selection for elevated gentamicin resistance resulted in two changes in chromosomal evolution ( Figure 4 ) . First , chromosomal evolution became dominated by mutations in a few key target genes , implying that many of the trajectories of chromosomal evolution followed during early adaptation ultimately led to evolutionary dead ends . For example , the correlation in mutation frequencies between early and late time points was very weak , in both WTA3 ( rho = −0 . 10 , p=0 . 58 ) and ΔintI1A3 ( rho = 0 . 20 , p=0 . 307 ) . In particular , we found evidence of extensive parallel evolution in mexZ , amgS , and rluD in both the WTA3 and ΔintI1A3 populations . At a functional level , the mutations found at ×1024 MIC are predominantly involved in antibiotic efflux , as opposed to translation . mexZ is a transcription factor that represses the expression of the mexXY multidrug efflux pump operon . Mutations inactivating mexZ cause a 2- to 16-fold increase in aminoglycoside resistance and have been widely identified in aminoglycoside-resistant P . aeruginosa isolates found in cystic fibrosis patients ( Vogne et al . , 2004 ) . AmgS is part of an envelope stress-responsive two-component system AmgRS ( Lau et al . , 2013 ) , and amgS mutations upregulate the mexXY multidrug efflux system in the presence of aminoglycosides ( Lau et al . , 2015 ) . Resistance to carbapenem antibiotics in P . aeruginosa has emerged as an important clinical threat; for example , the WHO has designated carbapenem-resistant P . aeruginosa as a ‘critical priority’ for the development of new antibiotics . Interestingly , mobile integrons carrying multiple blaVIM-1 carbapenemase cassettes have been found in clinical isolates of P . aeruginosa ( San Millan et al . , 2015b ) , suggesting that cassette duplications may play an important role in clinical settings . To test this idea , we challenged 30 populations of P . aeruginosa carrying a plasmid ( pAMBL-1 ) , which contains an integron carrying a single copy of blaVIM-1 followed by aadB , with increasing doses of meropenem using a similar evolutionary ramp experiment ( Figure 5 ) . PCR screening of populations that survived at ×2 MIC identified numerous cassette re-arrangements of both the blaVIM-1 and aadB cassettes , with potential blaVIM-1 duplications occurring in all 14 surviving populations ( Figure 5 ) . Short-read sequencing of clones isolated from two of these populations confirmed the presence of duplications , as demonstrated by increased copy number of blaVIM-1 per plasmid ( 2 . 0 copies/plasmid [95% CI: 1 . 90–2 . 10] and 2 . 67 copies/plasmid [95% CI: 2 . 53–2 . 78] ) . Although it is not possible to definitely prove the role of the integrase without control populations lacking a functional integrase , these results strongly support the idea that ‘copy and paste’ outcomes of cassette re-arrangements can drive the rapid evolution of elevated carbapenem resistance and was the mechanism behind the blaVIM-1 amplification observed in the plasmid pAMBL2 isolated in the same hospital ( San Millan et al . , 2015b ) .
Mobile integrons are widespread genetic platforms involved in the interchange and expression of antibiotic resistance cassettes in bacteria . Antibiotic-induced cassette re-shuffling mediated by the SOS response ( Barraud and Ploy , 2015; Cambray et al . , 2011; Guerin et al . , 2009 ) has been proposed to increase bacterial evolvability by providing ‘adaptation on demand’ to newly encountered antibiotics ( Engelstädter et al . , 2016; Escudero et al . , 2015 ) . We tested this hypothesis by quantifying the impact of integrase activity on adaptation to increasing doses of gentamicin in populations of P . aeruginosa carrying a customized integron on a broad host range plasmid . Crucially , integrase activity accelerated the evolution of gentamicin resistance through rapid and repeated re-shuffling of the aadB resistance cassette , providing experimental support for the ‘adaptation on demand’ hypothesis . We observed a diversity of cassette re-arrangements as a result of integrase activity , whose diverse prevalences can help us understand the evolutionary dynamics underlying cassette shuffling . Cassette shuffling and duplication were extremely frequent , both with the aadB and the blaVIM-1 cassettes . The semi-conservative nature of cassette excision ( Escudero et al . , 2015 ) implies that cassette re-shuffling can lead to either ‘cut and paste’ or ‘copy and paste’ outcomes . Strikingly , all of the aadB re-arrangements that we observed were ‘copy and paste’ outcomes , resulting in the duplication of aadB ( Figure 6 ) . A bias towards evolution by ‘copy and paste’ is expected if increased copy number of the cassette under selection is beneficial . In this case , the benefit provided by aadB is strongly dependent on position , suggesting that ‘copy and paste’ insertions are unlikely to have provided stronger benefits than ‘cut and paste’ re-arrangements in first position . Furthermore , we only observe a slight advantage provided by aadB-aadB as compared to aadB arrays , suggesting that secondary copies of aadB were mostly redundant in arrays containing an aadB cassette in first position . The presence of multiple integrons within the same cell may also create an apparent bias towards ‘copy and paste’ re-shuffling: if re-insertion of an excised cassette is equally likely between all the integron copies present in a cell , we would expect the chance of an excised aadB cassette to re-insert into its original array to be 25–30% given the copy number of the R388 plasmid ( 2–3 per cell [Fernández-López et al . , 2006] ) . However , the absence of any ‘cut and paste’ products in our experiments , instead of the expected 30% , suggests that the integrase enzyme may be inherently biased towards ‘copy and paste’ cassette re-arrangement , potentially by favoring the reinsertion of an excised cassette into the conserved array . Although duplicated cassettes are relatively common in large chromosomal integrons , such as the V . cholerae superintegron ( Escudero et al . , 2015 ) , they are rarely found in mobile integrons ( San Millan et al . , 2015b; Stokes and Hall , 1992 ) . For example , duplicate cassettes are only found in 5% of the integrons that contain two or more cassettes in the INTEGRALL database ( Moura et al . , 2009; Supplementary file 4 ) . Difficulties associated with resolving duplications from short-read sequencing data probably contribute to this ( Alkan et al . , 2011 ) , but it is clear that duplicate cassettes are rare . Interestingly , the duplicate copies of selected cassettes created by ‘copy and paste’ re-shuffling facilitate the loss of redundant cassettes . First , duplication of cassettes with highly recombinogenic attC sites , such as aadB , facilitates the integrase-mediated excision of redundant cassettes . In this case , the insertion of aadB in first position created the opportunity for the loss of the blaVEB-1-dfrA5 cassettes , through an attCaadB × attCdfrA5 reaction , or the excision of blaVEB-1-dfrA5-aadB , through an attCaadB x attCaadB recombination . Homologous recombination between duplicate cassettes or between copies of integrons on different plasmids provides a second mechanism for integron degeneration ( Andersson and Hughes , 2009 ) , in this case resulting in the formation of a single copy of the duplicate gene . This also highlights how extremely mobile cassettes , such as aadB , can compensate for the lack of mobility of other less recombinogenic cassettes , like blaVEB-1 ( Aubert et al . , 2012 ) . This bias toward integron degeneration may be detrimental over the long term by leading to the loss of potentially useful cassettes but may also promote the deletion of redundant cassettes with low excision rate , which could not otherwise be easily excised by the integrase . The constitutive expression of cassettes from the Pc promoter ensures that redundant cassettes impose a fitness cost ( Lacotte et al . , 2017 ) , implying that selection for cassette loss is likely to be common . Given this , we argue that semi-conservative nature of cassette excision is key to the evolutionary benefits of integrase activity , as it allows integrons to rapidly gain additional copies of selected cassettes and facilitates the subsequent elimination of costly redundant cassettes . In our experiments , mutations in the chromosome and in the integron made an important contribution to resistance evolution and are key to understanding the effective integrase evolutionary benefits . For example , integrase activity was associated with the loss of the redundant blaVEB-1 cassette , which is in line with the integrase-mediated loss of redundant cassettes during selection for elevated chloramphenicol resistance observed by Barraud and Ploy , 2015 . However , the blaVEB-1 cassette was also rapidly inactivated by mutations in ΔintI1 populations . While integrase activity offers additional evolutionary pathways to alleviate the cost of blaVEB-1 , the presence of numerous mutational targets achieving the same effect may offset the observable evolutionary benefit of the integrase . Similarly , mutations in the promoter of the aadB cassette provided an alternative mechanism to increase the expression of this cassette that did not depend on integrase activity . Moreover , P . aeruginosa has a strong potential to adapt to aminoglycosides via chromosomal mutations ( López-Causapé et al . , 2018; Schurek et al . , 2008 ) . Given our populations possess a sizable evolvability potential through mutations , even in the absence of integrase activity , we hypothesize an even stronger impact of the integrase should be observable in environments or species where evolution possibilities through mutations are more limited . Finally , we detected an interesting interplay between chromosomal evolution and integrase activity , with a divergence in the evolution of the two genotypes at the early ×4 MIC time point , which is potentially the most clinically relevant . Rapid chromosomal and plasmid evolution via mutations across a wide range of target genes allowed populations lacking a functional integrase to evolve resistance to increasing concentrations of gentamicin . The wild-type populations , on the other hand , showed a much-reduced mutation prevalence , as chromosomal mutations were potentially out-competed by the frequent and more efficient cassette re-arrangements . This accelerated chromosomal evolution in ΔintI1A3 populations was short lived , confirming that rapid evolution was driven by more effective selection for mutations in populations lacking an integrase , rather than a difference in mutation rate per se . All populations that were able to survive very high levels of gentamicin exposure evolved by mutations of a common suite of target genes involved in antibiotic efflux and ribosomal modification , highlighting the fact that integrase activity did not ultimately alter the mutational routes to high-level resistance , but can impact chromosomal evolution during the early evolution of resistance . Systems that upregulate the mutation rate under stress are widespread in bacteria ( Foster , 2011; MacLean et al . , 2013 ) , and beneficial mutations generated by these systems can accelerate adaptation to stress ( but see Torres-Barceló et al . , 2015 ) . However , most of the mutations generated by these systems will be deleterious , and stress-induced mutagenesis will therefore tend to reduce fitness , particularly if the deleterious effects of mutations are exacerbated by stress ( Kishony and Leibler , 2003 ) . The integron integrase is known to have off-target activity ( Recchia et al . , 1994 ) , and it has been argued that the costs associated with off-target recombination contribute to the cost of integrase expression ( Harms et al . , 2013 ) , limiting the evolutionary benefit of this system ( Engelstädter et al . , 2016 ) . Although our sequencing strategy had limited ability to detect rare re-arrangements ( we imposed a cut-off of >30% prevalence ) in any particular population , we compensated for this limitation by sequencing many replicate populations samples . We found no evidence of chromosomal re-arrangements or increased mutation rates that could be linked to integrase activity , with the exception of a single deletion in the R388 plasmid that was suggestive of integrase activity , suggesting a low rate of off-target recombination . It could be argued that the strong selection and regular population bottlenecking imposed by our experiment is likely to have been effective at purging populations from rare neutral and weakly deleterious variants produced by off-target integrase activity . However , a high rate of off-target integrase activity could have also resulted in an increased rate of fixation of mutations and/or re-arrangements , as a result of either positive selection ( beneficial mutations produced by integrase activity ) or hitch-hiking ( neutral integrase-produced mutations linked to other beneficial mutations ) . In summary , our results support the idea that integrase activity accelerates evolution by creating high levels of variation exclusively in a region of the genome containing genes involved in response to stress , allowing bacteria to benefit from increased diversity without compromising genomic integrity . This contrasts with stress-induced mutagenesis , where the vast majority of mutations occur in genes that are not related to stress response , but further work is required to improve our understanding of the rate and spectrum of off-target recombination generated by integrase activity . In conclusion , our study supports the view that integrons provide bacteria with an incredible opportunity to evolve in response to new antibiotic challenges by rapidly optimizing the expression of cassettes . Integrase activity allows bacteria to modulate cassette expression , rapidly gain additional copies of selected cassettes and eliminate redundant cassettes , while the high specificity of integrase-mediated recombination maintains genomic integrity , minimizing the costs of integrase activity . Given the importance of cassette re-shuffling , we argue that integrase activity will accelerate resistance evolution most for highly mobile cassettes that display strong positional effects , such as aadB . Given this , we argue that cassette re-shuffling will be most important in cases where bacteria have limited ability to adapt to antibiotics , for example when only a small number of mutations can increase resistance , or where resistance mutations carry large fitness costs . Integrase activity also provides bacteria with the opportunity to capture new resistance cassettes , an important challenge for future work will be to study the evolutionary processes driving cassette acquisition . Our work also supports the view that treatment strategies should seek to target integrons , for example by combining antibiotics with adjuvants that limit integrase activity by inhibiting the SOS response ( Hocquet et al . , 2012 ) , or by using combinations of antibiotics that impose conflicting selective pressures on the integron .
A complete list of strains and plasmids can be found in Supplementary file 1 . Unless stated , bacteria cultures were grown overnight at 37°C with shaking in Luria-Bertani ( LB ) Miller broth ( Sigma Aldrich ) and supplemented with 100 mg/L of ceftazidime when required to select for the integron-bearing plasmids . Six integron arrays covering all possible cassette orders were created using the plasmid R388 ( Avila and de la Cruz , 1988 ) as backbone and three resistance cassettes: aadB , blaVEB-1 , and dfrA5 . The blaVEB-1 and aadB cassettes were amplified from the integron of the E . coli MG-1 clinical isolate ( Poirel et al . , 1999 ) , while the dfrA5 cassette was obtained from an enteroinvasive E . coli strain isolated in Senegal ( Gassama et al . , 2004 ) . These cassettes were then assembled into custom integron arrays using Gibson assembly and inserted into the plasmid R388 while replacing its original dhfr–orf9–qacEΔ1–sul1 integron array ( Fernández-López et al . , 2006 ) . The original R388 strong PcS promoter variant ( high cassette expression but low integrase activity [Jové et al . , 2010] ) was replaced by the weaker PcW promoter from a clinical isolate to guarantee high integrase activity and represent the promoter most commonly found in class 1 integrons ( Jové et al . , 2010 ) . ΔintI1 mutants of these custom integrons were created by introducing a 948 bp deletion of the integrase IntI1 gene during array construction , deleting most of the integrase open-reading frame but conserving both the Pint and Pc promoters . The final arrays were then first transformed into chemically competent E . coli MG1655 . These plasmids were then conjugated into P . aeruginosa through filter mating using the previous E . coli strains as donors and PA01 as recipient . Bacteria were incubated overnight in LB broth with 100 mg/L of carbenicillin at 37°C for the donors and in LB Miller broth without antibiotic at 42°C for the recipient bacteria . The next day cells were spun down , washed , and re-suspended in LB broth , before mixing in a 1:4 donor to acceptor ratio . The mix , as well as pure donor and acceptor controls , were put on filters placed on LB agar without antibiotics and incubated at 37°C overnight . Afterwards , filters were placed in tubes containing LB media and agitated . The resulting supernatants were plated on LB agar supplemented with 50 mg/L of kanamycin and 25 mg/L of ceftazidime and incubated for 48 hr . As P . aeruginosa PA01 has a higher innate resistance to kanamycin than E . coli MG1655 due to a chromosomally encoded phosphotransferase ( Okii et al . , 1983 ) , kanamycin was used to select against the E . coli donors , while ceftazidime was used to select for the plasmid in the P . aeruginosa transconjugants . The final colonies were controlled by PCR for the presence of the plasmid and the absence of E . coli DNA . MIC were determined in cation-adjusted Mueller-Hinton Broth 2 ( MH2 ) , following the broth microdilution method from the Clinical and Laboratory Standards Institute guidelines ( CLSI , 2017 ) . Briefly 5×105 c . f . u bacteria inocula were prepared using individual colonies grown on selective agar in interlaced twofold-increasing concentrations of antibiotics and incubated for 20 hr in a shaking incubator at 37°C . The next day , plates’ optical density ( OD595 ) was read using a Biotek Synergy two-plate reader . Wells were considered empty when the overall was under 0 . 1 , and the MIC for each assay was defined as the minimal concentration in which growth was inhibited in all three technical replicates ( separate wells , but grown on the same day from the same inoculum ) . The final MICs values are the average of two to four replicate assays ( from separately prepared inocula , on different days ) . As antibiotics’ MICs vary depending on the size of the starting inoculum ( Brook , 1989 ) , we first determined MIC in densities similar to the experimental evolution experiment ( further called MICexp ) . Overnight cultures inoculated from two to three morphologically similar colonies grown on selective agar were incubated for 20 hr with shaking in MH2 media with antibiotics . These overnight cultures were then diluted 1/10 , 000 and supplemented with doubling concentrations of gentamicin in three replicates . MICexp were determined the next day after 20 hr of incubation using OD595 measurements . This process was repeated twice . In these conditions , the MICexp for PA01:WTA3 and PAO1:ΔintI1A3 were identical at 24 mg/L . At the start of the experiment , 90 individual colonies grown on selective agar of each strain ( PA01:WTA3 and PA01:ΔintI1A3 ) were inoculated in 200 μL of MH2 media supplemented with gentamicin at a concentration of 1/8 MICexp . WT and ΔintI1 strains were placed in a chequerboard pattern by interlacing the different genotypes to control easily for cross-contamination . Wells at the edge of every plate were kept bacteria free to avoid edge effects and identify contaminations . These 90 populations were passaged every day with a 1/10 , 000 dilution factor , and the antibiotic concentration was doubled , starting at 1/8 MICexp until reaching a concentration of ×1024 MICexp . Alongside these 90 populations per strain which were transferred in increasing antibiotic concentrations , 30 populations per strain were transferred as controls in constant conditions: 15 without antibiotic and 15 at a constant concentration of 1/8 MICexp . Each population’s OD595 was measured each day and a population was considered extinct when its OD595 fell below 0 . 1 after 20 hr of incubation . All populations were frozen in 15% glycerol every 2 days . Liquid cultures were grown from the frozen stock of all surviving PA01:WTA3 and PA01:ΔintI1 A3 populations at ×1024 MICexp in LB Miller media supplemented with gentamicin at ×128 MICexp , and were incubated for 24 hr with shaking . Six populations were inoculated from each control treatment in either LB Miller supplemented with a concentration of MICexp or with no antibiotic . For the ×4 MICexp time point , 26 populations of each PA01:WTA3 and PA01:ΔA3 genotype were regrown in ×2 MICexp concentration of gentamicin . Ancestral PA01:WTA3 and PA01:ΔintI1A3 populations were incubated in 100 mg/L of ceftazidime from the initial frozen stock . DNA extractions of the whole populations were carried out using the DNeasy Blood and Tissue Kit ( Qiagen ) on the QiaCube extraction platform ( Qiagen ) combined with RNAse treatment . DNA concentrations were determined using the Quantifluor dsDNA system ( Promega ) . At the ×1024 and ×4 MICexp transfers all surviving populations were controlled for cross-contamination by verifying the size of the integrase by PCR ( primers given in Supplementary file 1 ) . Starting materials were either 2 μL of extracted DNA ( ×1024 MICexp time point ) or 2 μL of inoculate previously incubated for 24 hr and then boiled for 10 min at 105°C ( ×4 MICexp time point ) . PCR were carried out using the GoTaq G2 DNA mastermix ( Promega ) and the following protocol: 30 s at 95°C , 30 s at 55°C , 3 min at 72°C for 30 cycles . Plate mishandling during the transfers resulted in the contamination of 40 and 34 wells of 90 ramping populations for each array . Areas of the plates where cross-contamination was detected in several wells in close proximity were excluded from the rest of the analysis , for a final population number of 65 per strain . The final populations at ×1024 MICexp were analyzed by PCR to determine the position of the aadB cassette relative to the start and the end of the array as well as identify any aadB duplications or inversions and deletions of the plasmid backbone . Library preparation and next-generation sequencing using the NovaSeq 6000 Sequencing System ( Illumina ) were carried out at the Oxford Genomics Centre at the Wellcome Centre for Human Genetics . Twenty-two PA01:WTA3 and 6 PA01:ΔintI1A3 populations were sequenced from the ×1024 MICexp time point . For each control treatment , six populations were pooled together and sequenced as one . For the ×4 MICexp time point , 26 PA01:WTA3 and 26 PA01:ΔintI1A3 were sequenced . PCR duplications and optical artifacts were removed using MarkDuplicates ( Broad Institute , 2019 ) and then low-quality bases and adaptors were trimmed from the sequencing reads using Trimmomatic v0 . 39 ( Bolger et al . , 2014 ) . Finally , overall read quality control was performed using FastQC ( Simon Andrews , 2010 ) and multiQC ( Ewels et al . , 2016 ) . During this process , one PA01:WTA3 sample from the ×1024 MIC time point was removed due to the presence of non-Pseudomonas DNA . Average coverage depth was 760 ( s . d . = 258 ) for the plasmid and 171 ( s . d . = 24 ) for the chromosome . Single-nucleotide polymorphism ( SNP ) , point insertion , and deletion identification were performed using the breseq ( Barrick et al . , 2014 ) pipeline in polymorphism mode . For each population , reads were first mapped to the P . aeruginosa PA01 complete genome NC_ 002516 . 2 and the predicted sequence of WTA3 . Non-mapped reads from the unevolved PA01:WTA3 population were then assembled de novo using SPAdes ( Bankevich et al . , 2012 ) , and any open-reading frame identified using Prokka ( Seemann , 2014 ) and further used as an additional reference to map unaligned reads from the other populations . The pipeline output was then further processed in MATLAB ( MathWorks ) . Variants present in the unevolved ancestor populations at any frequency were filtered out . We also excluded variants reaching a frequency of less than 30% within a single population . A 5% threshold was applied to the pooled controls , which allows the detection of any variant present in more than 30% of a single population . Final results were exported in table format and processed for visualization using Circos ( Krzywinski et al . , 2009 ) and Geneious ( Biomatters ) . Apart from the expected intI1 deletion , the PA01:ΔintI1A3 genome was shown to differ from PA01:WTA3 and PA01 by two SNPs likely to have arisen during the conjugation process: one in PA3734 and one in the phzM/phzA1 intergenic region . PA3734 is predicted to be a lipase involved in cell-wall metabolism ( Dettman et al . , 2015 ) and may be involved in quorum-sensing ( Levesque , 2006 ) , while phzM and phzA1 are involved in the production of pyocyanins ( Higgins et al . , 2018 ) . No literature linking those genes to aminoglycoside resistance was identified . Four PA01:ΔintI1A3 samples and one PA01:WTA3 sample from the ×4 MIC time point were removed from the analysis due to a wrong or mixed genotype from potential mislabeling or mishandling during DNA processing , leading to a final genomic dataset of 22 PA01:ΔintI1A3 and 24 PA01:WTA3 populations at ×4 MIC . All samples from the ×1024 MIC time point were of the correct genotype . Potential new junctions between distant regions of the reference genome were identified through the breseq software on the plasmid and on the chromosome ( Barrick et al . , 2014 ) . We detected any re-arrangement present in more than 5% of a population , but kept only re-arrangements above 30% for further analysis , which excluded four re-arrangements , all deletions of less than 1 kb . Copy number variants were identified with CNOGpro ( Brynildsrud , 2018 ) and used to confirm potential duplications or large-scale deletions . Finally , de novo assembly of the plasmids using plasmidSPAdes ( Antipov et al . , 2016 ) was carried out to provide additional evidence for the cassette rearrangements and visualized using Bandage ( Wick et al . , 2015 ) . The robustness of the detection of cassette re-arrangement from the sequencing data was tested by cross-referencing the predicted recombinations with the results from the PCR screen at ×1024 MIC: all predicted recombinations matched the bands of the PCR screen , and only two extraneous bands could not be explained in three populations ( Figure 3—figure supplement 2 ) , confirming the robustness of the bioinformatic analysis . The MICs of 15 PA01:WTA3 and 5 PA01:ΔintI1A3 evolved populations from the final ×1024 MIC time point were determined as described above using individual colonies plated on in LB Miller media supplemented with gentamicin at ×128 MICexp from the frozen stocks . To create populations containing the evolved plasmids in the ancestral chromosomal background , the plasmids of 15 PA01:WTA3 and 2 PA01:ΔintI1A3 ×1024 MIC populations containing a single type of array were first extracted using the QIAprep Miniprep ( Qiagen ) on the QiaCube extraction platform ( Qiagen ) from liquid culture in LB Miller supplemented with gentamicin at x64 MICexp . These plasmids were then electroporated back into the ancestral PA01 background using the protocol described in Choi and Schweizer , 2006 and plated on agar supplemented with gentamicin at 24 mg/L . The presence of the plasmid was controlled by PCR targeting the integrase , and MICs were performed following the protocol described above . Thirty colonies of PA01:pAMBL1 were inoculated in 100 μL of MH2 broth and transferred every day in doubling concentrations of meropenem with a 1/10 , 000 dilution and frozen in 15% glycerol every other day . Population survival was monitored each day after reaching a concentration of ×1 MICexp by plating every well on a MH2 agar plate without antibiotic using a pin replicator . Extinction of a population was defined as the absence of a visible colony after 24 hr incubation at 37°C . Surviving populations at ×2 MICexp were grown on a MH2 agar plate without antibiotic and used as substrate for PCR . Primers were used to identify potential duplications of the blaVIM cassette by PCR . PCR were carried out using the GoTaq G2 DNA mastermix ( Promega ) and the following protocol: 30 s at 95°C , 30 s at 55°C , 3 min at 72°C for 30 cycles . Single clones from two different populations were sequenced through whole genome sequencing and analyzed using the same protocol as described previously . Statistical analysis was carried out using R ( version 3 . 6 . 1 ) and RStudio ( Version 1 . 2 . 5 ) . Survival analysis using the log-rank test was performed using the survival ( Therneau , 2020 ) package to compare survival rates between populations with and without a functional integrase . | From urinary tract infections to bacterial pneumonia , many diseases can now be treated through a course of antibiotics . Yet bacteria have evolved to respond to this threat , gaining new antibiotic resistance genes that allow them to evade the drugs . Addressing this growing issue requires to either discover new antibiotics , or to stop resistance before it emerges – a strategy that can only work if scientists know exactly how this mechanism takes place . For bacteria , it is a waste of resources to produce the proteins that confer resistance if antibiotics are absent . In fact , doing so can decrease their chance to survive and reproduce . A genetic element known as an integron can help to manage that burden . This piece of genetic information is formed of a succession of ‘cassettes’ containing antibiotic resistance genes . More proteins are made from the genes present at the start of the integron , compared to the ones towards the end . When bacteria encounter antibiotics , an enzyme called integrase is activated , allowing the organisms to shuffle the order of their cassettes in the integron . It is thought – but not yet proven – that this mechanism helps bacteria to activate their resistance ‘on demand’ . To find out , Souque et al . engineered the bacteria Pseudomonas aeruginosa to carry a custom integron with three cassettes , each helping the organism to resist to a different antibiotic . In addition , only half of the bacteria had a working integrase and could therefore shuffle their gene cassettes . The organisms were then exposed to an increasing amount of the antibiotics for which the cassette in the last position provided resistance . The bacteria with a working integrase survived longer than those without , as they were able to shuffle their cassettes and move the useful antibiotic resistance gene into top position . In addition , the cassettes carrying the genes to resist to other types of antibiotics were excised from the genetic information and lost . Understanding integrons could guide future antibiotic treatment strategies , for instance by combining antibiotics with chemicals that block integrase activity . It might also be possible to force bacteria to delete resistance cassettes by cycling through different antibiotics . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology",
"microbiology",
"and",
"infectious",
"disease"
] | 2021 | Integron activity accelerates the evolution of antibiotic resistance |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.